if.c revision 249132
1/*-
2 * Copyright (c) 1980, 1986, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	@(#)if.c	8.5 (Berkeley) 1/9/95
30 * $FreeBSD: stable/9/sys/net/if.c 249132 2013-04-05 08:22:11Z mav $
31 */
32
33#include "opt_compat.h"
34#include "opt_inet6.h"
35#include "opt_inet.h"
36
37#include <sys/param.h>
38#include <sys/types.h>
39#include <sys/conf.h>
40#include <sys/malloc.h>
41#include <sys/sbuf.h>
42#include <sys/bus.h>
43#include <sys/mbuf.h>
44#include <sys/systm.h>
45#include <sys/priv.h>
46#include <sys/proc.h>
47#include <sys/socket.h>
48#include <sys/socketvar.h>
49#include <sys/protosw.h>
50#include <sys/kernel.h>
51#include <sys/lock.h>
52#include <sys/refcount.h>
53#include <sys/module.h>
54#include <sys/rwlock.h>
55#include <sys/sockio.h>
56#include <sys/syslog.h>
57#include <sys/sysctl.h>
58#include <sys/taskqueue.h>
59#include <sys/domain.h>
60#include <sys/jail.h>
61#include <sys/priv.h>
62
63#include <machine/stdarg.h>
64#include <vm/uma.h>
65
66#include <net/if.h>
67#include <net/if_arp.h>
68#include <net/if_clone.h>
69#include <net/if_dl.h>
70#include <net/if_types.h>
71#include <net/if_var.h>
72#include <net/radix.h>
73#include <net/route.h>
74#include <net/vnet.h>
75
76#if defined(INET) || defined(INET6)
77/*XXX*/
78#include <netinet/in.h>
79#include <netinet/in_var.h>
80#include <netinet/ip_carp.h>
81#ifdef INET6
82#include <netinet6/in6_var.h>
83#include <netinet6/in6_ifattach.h>
84#endif
85#endif
86#ifdef INET
87#include <netinet/if_ether.h>
88#endif
89
90#include <security/mac/mac_framework.h>
91
92#ifdef COMPAT_FREEBSD32
93#include <sys/mount.h>
94#include <compat/freebsd32/freebsd32.h>
95#endif
96
97struct ifindex_entry {
98	struct  ifnet *ife_ifnet;
99};
100
101SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
102SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
103
104TUNABLE_INT("net.link.ifqmaxlen", &ifqmaxlen);
105SYSCTL_INT(_net_link, OID_AUTO, ifqmaxlen, CTLFLAG_RDTUN,
106    &ifqmaxlen, 0, "max send queue size");
107
108/* Log link state change events */
109static int log_link_state_change = 1;
110
111SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
112	&log_link_state_change, 0,
113	"log interface link state change events");
114
115/* Interface description */
116static unsigned int ifdescr_maxlen = 1024;
117SYSCTL_UINT(_net, OID_AUTO, ifdescr_maxlen, CTLFLAG_RW,
118	&ifdescr_maxlen, 0,
119	"administrative maximum length for interface description");
120
121static MALLOC_DEFINE(M_IFDESCR, "ifdescr", "ifnet descriptions");
122
123/* global sx for non-critical path ifdescr */
124static struct sx ifdescr_sx;
125SX_SYSINIT(ifdescr_sx, &ifdescr_sx, "ifnet descr");
126
127void	(*bridge_linkstate_p)(struct ifnet *ifp);
128void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
129void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
130/* These are external hooks for CARP. */
131void	(*carp_linkstate_p)(struct ifnet *ifp);
132#if defined(INET) || defined(INET6)
133struct ifnet *(*carp_forus_p)(struct ifnet *ifp, u_char *dhost);
134int	(*carp_output_p)(struct ifnet *ifp, struct mbuf *m,
135    struct sockaddr *sa, struct rtentry *rt);
136#endif
137#ifdef INET
138int (*carp_iamatch_p)(struct ifnet *, struct in_ifaddr *, struct in_addr *,
139    u_int8_t **);
140#endif
141#ifdef INET6
142struct ifaddr *(*carp_iamatch6_p)(struct ifnet *ifp, struct in6_addr *taddr6);
143caddr_t (*carp_macmatch6_p)(struct ifnet *ifp, struct mbuf *m,
144    const struct in6_addr *taddr);
145#endif
146
147struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
148
149/*
150 * XXX: Style; these should be sorted alphabetically, and unprototyped
151 * static functions should be prototyped. Currently they are sorted by
152 * declaration order.
153 */
154static void	if_attachdomain(void *);
155static void	if_attachdomain1(struct ifnet *);
156static int	ifconf(u_long, caddr_t);
157static void	if_freemulti(struct ifmultiaddr *);
158static void	if_init(void *);
159static void	if_grow(void);
160static void	if_route(struct ifnet *, int flag, int fam);
161static int	if_setflag(struct ifnet *, int, int, int *, int);
162static int	if_transmit(struct ifnet *ifp, struct mbuf *m);
163static void	if_unroute(struct ifnet *, int flag, int fam);
164static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
165static int	if_rtdel(struct radix_node *, void *);
166static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
167static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
168static void	do_link_state_change(void *, int);
169static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
170static int	if_getgroupmembers(struct ifgroupreq *);
171static void	if_delgroups(struct ifnet *);
172static void	if_attach_internal(struct ifnet *, int);
173static void	if_detach_internal(struct ifnet *, int);
174
175#ifdef INET6
176/*
177 * XXX: declare here to avoid to include many inet6 related files..
178 * should be more generalized?
179 */
180extern void	nd6_setmtu(struct ifnet *);
181#endif
182
183VNET_DEFINE(int, if_index);
184int	ifqmaxlen = IFQ_MAXLEN;
185VNET_DEFINE(struct ifnethead, ifnet);	/* depend on static init XXX */
186VNET_DEFINE(struct ifgrouphead, ifg_head);
187
188static VNET_DEFINE(int, if_indexlim) = 8;
189
190/* Table of ifnet by index. */
191VNET_DEFINE(struct ifindex_entry *, ifindex_table);
192
193#define	V_if_indexlim		VNET(if_indexlim)
194#define	V_ifindex_table		VNET(ifindex_table)
195
196/*
197 * The global network interface list (V_ifnet) and related state (such as
198 * if_index, if_indexlim, and ifindex_table) are protected by an sxlock and
199 * an rwlock.  Either may be acquired shared to stablize the list, but both
200 * must be acquired writable to modify the list.  This model allows us to
201 * both stablize the interface list during interrupt thread processing, but
202 * also to stablize it over long-running ioctls, without introducing priority
203 * inversions and deadlocks.
204 */
205struct rwlock ifnet_rwlock;
206struct sx ifnet_sxlock;
207
208/*
209 * The allocation of network interfaces is a rather non-atomic affair; we
210 * need to select an index before we are ready to expose the interface for
211 * use, so will use this pointer value to indicate reservation.
212 */
213#define	IFNET_HOLD	(void *)(uintptr_t)(-1)
214
215static	if_com_alloc_t *if_com_alloc[256];
216static	if_com_free_t *if_com_free[256];
217
218static MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
219MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
220MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
221
222struct ifnet *
223ifnet_byindex_locked(u_short idx)
224{
225
226	if (idx > V_if_index)
227		return (NULL);
228	if (V_ifindex_table[idx].ife_ifnet == IFNET_HOLD)
229		return (NULL);
230	return (V_ifindex_table[idx].ife_ifnet);
231}
232
233struct ifnet *
234ifnet_byindex(u_short idx)
235{
236	struct ifnet *ifp;
237
238	IFNET_RLOCK_NOSLEEP();
239	ifp = ifnet_byindex_locked(idx);
240	IFNET_RUNLOCK_NOSLEEP();
241	return (ifp);
242}
243
244struct ifnet *
245ifnet_byindex_ref(u_short idx)
246{
247	struct ifnet *ifp;
248
249	IFNET_RLOCK_NOSLEEP();
250	ifp = ifnet_byindex_locked(idx);
251	if (ifp == NULL || (ifp->if_flags & IFF_DYING)) {
252		IFNET_RUNLOCK_NOSLEEP();
253		return (NULL);
254	}
255	if_ref(ifp);
256	IFNET_RUNLOCK_NOSLEEP();
257	return (ifp);
258}
259
260/*
261 * Allocate an ifindex array entry; return 0 on success or an error on
262 * failure.
263 */
264static int
265ifindex_alloc_locked(u_short *idxp)
266{
267	u_short idx;
268
269	IFNET_WLOCK_ASSERT();
270
271retry:
272	/*
273	 * Try to find an empty slot below V_if_index.  If we fail, take the
274	 * next slot.
275	 */
276	for (idx = 1; idx <= V_if_index; idx++) {
277		if (V_ifindex_table[idx].ife_ifnet == NULL)
278			break;
279	}
280
281	/* Catch if_index overflow. */
282	if (idx < 1)
283		return (ENOSPC);
284	if (idx >= V_if_indexlim) {
285		if_grow();
286		goto retry;
287	}
288	if (idx > V_if_index)
289		V_if_index = idx;
290	*idxp = idx;
291	return (0);
292}
293
294static void
295ifindex_free_locked(u_short idx)
296{
297
298	IFNET_WLOCK_ASSERT();
299
300	V_ifindex_table[idx].ife_ifnet = NULL;
301	while (V_if_index > 0 &&
302	    V_ifindex_table[V_if_index].ife_ifnet == NULL)
303		V_if_index--;
304}
305
306static void
307ifindex_free(u_short idx)
308{
309
310	IFNET_WLOCK();
311	ifindex_free_locked(idx);
312	IFNET_WUNLOCK();
313}
314
315static void
316ifnet_setbyindex_locked(u_short idx, struct ifnet *ifp)
317{
318
319	IFNET_WLOCK_ASSERT();
320
321	V_ifindex_table[idx].ife_ifnet = ifp;
322}
323
324static void
325ifnet_setbyindex(u_short idx, struct ifnet *ifp)
326{
327
328	IFNET_WLOCK();
329	ifnet_setbyindex_locked(idx, ifp);
330	IFNET_WUNLOCK();
331}
332
333struct ifaddr *
334ifaddr_byindex(u_short idx)
335{
336	struct ifaddr *ifa;
337
338	IFNET_RLOCK_NOSLEEP();
339	ifa = ifnet_byindex_locked(idx)->if_addr;
340	if (ifa != NULL)
341		ifa_ref(ifa);
342	IFNET_RUNLOCK_NOSLEEP();
343	return (ifa);
344}
345
346/*
347 * Network interface utility routines.
348 *
349 * Routines with ifa_ifwith* names take sockaddr *'s as
350 * parameters.
351 */
352
353static void
354vnet_if_init(const void *unused __unused)
355{
356
357	TAILQ_INIT(&V_ifnet);
358	TAILQ_INIT(&V_ifg_head);
359	IFNET_WLOCK();
360	if_grow();				/* create initial table */
361	IFNET_WUNLOCK();
362	vnet_if_clone_init();
363}
364VNET_SYSINIT(vnet_if_init, SI_SUB_INIT_IF, SI_ORDER_SECOND, vnet_if_init,
365    NULL);
366
367/* ARGSUSED*/
368static void
369if_init(void *dummy __unused)
370{
371
372	IFNET_LOCK_INIT();
373	if_clone_init();
374}
375SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL);
376
377
378#ifdef VIMAGE
379static void
380vnet_if_uninit(const void *unused __unused)
381{
382
383	VNET_ASSERT(TAILQ_EMPTY(&V_ifnet), ("%s:%d tailq &V_ifnet=%p "
384	    "not empty", __func__, __LINE__, &V_ifnet));
385	VNET_ASSERT(TAILQ_EMPTY(&V_ifg_head), ("%s:%d tailq &V_ifg_head=%p "
386	    "not empty", __func__, __LINE__, &V_ifg_head));
387
388	free((caddr_t)V_ifindex_table, M_IFNET);
389}
390VNET_SYSUNINIT(vnet_if_uninit, SI_SUB_INIT_IF, SI_ORDER_FIRST,
391    vnet_if_uninit, NULL);
392#endif
393
394static void
395if_grow(void)
396{
397	int oldlim;
398	u_int n;
399	struct ifindex_entry *e;
400
401	IFNET_WLOCK_ASSERT();
402	oldlim = V_if_indexlim;
403	IFNET_WUNLOCK();
404	n = (oldlim << 1) * sizeof(*e);
405	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
406	IFNET_WLOCK();
407	if (V_if_indexlim != oldlim) {
408		free(e, M_IFNET);
409		return;
410	}
411	if (V_ifindex_table != NULL) {
412		memcpy((caddr_t)e, (caddr_t)V_ifindex_table, n/2);
413		free((caddr_t)V_ifindex_table, M_IFNET);
414	}
415	V_if_indexlim <<= 1;
416	V_ifindex_table = e;
417}
418
419/*
420 * Allocate a struct ifnet and an index for an interface.  A layer 2
421 * common structure will also be allocated if an allocation routine is
422 * registered for the passed type.
423 */
424struct ifnet *
425if_alloc(u_char type)
426{
427	struct ifnet *ifp;
428	u_short idx;
429
430	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
431	IFNET_WLOCK();
432	if (ifindex_alloc_locked(&idx) != 0) {
433		IFNET_WUNLOCK();
434		free(ifp, M_IFNET);
435		return (NULL);
436	}
437	ifnet_setbyindex_locked(idx, IFNET_HOLD);
438	IFNET_WUNLOCK();
439	ifp->if_index = idx;
440	ifp->if_type = type;
441	ifp->if_alloctype = type;
442	if (if_com_alloc[type] != NULL) {
443		ifp->if_l2com = if_com_alloc[type](type, ifp);
444		if (ifp->if_l2com == NULL) {
445			free(ifp, M_IFNET);
446			ifindex_free(idx);
447			return (NULL);
448		}
449	}
450
451	IF_ADDR_LOCK_INIT(ifp);
452	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
453	ifp->if_afdata_initialized = 0;
454	IF_AFDATA_LOCK_INIT(ifp);
455	TAILQ_INIT(&ifp->if_addrhead);
456	TAILQ_INIT(&ifp->if_prefixhead);
457	TAILQ_INIT(&ifp->if_multiaddrs);
458	TAILQ_INIT(&ifp->if_groups);
459#ifdef MAC
460	mac_ifnet_init(ifp);
461#endif
462	ifq_init(&ifp->if_snd, ifp);
463
464	refcount_init(&ifp->if_refcount, 1);	/* Index reference. */
465	ifnet_setbyindex(ifp->if_index, ifp);
466	return (ifp);
467}
468
469/*
470 * Do the actual work of freeing a struct ifnet, and layer 2 common
471 * structure.  This call is made when the last reference to an
472 * interface is released.
473 */
474static void
475if_free_internal(struct ifnet *ifp)
476{
477
478	KASSERT((ifp->if_flags & IFF_DYING),
479	    ("if_free_internal: interface not dying"));
480
481	if (if_com_free[ifp->if_alloctype] != NULL)
482		if_com_free[ifp->if_alloctype](ifp->if_l2com,
483		    ifp->if_alloctype);
484
485#ifdef MAC
486	mac_ifnet_destroy(ifp);
487#endif /* MAC */
488	if (ifp->if_description != NULL)
489		free(ifp->if_description, M_IFDESCR);
490	IF_AFDATA_DESTROY(ifp);
491	IF_ADDR_LOCK_DESTROY(ifp);
492	ifq_delete(&ifp->if_snd);
493	free(ifp, M_IFNET);
494}
495
496/*
497 * This version should only be called by intefaces that switch their type
498 * after calling if_alloc().  if_free_type() will go away again now that we
499 * have if_alloctype to cache the original allocation type.  For now, assert
500 * that they match, since we require that in practice.
501 */
502void
503if_free_type(struct ifnet *ifp, u_char type)
504{
505
506	KASSERT(ifp->if_alloctype == type,
507	    ("if_free_type: type (%d) != alloctype (%d)", type,
508	    ifp->if_alloctype));
509
510	ifp->if_flags |= IFF_DYING;			/* XXX: Locking */
511
512	IFNET_WLOCK();
513	KASSERT(ifp == ifnet_byindex_locked(ifp->if_index),
514	    ("%s: freeing unallocated ifnet", ifp->if_xname));
515
516	ifindex_free_locked(ifp->if_index);
517	IFNET_WUNLOCK();
518
519	if (!refcount_release(&ifp->if_refcount))
520		return;
521	if_free_internal(ifp);
522}
523
524/*
525 * This is the normal version of if_free(), used by device drivers to free a
526 * detached network interface.  The contents of if_free_type() will move into
527 * here when if_free_type() goes away.
528 */
529void
530if_free(struct ifnet *ifp)
531{
532
533	if_free_type(ifp, ifp->if_alloctype);
534}
535
536/*
537 * Interfaces to keep an ifnet type-stable despite the possibility of the
538 * driver calling if_free().  If there are additional references, we defer
539 * freeing the underlying data structure.
540 */
541void
542if_ref(struct ifnet *ifp)
543{
544
545	/* We don't assert the ifnet list lock here, but arguably should. */
546	refcount_acquire(&ifp->if_refcount);
547}
548
549void
550if_rele(struct ifnet *ifp)
551{
552
553	if (!refcount_release(&ifp->if_refcount))
554		return;
555	if_free_internal(ifp);
556}
557
558void
559ifq_init(struct ifaltq *ifq, struct ifnet *ifp)
560{
561
562	mtx_init(&ifq->ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
563
564	if (ifq->ifq_maxlen == 0)
565		ifq->ifq_maxlen = ifqmaxlen;
566
567	ifq->altq_type = 0;
568	ifq->altq_disc = NULL;
569	ifq->altq_flags &= ALTQF_CANTCHANGE;
570	ifq->altq_tbr  = NULL;
571	ifq->altq_ifp  = ifp;
572}
573
574void
575ifq_delete(struct ifaltq *ifq)
576{
577	mtx_destroy(&ifq->ifq_mtx);
578}
579
580/*
581 * Perform generic interface initalization tasks and attach the interface
582 * to the list of "active" interfaces.  If vmove flag is set on entry
583 * to if_attach_internal(), perform only a limited subset of initialization
584 * tasks, given that we are moving from one vnet to another an ifnet which
585 * has already been fully initialized.
586 *
587 * XXX:
588 *  - The decision to return void and thus require this function to
589 *    succeed is questionable.
590 *  - We should probably do more sanity checking.  For instance we don't
591 *    do anything to insure if_xname is unique or non-empty.
592 */
593void
594if_attach(struct ifnet *ifp)
595{
596
597	if_attach_internal(ifp, 0);
598}
599
600static void
601if_attach_internal(struct ifnet *ifp, int vmove)
602{
603	unsigned socksize, ifasize;
604	int namelen, masklen;
605	struct sockaddr_dl *sdl;
606	struct ifaddr *ifa;
607
608	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
609		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
610		    ifp->if_xname);
611
612#ifdef VIMAGE
613	ifp->if_vnet = curvnet;
614	if (ifp->if_home_vnet == NULL)
615		ifp->if_home_vnet = curvnet;
616#endif
617
618	if_addgroup(ifp, IFG_ALL);
619
620	getmicrotime(&ifp->if_lastchange);
621	ifp->if_data.ifi_epoch = time_uptime;
622	ifp->if_data.ifi_datalen = sizeof(struct if_data);
623
624	KASSERT((ifp->if_transmit == NULL && ifp->if_qflush == NULL) ||
625	    (ifp->if_transmit != NULL && ifp->if_qflush != NULL),
626	    ("transmit and qflush must both either be set or both be NULL"));
627	if (ifp->if_transmit == NULL) {
628		ifp->if_transmit = if_transmit;
629		ifp->if_qflush = if_qflush;
630	}
631
632	if (!vmove) {
633#ifdef MAC
634		mac_ifnet_create(ifp);
635#endif
636
637		/*
638		 * Create a Link Level name for this device.
639		 */
640		namelen = strlen(ifp->if_xname);
641		/*
642		 * Always save enough space for any possiable name so we
643		 * can do a rename in place later.
644		 */
645		masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
646		socksize = masklen + ifp->if_addrlen;
647		if (socksize < sizeof(*sdl))
648			socksize = sizeof(*sdl);
649		socksize = roundup2(socksize, sizeof(long));
650		ifasize = sizeof(*ifa) + 2 * socksize;
651		ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
652		ifa_init(ifa);
653		sdl = (struct sockaddr_dl *)(ifa + 1);
654		sdl->sdl_len = socksize;
655		sdl->sdl_family = AF_LINK;
656		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
657		sdl->sdl_nlen = namelen;
658		sdl->sdl_index = ifp->if_index;
659		sdl->sdl_type = ifp->if_type;
660		ifp->if_addr = ifa;
661		ifa->ifa_ifp = ifp;
662		ifa->ifa_rtrequest = link_rtrequest;
663		ifa->ifa_addr = (struct sockaddr *)sdl;
664		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
665		ifa->ifa_netmask = (struct sockaddr *)sdl;
666		sdl->sdl_len = masklen;
667		while (namelen != 0)
668			sdl->sdl_data[--namelen] = 0xff;
669		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
670		/* Reliably crash if used uninitialized. */
671		ifp->if_broadcastaddr = NULL;
672	}
673#ifdef VIMAGE
674	else {
675		/*
676		 * Update the interface index in the link layer address
677		 * of the interface.
678		 */
679		for (ifa = ifp->if_addr; ifa != NULL;
680		    ifa = TAILQ_NEXT(ifa, ifa_link)) {
681			if (ifa->ifa_addr->sa_family == AF_LINK) {
682				sdl = (struct sockaddr_dl *)ifa->ifa_addr;
683				sdl->sdl_index = ifp->if_index;
684			}
685		}
686	}
687#endif
688
689	IFNET_WLOCK();
690	TAILQ_INSERT_TAIL(&V_ifnet, ifp, if_link);
691#ifdef VIMAGE
692	curvnet->vnet_ifcnt++;
693#endif
694	IFNET_WUNLOCK();
695
696	if (domain_init_status >= 2)
697		if_attachdomain1(ifp);
698
699	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
700	if (IS_DEFAULT_VNET(curvnet))
701		devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
702
703	/* Announce the interface. */
704	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
705}
706
707static void
708if_attachdomain(void *dummy)
709{
710	struct ifnet *ifp;
711	int s;
712
713	s = splnet();
714	TAILQ_FOREACH(ifp, &V_ifnet, if_link)
715		if_attachdomain1(ifp);
716	splx(s);
717}
718SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
719    if_attachdomain, NULL);
720
721static void
722if_attachdomain1(struct ifnet *ifp)
723{
724	struct domain *dp;
725	int s;
726
727	s = splnet();
728
729	/*
730	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
731	 * cannot lock ifp->if_afdata initialization, entirely.
732	 */
733	if (IF_AFDATA_TRYLOCK(ifp) == 0) {
734		splx(s);
735		return;
736	}
737	if (ifp->if_afdata_initialized >= domain_init_status) {
738		IF_AFDATA_UNLOCK(ifp);
739		splx(s);
740		printf("if_attachdomain called more than once on %s\n",
741		    ifp->if_xname);
742		return;
743	}
744	ifp->if_afdata_initialized = domain_init_status;
745	IF_AFDATA_UNLOCK(ifp);
746
747	/* address family dependent data region */
748	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
749	for (dp = domains; dp; dp = dp->dom_next) {
750		if (dp->dom_ifattach)
751			ifp->if_afdata[dp->dom_family] =
752			    (*dp->dom_ifattach)(ifp);
753	}
754
755	splx(s);
756}
757
758/*
759 * Remove any unicast or broadcast network addresses from an interface.
760 */
761void
762if_purgeaddrs(struct ifnet *ifp)
763{
764	struct ifaddr *ifa, *next;
765
766	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
767		if (ifa->ifa_addr->sa_family == AF_LINK)
768			continue;
769#ifdef INET
770		/* XXX: Ugly!! ad hoc just for INET */
771		if (ifa->ifa_addr->sa_family == AF_INET) {
772			struct ifaliasreq ifr;
773
774			bzero(&ifr, sizeof(ifr));
775			ifr.ifra_addr = *ifa->ifa_addr;
776			if (ifa->ifa_dstaddr)
777				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
778			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
779			    NULL) == 0)
780				continue;
781		}
782#endif /* INET */
783#ifdef INET6
784		if (ifa->ifa_addr->sa_family == AF_INET6) {
785			in6_purgeaddr(ifa);
786			/* ifp_addrhead is already updated */
787			continue;
788		}
789#endif /* INET6 */
790		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
791		ifa_free(ifa);
792	}
793}
794
795/*
796 * Remove any multicast network addresses from an interface when an ifnet
797 * is going away.
798 */
799static void
800if_purgemaddrs(struct ifnet *ifp)
801{
802	struct ifmultiaddr *ifma;
803	struct ifmultiaddr *next;
804
805	IF_ADDR_WLOCK(ifp);
806	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
807		if_delmulti_locked(ifp, ifma, 1);
808	IF_ADDR_WUNLOCK(ifp);
809}
810
811/*
812 * Detach an interface, removing it from the list of "active" interfaces.
813 * If vmove flag is set on entry to if_detach_internal(), perform only a
814 * limited subset of cleanup tasks, given that we are moving an ifnet from
815 * one vnet to another, where it must be fully operational.
816 *
817 * XXXRW: There are some significant questions about event ordering, and
818 * how to prevent things from starting to use the interface during detach.
819 */
820void
821if_detach(struct ifnet *ifp)
822{
823
824	if_detach_internal(ifp, 0);
825}
826
827static void
828if_detach_internal(struct ifnet *ifp, int vmove)
829{
830	struct ifaddr *ifa;
831	struct radix_node_head	*rnh;
832	int i, j;
833	struct domain *dp;
834 	struct ifnet *iter;
835 	int found = 0;
836
837	IFNET_WLOCK();
838	TAILQ_FOREACH(iter, &V_ifnet, if_link)
839		if (iter == ifp) {
840			TAILQ_REMOVE(&V_ifnet, ifp, if_link);
841			found = 1;
842			break;
843		}
844#ifdef VIMAGE
845	if (found)
846		curvnet->vnet_ifcnt--;
847#endif
848	IFNET_WUNLOCK();
849	if (!found) {
850		if (vmove)
851			panic("%s: ifp=%p not on the ifnet tailq %p",
852			    __func__, ifp, &V_ifnet);
853		else
854			return; /* XXX this should panic as well? */
855	}
856
857	/*
858	 * Remove/wait for pending events.
859	 */
860	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
861
862	/*
863	 * Remove routes and flush queues.
864	 */
865	if_down(ifp);
866#ifdef ALTQ
867	if (ALTQ_IS_ENABLED(&ifp->if_snd))
868		altq_disable(&ifp->if_snd);
869	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
870		altq_detach(&ifp->if_snd);
871#endif
872
873	if_purgeaddrs(ifp);
874
875#ifdef INET
876	in_ifdetach(ifp);
877#endif
878
879#ifdef INET6
880	/*
881	 * Remove all IPv6 kernel structs related to ifp.  This should be done
882	 * before removing routing entries below, since IPv6 interface direct
883	 * routes are expected to be removed by the IPv6-specific kernel API.
884	 * Otherwise, the kernel will detect some inconsistency and bark it.
885	 */
886	in6_ifdetach(ifp);
887#endif
888	if_purgemaddrs(ifp);
889
890	if (!vmove) {
891		/*
892		 * Prevent further calls into the device driver via ifnet.
893		 */
894		if_dead(ifp);
895
896		/*
897		 * Remove link ifaddr pointer and maybe decrement if_index.
898		 * Clean up all addresses.
899		 */
900		ifp->if_addr = NULL;
901
902		/* We can now free link ifaddr. */
903		if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
904			ifa = TAILQ_FIRST(&ifp->if_addrhead);
905			TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
906			ifa_free(ifa);
907		}
908	}
909
910	/*
911	 * Delete all remaining routes using this interface
912	 * Unfortuneatly the only way to do this is to slog through
913	 * the entire routing table looking for routes which point
914	 * to this interface...oh well...
915	 */
916	for (i = 1; i <= AF_MAX; i++) {
917		for (j = 0; j < rt_numfibs; j++) {
918			rnh = rt_tables_get_rnh(j, i);
919			if (rnh == NULL)
920				continue;
921			RADIX_NODE_HEAD_LOCK(rnh);
922			(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
923			RADIX_NODE_HEAD_UNLOCK(rnh);
924		}
925	}
926
927	/* Announce that the interface is gone. */
928	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
929	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
930	if (IS_DEFAULT_VNET(curvnet))
931		devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
932	if_delgroups(ifp);
933
934	/*
935	 * We cannot hold the lock over dom_ifdetach calls as they might
936	 * sleep, for example trying to drain a callout, thus open up the
937	 * theoretical race with re-attaching.
938	 */
939	IF_AFDATA_LOCK(ifp);
940	i = ifp->if_afdata_initialized;
941	ifp->if_afdata_initialized = 0;
942	IF_AFDATA_UNLOCK(ifp);
943	for (dp = domains; i > 0 && dp; dp = dp->dom_next) {
944		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
945			(*dp->dom_ifdetach)(ifp,
946			    ifp->if_afdata[dp->dom_family]);
947	}
948}
949
950#ifdef VIMAGE
951/*
952 * if_vmove() performs a limited version of if_detach() in current
953 * vnet and if_attach()es the ifnet to the vnet specified as 2nd arg.
954 * An attempt is made to shrink if_index in current vnet, find an
955 * unused if_index in target vnet and calls if_grow() if necessary,
956 * and finally find an unused if_xname for the target vnet.
957 */
958void
959if_vmove(struct ifnet *ifp, struct vnet *new_vnet)
960{
961	u_short idx;
962
963	/*
964	 * Detach from current vnet, but preserve LLADDR info, do not
965	 * mark as dead etc. so that the ifnet can be reattached later.
966	 */
967	if_detach_internal(ifp, 1);
968
969	/*
970	 * Unlink the ifnet from ifindex_table[] in current vnet, and shrink
971	 * the if_index for that vnet if possible.
972	 *
973	 * NOTE: IFNET_WLOCK/IFNET_WUNLOCK() are assumed to be unvirtualized,
974	 * or we'd lock on one vnet and unlock on another.
975	 */
976	IFNET_WLOCK();
977	ifindex_free_locked(ifp->if_index);
978	IFNET_WUNLOCK();
979
980	/*
981	 * Perform interface-specific reassignment tasks, if provided by
982	 * the driver.
983	 */
984	if (ifp->if_reassign != NULL)
985		ifp->if_reassign(ifp, new_vnet, NULL);
986
987	/*
988	 * Switch to the context of the target vnet.
989	 */
990	CURVNET_SET_QUIET(new_vnet);
991
992	IFNET_WLOCK();
993	if (ifindex_alloc_locked(&idx) != 0) {
994		IFNET_WUNLOCK();
995		panic("if_index overflow");
996	}
997	ifp->if_index = idx;
998	ifnet_setbyindex_locked(ifp->if_index, ifp);
999	IFNET_WUNLOCK();
1000
1001	if_attach_internal(ifp, 1);
1002
1003	CURVNET_RESTORE();
1004}
1005
1006/*
1007 * Move an ifnet to or from another child prison/vnet, specified by the jail id.
1008 */
1009static int
1010if_vmove_loan(struct thread *td, struct ifnet *ifp, char *ifname, int jid)
1011{
1012	struct prison *pr;
1013	struct ifnet *difp;
1014
1015	/* Try to find the prison within our visibility. */
1016	sx_slock(&allprison_lock);
1017	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1018	sx_sunlock(&allprison_lock);
1019	if (pr == NULL)
1020		return (ENXIO);
1021	prison_hold_locked(pr);
1022	mtx_unlock(&pr->pr_mtx);
1023
1024	/* Do not try to move the iface from and to the same prison. */
1025	if (pr->pr_vnet == ifp->if_vnet) {
1026		prison_free(pr);
1027		return (EEXIST);
1028	}
1029
1030	/* Make sure the named iface does not exists in the dst. prison/vnet. */
1031	/* XXX Lock interfaces to avoid races. */
1032	CURVNET_SET_QUIET(pr->pr_vnet);
1033	difp = ifunit(ifname);
1034	CURVNET_RESTORE();
1035	if (difp != NULL) {
1036		prison_free(pr);
1037		return (EEXIST);
1038	}
1039
1040	/* Move the interface into the child jail/vnet. */
1041	if_vmove(ifp, pr->pr_vnet);
1042
1043	/* Report the new if_xname back to the userland. */
1044	sprintf(ifname, "%s", ifp->if_xname);
1045
1046	prison_free(pr);
1047	return (0);
1048}
1049
1050static int
1051if_vmove_reclaim(struct thread *td, char *ifname, int jid)
1052{
1053	struct prison *pr;
1054	struct vnet *vnet_dst;
1055	struct ifnet *ifp;
1056
1057	/* Try to find the prison within our visibility. */
1058	sx_slock(&allprison_lock);
1059	pr = prison_find_child(td->td_ucred->cr_prison, jid);
1060	sx_sunlock(&allprison_lock);
1061	if (pr == NULL)
1062		return (ENXIO);
1063	prison_hold_locked(pr);
1064	mtx_unlock(&pr->pr_mtx);
1065
1066	/* Make sure the named iface exists in the source prison/vnet. */
1067	CURVNET_SET(pr->pr_vnet);
1068	ifp = ifunit(ifname);		/* XXX Lock to avoid races. */
1069	if (ifp == NULL) {
1070		CURVNET_RESTORE();
1071		prison_free(pr);
1072		return (ENXIO);
1073	}
1074
1075	/* Do not try to move the iface from and to the same prison. */
1076	vnet_dst = TD_TO_VNET(td);
1077	if (vnet_dst == ifp->if_vnet) {
1078		CURVNET_RESTORE();
1079		prison_free(pr);
1080		return (EEXIST);
1081	}
1082
1083	/* Get interface back from child jail/vnet. */
1084	if_vmove(ifp, vnet_dst);
1085	CURVNET_RESTORE();
1086
1087	/* Report the new if_xname back to the userland. */
1088	sprintf(ifname, "%s", ifp->if_xname);
1089
1090	prison_free(pr);
1091	return (0);
1092}
1093#endif /* VIMAGE */
1094
1095/*
1096 * Add a group to an interface
1097 */
1098int
1099if_addgroup(struct ifnet *ifp, const char *groupname)
1100{
1101	struct ifg_list		*ifgl;
1102	struct ifg_group	*ifg = NULL;
1103	struct ifg_member	*ifgm;
1104
1105	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
1106	    groupname[strlen(groupname) - 1] <= '9')
1107		return (EINVAL);
1108
1109	IFNET_WLOCK();
1110	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1111		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
1112			IFNET_WUNLOCK();
1113			return (EEXIST);
1114		}
1115
1116	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
1117	    M_NOWAIT)) == NULL) {
1118	    	IFNET_WUNLOCK();
1119		return (ENOMEM);
1120	}
1121
1122	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
1123	    M_TEMP, M_NOWAIT)) == NULL) {
1124		free(ifgl, M_TEMP);
1125		IFNET_WUNLOCK();
1126		return (ENOMEM);
1127	}
1128
1129	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1130		if (!strcmp(ifg->ifg_group, groupname))
1131			break;
1132
1133	if (ifg == NULL) {
1134		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
1135		    M_TEMP, M_NOWAIT)) == NULL) {
1136			free(ifgl, M_TEMP);
1137			free(ifgm, M_TEMP);
1138			IFNET_WUNLOCK();
1139			return (ENOMEM);
1140		}
1141		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
1142		ifg->ifg_refcnt = 0;
1143		TAILQ_INIT(&ifg->ifg_members);
1144		EVENTHANDLER_INVOKE(group_attach_event, ifg);
1145		TAILQ_INSERT_TAIL(&V_ifg_head, ifg, ifg_next);
1146	}
1147
1148	ifg->ifg_refcnt++;
1149	ifgl->ifgl_group = ifg;
1150	ifgm->ifgm_ifp = ifp;
1151
1152	IF_ADDR_WLOCK(ifp);
1153	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
1154	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
1155	IF_ADDR_WUNLOCK(ifp);
1156
1157	IFNET_WUNLOCK();
1158
1159	EVENTHANDLER_INVOKE(group_change_event, groupname);
1160
1161	return (0);
1162}
1163
1164/*
1165 * Remove a group from an interface
1166 */
1167int
1168if_delgroup(struct ifnet *ifp, const char *groupname)
1169{
1170	struct ifg_list		*ifgl;
1171	struct ifg_member	*ifgm;
1172
1173	IFNET_WLOCK();
1174	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1175		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
1176			break;
1177	if (ifgl == NULL) {
1178		IFNET_WUNLOCK();
1179		return (ENOENT);
1180	}
1181
1182	IF_ADDR_WLOCK(ifp);
1183	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1184	IF_ADDR_WUNLOCK(ifp);
1185
1186	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1187		if (ifgm->ifgm_ifp == ifp)
1188			break;
1189
1190	if (ifgm != NULL) {
1191		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
1192		free(ifgm, M_TEMP);
1193	}
1194
1195	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1196		TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1197		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
1198		free(ifgl->ifgl_group, M_TEMP);
1199	}
1200	IFNET_WUNLOCK();
1201
1202	free(ifgl, M_TEMP);
1203
1204	EVENTHANDLER_INVOKE(group_change_event, groupname);
1205
1206	return (0);
1207}
1208
1209/*
1210 * Remove an interface from all groups
1211 */
1212static void
1213if_delgroups(struct ifnet *ifp)
1214{
1215	struct ifg_list		*ifgl;
1216	struct ifg_member	*ifgm;
1217	char groupname[IFNAMSIZ];
1218
1219	IFNET_WLOCK();
1220	while (!TAILQ_EMPTY(&ifp->if_groups)) {
1221		ifgl = TAILQ_FIRST(&ifp->if_groups);
1222
1223		strlcpy(groupname, ifgl->ifgl_group->ifg_group, IFNAMSIZ);
1224
1225		IF_ADDR_WLOCK(ifp);
1226		TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
1227		IF_ADDR_WUNLOCK(ifp);
1228
1229		TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
1230			if (ifgm->ifgm_ifp == ifp)
1231				break;
1232
1233		if (ifgm != NULL) {
1234			TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm,
1235			    ifgm_next);
1236			free(ifgm, M_TEMP);
1237		}
1238
1239		if (--ifgl->ifgl_group->ifg_refcnt == 0) {
1240			TAILQ_REMOVE(&V_ifg_head, ifgl->ifgl_group, ifg_next);
1241			EVENTHANDLER_INVOKE(group_detach_event,
1242			    ifgl->ifgl_group);
1243			free(ifgl->ifgl_group, M_TEMP);
1244		}
1245		IFNET_WUNLOCK();
1246
1247		free(ifgl, M_TEMP);
1248
1249		EVENTHANDLER_INVOKE(group_change_event, groupname);
1250
1251		IFNET_WLOCK();
1252	}
1253	IFNET_WUNLOCK();
1254}
1255
1256/*
1257 * Stores all groups from an interface in memory pointed
1258 * to by data
1259 */
1260static int
1261if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
1262{
1263	int			 len, error;
1264	struct ifg_list		*ifgl;
1265	struct ifg_req		 ifgrq, *ifgp;
1266	struct ifgroupreq	*ifgr = data;
1267
1268	if (ifgr->ifgr_len == 0) {
1269		IF_ADDR_RLOCK(ifp);
1270		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
1271			ifgr->ifgr_len += sizeof(struct ifg_req);
1272		IF_ADDR_RUNLOCK(ifp);
1273		return (0);
1274	}
1275
1276	len = ifgr->ifgr_len;
1277	ifgp = ifgr->ifgr_groups;
1278	/* XXX: wire */
1279	IF_ADDR_RLOCK(ifp);
1280	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
1281		if (len < sizeof(ifgrq)) {
1282			IF_ADDR_RUNLOCK(ifp);
1283			return (EINVAL);
1284		}
1285		bzero(&ifgrq, sizeof ifgrq);
1286		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
1287		    sizeof(ifgrq.ifgrq_group));
1288		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1289		    	IF_ADDR_RUNLOCK(ifp);
1290			return (error);
1291		}
1292		len -= sizeof(ifgrq);
1293		ifgp++;
1294	}
1295	IF_ADDR_RUNLOCK(ifp);
1296
1297	return (0);
1298}
1299
1300/*
1301 * Stores all members of a group in memory pointed to by data
1302 */
1303static int
1304if_getgroupmembers(struct ifgroupreq *data)
1305{
1306	struct ifgroupreq	*ifgr = data;
1307	struct ifg_group	*ifg;
1308	struct ifg_member	*ifgm;
1309	struct ifg_req		 ifgrq, *ifgp;
1310	int			 len, error;
1311
1312	IFNET_RLOCK();
1313	TAILQ_FOREACH(ifg, &V_ifg_head, ifg_next)
1314		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
1315			break;
1316	if (ifg == NULL) {
1317		IFNET_RUNLOCK();
1318		return (ENOENT);
1319	}
1320
1321	if (ifgr->ifgr_len == 0) {
1322		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
1323			ifgr->ifgr_len += sizeof(ifgrq);
1324		IFNET_RUNLOCK();
1325		return (0);
1326	}
1327
1328	len = ifgr->ifgr_len;
1329	ifgp = ifgr->ifgr_groups;
1330	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
1331		if (len < sizeof(ifgrq)) {
1332			IFNET_RUNLOCK();
1333			return (EINVAL);
1334		}
1335		bzero(&ifgrq, sizeof ifgrq);
1336		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
1337		    sizeof(ifgrq.ifgrq_member));
1338		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
1339			IFNET_RUNLOCK();
1340			return (error);
1341		}
1342		len -= sizeof(ifgrq);
1343		ifgp++;
1344	}
1345	IFNET_RUNLOCK();
1346
1347	return (0);
1348}
1349
1350/*
1351 * Delete Routes for a Network Interface
1352 *
1353 * Called for each routing entry via the rnh->rnh_walktree() call above
1354 * to delete all route entries referencing a detaching network interface.
1355 *
1356 * Arguments:
1357 *	rn	pointer to node in the routing table
1358 *	arg	argument passed to rnh->rnh_walktree() - detaching interface
1359 *
1360 * Returns:
1361 *	0	successful
1362 *	errno	failed - reason indicated
1363 *
1364 */
1365static int
1366if_rtdel(struct radix_node *rn, void *arg)
1367{
1368	struct rtentry	*rt = (struct rtentry *)rn;
1369	struct ifnet	*ifp = arg;
1370	int		err;
1371
1372	if (rt->rt_ifp == ifp) {
1373
1374		/*
1375		 * Protect (sorta) against walktree recursion problems
1376		 * with cloned routes
1377		 */
1378		if ((rt->rt_flags & RTF_UP) == 0)
1379			return (0);
1380
1381		err = rtrequest_fib(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1382				rt_mask(rt),
1383				rt->rt_flags|RTF_RNH_LOCKED|RTF_PINNED,
1384				(struct rtentry **) NULL, rt->rt_fibnum);
1385		if (err) {
1386			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1387		}
1388	}
1389
1390	return (0);
1391}
1392
1393/*
1394 * Wrapper functions for struct ifnet address list locking macros.  These are
1395 * used by kernel modules to avoid encoding programming interface or binary
1396 * interface assumptions that may be violated when kernel-internal locking
1397 * approaches change.
1398 */
1399void
1400if_addr_rlock(struct ifnet *ifp)
1401{
1402
1403	IF_ADDR_RLOCK(ifp);
1404}
1405
1406void
1407if_addr_runlock(struct ifnet *ifp)
1408{
1409
1410	IF_ADDR_RUNLOCK(ifp);
1411}
1412
1413void
1414if_maddr_rlock(struct ifnet *ifp)
1415{
1416
1417	IF_ADDR_RLOCK(ifp);
1418}
1419
1420void
1421if_maddr_runlock(struct ifnet *ifp)
1422{
1423
1424	IF_ADDR_RUNLOCK(ifp);
1425}
1426
1427/*
1428 * Reference count functions for ifaddrs.
1429 */
1430void
1431ifa_init(struct ifaddr *ifa)
1432{
1433
1434	mtx_init(&ifa->ifa_mtx, "ifaddr", NULL, MTX_DEF);
1435	refcount_init(&ifa->ifa_refcnt, 1);
1436}
1437
1438void
1439ifa_ref(struct ifaddr *ifa)
1440{
1441
1442	refcount_acquire(&ifa->ifa_refcnt);
1443}
1444
1445void
1446ifa_free(struct ifaddr *ifa)
1447{
1448
1449	if (refcount_release(&ifa->ifa_refcnt)) {
1450		mtx_destroy(&ifa->ifa_mtx);
1451		free(ifa, M_IFADDR);
1452	}
1453}
1454
1455int
1456ifa_add_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1457{
1458	int error = 0;
1459	struct rtentry *rt = NULL;
1460	struct rt_addrinfo info;
1461	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
1462
1463	bzero(&info, sizeof(info));
1464	info.rti_ifp = V_loif;
1465	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1466	info.rti_info[RTAX_DST] = ia;
1467	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1468	error = rtrequest1_fib(RTM_ADD, &info, &rt, 0);
1469
1470	if (error == 0 && rt != NULL) {
1471		RT_LOCK(rt);
1472		((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
1473			ifa->ifa_ifp->if_type;
1474		((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
1475			ifa->ifa_ifp->if_index;
1476		RT_REMREF(rt);
1477		RT_UNLOCK(rt);
1478	} else if (error != 0)
1479		log(LOG_INFO, "ifa_add_loopback_route: insertion failed\n");
1480
1481	return (error);
1482}
1483
1484int
1485ifa_del_loopback_route(struct ifaddr *ifa, struct sockaddr *ia)
1486{
1487	int error = 0;
1488	struct rt_addrinfo info;
1489	struct sockaddr_dl null_sdl;
1490
1491	bzero(&null_sdl, sizeof(null_sdl));
1492	null_sdl.sdl_len = sizeof(null_sdl);
1493	null_sdl.sdl_family = AF_LINK;
1494	null_sdl.sdl_type = ifa->ifa_ifp->if_type;
1495	null_sdl.sdl_index = ifa->ifa_ifp->if_index;
1496	bzero(&info, sizeof(info));
1497	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC;
1498	info.rti_info[RTAX_DST] = ia;
1499	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&null_sdl;
1500	error = rtrequest1_fib(RTM_DELETE, &info, NULL, 0);
1501
1502	if (error != 0)
1503		log(LOG_INFO, "ifa_del_loopback_route: deletion failed\n");
1504
1505	return (error);
1506}
1507
1508/*
1509 * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1510 * structs used to represent other address families, it is necessary
1511 * to perform a different comparison.
1512 */
1513
1514#define	sa_equal(a1, a2)	\
1515	(bcmp((a1), (a2), ((a1))->sa_len) == 0)
1516
1517#define	sa_dl_equal(a1, a2)	\
1518	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1519	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1520	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1521	       LLADDR((struct sockaddr_dl *)(a2)),			\
1522	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1523
1524/*
1525 * Locate an interface based on a complete address.
1526 */
1527/*ARGSUSED*/
1528static struct ifaddr *
1529ifa_ifwithaddr_internal(struct sockaddr *addr, int getref)
1530{
1531	struct ifnet *ifp;
1532	struct ifaddr *ifa;
1533
1534	IFNET_RLOCK_NOSLEEP();
1535	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1536		IF_ADDR_RLOCK(ifp);
1537		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1538			if (ifa->ifa_addr->sa_family != addr->sa_family)
1539				continue;
1540			if (sa_equal(addr, ifa->ifa_addr)) {
1541				if (getref)
1542					ifa_ref(ifa);
1543				IF_ADDR_RUNLOCK(ifp);
1544				goto done;
1545			}
1546			/* IP6 doesn't have broadcast */
1547			if ((ifp->if_flags & IFF_BROADCAST) &&
1548			    ifa->ifa_broadaddr &&
1549			    ifa->ifa_broadaddr->sa_len != 0 &&
1550			    sa_equal(ifa->ifa_broadaddr, addr)) {
1551				if (getref)
1552					ifa_ref(ifa);
1553				IF_ADDR_RUNLOCK(ifp);
1554				goto done;
1555			}
1556		}
1557		IF_ADDR_RUNLOCK(ifp);
1558	}
1559	ifa = NULL;
1560done:
1561	IFNET_RUNLOCK_NOSLEEP();
1562	return (ifa);
1563}
1564
1565struct ifaddr *
1566ifa_ifwithaddr(struct sockaddr *addr)
1567{
1568
1569	return (ifa_ifwithaddr_internal(addr, 1));
1570}
1571
1572int
1573ifa_ifwithaddr_check(struct sockaddr *addr)
1574{
1575
1576	return (ifa_ifwithaddr_internal(addr, 0) != NULL);
1577}
1578
1579/*
1580 * Locate an interface based on the broadcast address.
1581 */
1582/* ARGSUSED */
1583struct ifaddr *
1584ifa_ifwithbroadaddr(struct sockaddr *addr)
1585{
1586	struct ifnet *ifp;
1587	struct ifaddr *ifa;
1588
1589	IFNET_RLOCK_NOSLEEP();
1590	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1591		IF_ADDR_RLOCK(ifp);
1592		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1593			if (ifa->ifa_addr->sa_family != addr->sa_family)
1594				continue;
1595			if ((ifp->if_flags & IFF_BROADCAST) &&
1596			    ifa->ifa_broadaddr &&
1597			    ifa->ifa_broadaddr->sa_len != 0 &&
1598			    sa_equal(ifa->ifa_broadaddr, addr)) {
1599				ifa_ref(ifa);
1600				IF_ADDR_RUNLOCK(ifp);
1601				goto done;
1602			}
1603		}
1604		IF_ADDR_RUNLOCK(ifp);
1605	}
1606	ifa = NULL;
1607done:
1608	IFNET_RUNLOCK_NOSLEEP();
1609	return (ifa);
1610}
1611
1612/*
1613 * Locate the point to point interface with a given destination address.
1614 */
1615/*ARGSUSED*/
1616struct ifaddr *
1617ifa_ifwithdstaddr(struct sockaddr *addr)
1618{
1619	struct ifnet *ifp;
1620	struct ifaddr *ifa;
1621
1622	IFNET_RLOCK_NOSLEEP();
1623	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1624		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1625			continue;
1626		IF_ADDR_RLOCK(ifp);
1627		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1628			if (ifa->ifa_addr->sa_family != addr->sa_family)
1629				continue;
1630			if (ifa->ifa_dstaddr != NULL &&
1631			    sa_equal(addr, ifa->ifa_dstaddr)) {
1632				ifa_ref(ifa);
1633				IF_ADDR_RUNLOCK(ifp);
1634				goto done;
1635			}
1636		}
1637		IF_ADDR_RUNLOCK(ifp);
1638	}
1639	ifa = NULL;
1640done:
1641	IFNET_RUNLOCK_NOSLEEP();
1642	return (ifa);
1643}
1644
1645/*
1646 * Find an interface on a specific network.  If many, choice
1647 * is most specific found.
1648 */
1649struct ifaddr *
1650ifa_ifwithnet(struct sockaddr *addr, int ignore_ptp)
1651{
1652	struct ifnet *ifp;
1653	struct ifaddr *ifa;
1654	struct ifaddr *ifa_maybe = NULL;
1655	u_int af = addr->sa_family;
1656	char *addr_data = addr->sa_data, *cplim;
1657
1658	/*
1659	 * AF_LINK addresses can be looked up directly by their index number,
1660	 * so do that if we can.
1661	 */
1662	if (af == AF_LINK) {
1663	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1664	    if (sdl->sdl_index && sdl->sdl_index <= V_if_index)
1665		return (ifaddr_byindex(sdl->sdl_index));
1666	}
1667
1668	/*
1669	 * Scan though each interface, looking for ones that have addresses
1670	 * in this address family.  Maintain a reference on ifa_maybe once
1671	 * we find one, as we release the IF_ADDR_RLOCK() that kept it stable
1672	 * when we move onto the next interface.
1673	 */
1674	IFNET_RLOCK_NOSLEEP();
1675	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1676		IF_ADDR_RLOCK(ifp);
1677		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1678			char *cp, *cp2, *cp3;
1679
1680			if (ifa->ifa_addr->sa_family != af)
1681next:				continue;
1682			if (af == AF_INET &&
1683			    ifp->if_flags & IFF_POINTOPOINT && !ignore_ptp) {
1684				/*
1685				 * This is a bit broken as it doesn't
1686				 * take into account that the remote end may
1687				 * be a single node in the network we are
1688				 * looking for.
1689				 * The trouble is that we don't know the
1690				 * netmask for the remote end.
1691				 */
1692				if (ifa->ifa_dstaddr != NULL &&
1693				    sa_equal(addr, ifa->ifa_dstaddr)) {
1694					ifa_ref(ifa);
1695					IF_ADDR_RUNLOCK(ifp);
1696					goto done;
1697				}
1698			} else {
1699				/*
1700				 * if we have a special address handler,
1701				 * then use it instead of the generic one.
1702				 */
1703				if (ifa->ifa_claim_addr) {
1704					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
1705						ifa_ref(ifa);
1706						IF_ADDR_RUNLOCK(ifp);
1707						goto done;
1708					}
1709					continue;
1710				}
1711
1712				/*
1713				 * Scan all the bits in the ifa's address.
1714				 * If a bit dissagrees with what we are
1715				 * looking for, mask it with the netmask
1716				 * to see if it really matters.
1717				 * (A byte at a time)
1718				 */
1719				if (ifa->ifa_netmask == 0)
1720					continue;
1721				cp = addr_data;
1722				cp2 = ifa->ifa_addr->sa_data;
1723				cp3 = ifa->ifa_netmask->sa_data;
1724				cplim = ifa->ifa_netmask->sa_len
1725					+ (char *)ifa->ifa_netmask;
1726				while (cp3 < cplim)
1727					if ((*cp++ ^ *cp2++) & *cp3++)
1728						goto next; /* next address! */
1729				/*
1730				 * If the netmask of what we just found
1731				 * is more specific than what we had before
1732				 * (if we had one) then remember the new one
1733				 * before continuing to search
1734				 * for an even better one.
1735				 */
1736				if (ifa_maybe == NULL ||
1737				    rn_refines((caddr_t)ifa->ifa_netmask,
1738				    (caddr_t)ifa_maybe->ifa_netmask)) {
1739					if (ifa_maybe != NULL)
1740						ifa_free(ifa_maybe);
1741					ifa_maybe = ifa;
1742					ifa_ref(ifa_maybe);
1743				}
1744			}
1745		}
1746		IF_ADDR_RUNLOCK(ifp);
1747	}
1748	ifa = ifa_maybe;
1749	ifa_maybe = NULL;
1750done:
1751	IFNET_RUNLOCK_NOSLEEP();
1752	if (ifa_maybe != NULL)
1753		ifa_free(ifa_maybe);
1754	return (ifa);
1755}
1756
1757/*
1758 * Find an interface address specific to an interface best matching
1759 * a given address.
1760 */
1761struct ifaddr *
1762ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1763{
1764	struct ifaddr *ifa;
1765	char *cp, *cp2, *cp3;
1766	char *cplim;
1767	struct ifaddr *ifa_maybe = NULL;
1768	u_int af = addr->sa_family;
1769
1770	if (af >= AF_MAX)
1771		return (NULL);
1772	IF_ADDR_RLOCK(ifp);
1773	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1774		if (ifa->ifa_addr->sa_family != af)
1775			continue;
1776		if (ifa_maybe == NULL)
1777			ifa_maybe = ifa;
1778		if (ifa->ifa_netmask == 0) {
1779			if (sa_equal(addr, ifa->ifa_addr) ||
1780			    (ifa->ifa_dstaddr &&
1781			    sa_equal(addr, ifa->ifa_dstaddr)))
1782				goto done;
1783			continue;
1784		}
1785		if (ifp->if_flags & IFF_POINTOPOINT) {
1786			if (sa_equal(addr, ifa->ifa_dstaddr))
1787				goto done;
1788		} else {
1789			cp = addr->sa_data;
1790			cp2 = ifa->ifa_addr->sa_data;
1791			cp3 = ifa->ifa_netmask->sa_data;
1792			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1793			for (; cp3 < cplim; cp3++)
1794				if ((*cp++ ^ *cp2++) & *cp3)
1795					break;
1796			if (cp3 == cplim)
1797				goto done;
1798		}
1799	}
1800	ifa = ifa_maybe;
1801done:
1802	if (ifa != NULL)
1803		ifa_ref(ifa);
1804	IF_ADDR_RUNLOCK(ifp);
1805	return (ifa);
1806}
1807
1808#include <net/if_llatbl.h>
1809
1810/*
1811 * Default action when installing a route with a Link Level gateway.
1812 * Lookup an appropriate real ifa to point to.
1813 * This should be moved to /sys/net/link.c eventually.
1814 */
1815static void
1816link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1817{
1818	struct ifaddr *ifa, *oifa;
1819	struct sockaddr *dst;
1820	struct ifnet *ifp;
1821
1822	RT_LOCK_ASSERT(rt);
1823
1824	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1825	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1826		return;
1827	ifa = ifaof_ifpforaddr(dst, ifp);
1828	if (ifa) {
1829		oifa = rt->rt_ifa;
1830		rt->rt_ifa = ifa;
1831		ifa_free(oifa);
1832		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1833			ifa->ifa_rtrequest(cmd, rt, info);
1834	}
1835}
1836
1837/*
1838 * Mark an interface down and notify protocols of
1839 * the transition.
1840 * NOTE: must be called at splnet or eqivalent.
1841 */
1842static void
1843if_unroute(struct ifnet *ifp, int flag, int fam)
1844{
1845	struct ifaddr *ifa;
1846
1847	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1848
1849	ifp->if_flags &= ~flag;
1850	getmicrotime(&ifp->if_lastchange);
1851	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1852		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1853			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1854	ifp->if_qflush(ifp);
1855
1856	if (ifp->if_carp)
1857		(*carp_linkstate_p)(ifp);
1858	rt_ifmsg(ifp);
1859}
1860
1861/*
1862 * Mark an interface up and notify protocols of
1863 * the transition.
1864 * NOTE: must be called at splnet or eqivalent.
1865 */
1866static void
1867if_route(struct ifnet *ifp, int flag, int fam)
1868{
1869	struct ifaddr *ifa;
1870
1871	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1872
1873	ifp->if_flags |= flag;
1874	getmicrotime(&ifp->if_lastchange);
1875	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1876		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1877			pfctlinput(PRC_IFUP, ifa->ifa_addr);
1878	if (ifp->if_carp)
1879		(*carp_linkstate_p)(ifp);
1880	rt_ifmsg(ifp);
1881#ifdef INET6
1882	in6_if_up(ifp);
1883#endif
1884}
1885
1886void	(*vlan_link_state_p)(struct ifnet *);	/* XXX: private from if_vlan */
1887void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
1888struct ifnet *(*vlan_trunkdev_p)(struct ifnet *);
1889struct	ifnet *(*vlan_devat_p)(struct ifnet *, uint16_t);
1890int	(*vlan_tag_p)(struct ifnet *, uint16_t *);
1891int	(*vlan_setcookie_p)(struct ifnet *, void *);
1892void	*(*vlan_cookie_p)(struct ifnet *);
1893
1894/*
1895 * Handle a change in the interface link state. To avoid LORs
1896 * between driver lock and upper layer locks, as well as possible
1897 * recursions, we post event to taskqueue, and all job
1898 * is done in static do_link_state_change().
1899 */
1900void
1901if_link_state_change(struct ifnet *ifp, int link_state)
1902{
1903	/* Return if state hasn't changed. */
1904	if (ifp->if_link_state == link_state)
1905		return;
1906
1907	ifp->if_link_state = link_state;
1908
1909	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1910}
1911
1912static void
1913do_link_state_change(void *arg, int pending)
1914{
1915	struct ifnet *ifp = (struct ifnet *)arg;
1916	int link_state = ifp->if_link_state;
1917	CURVNET_SET(ifp->if_vnet);
1918
1919	/* Notify that the link state has changed. */
1920	rt_ifmsg(ifp);
1921	if (ifp->if_vlantrunk != NULL)
1922		(*vlan_link_state_p)(ifp);
1923
1924	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1925	    IFP2AC(ifp)->ac_netgraph != NULL)
1926		(*ng_ether_link_state_p)(ifp, link_state);
1927	if (ifp->if_carp)
1928		(*carp_linkstate_p)(ifp);
1929	if (ifp->if_bridge)
1930		(*bridge_linkstate_p)(ifp);
1931	if (ifp->if_lagg)
1932		(*lagg_linkstate_p)(ifp, link_state);
1933
1934	if (IS_DEFAULT_VNET(curvnet))
1935		devctl_notify("IFNET", ifp->if_xname,
1936		    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN",
1937		    NULL);
1938	if (pending > 1)
1939		if_printf(ifp, "%d link states coalesced\n", pending);
1940	if (log_link_state_change)
1941		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1942		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1943	EVENTHANDLER_INVOKE(ifnet_link_event, ifp, ifp->if_link_state);
1944	CURVNET_RESTORE();
1945}
1946
1947/*
1948 * Mark an interface down and notify protocols of
1949 * the transition.
1950 * NOTE: must be called at splnet or eqivalent.
1951 */
1952void
1953if_down(struct ifnet *ifp)
1954{
1955
1956	if_unroute(ifp, IFF_UP, AF_UNSPEC);
1957}
1958
1959/*
1960 * Mark an interface up and notify protocols of
1961 * the transition.
1962 * NOTE: must be called at splnet or eqivalent.
1963 */
1964void
1965if_up(struct ifnet *ifp)
1966{
1967
1968	if_route(ifp, IFF_UP, AF_UNSPEC);
1969}
1970
1971/*
1972 * Flush an interface queue.
1973 */
1974void
1975if_qflush(struct ifnet *ifp)
1976{
1977	struct mbuf *m, *n;
1978	struct ifaltq *ifq;
1979
1980	ifq = &ifp->if_snd;
1981	IFQ_LOCK(ifq);
1982#ifdef ALTQ
1983	if (ALTQ_IS_ENABLED(ifq))
1984		ALTQ_PURGE(ifq);
1985#endif
1986	n = ifq->ifq_head;
1987	while ((m = n) != 0) {
1988		n = m->m_act;
1989		m_freem(m);
1990	}
1991	ifq->ifq_head = 0;
1992	ifq->ifq_tail = 0;
1993	ifq->ifq_len = 0;
1994	IFQ_UNLOCK(ifq);
1995}
1996
1997/*
1998 * Map interface name to interface structure pointer, with or without
1999 * returning a reference.
2000 */
2001struct ifnet *
2002ifunit_ref(const char *name)
2003{
2004	struct ifnet *ifp;
2005
2006	IFNET_RLOCK_NOSLEEP();
2007	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2008		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0 &&
2009		    !(ifp->if_flags & IFF_DYING))
2010			break;
2011	}
2012	if (ifp != NULL)
2013		if_ref(ifp);
2014	IFNET_RUNLOCK_NOSLEEP();
2015	return (ifp);
2016}
2017
2018struct ifnet *
2019ifunit(const char *name)
2020{
2021	struct ifnet *ifp;
2022
2023	IFNET_RLOCK_NOSLEEP();
2024	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2025		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
2026			break;
2027	}
2028	IFNET_RUNLOCK_NOSLEEP();
2029	return (ifp);
2030}
2031
2032/*
2033 * Hardware specific interface ioctls.
2034 */
2035static int
2036ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
2037{
2038	struct ifreq *ifr;
2039	struct ifstat *ifs;
2040	int error = 0;
2041	int new_flags, temp_flags;
2042	size_t namelen, onamelen;
2043	size_t descrlen;
2044	char *descrbuf, *odescrbuf;
2045	char new_name[IFNAMSIZ];
2046	struct ifaddr *ifa;
2047	struct sockaddr_dl *sdl;
2048
2049	ifr = (struct ifreq *)data;
2050	switch (cmd) {
2051	case SIOCGIFINDEX:
2052		ifr->ifr_index = ifp->if_index;
2053		break;
2054
2055	case SIOCGIFFLAGS:
2056		temp_flags = ifp->if_flags | ifp->if_drv_flags;
2057		ifr->ifr_flags = temp_flags & 0xffff;
2058		ifr->ifr_flagshigh = temp_flags >> 16;
2059		break;
2060
2061	case SIOCGIFCAP:
2062		ifr->ifr_reqcap = ifp->if_capabilities;
2063		ifr->ifr_curcap = ifp->if_capenable;
2064		break;
2065
2066#ifdef MAC
2067	case SIOCGIFMAC:
2068		error = mac_ifnet_ioctl_get(td->td_ucred, ifr, ifp);
2069		break;
2070#endif
2071
2072	case SIOCGIFMETRIC:
2073		ifr->ifr_metric = ifp->if_metric;
2074		break;
2075
2076	case SIOCGIFMTU:
2077		ifr->ifr_mtu = ifp->if_mtu;
2078		break;
2079
2080	case SIOCGIFPHYS:
2081		ifr->ifr_phys = ifp->if_physical;
2082		break;
2083
2084	case SIOCGIFDESCR:
2085		error = 0;
2086		sx_slock(&ifdescr_sx);
2087		if (ifp->if_description == NULL)
2088			error = ENOMSG;
2089		else {
2090			/* space for terminating nul */
2091			descrlen = strlen(ifp->if_description) + 1;
2092			if (ifr->ifr_buffer.length < descrlen)
2093				ifr->ifr_buffer.buffer = NULL;
2094			else
2095				error = copyout(ifp->if_description,
2096				    ifr->ifr_buffer.buffer, descrlen);
2097			ifr->ifr_buffer.length = descrlen;
2098		}
2099		sx_sunlock(&ifdescr_sx);
2100		break;
2101
2102	case SIOCSIFDESCR:
2103		error = priv_check(td, PRIV_NET_SETIFDESCR);
2104		if (error)
2105			return (error);
2106
2107		/*
2108		 * Copy only (length-1) bytes to make sure that
2109		 * if_description is always nul terminated.  The
2110		 * length parameter is supposed to count the
2111		 * terminating nul in.
2112		 */
2113		if (ifr->ifr_buffer.length > ifdescr_maxlen)
2114			return (ENAMETOOLONG);
2115		else if (ifr->ifr_buffer.length == 0)
2116			descrbuf = NULL;
2117		else {
2118			descrbuf = malloc(ifr->ifr_buffer.length, M_IFDESCR,
2119			    M_WAITOK | M_ZERO);
2120			error = copyin(ifr->ifr_buffer.buffer, descrbuf,
2121			    ifr->ifr_buffer.length - 1);
2122			if (error) {
2123				free(descrbuf, M_IFDESCR);
2124				break;
2125			}
2126		}
2127
2128		sx_xlock(&ifdescr_sx);
2129		odescrbuf = ifp->if_description;
2130		ifp->if_description = descrbuf;
2131		sx_xunlock(&ifdescr_sx);
2132
2133		getmicrotime(&ifp->if_lastchange);
2134		free(odescrbuf, M_IFDESCR);
2135		break;
2136
2137	case SIOCGIFFIB:
2138		ifr->ifr_fib = ifp->if_fib;
2139		break;
2140
2141	case SIOCSIFFIB:
2142		error = priv_check(td, PRIV_NET_SETIFFIB);
2143		if (error)
2144			return (error);
2145		if (ifr->ifr_fib >= rt_numfibs)
2146			return (EINVAL);
2147
2148		ifp->if_fib = ifr->ifr_fib;
2149		break;
2150
2151	case SIOCSIFFLAGS:
2152		error = priv_check(td, PRIV_NET_SETIFFLAGS);
2153		if (error)
2154			return (error);
2155		/*
2156		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
2157		 * check, so we don't need special handling here yet.
2158		 */
2159		new_flags = (ifr->ifr_flags & 0xffff) |
2160		    (ifr->ifr_flagshigh << 16);
2161		if (ifp->if_flags & IFF_SMART) {
2162			/* Smart drivers twiddle their own routes */
2163		} else if (ifp->if_flags & IFF_UP &&
2164		    (new_flags & IFF_UP) == 0) {
2165			int s = splimp();
2166			if_down(ifp);
2167			splx(s);
2168		} else if (new_flags & IFF_UP &&
2169		    (ifp->if_flags & IFF_UP) == 0) {
2170			int s = splimp();
2171			if_up(ifp);
2172			splx(s);
2173		}
2174		/* See if permanently promiscuous mode bit is about to flip */
2175		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
2176			if (new_flags & IFF_PPROMISC)
2177				ifp->if_flags |= IFF_PROMISC;
2178			else if (ifp->if_pcount == 0)
2179				ifp->if_flags &= ~IFF_PROMISC;
2180			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
2181			    ifp->if_xname,
2182			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
2183		}
2184		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
2185			(new_flags &~ IFF_CANTCHANGE);
2186		if (ifp->if_ioctl) {
2187			(void) (*ifp->if_ioctl)(ifp, cmd, data);
2188		}
2189		getmicrotime(&ifp->if_lastchange);
2190		break;
2191
2192	case SIOCSIFCAP:
2193		error = priv_check(td, PRIV_NET_SETIFCAP);
2194		if (error)
2195			return (error);
2196		if (ifp->if_ioctl == NULL)
2197			return (EOPNOTSUPP);
2198		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
2199			return (EINVAL);
2200		error = (*ifp->if_ioctl)(ifp, cmd, data);
2201		if (error == 0)
2202			getmicrotime(&ifp->if_lastchange);
2203		break;
2204
2205#ifdef MAC
2206	case SIOCSIFMAC:
2207		error = mac_ifnet_ioctl_set(td->td_ucred, ifr, ifp);
2208		break;
2209#endif
2210
2211	case SIOCSIFNAME:
2212		error = priv_check(td, PRIV_NET_SETIFNAME);
2213		if (error)
2214			return (error);
2215		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
2216		if (error != 0)
2217			return (error);
2218		if (new_name[0] == '\0')
2219			return (EINVAL);
2220		if (ifunit(new_name) != NULL)
2221			return (EEXIST);
2222
2223		/*
2224		 * XXX: Locking.  Nothing else seems to lock if_flags,
2225		 * and there are numerous other races with the
2226		 * ifunit() checks not being atomic with namespace
2227		 * changes (renames, vmoves, if_attach, etc).
2228		 */
2229		ifp->if_flags |= IFF_RENAMING;
2230
2231		/* Announce the departure of the interface. */
2232		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
2233		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
2234
2235		log(LOG_INFO, "%s: changing name to '%s'\n",
2236		    ifp->if_xname, new_name);
2237
2238		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
2239		ifa = ifp->if_addr;
2240		IFA_LOCK(ifa);
2241		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2242		namelen = strlen(new_name);
2243		onamelen = sdl->sdl_nlen;
2244		/*
2245		 * Move the address if needed.  This is safe because we
2246		 * allocate space for a name of length IFNAMSIZ when we
2247		 * create this in if_attach().
2248		 */
2249		if (namelen != onamelen) {
2250			bcopy(sdl->sdl_data + onamelen,
2251			    sdl->sdl_data + namelen, sdl->sdl_alen);
2252		}
2253		bcopy(new_name, sdl->sdl_data, namelen);
2254		sdl->sdl_nlen = namelen;
2255		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
2256		bzero(sdl->sdl_data, onamelen);
2257		while (namelen != 0)
2258			sdl->sdl_data[--namelen] = 0xff;
2259		IFA_UNLOCK(ifa);
2260
2261		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
2262		/* Announce the return of the interface. */
2263		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
2264
2265		ifp->if_flags &= ~IFF_RENAMING;
2266		break;
2267
2268#ifdef VIMAGE
2269	case SIOCSIFVNET:
2270		error = priv_check(td, PRIV_NET_SETIFVNET);
2271		if (error)
2272			return (error);
2273		error = if_vmove_loan(td, ifp, ifr->ifr_name, ifr->ifr_jid);
2274		break;
2275#endif
2276
2277	case SIOCSIFMETRIC:
2278		error = priv_check(td, PRIV_NET_SETIFMETRIC);
2279		if (error)
2280			return (error);
2281		ifp->if_metric = ifr->ifr_metric;
2282		getmicrotime(&ifp->if_lastchange);
2283		break;
2284
2285	case SIOCSIFPHYS:
2286		error = priv_check(td, PRIV_NET_SETIFPHYS);
2287		if (error)
2288			return (error);
2289		if (ifp->if_ioctl == NULL)
2290			return (EOPNOTSUPP);
2291		error = (*ifp->if_ioctl)(ifp, cmd, data);
2292		if (error == 0)
2293			getmicrotime(&ifp->if_lastchange);
2294		break;
2295
2296	case SIOCSIFMTU:
2297	{
2298		u_long oldmtu = ifp->if_mtu;
2299
2300		error = priv_check(td, PRIV_NET_SETIFMTU);
2301		if (error)
2302			return (error);
2303		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
2304			return (EINVAL);
2305		if (ifp->if_ioctl == NULL)
2306			return (EOPNOTSUPP);
2307		error = (*ifp->if_ioctl)(ifp, cmd, data);
2308		if (error == 0) {
2309			getmicrotime(&ifp->if_lastchange);
2310			rt_ifmsg(ifp);
2311		}
2312		/*
2313		 * If the link MTU changed, do network layer specific procedure.
2314		 */
2315		if (ifp->if_mtu != oldmtu) {
2316#ifdef INET6
2317			nd6_setmtu(ifp);
2318#endif
2319		}
2320		break;
2321	}
2322
2323	case SIOCADDMULTI:
2324	case SIOCDELMULTI:
2325		if (cmd == SIOCADDMULTI)
2326			error = priv_check(td, PRIV_NET_ADDMULTI);
2327		else
2328			error = priv_check(td, PRIV_NET_DELMULTI);
2329		if (error)
2330			return (error);
2331
2332		/* Don't allow group membership on non-multicast interfaces. */
2333		if ((ifp->if_flags & IFF_MULTICAST) == 0)
2334			return (EOPNOTSUPP);
2335
2336		/* Don't let users screw up protocols' entries. */
2337		if (ifr->ifr_addr.sa_family != AF_LINK)
2338			return (EINVAL);
2339
2340		if (cmd == SIOCADDMULTI) {
2341			struct ifmultiaddr *ifma;
2342
2343			/*
2344			 * Userland is only permitted to join groups once
2345			 * via the if_addmulti() KPI, because it cannot hold
2346			 * struct ifmultiaddr * between calls. It may also
2347			 * lose a race while we check if the membership
2348			 * already exists.
2349			 */
2350			IF_ADDR_RLOCK(ifp);
2351			ifma = if_findmulti(ifp, &ifr->ifr_addr);
2352			IF_ADDR_RUNLOCK(ifp);
2353			if (ifma != NULL)
2354				error = EADDRINUSE;
2355			else
2356				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
2357		} else {
2358			error = if_delmulti(ifp, &ifr->ifr_addr);
2359		}
2360		if (error == 0)
2361			getmicrotime(&ifp->if_lastchange);
2362		break;
2363
2364	case SIOCSIFPHYADDR:
2365	case SIOCDIFPHYADDR:
2366#ifdef INET6
2367	case SIOCSIFPHYADDR_IN6:
2368#endif
2369	case SIOCSLIFPHYADDR:
2370	case SIOCSIFMEDIA:
2371	case SIOCSIFGENERIC:
2372		error = priv_check(td, PRIV_NET_HWIOCTL);
2373		if (error)
2374			return (error);
2375		if (ifp->if_ioctl == NULL)
2376			return (EOPNOTSUPP);
2377		error = (*ifp->if_ioctl)(ifp, cmd, data);
2378		if (error == 0)
2379			getmicrotime(&ifp->if_lastchange);
2380		break;
2381
2382	case SIOCGIFSTATUS:
2383		ifs = (struct ifstat *)data;
2384		ifs->ascii[0] = '\0';
2385
2386	case SIOCGIFPSRCADDR:
2387	case SIOCGIFPDSTADDR:
2388	case SIOCGLIFPHYADDR:
2389	case SIOCGIFMEDIA:
2390	case SIOCGIFGENERIC:
2391		if (ifp->if_ioctl == NULL)
2392			return (EOPNOTSUPP);
2393		error = (*ifp->if_ioctl)(ifp, cmd, data);
2394		break;
2395
2396	case SIOCSIFLLADDR:
2397		error = priv_check(td, PRIV_NET_SETLLADDR);
2398		if (error)
2399			return (error);
2400		error = if_setlladdr(ifp,
2401		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
2402		EVENTHANDLER_INVOKE(iflladdr_event, ifp);
2403		break;
2404
2405	case SIOCAIFGROUP:
2406	{
2407		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2408
2409		error = priv_check(td, PRIV_NET_ADDIFGROUP);
2410		if (error)
2411			return (error);
2412		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
2413			return (error);
2414		break;
2415	}
2416
2417	case SIOCGIFGROUP:
2418		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
2419			return (error);
2420		break;
2421
2422	case SIOCDIFGROUP:
2423	{
2424		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
2425
2426		error = priv_check(td, PRIV_NET_DELIFGROUP);
2427		if (error)
2428			return (error);
2429		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
2430			return (error);
2431		break;
2432	}
2433
2434	default:
2435		error = ENOIOCTL;
2436		break;
2437	}
2438	return (error);
2439}
2440
2441#ifdef COMPAT_FREEBSD32
2442struct ifconf32 {
2443	int32_t	ifc_len;
2444	union {
2445		uint32_t	ifcu_buf;
2446		uint32_t	ifcu_req;
2447	} ifc_ifcu;
2448};
2449#define	SIOCGIFCONF32	_IOWR('i', 36, struct ifconf32)
2450#endif
2451
2452/*
2453 * Interface ioctls.
2454 */
2455int
2456ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
2457{
2458	struct ifnet *ifp;
2459	struct ifreq *ifr;
2460	int error;
2461	int oif_flags;
2462
2463	CURVNET_SET(so->so_vnet);
2464	switch (cmd) {
2465	case SIOCGIFCONF:
2466	case OSIOCGIFCONF:
2467		error = ifconf(cmd, data);
2468		CURVNET_RESTORE();
2469		return (error);
2470
2471#ifdef COMPAT_FREEBSD32
2472	case SIOCGIFCONF32:
2473		{
2474			struct ifconf32 *ifc32;
2475			struct ifconf ifc;
2476
2477			ifc32 = (struct ifconf32 *)data;
2478			ifc.ifc_len = ifc32->ifc_len;
2479			ifc.ifc_buf = PTRIN(ifc32->ifc_buf);
2480
2481			error = ifconf(SIOCGIFCONF, (void *)&ifc);
2482			CURVNET_RESTORE();
2483			if (error == 0)
2484				ifc32->ifc_len = ifc.ifc_len;
2485			return (error);
2486		}
2487#endif
2488	}
2489	ifr = (struct ifreq *)data;
2490
2491	switch (cmd) {
2492#ifdef VIMAGE
2493	case SIOCSIFRVNET:
2494		error = priv_check(td, PRIV_NET_SETIFVNET);
2495		if (error == 0)
2496			error = if_vmove_reclaim(td, ifr->ifr_name,
2497			    ifr->ifr_jid);
2498		CURVNET_RESTORE();
2499		return (error);
2500#endif
2501	case SIOCIFCREATE:
2502	case SIOCIFCREATE2:
2503		error = priv_check(td, PRIV_NET_IFCREATE);
2504		if (error == 0)
2505			error = if_clone_create(ifr->ifr_name,
2506			    sizeof(ifr->ifr_name),
2507			    cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL);
2508		CURVNET_RESTORE();
2509		return (error);
2510	case SIOCIFDESTROY:
2511		error = priv_check(td, PRIV_NET_IFDESTROY);
2512		if (error == 0)
2513			error = if_clone_destroy(ifr->ifr_name);
2514		CURVNET_RESTORE();
2515		return (error);
2516
2517	case SIOCIFGCLONERS:
2518		error = if_clone_list((struct if_clonereq *)data);
2519		CURVNET_RESTORE();
2520		return (error);
2521	case SIOCGIFGMEMB:
2522		error = if_getgroupmembers((struct ifgroupreq *)data);
2523		CURVNET_RESTORE();
2524		return (error);
2525	}
2526
2527	ifp = ifunit_ref(ifr->ifr_name);
2528	if (ifp == NULL) {
2529		CURVNET_RESTORE();
2530		return (ENXIO);
2531	}
2532
2533	error = ifhwioctl(cmd, ifp, data, td);
2534	if (error != ENOIOCTL) {
2535		if_rele(ifp);
2536		CURVNET_RESTORE();
2537		return (error);
2538	}
2539
2540	oif_flags = ifp->if_flags;
2541	if (so->so_proto == NULL) {
2542		if_rele(ifp);
2543		CURVNET_RESTORE();
2544		return (EOPNOTSUPP);
2545	}
2546#ifndef COMPAT_43
2547	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
2548								 data,
2549								 ifp, td));
2550	if (error == EOPNOTSUPP && ifp != NULL && ifp->if_ioctl != NULL)
2551		error = (*ifp->if_ioctl)(ifp, cmd, data);
2552#else
2553	{
2554		u_long ocmd = cmd;
2555
2556		switch (cmd) {
2557
2558		case SIOCSIFDSTADDR:
2559		case SIOCSIFADDR:
2560		case SIOCSIFBRDADDR:
2561		case SIOCSIFNETMASK:
2562#if BYTE_ORDER != BIG_ENDIAN
2563			if (ifr->ifr_addr.sa_family == 0 &&
2564			    ifr->ifr_addr.sa_len < 16) {
2565				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
2566				ifr->ifr_addr.sa_len = 16;
2567			}
2568#else
2569			if (ifr->ifr_addr.sa_len == 0)
2570				ifr->ifr_addr.sa_len = 16;
2571#endif
2572			break;
2573
2574		case OSIOCGIFADDR:
2575			cmd = SIOCGIFADDR;
2576			break;
2577
2578		case OSIOCGIFDSTADDR:
2579			cmd = SIOCGIFDSTADDR;
2580			break;
2581
2582		case OSIOCGIFBRDADDR:
2583			cmd = SIOCGIFBRDADDR;
2584			break;
2585
2586		case OSIOCGIFNETMASK:
2587			cmd = SIOCGIFNETMASK;
2588		}
2589		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
2590								   cmd,
2591								   data,
2592								   ifp, td));
2593		if (error == EOPNOTSUPP && ifp != NULL &&
2594		    ifp->if_ioctl != NULL)
2595			error = (*ifp->if_ioctl)(ifp, cmd, data);
2596		switch (ocmd) {
2597
2598		case OSIOCGIFADDR:
2599		case OSIOCGIFDSTADDR:
2600		case OSIOCGIFBRDADDR:
2601		case OSIOCGIFNETMASK:
2602			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
2603
2604		}
2605	}
2606#endif /* COMPAT_43 */
2607
2608	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
2609#ifdef INET6
2610		if (ifp->if_flags & IFF_UP) {
2611			int s = splimp();
2612			in6_if_up(ifp);
2613			splx(s);
2614		}
2615#endif
2616	}
2617	if_rele(ifp);
2618	CURVNET_RESTORE();
2619	return (error);
2620}
2621
2622/*
2623 * The code common to handling reference counted flags,
2624 * e.g., in ifpromisc() and if_allmulti().
2625 * The "pflag" argument can specify a permanent mode flag to check,
2626 * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
2627 *
2628 * Only to be used on stack-owned flags, not driver-owned flags.
2629 */
2630static int
2631if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
2632{
2633	struct ifreq ifr;
2634	int error;
2635	int oldflags, oldcount;
2636
2637	/* Sanity checks to catch programming errors */
2638	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
2639	    ("%s: setting driver-owned flag %d", __func__, flag));
2640
2641	if (onswitch)
2642		KASSERT(*refcount >= 0,
2643		    ("%s: increment negative refcount %d for flag %d",
2644		    __func__, *refcount, flag));
2645	else
2646		KASSERT(*refcount > 0,
2647		    ("%s: decrement non-positive refcount %d for flag %d",
2648		    __func__, *refcount, flag));
2649
2650	/* In case this mode is permanent, just touch refcount */
2651	if (ifp->if_flags & pflag) {
2652		*refcount += onswitch ? 1 : -1;
2653		return (0);
2654	}
2655
2656	/* Save ifnet parameters for if_ioctl() may fail */
2657	oldcount = *refcount;
2658	oldflags = ifp->if_flags;
2659
2660	/*
2661	 * See if we aren't the only and touching refcount is enough.
2662	 * Actually toggle interface flag if we are the first or last.
2663	 */
2664	if (onswitch) {
2665		if ((*refcount)++)
2666			return (0);
2667		ifp->if_flags |= flag;
2668	} else {
2669		if (--(*refcount))
2670			return (0);
2671		ifp->if_flags &= ~flag;
2672	}
2673
2674	/* Call down the driver since we've changed interface flags */
2675	if (ifp->if_ioctl == NULL) {
2676		error = EOPNOTSUPP;
2677		goto recover;
2678	}
2679	ifr.ifr_flags = ifp->if_flags & 0xffff;
2680	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2681	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2682	if (error)
2683		goto recover;
2684	/* Notify userland that interface flags have changed */
2685	rt_ifmsg(ifp);
2686	return (0);
2687
2688recover:
2689	/* Recover after driver error */
2690	*refcount = oldcount;
2691	ifp->if_flags = oldflags;
2692	return (error);
2693}
2694
2695/*
2696 * Set/clear promiscuous mode on interface ifp based on the truth value
2697 * of pswitch.  The calls are reference counted so that only the first
2698 * "on" request actually has an effect, as does the final "off" request.
2699 * Results are undefined if the "off" and "on" requests are not matched.
2700 */
2701int
2702ifpromisc(struct ifnet *ifp, int pswitch)
2703{
2704	int error;
2705	int oldflags = ifp->if_flags;
2706
2707	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2708			   &ifp->if_pcount, pswitch);
2709	/* If promiscuous mode status has changed, log a message */
2710	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2711		log(LOG_INFO, "%s: promiscuous mode %s\n",
2712		    ifp->if_xname,
2713		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2714	return (error);
2715}
2716
2717/*
2718 * Return interface configuration
2719 * of system.  List may be used
2720 * in later ioctl's (above) to get
2721 * other information.
2722 */
2723/*ARGSUSED*/
2724static int
2725ifconf(u_long cmd, caddr_t data)
2726{
2727	struct ifconf *ifc = (struct ifconf *)data;
2728	struct ifnet *ifp;
2729	struct ifaddr *ifa;
2730	struct ifreq ifr;
2731	struct sbuf *sb;
2732	int error, full = 0, valid_len, max_len;
2733
2734	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2735	max_len = MAXPHYS - 1;
2736
2737	/* Prevent hostile input from being able to crash the system */
2738	if (ifc->ifc_len <= 0)
2739		return (EINVAL);
2740
2741again:
2742	if (ifc->ifc_len <= max_len) {
2743		max_len = ifc->ifc_len;
2744		full = 1;
2745	}
2746	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2747	max_len = 0;
2748	valid_len = 0;
2749
2750	IFNET_RLOCK();
2751	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2752		int addrs;
2753
2754		/*
2755		 * Zero the ifr_name buffer to make sure we don't
2756		 * disclose the contents of the stack.
2757		 */
2758		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2759
2760		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2761		    >= sizeof(ifr.ifr_name)) {
2762			sbuf_delete(sb);
2763			IFNET_RUNLOCK();
2764			return (ENAMETOOLONG);
2765		}
2766
2767		addrs = 0;
2768		IF_ADDR_RLOCK(ifp);
2769		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2770			struct sockaddr *sa = ifa->ifa_addr;
2771
2772			if (prison_if(curthread->td_ucred, sa) != 0)
2773				continue;
2774			addrs++;
2775#ifdef COMPAT_43
2776			if (cmd == OSIOCGIFCONF) {
2777				struct osockaddr *osa =
2778					 (struct osockaddr *)&ifr.ifr_addr;
2779				ifr.ifr_addr = *sa;
2780				osa->sa_family = sa->sa_family;
2781				sbuf_bcat(sb, &ifr, sizeof(ifr));
2782				max_len += sizeof(ifr);
2783			} else
2784#endif
2785			if (sa->sa_len <= sizeof(*sa)) {
2786				ifr.ifr_addr = *sa;
2787				sbuf_bcat(sb, &ifr, sizeof(ifr));
2788				max_len += sizeof(ifr);
2789			} else {
2790				sbuf_bcat(sb, &ifr,
2791				    offsetof(struct ifreq, ifr_addr));
2792				max_len += offsetof(struct ifreq, ifr_addr);
2793				sbuf_bcat(sb, sa, sa->sa_len);
2794				max_len += sa->sa_len;
2795			}
2796
2797			if (sbuf_error(sb) == 0)
2798				valid_len = sbuf_len(sb);
2799		}
2800		IF_ADDR_RUNLOCK(ifp);
2801		if (addrs == 0) {
2802			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2803			sbuf_bcat(sb, &ifr, sizeof(ifr));
2804			max_len += sizeof(ifr);
2805
2806			if (sbuf_error(sb) == 0)
2807				valid_len = sbuf_len(sb);
2808		}
2809	}
2810	IFNET_RUNLOCK();
2811
2812	/*
2813	 * If we didn't allocate enough space (uncommon), try again.  If
2814	 * we have already allocated as much space as we are allowed,
2815	 * return what we've got.
2816	 */
2817	if (valid_len != max_len && !full) {
2818		sbuf_delete(sb);
2819		goto again;
2820	}
2821
2822	ifc->ifc_len = valid_len;
2823	sbuf_finish(sb);
2824	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2825	sbuf_delete(sb);
2826	return (error);
2827}
2828
2829/*
2830 * Just like ifpromisc(), but for all-multicast-reception mode.
2831 */
2832int
2833if_allmulti(struct ifnet *ifp, int onswitch)
2834{
2835
2836	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2837}
2838
2839struct ifmultiaddr *
2840if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2841{
2842	struct ifmultiaddr *ifma;
2843
2844	IF_ADDR_LOCK_ASSERT(ifp);
2845
2846	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2847		if (sa->sa_family == AF_LINK) {
2848			if (sa_dl_equal(ifma->ifma_addr, sa))
2849				break;
2850		} else {
2851			if (sa_equal(ifma->ifma_addr, sa))
2852				break;
2853		}
2854	}
2855
2856	return ifma;
2857}
2858
2859/*
2860 * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2861 * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2862 * the ifnet multicast address list here, so the caller must do that and
2863 * other setup work (such as notifying the device driver).  The reference
2864 * count is initialized to 1.
2865 */
2866static struct ifmultiaddr *
2867if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2868    int mflags)
2869{
2870	struct ifmultiaddr *ifma;
2871	struct sockaddr *dupsa;
2872
2873	ifma = malloc(sizeof *ifma, M_IFMADDR, mflags |
2874	    M_ZERO);
2875	if (ifma == NULL)
2876		return (NULL);
2877
2878	dupsa = malloc(sa->sa_len, M_IFMADDR, mflags);
2879	if (dupsa == NULL) {
2880		free(ifma, M_IFMADDR);
2881		return (NULL);
2882	}
2883	bcopy(sa, dupsa, sa->sa_len);
2884	ifma->ifma_addr = dupsa;
2885
2886	ifma->ifma_ifp = ifp;
2887	ifma->ifma_refcount = 1;
2888	ifma->ifma_protospec = NULL;
2889
2890	if (llsa == NULL) {
2891		ifma->ifma_lladdr = NULL;
2892		return (ifma);
2893	}
2894
2895	dupsa = malloc(llsa->sa_len, M_IFMADDR, mflags);
2896	if (dupsa == NULL) {
2897		free(ifma->ifma_addr, M_IFMADDR);
2898		free(ifma, M_IFMADDR);
2899		return (NULL);
2900	}
2901	bcopy(llsa, dupsa, llsa->sa_len);
2902	ifma->ifma_lladdr = dupsa;
2903
2904	return (ifma);
2905}
2906
2907/*
2908 * if_freemulti: free ifmultiaddr structure and possibly attached related
2909 * addresses.  The caller is responsible for implementing reference
2910 * counting, notifying the driver, handling routing messages, and releasing
2911 * any dependent link layer state.
2912 */
2913static void
2914if_freemulti(struct ifmultiaddr *ifma)
2915{
2916
2917	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2918	    ifma->ifma_refcount));
2919	KASSERT(ifma->ifma_protospec == NULL,
2920	    ("if_freemulti: protospec not NULL"));
2921
2922	if (ifma->ifma_lladdr != NULL)
2923		free(ifma->ifma_lladdr, M_IFMADDR);
2924	free(ifma->ifma_addr, M_IFMADDR);
2925	free(ifma, M_IFMADDR);
2926}
2927
2928/*
2929 * Register an additional multicast address with a network interface.
2930 *
2931 * - If the address is already present, bump the reference count on the
2932 *   address and return.
2933 * - If the address is not link-layer, look up a link layer address.
2934 * - Allocate address structures for one or both addresses, and attach to the
2935 *   multicast address list on the interface.  If automatically adding a link
2936 *   layer address, the protocol address will own a reference to the link
2937 *   layer address, to be freed when it is freed.
2938 * - Notify the network device driver of an addition to the multicast address
2939 *   list.
2940 *
2941 * 'sa' points to caller-owned memory with the desired multicast address.
2942 *
2943 * 'retifma' will be used to return a pointer to the resulting multicast
2944 * address reference, if desired.
2945 */
2946int
2947if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2948    struct ifmultiaddr **retifma)
2949{
2950	struct ifmultiaddr *ifma, *ll_ifma;
2951	struct sockaddr *llsa;
2952	int error;
2953
2954	/*
2955	 * If the address is already present, return a new reference to it;
2956	 * otherwise, allocate storage and set up a new address.
2957	 */
2958	IF_ADDR_WLOCK(ifp);
2959	ifma = if_findmulti(ifp, sa);
2960	if (ifma != NULL) {
2961		ifma->ifma_refcount++;
2962		if (retifma != NULL)
2963			*retifma = ifma;
2964		IF_ADDR_WUNLOCK(ifp);
2965		return (0);
2966	}
2967
2968	/*
2969	 * The address isn't already present; resolve the protocol address
2970	 * into a link layer address, and then look that up, bump its
2971	 * refcount or allocate an ifma for that also.  If 'llsa' was
2972	 * returned, we will need to free it later.
2973	 */
2974	llsa = NULL;
2975	ll_ifma = NULL;
2976	if (ifp->if_resolvemulti != NULL) {
2977		error = ifp->if_resolvemulti(ifp, &llsa, sa);
2978		if (error)
2979			goto unlock_out;
2980	}
2981
2982	/*
2983	 * Allocate the new address.  Don't hook it up yet, as we may also
2984	 * need to allocate a link layer multicast address.
2985	 */
2986	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
2987	if (ifma == NULL) {
2988		error = ENOMEM;
2989		goto free_llsa_out;
2990	}
2991
2992	/*
2993	 * If a link layer address is found, we'll need to see if it's
2994	 * already present in the address list, or allocate is as well.
2995	 * When this block finishes, the link layer address will be on the
2996	 * list.
2997	 */
2998	if (llsa != NULL) {
2999		ll_ifma = if_findmulti(ifp, llsa);
3000		if (ll_ifma == NULL) {
3001			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
3002			if (ll_ifma == NULL) {
3003				--ifma->ifma_refcount;
3004				if_freemulti(ifma);
3005				error = ENOMEM;
3006				goto free_llsa_out;
3007			}
3008			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
3009			    ifma_link);
3010		} else
3011			ll_ifma->ifma_refcount++;
3012		ifma->ifma_llifma = ll_ifma;
3013	}
3014
3015	/*
3016	 * We now have a new multicast address, ifma, and possibly a new or
3017	 * referenced link layer address.  Add the primary address to the
3018	 * ifnet address list.
3019	 */
3020	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
3021
3022	if (retifma != NULL)
3023		*retifma = ifma;
3024
3025	/*
3026	 * Must generate the message while holding the lock so that 'ifma'
3027	 * pointer is still valid.
3028	 */
3029	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
3030	IF_ADDR_WUNLOCK(ifp);
3031
3032	/*
3033	 * We are certain we have added something, so call down to the
3034	 * interface to let them know about it.
3035	 */
3036	if (ifp->if_ioctl != NULL) {
3037		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
3038	}
3039
3040	if (llsa != NULL)
3041		free(llsa, M_IFMADDR);
3042
3043	return (0);
3044
3045free_llsa_out:
3046	if (llsa != NULL)
3047		free(llsa, M_IFMADDR);
3048
3049unlock_out:
3050	IF_ADDR_WUNLOCK(ifp);
3051	return (error);
3052}
3053
3054/*
3055 * Delete a multicast group membership by network-layer group address.
3056 *
3057 * Returns ENOENT if the entry could not be found. If ifp no longer
3058 * exists, results are undefined. This entry point should only be used
3059 * from subsystems which do appropriate locking to hold ifp for the
3060 * duration of the call.
3061 * Network-layer protocol domains must use if_delmulti_ifma().
3062 */
3063int
3064if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
3065{
3066	struct ifmultiaddr *ifma;
3067	int lastref;
3068#ifdef INVARIANTS
3069	struct ifnet *oifp;
3070
3071	IFNET_RLOCK_NOSLEEP();
3072	TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3073		if (ifp == oifp)
3074			break;
3075	if (ifp != oifp)
3076		ifp = NULL;
3077	IFNET_RUNLOCK_NOSLEEP();
3078
3079	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
3080#endif
3081	if (ifp == NULL)
3082		return (ENOENT);
3083
3084	IF_ADDR_WLOCK(ifp);
3085	lastref = 0;
3086	ifma = if_findmulti(ifp, sa);
3087	if (ifma != NULL)
3088		lastref = if_delmulti_locked(ifp, ifma, 0);
3089	IF_ADDR_WUNLOCK(ifp);
3090
3091	if (ifma == NULL)
3092		return (ENOENT);
3093
3094	if (lastref && ifp->if_ioctl != NULL) {
3095		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3096	}
3097
3098	return (0);
3099}
3100
3101/*
3102 * Delete all multicast group membership for an interface.
3103 * Should be used to quickly flush all multicast filters.
3104 */
3105void
3106if_delallmulti(struct ifnet *ifp)
3107{
3108	struct ifmultiaddr *ifma;
3109	struct ifmultiaddr *next;
3110
3111	IF_ADDR_WLOCK(ifp);
3112	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
3113		if_delmulti_locked(ifp, ifma, 0);
3114	IF_ADDR_WUNLOCK(ifp);
3115}
3116
3117/*
3118 * Delete a multicast group membership by group membership pointer.
3119 * Network-layer protocol domains must use this routine.
3120 *
3121 * It is safe to call this routine if the ifp disappeared.
3122 */
3123void
3124if_delmulti_ifma(struct ifmultiaddr *ifma)
3125{
3126	struct ifnet *ifp;
3127	int lastref;
3128
3129	ifp = ifma->ifma_ifp;
3130#ifdef DIAGNOSTIC
3131	if (ifp == NULL) {
3132		printf("%s: ifma_ifp seems to be detached\n", __func__);
3133	} else {
3134		struct ifnet *oifp;
3135
3136		IFNET_RLOCK_NOSLEEP();
3137		TAILQ_FOREACH(oifp, &V_ifnet, if_link)
3138			if (ifp == oifp)
3139				break;
3140		if (ifp != oifp) {
3141			printf("%s: ifnet %p disappeared\n", __func__, ifp);
3142			ifp = NULL;
3143		}
3144		IFNET_RUNLOCK_NOSLEEP();
3145	}
3146#endif
3147	/*
3148	 * If and only if the ifnet instance exists: Acquire the address lock.
3149	 */
3150	if (ifp != NULL)
3151		IF_ADDR_WLOCK(ifp);
3152
3153	lastref = if_delmulti_locked(ifp, ifma, 0);
3154
3155	if (ifp != NULL) {
3156		/*
3157		 * If and only if the ifnet instance exists:
3158		 *  Release the address lock.
3159		 *  If the group was left: update the hardware hash filter.
3160		 */
3161		IF_ADDR_WUNLOCK(ifp);
3162		if (lastref && ifp->if_ioctl != NULL) {
3163			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
3164		}
3165	}
3166}
3167
3168/*
3169 * Perform deletion of network-layer and/or link-layer multicast address.
3170 *
3171 * Return 0 if the reference count was decremented.
3172 * Return 1 if the final reference was released, indicating that the
3173 * hardware hash filter should be reprogrammed.
3174 */
3175static int
3176if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
3177{
3178	struct ifmultiaddr *ll_ifma;
3179
3180	if (ifp != NULL && ifma->ifma_ifp != NULL) {
3181		KASSERT(ifma->ifma_ifp == ifp,
3182		    ("%s: inconsistent ifp %p", __func__, ifp));
3183		IF_ADDR_WLOCK_ASSERT(ifp);
3184	}
3185
3186	ifp = ifma->ifma_ifp;
3187
3188	/*
3189	 * If the ifnet is detaching, null out references to ifnet,
3190	 * so that upper protocol layers will notice, and not attempt
3191	 * to obtain locks for an ifnet which no longer exists. The
3192	 * routing socket announcement must happen before the ifnet
3193	 * instance is detached from the system.
3194	 */
3195	if (detaching) {
3196#ifdef DIAGNOSTIC
3197		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
3198#endif
3199		/*
3200		 * ifp may already be nulled out if we are being reentered
3201		 * to delete the ll_ifma.
3202		 */
3203		if (ifp != NULL) {
3204			rt_newmaddrmsg(RTM_DELMADDR, ifma);
3205			ifma->ifma_ifp = NULL;
3206		}
3207	}
3208
3209	if (--ifma->ifma_refcount > 0)
3210		return 0;
3211
3212	/*
3213	 * If this ifma is a network-layer ifma, a link-layer ifma may
3214	 * have been associated with it. Release it first if so.
3215	 */
3216	ll_ifma = ifma->ifma_llifma;
3217	if (ll_ifma != NULL) {
3218		KASSERT(ifma->ifma_lladdr != NULL,
3219		    ("%s: llifma w/o lladdr", __func__));
3220		if (detaching)
3221			ll_ifma->ifma_ifp = NULL;	/* XXX */
3222		if (--ll_ifma->ifma_refcount == 0) {
3223			if (ifp != NULL) {
3224				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
3225				    ifma_link);
3226			}
3227			if_freemulti(ll_ifma);
3228		}
3229	}
3230
3231	if (ifp != NULL)
3232		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
3233
3234	if_freemulti(ifma);
3235
3236	/*
3237	 * The last reference to this instance of struct ifmultiaddr
3238	 * was released; the hardware should be notified of this change.
3239	 */
3240	return 1;
3241}
3242
3243/*
3244 * Set the link layer address on an interface.
3245 *
3246 * At this time we only support certain types of interfaces,
3247 * and we don't allow the length of the address to change.
3248 */
3249int
3250if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
3251{
3252	struct sockaddr_dl *sdl;
3253	struct ifaddr *ifa;
3254	struct ifreq ifr;
3255
3256	IF_ADDR_RLOCK(ifp);
3257	ifa = ifp->if_addr;
3258	if (ifa == NULL) {
3259		IF_ADDR_RUNLOCK(ifp);
3260		return (EINVAL);
3261	}
3262	ifa_ref(ifa);
3263	IF_ADDR_RUNLOCK(ifp);
3264	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
3265	if (sdl == NULL) {
3266		ifa_free(ifa);
3267		return (EINVAL);
3268	}
3269	if (len != sdl->sdl_alen) {	/* don't allow length to change */
3270		ifa_free(ifa);
3271		return (EINVAL);
3272	}
3273	switch (ifp->if_type) {
3274	case IFT_ETHER:
3275	case IFT_FDDI:
3276	case IFT_XETHER:
3277	case IFT_ISO88025:
3278	case IFT_L2VLAN:
3279	case IFT_BRIDGE:
3280	case IFT_ARCNET:
3281	case IFT_IEEE8023ADLAG:
3282	case IFT_IEEE80211:
3283		bcopy(lladdr, LLADDR(sdl), len);
3284		ifa_free(ifa);
3285		break;
3286	default:
3287		ifa_free(ifa);
3288		return (ENODEV);
3289	}
3290
3291	/*
3292	 * If the interface is already up, we need
3293	 * to re-init it in order to reprogram its
3294	 * address filter.
3295	 */
3296	if ((ifp->if_flags & IFF_UP) != 0) {
3297		if (ifp->if_ioctl) {
3298			ifp->if_flags &= ~IFF_UP;
3299			ifr.ifr_flags = ifp->if_flags & 0xffff;
3300			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3301			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3302			ifp->if_flags |= IFF_UP;
3303			ifr.ifr_flags = ifp->if_flags & 0xffff;
3304			ifr.ifr_flagshigh = ifp->if_flags >> 16;
3305			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
3306		}
3307#ifdef INET
3308		/*
3309		 * Also send gratuitous ARPs to notify other nodes about
3310		 * the address change.
3311		 */
3312		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
3313			if (ifa->ifa_addr->sa_family == AF_INET)
3314				arp_ifinit(ifp, ifa);
3315		}
3316#endif
3317	}
3318	return (0);
3319}
3320
3321/*
3322 * The name argument must be a pointer to storage which will last as
3323 * long as the interface does.  For physical devices, the result of
3324 * device_get_name(dev) is a good choice and for pseudo-devices a
3325 * static string works well.
3326 */
3327void
3328if_initname(struct ifnet *ifp, const char *name, int unit)
3329{
3330	ifp->if_dname = name;
3331	ifp->if_dunit = unit;
3332	if (unit != IF_DUNIT_NONE)
3333		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
3334	else
3335		strlcpy(ifp->if_xname, name, IFNAMSIZ);
3336}
3337
3338int
3339if_printf(struct ifnet *ifp, const char * fmt, ...)
3340{
3341	va_list ap;
3342	int retval;
3343
3344	retval = printf("%s: ", ifp->if_xname);
3345	va_start(ap, fmt);
3346	retval += vprintf(fmt, ap);
3347	va_end(ap);
3348	return (retval);
3349}
3350
3351void
3352if_start(struct ifnet *ifp)
3353{
3354
3355	(*(ifp)->if_start)(ifp);
3356}
3357
3358/*
3359 * Backwards compatibility interface for drivers
3360 * that have not implemented it
3361 */
3362static int
3363if_transmit(struct ifnet *ifp, struct mbuf *m)
3364{
3365	int error;
3366
3367	IFQ_HANDOFF(ifp, m, error);
3368	return (error);
3369}
3370
3371int
3372if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
3373{
3374	int active = 0;
3375
3376	IF_LOCK(ifq);
3377	if (_IF_QFULL(ifq)) {
3378		_IF_DROP(ifq);
3379		IF_UNLOCK(ifq);
3380		m_freem(m);
3381		return (0);
3382	}
3383	if (ifp != NULL) {
3384		ifp->if_obytes += m->m_pkthdr.len + adjust;
3385		if (m->m_flags & (M_BCAST|M_MCAST))
3386			ifp->if_omcasts++;
3387		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
3388	}
3389	_IF_ENQUEUE(ifq, m);
3390	IF_UNLOCK(ifq);
3391	if (ifp != NULL && !active)
3392		(*(ifp)->if_start)(ifp);
3393	return (1);
3394}
3395
3396void
3397if_register_com_alloc(u_char type,
3398    if_com_alloc_t *a, if_com_free_t *f)
3399{
3400
3401	KASSERT(if_com_alloc[type] == NULL,
3402	    ("if_register_com_alloc: %d already registered", type));
3403	KASSERT(if_com_free[type] == NULL,
3404	    ("if_register_com_alloc: %d free already registered", type));
3405
3406	if_com_alloc[type] = a;
3407	if_com_free[type] = f;
3408}
3409
3410void
3411if_deregister_com_alloc(u_char type)
3412{
3413
3414	KASSERT(if_com_alloc[type] != NULL,
3415	    ("if_deregister_com_alloc: %d not registered", type));
3416	KASSERT(if_com_free[type] != NULL,
3417	    ("if_deregister_com_alloc: %d free not registered", type));
3418	if_com_alloc[type] = NULL;
3419	if_com_free[type] = NULL;
3420}
3421