xref: /freebsd/sys/net/if.c (revision 282a3889ebf826db9839be296ff1dd903f6d6d6e)
1 /*-
2  * Copyright (c) 1980, 1986, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if.c	8.5 (Berkeley) 1/9/95
30  * $FreeBSD$
31  */
32 
33 #include "opt_compat.h"
34 #include "opt_inet6.h"
35 #include "opt_inet.h"
36 #include "opt_mac.h"
37 #include "opt_carp.h"
38 
39 #include <sys/param.h>
40 #include <sys/types.h>
41 #include <sys/conf.h>
42 #include <sys/malloc.h>
43 #include <sys/sbuf.h>
44 #include <sys/bus.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/sockio.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 #include <sys/taskqueue.h>
57 #include <sys/domain.h>
58 #include <sys/jail.h>
59 #include <machine/stdarg.h>
60 
61 #include <net/if.h>
62 #include <net/if_clone.h>
63 #include <net/if_dl.h>
64 #include <net/if_types.h>
65 #include <net/if_var.h>
66 #include <net/radix.h>
67 #include <net/route.h>
68 
69 #if defined(INET) || defined(INET6)
70 /*XXX*/
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #ifdef INET6
74 #include <netinet6/in6_var.h>
75 #include <netinet6/in6_ifattach.h>
76 #endif
77 #endif
78 #ifdef INET
79 #include <netinet/if_ether.h>
80 #endif
81 #ifdef DEV_CARP
82 #include <netinet/ip_carp.h>
83 #endif
84 
85 #include <security/mac/mac_framework.h>
86 
87 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
88 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
89 
90 /* Log link state change events */
91 static int log_link_state_change = 1;
92 
93 SYSCTL_INT(_net_link, OID_AUTO, log_link_state_change, CTLFLAG_RW,
94 	&log_link_state_change, 0,
95 	"log interface link state change events");
96 
97 void	(*bstp_linkstate_p)(struct ifnet *ifp, int state);
98 void	(*ng_ether_link_state_p)(struct ifnet *ifp, int state);
99 void	(*lagg_linkstate_p)(struct ifnet *ifp, int state);
100 
101 struct mbuf *(*tbr_dequeue_ptr)(struct ifaltq *, int) = NULL;
102 
103 /*
104  * XXX: Style; these should be sorted alphabetically, and unprototyped
105  * static functions should be prototyped. Currently they are sorted by
106  * declaration order.
107  */
108 static void	if_attachdomain(void *);
109 static void	if_attachdomain1(struct ifnet *);
110 static void	if_purgemaddrs(struct ifnet *);
111 static int	ifconf(u_long, caddr_t);
112 static void	if_freemulti(struct ifmultiaddr *);
113 static void	if_grow(void);
114 static void	if_init(void *);
115 static void	if_check(void *);
116 static void	if_qflush(struct ifaltq *);
117 static void	if_route(struct ifnet *, int flag, int fam);
118 static int	if_setflag(struct ifnet *, int, int, int *, int);
119 static void	if_slowtimo(void *);
120 static void	if_unroute(struct ifnet *, int flag, int fam);
121 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
122 static int	if_rtdel(struct radix_node *, void *);
123 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
124 static int	if_delmulti_locked(struct ifnet *, struct ifmultiaddr *, int);
125 static void	if_start_deferred(void *context, int pending);
126 static void	do_link_state_change(void *, int);
127 static int	if_getgroup(struct ifgroupreq *, struct ifnet *);
128 static int	if_getgroupmembers(struct ifgroupreq *);
129 #ifdef INET6
130 /*
131  * XXX: declare here to avoid to include many inet6 related files..
132  * should be more generalized?
133  */
134 extern void	nd6_setmtu(struct ifnet *);
135 #endif
136 
137 int	if_index = 0;
138 struct	ifindex_entry *ifindex_table = NULL;
139 int	ifqmaxlen = IFQ_MAXLEN;
140 struct	ifnethead ifnet;	/* depend on static init XXX */
141 struct	ifgrouphead ifg_head;
142 struct	mtx ifnet_lock;
143 static	if_com_alloc_t *if_com_alloc[256];
144 static	if_com_free_t *if_com_free[256];
145 
146 static int	if_indexlim = 8;
147 static struct	knlist ifklist;
148 
149 static void	filt_netdetach(struct knote *kn);
150 static int	filt_netdev(struct knote *kn, long hint);
151 
152 static struct filterops netdev_filtops =
153     { 1, NULL, filt_netdetach, filt_netdev };
154 
155 /*
156  * System initialization
157  */
158 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL)
159 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL)
160 
161 MALLOC_DEFINE(M_IFNET, "ifnet", "interface internals");
162 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
163 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
164 
165 static d_open_t		netopen;
166 static d_close_t	netclose;
167 static d_ioctl_t	netioctl;
168 static d_kqfilter_t	netkqfilter;
169 
170 static struct cdevsw net_cdevsw = {
171 	.d_version =	D_VERSION,
172 	.d_flags =	D_NEEDGIANT,
173 	.d_open =	netopen,
174 	.d_close =	netclose,
175 	.d_ioctl =	netioctl,
176 	.d_name =	"net",
177 	.d_kqfilter =	netkqfilter,
178 };
179 
180 static int
181 netopen(struct cdev *dev, int flag, int mode, struct thread *td)
182 {
183 	return (0);
184 }
185 
186 static int
187 netclose(struct cdev *dev, int flags, int fmt, struct thread *td)
188 {
189 	return (0);
190 }
191 
192 static int
193 netioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
194 {
195 	struct ifnet *ifp;
196 	int error, idx;
197 
198 	/* only support interface specific ioctls */
199 	if (IOCGROUP(cmd) != 'i')
200 		return (EOPNOTSUPP);
201 	idx = minor(dev);
202 	if (idx == 0) {
203 		/*
204 		 * special network device, not interface.
205 		 */
206 		if (cmd == SIOCGIFCONF)
207 			return (ifconf(cmd, data));	/* XXX remove cmd */
208 #ifdef __amd64__
209 		if (cmd == SIOCGIFCONF32)
210 			return (ifconf(cmd, data));	/* XXX remove cmd */
211 #endif
212 		return (EOPNOTSUPP);
213 	}
214 
215 	ifp = ifnet_byindex(idx);
216 	if (ifp == NULL)
217 		return (ENXIO);
218 
219 	error = ifhwioctl(cmd, ifp, data, td);
220 	if (error == ENOIOCTL)
221 		error = EOPNOTSUPP;
222 	return (error);
223 }
224 
225 static int
226 netkqfilter(struct cdev *dev, struct knote *kn)
227 {
228 	struct knlist *klist;
229 	struct ifnet *ifp;
230 	int idx;
231 
232 	switch (kn->kn_filter) {
233 	case EVFILT_NETDEV:
234 		kn->kn_fop = &netdev_filtops;
235 		break;
236 	default:
237 		return (EINVAL);
238 	}
239 
240 	idx = minor(dev);
241 	if (idx == 0) {
242 		klist = &ifklist;
243 	} else {
244 		ifp = ifnet_byindex(idx);
245 		if (ifp == NULL)
246 			return (1);
247 		klist = &ifp->if_klist;
248 	}
249 
250 	kn->kn_hook = (caddr_t)klist;
251 
252 	knlist_add(klist, kn, 0);
253 
254 	return (0);
255 }
256 
257 static void
258 filt_netdetach(struct knote *kn)
259 {
260 	struct knlist *klist = (struct knlist *)kn->kn_hook;
261 
262 	knlist_remove(klist, kn, 0);
263 }
264 
265 static int
266 filt_netdev(struct knote *kn, long hint)
267 {
268 	struct knlist *klist = (struct knlist *)kn->kn_hook;
269 
270 	/*
271 	 * Currently NOTE_EXIT is abused to indicate device detach.
272 	 */
273 	if (hint == NOTE_EXIT) {
274 		kn->kn_data = NOTE_LINKINV;
275 		kn->kn_flags |= (EV_EOF | EV_ONESHOT);
276 		knlist_remove_inevent(klist, kn);
277 		return (1);
278 	}
279 	if (hint != 0)
280 		kn->kn_data = hint;			/* current status */
281 	if (kn->kn_sfflags & hint)
282 		kn->kn_fflags |= hint;
283 	return (kn->kn_fflags != 0);
284 }
285 
286 /*
287  * Network interface utility routines.
288  *
289  * Routines with ifa_ifwith* names take sockaddr *'s as
290  * parameters.
291  */
292 
293 /* ARGSUSED*/
294 static void
295 if_init(void *dummy __unused)
296 {
297 
298 	IFNET_LOCK_INIT();
299 	TAILQ_INIT(&ifnet);
300 	TAILQ_INIT(&ifg_head);
301 	knlist_init(&ifklist, NULL, NULL, NULL, NULL);
302 	if_grow();				/* create initial table */
303 	ifdev_byindex(0) = make_dev(&net_cdevsw, 0,
304 	    UID_ROOT, GID_WHEEL, 0600, "network");
305 	if_clone_init();
306 }
307 
308 static void
309 if_grow(void)
310 {
311 	u_int n;
312 	struct ifindex_entry *e;
313 
314 	if_indexlim <<= 1;
315 	n = if_indexlim * sizeof(*e);
316 	e = malloc(n, M_IFNET, M_WAITOK | M_ZERO);
317 	if (ifindex_table != NULL) {
318 		memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2);
319 		free((caddr_t)ifindex_table, M_IFNET);
320 	}
321 	ifindex_table = e;
322 }
323 
324 /* ARGSUSED*/
325 static void
326 if_check(void *dummy __unused)
327 {
328 	struct ifnet *ifp;
329 	int s;
330 
331 	s = splimp();
332 	IFNET_RLOCK();	/* could sleep on rare error; mostly okay XXX */
333 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
334 		if (ifp->if_snd.ifq_maxlen == 0) {
335 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
336 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
337 		}
338 		if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) {
339 			if_printf(ifp,
340 			    "XXX: driver didn't initialize queue mtx\n");
341 			mtx_init(&ifp->if_snd.ifq_mtx, "unknown",
342 			    MTX_NETWORK_LOCK, MTX_DEF);
343 		}
344 	}
345 	IFNET_RUNLOCK();
346 	splx(s);
347 	if_slowtimo(0);
348 }
349 
350 /*
351  * Allocate a struct ifnet and an index for an interface.  A layer 2
352  * common structure will also be allocated if an allocation routine is
353  * registered for the passed type.
354  */
355 struct ifnet*
356 if_alloc(u_char type)
357 {
358 	struct ifnet *ifp;
359 
360 	ifp = malloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
361 
362 	/*
363 	 * Try to find an empty slot below if_index.  If we fail, take
364 	 * the next slot.
365 	 *
366 	 * XXX: should be locked!
367 	 */
368 	for (ifp->if_index = 1; ifp->if_index <= if_index; ifp->if_index++) {
369 		if (ifnet_byindex(ifp->if_index) == NULL)
370 			break;
371 	}
372 	/* Catch if_index overflow. */
373 	if (ifp->if_index < 1) {
374 		free(ifp, M_IFNET);
375 		return (NULL);
376 	}
377 	if (ifp->if_index > if_index)
378 		if_index = ifp->if_index;
379 	if (if_index >= if_indexlim)
380 		if_grow();
381 	ifnet_byindex(ifp->if_index) = ifp;
382 
383 	ifp->if_type = type;
384 
385 	if (if_com_alloc[type] != NULL) {
386 		ifp->if_l2com = if_com_alloc[type](type, ifp);
387 		if (ifp->if_l2com == NULL) {
388 			free(ifp, M_IFNET);
389 			return (NULL);
390 		}
391 	}
392 	IF_ADDR_LOCK_INIT(ifp);
393 
394 	return (ifp);
395 }
396 
397 /*
398  * Free the struct ifnet, the associated index, and the layer 2 common
399  * structure if needed.  All the work is done in if_free_type().
400  *
401  * Do not add code to this function!  Add it to if_free_type().
402  */
403 void
404 if_free(struct ifnet *ifp)
405 {
406 
407 	if_free_type(ifp, ifp->if_type);
408 }
409 
410 /*
411  * Do the actual work of freeing a struct ifnet, associated index, and
412  * layer 2 common structure.  This version should only be called by
413  * intefaces that switch their type after calling if_alloc().
414  */
415 void
416 if_free_type(struct ifnet *ifp, u_char type)
417 {
418 
419 	if (ifp != ifnet_byindex(ifp->if_index)) {
420 		if_printf(ifp, "%s: value was not if_alloced, skipping\n",
421 		    __func__);
422 		return;
423 	}
424 
425 	IF_ADDR_LOCK_DESTROY(ifp);
426 
427 	ifnet_byindex(ifp->if_index) = NULL;
428 
429 	/* XXX: should be locked with if_findindex() */
430 	while (if_index > 0 && ifnet_byindex(if_index) == NULL)
431 		if_index--;
432 
433 	if (if_com_free[type] != NULL)
434 		if_com_free[type](ifp->if_l2com, type);
435 
436 	free(ifp, M_IFNET);
437 };
438 
439 /*
440  * Perform generic interface initalization tasks and attach the interface
441  * to the list of "active" interfaces.
442  *
443  * XXX:
444  *  - The decision to return void and thus require this function to
445  *    succeed is questionable.
446  *  - We do more initialization here then is probably a good idea.
447  *    Some of this should probably move to if_alloc().
448  *  - We should probably do more sanity checking.  For instance we don't
449  *    do anything to insure if_xname is unique or non-empty.
450  */
451 void
452 if_attach(struct ifnet *ifp)
453 {
454 	unsigned socksize, ifasize;
455 	int namelen, masklen;
456 	struct sockaddr_dl *sdl;
457 	struct ifaddr *ifa;
458 
459 	if (ifp->if_index == 0 || ifp != ifnet_byindex(ifp->if_index))
460 		panic ("%s: BUG: if_attach called without if_alloc'd input()\n",
461 		    ifp->if_xname);
462 
463 	TASK_INIT(&ifp->if_starttask, 0, if_start_deferred, ifp);
464 	TASK_INIT(&ifp->if_linktask, 0, do_link_state_change, ifp);
465 	IF_AFDATA_LOCK_INIT(ifp);
466 	ifp->if_afdata_initialized = 0;
467 
468 	TAILQ_INIT(&ifp->if_addrhead);
469 	TAILQ_INIT(&ifp->if_prefixhead);
470 	TAILQ_INIT(&ifp->if_multiaddrs);
471 	TAILQ_INIT(&ifp->if_groups);
472 
473 	if_addgroup(ifp, IFG_ALL);
474 
475 	knlist_init(&ifp->if_klist, NULL, NULL, NULL, NULL);
476 	getmicrotime(&ifp->if_lastchange);
477 	ifp->if_data.ifi_epoch = time_uptime;
478 	ifp->if_data.ifi_datalen = sizeof(struct if_data);
479 
480 #ifdef MAC
481 	mac_init_ifnet(ifp);
482 	mac_create_ifnet(ifp);
483 #endif
484 
485 	ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw,
486 	    unit2minor(ifp->if_index),
487 	    UID_ROOT, GID_WHEEL, 0600, "%s/%s",
488 	    net_cdevsw.d_name, ifp->if_xname);
489 	make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d",
490 	    net_cdevsw.d_name, ifp->if_index);
491 
492 	mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_xname, "if send queue", MTX_DEF);
493 
494 	/*
495 	 * create a Link Level name for this device
496 	 */
497 	namelen = strlen(ifp->if_xname);
498 	/*
499 	 * Always save enough space for any possiable name so we can do
500 	 * a rename in place later.
501 	 */
502 	masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + IFNAMSIZ;
503 	socksize = masklen + ifp->if_addrlen;
504 	if (socksize < sizeof(*sdl))
505 		socksize = sizeof(*sdl);
506 	socksize = roundup2(socksize, sizeof(long));
507 	ifasize = sizeof(*ifa) + 2 * socksize;
508 	ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
509 	IFA_LOCK_INIT(ifa);
510 	sdl = (struct sockaddr_dl *)(ifa + 1);
511 	sdl->sdl_len = socksize;
512 	sdl->sdl_family = AF_LINK;
513 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
514 	sdl->sdl_nlen = namelen;
515 	sdl->sdl_index = ifp->if_index;
516 	sdl->sdl_type = ifp->if_type;
517 	ifp->if_addr = ifa;
518 	ifa->ifa_ifp = ifp;
519 	ifa->ifa_rtrequest = link_rtrequest;
520 	ifa->ifa_addr = (struct sockaddr *)sdl;
521 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
522 	ifa->ifa_netmask = (struct sockaddr *)sdl;
523 	sdl->sdl_len = masklen;
524 	while (namelen != 0)
525 		sdl->sdl_data[--namelen] = 0xff;
526 	ifa->ifa_refcnt = 1;
527 	TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
528 	ifp->if_broadcastaddr = NULL; /* reliably crash if used uninitialized */
529 	ifp->if_snd.altq_type = 0;
530 	ifp->if_snd.altq_disc = NULL;
531 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
532 	ifp->if_snd.altq_tbr  = NULL;
533 	ifp->if_snd.altq_ifp  = ifp;
534 
535 	IFNET_WLOCK();
536 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
537 	IFNET_WUNLOCK();
538 
539 	if (domain_init_status >= 2)
540 		if_attachdomain1(ifp);
541 
542 	EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
543 	devctl_notify("IFNET", ifp->if_xname, "ATTACH", NULL);
544 
545 	/* Announce the interface. */
546 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
547 
548 	if (ifp->if_watchdog != NULL)
549 		if_printf(ifp, "using obsoleted if_watchdog interface\n");
550 }
551 
552 static void
553 if_attachdomain(void *dummy)
554 {
555 	struct ifnet *ifp;
556 	int s;
557 
558 	s = splnet();
559 	TAILQ_FOREACH(ifp, &ifnet, if_link)
560 		if_attachdomain1(ifp);
561 	splx(s);
562 }
563 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_SECOND,
564     if_attachdomain, NULL);
565 
566 static void
567 if_attachdomain1(struct ifnet *ifp)
568 {
569 	struct domain *dp;
570 	int s;
571 
572 	s = splnet();
573 
574 	/*
575 	 * Since dp->dom_ifattach calls malloc() with M_WAITOK, we
576 	 * cannot lock ifp->if_afdata initialization, entirely.
577 	 */
578 	if (IF_AFDATA_TRYLOCK(ifp) == 0) {
579 		splx(s);
580 		return;
581 	}
582 	if (ifp->if_afdata_initialized >= domain_init_status) {
583 		IF_AFDATA_UNLOCK(ifp);
584 		splx(s);
585 		printf("if_attachdomain called more than once on %s\n",
586 		    ifp->if_xname);
587 		return;
588 	}
589 	ifp->if_afdata_initialized = domain_init_status;
590 	IF_AFDATA_UNLOCK(ifp);
591 
592 	/* address family dependent data region */
593 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
594 	for (dp = domains; dp; dp = dp->dom_next) {
595 		if (dp->dom_ifattach)
596 			ifp->if_afdata[dp->dom_family] =
597 			    (*dp->dom_ifattach)(ifp);
598 	}
599 
600 	splx(s);
601 }
602 
603 /*
604  * Remove any unicast or broadcast network addresses from an interface.
605  */
606 void
607 if_purgeaddrs(struct ifnet *ifp)
608 {
609 	struct ifaddr *ifa, *next;
610 
611 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
612 		if (ifa->ifa_addr->sa_family == AF_LINK)
613 			continue;
614 #ifdef INET
615 		/* XXX: Ugly!! ad hoc just for INET */
616 		if (ifa->ifa_addr->sa_family == AF_INET) {
617 			struct ifaliasreq ifr;
618 
619 			bzero(&ifr, sizeof(ifr));
620 			ifr.ifra_addr = *ifa->ifa_addr;
621 			if (ifa->ifa_dstaddr)
622 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
623 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
624 			    NULL) == 0)
625 				continue;
626 		}
627 #endif /* INET */
628 #ifdef INET6
629 		if (ifa->ifa_addr->sa_family == AF_INET6) {
630 			in6_purgeaddr(ifa);
631 			/* ifp_addrhead is already updated */
632 			continue;
633 		}
634 #endif /* INET6 */
635 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
636 		IFAFREE(ifa);
637 	}
638 }
639 
640 /*
641  * Remove any multicast network addresses from an interface.
642  */
643 static void
644 if_purgemaddrs(struct ifnet *ifp)
645 {
646 	struct ifmultiaddr *ifma;
647 	struct ifmultiaddr *next;
648 
649 	IF_ADDR_LOCK(ifp);
650 	TAILQ_FOREACH_SAFE(ifma, &ifp->if_multiaddrs, ifma_link, next)
651 		if_delmulti_locked(ifp, ifma, 1);
652 	IF_ADDR_UNLOCK(ifp);
653 }
654 
655 /*
656  * Detach an interface, removing it from the
657  * list of "active" interfaces.
658  *
659  * XXXRW: There are some significant questions about event ordering, and
660  * how to prevent things from starting to use the interface during detach.
661  */
662 void
663 if_detach(struct ifnet *ifp)
664 {
665 	struct ifaddr *ifa;
666 	struct radix_node_head	*rnh;
667 	int s;
668 	int i;
669 	struct domain *dp;
670  	struct ifnet *iter;
671  	int found = 0;
672 
673 	IFNET_WLOCK();
674 	TAILQ_FOREACH(iter, &ifnet, if_link)
675 		if (iter == ifp) {
676 			TAILQ_REMOVE(&ifnet, ifp, if_link);
677 			found = 1;
678 			break;
679 		}
680 	IFNET_WUNLOCK();
681 	if (!found)
682 		return;
683 
684 	/*
685 	 * Remove/wait for pending events.
686 	 */
687 	taskqueue_drain(taskqueue_swi, &ifp->if_linktask);
688 
689 	/*
690 	 * Remove routes and flush queues.
691 	 */
692 	s = splnet();
693 	if_down(ifp);
694 #ifdef ALTQ
695 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
696 		altq_disable(&ifp->if_snd);
697 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
698 		altq_detach(&ifp->if_snd);
699 #endif
700 
701 	if_purgeaddrs(ifp);
702 
703 #ifdef INET
704 	in_ifdetach(ifp);
705 #endif
706 
707 #ifdef INET6
708 	/*
709 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
710 	 * before removing routing entries below, since IPv6 interface direct
711 	 * routes are expected to be removed by the IPv6-specific kernel API.
712 	 * Otherwise, the kernel will detect some inconsistency and bark it.
713 	 */
714 	in6_ifdetach(ifp);
715 #endif
716 	if_purgemaddrs(ifp);
717 
718 	/*
719 	 * Remove link ifaddr pointer and maybe decrement if_index.
720 	 * Clean up all addresses.
721 	 */
722 	ifp->if_addr = NULL;
723 	destroy_dev(ifdev_byindex(ifp->if_index));
724 	ifdev_byindex(ifp->if_index) = NULL;
725 
726 	/* We can now free link ifaddr. */
727 	if (!TAILQ_EMPTY(&ifp->if_addrhead)) {
728 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
729 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
730 		IFAFREE(ifa);
731 	}
732 
733 	/*
734 	 * Delete all remaining routes using this interface
735 	 * Unfortuneatly the only way to do this is to slog through
736 	 * the entire routing table looking for routes which point
737 	 * to this interface...oh well...
738 	 */
739 	for (i = 1; i <= AF_MAX; i++) {
740 		if ((rnh = rt_tables[i]) == NULL)
741 			continue;
742 		RADIX_NODE_HEAD_LOCK(rnh);
743 		(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
744 		RADIX_NODE_HEAD_UNLOCK(rnh);
745 	}
746 
747 	/* Announce that the interface is gone. */
748 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
749 	EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
750 	devctl_notify("IFNET", ifp->if_xname, "DETACH", NULL);
751 
752 	IF_AFDATA_LOCK(ifp);
753 	for (dp = domains; dp; dp = dp->dom_next) {
754 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
755 			(*dp->dom_ifdetach)(ifp,
756 			    ifp->if_afdata[dp->dom_family]);
757 	}
758 	IF_AFDATA_UNLOCK(ifp);
759 
760 #ifdef MAC
761 	mac_destroy_ifnet(ifp);
762 #endif /* MAC */
763 	KNOTE_UNLOCKED(&ifp->if_klist, NOTE_EXIT);
764 	knlist_clear(&ifp->if_klist, 0);
765 	knlist_destroy(&ifp->if_klist);
766 	mtx_destroy(&ifp->if_snd.ifq_mtx);
767 	IF_AFDATA_DESTROY(ifp);
768 	splx(s);
769 }
770 
771 /*
772  * Add a group to an interface
773  */
774 int
775 if_addgroup(struct ifnet *ifp, const char *groupname)
776 {
777 	struct ifg_list		*ifgl;
778 	struct ifg_group	*ifg = NULL;
779 	struct ifg_member	*ifgm;
780 
781 	if (groupname[0] && groupname[strlen(groupname) - 1] >= '0' &&
782 	    groupname[strlen(groupname) - 1] <= '9')
783 		return (EINVAL);
784 
785 	IFNET_WLOCK();
786 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
787 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname)) {
788 			IFNET_WUNLOCK();
789 			return (EEXIST);
790 		}
791 
792 	if ((ifgl = (struct ifg_list *)malloc(sizeof(struct ifg_list), M_TEMP,
793 	    M_NOWAIT)) == NULL) {
794 	    	IFNET_WUNLOCK();
795 		return (ENOMEM);
796 	}
797 
798 	if ((ifgm = (struct ifg_member *)malloc(sizeof(struct ifg_member),
799 	    M_TEMP, M_NOWAIT)) == NULL) {
800 		free(ifgl, M_TEMP);
801 		IFNET_WUNLOCK();
802 		return (ENOMEM);
803 	}
804 
805 	TAILQ_FOREACH(ifg, &ifg_head, ifg_next)
806 		if (!strcmp(ifg->ifg_group, groupname))
807 			break;
808 
809 	if (ifg == NULL) {
810 		if ((ifg = (struct ifg_group *)malloc(sizeof(struct ifg_group),
811 		    M_TEMP, M_NOWAIT)) == NULL) {
812 			free(ifgl, M_TEMP);
813 			free(ifgm, M_TEMP);
814 			IFNET_WUNLOCK();
815 			return (ENOMEM);
816 		}
817 		strlcpy(ifg->ifg_group, groupname, sizeof(ifg->ifg_group));
818 		ifg->ifg_refcnt = 0;
819 		TAILQ_INIT(&ifg->ifg_members);
820 		EVENTHANDLER_INVOKE(group_attach_event, ifg);
821 		TAILQ_INSERT_TAIL(&ifg_head, ifg, ifg_next);
822 	}
823 
824 	ifg->ifg_refcnt++;
825 	ifgl->ifgl_group = ifg;
826 	ifgm->ifgm_ifp = ifp;
827 
828 	IF_ADDR_LOCK(ifp);
829 	TAILQ_INSERT_TAIL(&ifg->ifg_members, ifgm, ifgm_next);
830 	TAILQ_INSERT_TAIL(&ifp->if_groups, ifgl, ifgl_next);
831 	IF_ADDR_UNLOCK(ifp);
832 
833 	IFNET_WUNLOCK();
834 
835 	EVENTHANDLER_INVOKE(group_change_event, groupname);
836 
837 	return (0);
838 }
839 
840 /*
841  * Remove a group from an interface
842  */
843 int
844 if_delgroup(struct ifnet *ifp, const char *groupname)
845 {
846 	struct ifg_list		*ifgl;
847 	struct ifg_member	*ifgm;
848 
849 	IFNET_WLOCK();
850 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
851 		if (!strcmp(ifgl->ifgl_group->ifg_group, groupname))
852 			break;
853 	if (ifgl == NULL) {
854 		IFNET_WUNLOCK();
855 		return (ENOENT);
856 	}
857 
858 	IF_ADDR_LOCK(ifp);
859 	TAILQ_REMOVE(&ifp->if_groups, ifgl, ifgl_next);
860 	IF_ADDR_UNLOCK(ifp);
861 
862 	TAILQ_FOREACH(ifgm, &ifgl->ifgl_group->ifg_members, ifgm_next)
863 		if (ifgm->ifgm_ifp == ifp)
864 			break;
865 
866 	if (ifgm != NULL) {
867 		TAILQ_REMOVE(&ifgl->ifgl_group->ifg_members, ifgm, ifgm_next);
868 		free(ifgm, M_TEMP);
869 	}
870 
871 	if (--ifgl->ifgl_group->ifg_refcnt == 0) {
872 		TAILQ_REMOVE(&ifg_head, ifgl->ifgl_group, ifg_next);
873 		EVENTHANDLER_INVOKE(group_detach_event, ifgl->ifgl_group);
874 		free(ifgl->ifgl_group, M_TEMP);
875 	}
876 	IFNET_WUNLOCK();
877 
878 	free(ifgl, M_TEMP);
879 
880 	EVENTHANDLER_INVOKE(group_change_event, groupname);
881 
882 	return (0);
883 }
884 
885 /*
886  * Stores all groups from an interface in memory pointed
887  * to by data
888  */
889 static int
890 if_getgroup(struct ifgroupreq *data, struct ifnet *ifp)
891 {
892 	int			 len, error;
893 	struct ifg_list		*ifgl;
894 	struct ifg_req		 ifgrq, *ifgp;
895 	struct ifgroupreq	*ifgr = data;
896 
897 	if (ifgr->ifgr_len == 0) {
898 		IF_ADDR_LOCK(ifp);
899 		TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next)
900 			ifgr->ifgr_len += sizeof(struct ifg_req);
901 		IF_ADDR_UNLOCK(ifp);
902 		return (0);
903 	}
904 
905 	len = ifgr->ifgr_len;
906 	ifgp = ifgr->ifgr_groups;
907 	/* XXX: wire */
908 	IF_ADDR_LOCK(ifp);
909 	TAILQ_FOREACH(ifgl, &ifp->if_groups, ifgl_next) {
910 		if (len < sizeof(ifgrq)) {
911 			IF_ADDR_UNLOCK(ifp);
912 			return (EINVAL);
913 		}
914 		bzero(&ifgrq, sizeof ifgrq);
915 		strlcpy(ifgrq.ifgrq_group, ifgl->ifgl_group->ifg_group,
916 		    sizeof(ifgrq.ifgrq_group));
917 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
918 		    	IF_ADDR_UNLOCK(ifp);
919 			return (error);
920 		}
921 		len -= sizeof(ifgrq);
922 		ifgp++;
923 	}
924 	IF_ADDR_UNLOCK(ifp);
925 
926 	return (0);
927 }
928 
929 /*
930  * Stores all members of a group in memory pointed to by data
931  */
932 static int
933 if_getgroupmembers(struct ifgroupreq *data)
934 {
935 	struct ifgroupreq	*ifgr = data;
936 	struct ifg_group	*ifg;
937 	struct ifg_member	*ifgm;
938 	struct ifg_req		 ifgrq, *ifgp;
939 	int			 len, error;
940 
941 	IFNET_RLOCK();
942 	TAILQ_FOREACH(ifg, &ifg_head, ifg_next)
943 		if (!strcmp(ifg->ifg_group, ifgr->ifgr_name))
944 			break;
945 	if (ifg == NULL) {
946 		IFNET_RUNLOCK();
947 		return (ENOENT);
948 	}
949 
950 	if (ifgr->ifgr_len == 0) {
951 		TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next)
952 			ifgr->ifgr_len += sizeof(ifgrq);
953 		IFNET_RUNLOCK();
954 		return (0);
955 	}
956 
957 	len = ifgr->ifgr_len;
958 	ifgp = ifgr->ifgr_groups;
959 	TAILQ_FOREACH(ifgm, &ifg->ifg_members, ifgm_next) {
960 		if (len < sizeof(ifgrq)) {
961 			IFNET_RUNLOCK();
962 			return (EINVAL);
963 		}
964 		bzero(&ifgrq, sizeof ifgrq);
965 		strlcpy(ifgrq.ifgrq_member, ifgm->ifgm_ifp->if_xname,
966 		    sizeof(ifgrq.ifgrq_member));
967 		if ((error = copyout(&ifgrq, ifgp, sizeof(struct ifg_req)))) {
968 			IFNET_RUNLOCK();
969 			return (error);
970 		}
971 		len -= sizeof(ifgrq);
972 		ifgp++;
973 	}
974 	IFNET_RUNLOCK();
975 
976 	return (0);
977 }
978 
979 /*
980  * Delete Routes for a Network Interface
981  *
982  * Called for each routing entry via the rnh->rnh_walktree() call above
983  * to delete all route entries referencing a detaching network interface.
984  *
985  * Arguments:
986  *	rn	pointer to node in the routing table
987  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
988  *
989  * Returns:
990  *	0	successful
991  *	errno	failed - reason indicated
992  *
993  */
994 static int
995 if_rtdel(struct radix_node *rn, void *arg)
996 {
997 	struct rtentry	*rt = (struct rtentry *)rn;
998 	struct ifnet	*ifp = arg;
999 	int		err;
1000 
1001 	if (rt->rt_ifp == ifp) {
1002 
1003 		/*
1004 		 * Protect (sorta) against walktree recursion problems
1005 		 * with cloned routes
1006 		 */
1007 		if ((rt->rt_flags & RTF_UP) == 0)
1008 			return (0);
1009 
1010 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
1011 				rt_mask(rt), rt->rt_flags,
1012 				(struct rtentry **) NULL);
1013 		if (err) {
1014 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
1015 		}
1016 	}
1017 
1018 	return (0);
1019 }
1020 
1021 /*
1022  * XXX: Because sockaddr_dl has deeper structure than the sockaddr
1023  * structs used to represent other address families, it is necessary
1024  * to perform a different comparison.
1025  */
1026 
1027 #define	sa_equal(a1, a2)	\
1028 	(bcmp((a1), (a2), ((a1))->sa_len) == 0)
1029 
1030 #define	sa_dl_equal(a1, a2)	\
1031 	((((struct sockaddr_dl *)(a1))->sdl_len ==			\
1032 	 ((struct sockaddr_dl *)(a2))->sdl_len) &&			\
1033 	 (bcmp(LLADDR((struct sockaddr_dl *)(a1)),			\
1034 	       LLADDR((struct sockaddr_dl *)(a2)),			\
1035 	       ((struct sockaddr_dl *)(a1))->sdl_alen) == 0))
1036 
1037 /*
1038  * Locate an interface based on a complete address.
1039  */
1040 /*ARGSUSED*/
1041 struct ifaddr *
1042 ifa_ifwithaddr(struct sockaddr *addr)
1043 {
1044 	struct ifnet *ifp;
1045 	struct ifaddr *ifa;
1046 
1047 	IFNET_RLOCK();
1048 	TAILQ_FOREACH(ifp, &ifnet, if_link)
1049 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1050 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1051 				continue;
1052 			if (sa_equal(addr, ifa->ifa_addr))
1053 				goto done;
1054 			/* IP6 doesn't have broadcast */
1055 			if ((ifp->if_flags & IFF_BROADCAST) &&
1056 			    ifa->ifa_broadaddr &&
1057 			    ifa->ifa_broadaddr->sa_len != 0 &&
1058 			    sa_equal(ifa->ifa_broadaddr, addr))
1059 				goto done;
1060 		}
1061 	ifa = NULL;
1062 done:
1063 	IFNET_RUNLOCK();
1064 	return (ifa);
1065 }
1066 
1067 /*
1068  * Locate an interface based on the broadcast address.
1069  */
1070 /* ARGSUSED */
1071 struct ifaddr *
1072 ifa_ifwithbroadaddr(struct sockaddr *addr)
1073 {
1074 	struct ifnet *ifp;
1075 	struct ifaddr *ifa;
1076 
1077 	IFNET_RLOCK();
1078 	TAILQ_FOREACH(ifp, &ifnet, if_link)
1079 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1080 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1081 				continue;
1082 			if ((ifp->if_flags & IFF_BROADCAST) &&
1083 			    ifa->ifa_broadaddr &&
1084 			    ifa->ifa_broadaddr->sa_len != 0 &&
1085 			    sa_equal(ifa->ifa_broadaddr, addr))
1086 				goto done;
1087 		}
1088 	ifa = NULL;
1089 done:
1090 	IFNET_RUNLOCK();
1091 	return (ifa);
1092 }
1093 
1094 /*
1095  * Locate the point to point interface with a given destination address.
1096  */
1097 /*ARGSUSED*/
1098 struct ifaddr *
1099 ifa_ifwithdstaddr(struct sockaddr *addr)
1100 {
1101 	struct ifnet *ifp;
1102 	struct ifaddr *ifa;
1103 
1104 	IFNET_RLOCK();
1105 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1106 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
1107 			continue;
1108 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1109 			if (ifa->ifa_addr->sa_family != addr->sa_family)
1110 				continue;
1111 			if (ifa->ifa_dstaddr &&
1112 			    sa_equal(addr, ifa->ifa_dstaddr))
1113 				goto done;
1114 		}
1115 	}
1116 	ifa = NULL;
1117 done:
1118 	IFNET_RUNLOCK();
1119 	return (ifa);
1120 }
1121 
1122 /*
1123  * Find an interface on a specific network.  If many, choice
1124  * is most specific found.
1125  */
1126 struct ifaddr *
1127 ifa_ifwithnet(struct sockaddr *addr)
1128 {
1129 	struct ifnet *ifp;
1130 	struct ifaddr *ifa;
1131 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
1132 	u_int af = addr->sa_family;
1133 	char *addr_data = addr->sa_data, *cplim;
1134 
1135 	/*
1136 	 * AF_LINK addresses can be looked up directly by their index number,
1137 	 * so do that if we can.
1138 	 */
1139 	if (af == AF_LINK) {
1140 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
1141 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
1142 		return (ifaddr_byindex(sdl->sdl_index));
1143 	}
1144 
1145 	/*
1146 	 * Scan though each interface, looking for ones that have
1147 	 * addresses in this address family.
1148 	 */
1149 	IFNET_RLOCK();
1150 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1151 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1152 			char *cp, *cp2, *cp3;
1153 
1154 			if (ifa->ifa_addr->sa_family != af)
1155 next:				continue;
1156 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
1157 				/*
1158 				 * This is a bit broken as it doesn't
1159 				 * take into account that the remote end may
1160 				 * be a single node in the network we are
1161 				 * looking for.
1162 				 * The trouble is that we don't know the
1163 				 * netmask for the remote end.
1164 				 */
1165 				if (ifa->ifa_dstaddr != 0 &&
1166 				    sa_equal(addr, ifa->ifa_dstaddr))
1167 					goto done;
1168 			} else {
1169 				/*
1170 				 * if we have a special address handler,
1171 				 * then use it instead of the generic one.
1172 				 */
1173 				if (ifa->ifa_claim_addr) {
1174 					if ((*ifa->ifa_claim_addr)(ifa, addr))
1175 						goto done;
1176 					continue;
1177 				}
1178 
1179 				/*
1180 				 * Scan all the bits in the ifa's address.
1181 				 * If a bit dissagrees with what we are
1182 				 * looking for, mask it with the netmask
1183 				 * to see if it really matters.
1184 				 * (A byte at a time)
1185 				 */
1186 				if (ifa->ifa_netmask == 0)
1187 					continue;
1188 				cp = addr_data;
1189 				cp2 = ifa->ifa_addr->sa_data;
1190 				cp3 = ifa->ifa_netmask->sa_data;
1191 				cplim = ifa->ifa_netmask->sa_len
1192 					+ (char *)ifa->ifa_netmask;
1193 				while (cp3 < cplim)
1194 					if ((*cp++ ^ *cp2++) & *cp3++)
1195 						goto next; /* next address! */
1196 				/*
1197 				 * If the netmask of what we just found
1198 				 * is more specific than what we had before
1199 				 * (if we had one) then remember the new one
1200 				 * before continuing to search
1201 				 * for an even better one.
1202 				 */
1203 				if (ifa_maybe == 0 ||
1204 				    rn_refines((caddr_t)ifa->ifa_netmask,
1205 				    (caddr_t)ifa_maybe->ifa_netmask))
1206 					ifa_maybe = ifa;
1207 			}
1208 		}
1209 	}
1210 	ifa = ifa_maybe;
1211 done:
1212 	IFNET_RUNLOCK();
1213 	return (ifa);
1214 }
1215 
1216 /*
1217  * Find an interface address specific to an interface best matching
1218  * a given address.
1219  */
1220 struct ifaddr *
1221 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
1222 {
1223 	struct ifaddr *ifa;
1224 	char *cp, *cp2, *cp3;
1225 	char *cplim;
1226 	struct ifaddr *ifa_maybe = 0;
1227 	u_int af = addr->sa_family;
1228 
1229 	if (af >= AF_MAX)
1230 		return (0);
1231 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1232 		if (ifa->ifa_addr->sa_family != af)
1233 			continue;
1234 		if (ifa_maybe == 0)
1235 			ifa_maybe = ifa;
1236 		if (ifa->ifa_netmask == 0) {
1237 			if (sa_equal(addr, ifa->ifa_addr) ||
1238 			    (ifa->ifa_dstaddr &&
1239 			    sa_equal(addr, ifa->ifa_dstaddr)))
1240 				goto done;
1241 			continue;
1242 		}
1243 		if (ifp->if_flags & IFF_POINTOPOINT) {
1244 			if (sa_equal(addr, ifa->ifa_dstaddr))
1245 				goto done;
1246 		} else {
1247 			cp = addr->sa_data;
1248 			cp2 = ifa->ifa_addr->sa_data;
1249 			cp3 = ifa->ifa_netmask->sa_data;
1250 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1251 			for (; cp3 < cplim; cp3++)
1252 				if ((*cp++ ^ *cp2++) & *cp3)
1253 					break;
1254 			if (cp3 == cplim)
1255 				goto done;
1256 		}
1257 	}
1258 	ifa = ifa_maybe;
1259 done:
1260 	return (ifa);
1261 }
1262 
1263 #include <net/route.h>
1264 
1265 /*
1266  * Default action when installing a route with a Link Level gateway.
1267  * Lookup an appropriate real ifa to point to.
1268  * This should be moved to /sys/net/link.c eventually.
1269  */
1270 static void
1271 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
1272 {
1273 	struct ifaddr *ifa, *oifa;
1274 	struct sockaddr *dst;
1275 	struct ifnet *ifp;
1276 
1277 	RT_LOCK_ASSERT(rt);
1278 
1279 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1280 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1281 		return;
1282 	ifa = ifaof_ifpforaddr(dst, ifp);
1283 	if (ifa) {
1284 		IFAREF(ifa);		/* XXX */
1285 		oifa = rt->rt_ifa;
1286 		rt->rt_ifa = ifa;
1287 		IFAFREE(oifa);
1288 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1289 			ifa->ifa_rtrequest(cmd, rt, info);
1290 	}
1291 }
1292 
1293 /*
1294  * Mark an interface down and notify protocols of
1295  * the transition.
1296  * NOTE: must be called at splnet or eqivalent.
1297  */
1298 static void
1299 if_unroute(struct ifnet *ifp, int flag, int fam)
1300 {
1301 	struct ifaddr *ifa;
1302 
1303 	KASSERT(flag == IFF_UP, ("if_unroute: flag != IFF_UP"));
1304 
1305 	ifp->if_flags &= ~flag;
1306 	getmicrotime(&ifp->if_lastchange);
1307 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1308 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1309 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1310 	if_qflush(&ifp->if_snd);
1311 #ifdef DEV_CARP
1312 	if (ifp->if_carp)
1313 		carp_carpdev_state(ifp->if_carp);
1314 #endif
1315 	rt_ifmsg(ifp);
1316 }
1317 
1318 /*
1319  * Mark an interface up and notify protocols of
1320  * the transition.
1321  * NOTE: must be called at splnet or eqivalent.
1322  */
1323 static void
1324 if_route(struct ifnet *ifp, int flag, int fam)
1325 {
1326 	struct ifaddr *ifa;
1327 
1328 	KASSERT(flag == IFF_UP, ("if_route: flag != IFF_UP"));
1329 
1330 	ifp->if_flags |= flag;
1331 	getmicrotime(&ifp->if_lastchange);
1332 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1333 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1334 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
1335 #ifdef DEV_CARP
1336 	if (ifp->if_carp)
1337 		carp_carpdev_state(ifp->if_carp);
1338 #endif
1339 	rt_ifmsg(ifp);
1340 #ifdef INET6
1341 	in6_if_up(ifp);
1342 #endif
1343 }
1344 
1345 void	(*vlan_link_state_p)(struct ifnet *, int);	/* XXX: private from if_vlan */
1346 void	(*vlan_trunk_cap_p)(struct ifnet *);		/* XXX: private from if_vlan */
1347 
1348 /*
1349  * Handle a change in the interface link state. To avoid LORs
1350  * between driver lock and upper layer locks, as well as possible
1351  * recursions, we post event to taskqueue, and all job
1352  * is done in static do_link_state_change().
1353  */
1354 void
1355 if_link_state_change(struct ifnet *ifp, int link_state)
1356 {
1357 	/* Return if state hasn't changed. */
1358 	if (ifp->if_link_state == link_state)
1359 		return;
1360 
1361 	ifp->if_link_state = link_state;
1362 
1363 	taskqueue_enqueue(taskqueue_swi, &ifp->if_linktask);
1364 }
1365 
1366 static void
1367 do_link_state_change(void *arg, int pending)
1368 {
1369 	struct ifnet *ifp = (struct ifnet *)arg;
1370 	int link_state = ifp->if_link_state;
1371 	int link;
1372 
1373 	/* Notify that the link state has changed. */
1374 	rt_ifmsg(ifp);
1375 	if (link_state == LINK_STATE_UP)
1376 		link = NOTE_LINKUP;
1377 	else if (link_state == LINK_STATE_DOWN)
1378 		link = NOTE_LINKDOWN;
1379 	else
1380 		link = NOTE_LINKINV;
1381 	KNOTE_UNLOCKED(&ifp->if_klist, link);
1382 	if (ifp->if_vlantrunk != NULL)
1383 		(*vlan_link_state_p)(ifp, link);
1384 
1385 	if ((ifp->if_type == IFT_ETHER || ifp->if_type == IFT_L2VLAN) &&
1386 	    IFP2AC(ifp)->ac_netgraph != NULL)
1387 		(*ng_ether_link_state_p)(ifp, link_state);
1388 #ifdef DEV_CARP
1389 	if (ifp->if_carp)
1390 		carp_carpdev_state(ifp->if_carp);
1391 #endif
1392 	if (ifp->if_bridge) {
1393 		KASSERT(bstp_linkstate_p != NULL,("if_bridge bstp not loaded!"));
1394 		(*bstp_linkstate_p)(ifp, link_state);
1395 	}
1396 	if (ifp->if_lagg) {
1397 		KASSERT(lagg_linkstate_p != NULL,("if_lagg not loaded!"));
1398 		(*lagg_linkstate_p)(ifp, link_state);
1399 	}
1400 
1401 	devctl_notify("IFNET", ifp->if_xname,
1402 	    (link_state == LINK_STATE_UP) ? "LINK_UP" : "LINK_DOWN", NULL);
1403 	if (pending > 1)
1404 		if_printf(ifp, "%d link states coalesced\n", pending);
1405 	if (log_link_state_change)
1406 		log(LOG_NOTICE, "%s: link state changed to %s\n", ifp->if_xname,
1407 		    (link_state == LINK_STATE_UP) ? "UP" : "DOWN" );
1408 }
1409 
1410 /*
1411  * Mark an interface down and notify protocols of
1412  * the transition.
1413  * NOTE: must be called at splnet or eqivalent.
1414  */
1415 void
1416 if_down(struct ifnet *ifp)
1417 {
1418 
1419 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
1420 }
1421 
1422 /*
1423  * Mark an interface up and notify protocols of
1424  * the transition.
1425  * NOTE: must be called at splnet or eqivalent.
1426  */
1427 void
1428 if_up(struct ifnet *ifp)
1429 {
1430 
1431 	if_route(ifp, IFF_UP, AF_UNSPEC);
1432 }
1433 
1434 /*
1435  * Flush an interface queue.
1436  */
1437 static void
1438 if_qflush(struct ifaltq *ifq)
1439 {
1440 	struct mbuf *m, *n;
1441 
1442 	IFQ_LOCK(ifq);
1443 #ifdef ALTQ
1444 	if (ALTQ_IS_ENABLED(ifq))
1445 		ALTQ_PURGE(ifq);
1446 #endif
1447 	n = ifq->ifq_head;
1448 	while ((m = n) != 0) {
1449 		n = m->m_act;
1450 		m_freem(m);
1451 	}
1452 	ifq->ifq_head = 0;
1453 	ifq->ifq_tail = 0;
1454 	ifq->ifq_len = 0;
1455 	IFQ_UNLOCK(ifq);
1456 }
1457 
1458 /*
1459  * Handle interface watchdog timer routines.  Called
1460  * from softclock, we decrement timers (if set) and
1461  * call the appropriate interface routine on expiration.
1462  *
1463  * XXXRW: Note that because timeouts run with Giant, if_watchdog() is called
1464  * holding Giant.  If we switch to an MPSAFE callout, we likely need to grab
1465  * Giant before entering if_watchdog() on an IFF_NEEDSGIANT interface.
1466  */
1467 static void
1468 if_slowtimo(void *arg)
1469 {
1470 	struct ifnet *ifp;
1471 	int s = splimp();
1472 
1473 	IFNET_RLOCK();
1474 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1475 		if (ifp->if_timer == 0 || --ifp->if_timer)
1476 			continue;
1477 		if (ifp->if_watchdog)
1478 			(*ifp->if_watchdog)(ifp);
1479 	}
1480 	IFNET_RUNLOCK();
1481 	splx(s);
1482 	timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
1483 }
1484 
1485 /*
1486  * Map interface name to
1487  * interface structure pointer.
1488  */
1489 struct ifnet *
1490 ifunit(const char *name)
1491 {
1492 	struct ifnet *ifp;
1493 
1494 	IFNET_RLOCK();
1495 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1496 		if (strncmp(name, ifp->if_xname, IFNAMSIZ) == 0)
1497 			break;
1498 	}
1499 	IFNET_RUNLOCK();
1500 	return (ifp);
1501 }
1502 
1503 /*
1504  * Hardware specific interface ioctls.
1505  */
1506 static int
1507 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
1508 {
1509 	struct ifreq *ifr;
1510 	struct ifstat *ifs;
1511 	int error = 0;
1512 	int new_flags, temp_flags;
1513 	size_t namelen, onamelen;
1514 	char new_name[IFNAMSIZ];
1515 	struct ifaddr *ifa;
1516 	struct sockaddr_dl *sdl;
1517 
1518 	ifr = (struct ifreq *)data;
1519 	switch (cmd) {
1520 	case SIOCGIFINDEX:
1521 		ifr->ifr_index = ifp->if_index;
1522 		break;
1523 
1524 	case SIOCGIFFLAGS:
1525 		temp_flags = ifp->if_flags | ifp->if_drv_flags;
1526 		ifr->ifr_flags = temp_flags & 0xffff;
1527 		ifr->ifr_flagshigh = temp_flags >> 16;
1528 		break;
1529 
1530 	case SIOCGIFCAP:
1531 		ifr->ifr_reqcap = ifp->if_capabilities;
1532 		ifr->ifr_curcap = ifp->if_capenable;
1533 		break;
1534 
1535 #ifdef MAC
1536 	case SIOCGIFMAC:
1537 		error = mac_ioctl_ifnet_get(td->td_ucred, ifr, ifp);
1538 		break;
1539 #endif
1540 
1541 	case SIOCGIFMETRIC:
1542 		ifr->ifr_metric = ifp->if_metric;
1543 		break;
1544 
1545 	case SIOCGIFMTU:
1546 		ifr->ifr_mtu = ifp->if_mtu;
1547 		break;
1548 
1549 	case SIOCGIFPHYS:
1550 		ifr->ifr_phys = ifp->if_physical;
1551 		break;
1552 
1553 	case SIOCSIFFLAGS:
1554 		error = priv_check(td, PRIV_NET_SETIFFLAGS);
1555 		if (error)
1556 			return (error);
1557 		/*
1558 		 * Currently, no driver owned flags pass the IFF_CANTCHANGE
1559 		 * check, so we don't need special handling here yet.
1560 		 */
1561 		new_flags = (ifr->ifr_flags & 0xffff) |
1562 		    (ifr->ifr_flagshigh << 16);
1563 		if (ifp->if_flags & IFF_SMART) {
1564 			/* Smart drivers twiddle their own routes */
1565 		} else if (ifp->if_flags & IFF_UP &&
1566 		    (new_flags & IFF_UP) == 0) {
1567 			int s = splimp();
1568 			if_down(ifp);
1569 			splx(s);
1570 		} else if (new_flags & IFF_UP &&
1571 		    (ifp->if_flags & IFF_UP) == 0) {
1572 			int s = splimp();
1573 			if_up(ifp);
1574 			splx(s);
1575 		}
1576 		/* See if permanently promiscuous mode bit is about to flip */
1577 		if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
1578 			if (new_flags & IFF_PPROMISC)
1579 				ifp->if_flags |= IFF_PROMISC;
1580 			else if (ifp->if_pcount == 0)
1581 				ifp->if_flags &= ~IFF_PROMISC;
1582 			log(LOG_INFO, "%s: permanently promiscuous mode %s\n",
1583 			    ifp->if_xname,
1584 			    (new_flags & IFF_PPROMISC) ? "enabled" : "disabled");
1585 		}
1586 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1587 			(new_flags &~ IFF_CANTCHANGE);
1588 		if (ifp->if_ioctl) {
1589 			IFF_LOCKGIANT(ifp);
1590 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1591 			IFF_UNLOCKGIANT(ifp);
1592 		}
1593 		getmicrotime(&ifp->if_lastchange);
1594 		break;
1595 
1596 	case SIOCSIFCAP:
1597 		error = priv_check(td, PRIV_NET_SETIFCAP);
1598 		if (error)
1599 			return (error);
1600 		if (ifp->if_ioctl == NULL)
1601 			return (EOPNOTSUPP);
1602 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1603 			return (EINVAL);
1604 		IFF_LOCKGIANT(ifp);
1605 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1606 		IFF_UNLOCKGIANT(ifp);
1607 		if (error == 0)
1608 			getmicrotime(&ifp->if_lastchange);
1609 		break;
1610 
1611 #ifdef MAC
1612 	case SIOCSIFMAC:
1613 		error = mac_ioctl_ifnet_set(td->td_ucred, ifr, ifp);
1614 		break;
1615 #endif
1616 
1617 	case SIOCSIFNAME:
1618 		error = priv_check(td, PRIV_NET_SETIFNAME);
1619 		if (error)
1620 			return (error);
1621 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1622 		if (error != 0)
1623 			return (error);
1624 		if (new_name[0] == '\0')
1625 			return (EINVAL);
1626 		if (ifunit(new_name) != NULL)
1627 			return (EEXIST);
1628 
1629 		/* Announce the departure of the interface. */
1630 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1631 		EVENTHANDLER_INVOKE(ifnet_departure_event, ifp);
1632 
1633 		log(LOG_INFO, "%s: changing name to '%s'\n",
1634 		    ifp->if_xname, new_name);
1635 
1636 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1637 		ifa = ifp->if_addr;
1638 		IFA_LOCK(ifa);
1639 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1640 		namelen = strlen(new_name);
1641 		onamelen = sdl->sdl_nlen;
1642 		/*
1643 		 * Move the address if needed.  This is safe because we
1644 		 * allocate space for a name of length IFNAMSIZ when we
1645 		 * create this in if_attach().
1646 		 */
1647 		if (namelen != onamelen) {
1648 			bcopy(sdl->sdl_data + onamelen,
1649 			    sdl->sdl_data + namelen, sdl->sdl_alen);
1650 		}
1651 		bcopy(new_name, sdl->sdl_data, namelen);
1652 		sdl->sdl_nlen = namelen;
1653 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1654 		bzero(sdl->sdl_data, onamelen);
1655 		while (namelen != 0)
1656 			sdl->sdl_data[--namelen] = 0xff;
1657 		IFA_UNLOCK(ifa);
1658 
1659 		EVENTHANDLER_INVOKE(ifnet_arrival_event, ifp);
1660 		/* Announce the return of the interface. */
1661 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1662 		break;
1663 
1664 	case SIOCSIFMETRIC:
1665 		error = priv_check(td, PRIV_NET_SETIFMETRIC);
1666 		if (error)
1667 			return (error);
1668 		ifp->if_metric = ifr->ifr_metric;
1669 		getmicrotime(&ifp->if_lastchange);
1670 		break;
1671 
1672 	case SIOCSIFPHYS:
1673 		error = priv_check(td, PRIV_NET_SETIFPHYS);
1674 		if (error)
1675 			return (error);
1676 		if (ifp->if_ioctl == NULL)
1677 			return (EOPNOTSUPP);
1678 		IFF_LOCKGIANT(ifp);
1679 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1680 		IFF_UNLOCKGIANT(ifp);
1681 		if (error == 0)
1682 			getmicrotime(&ifp->if_lastchange);
1683 		break;
1684 
1685 	case SIOCSIFMTU:
1686 	{
1687 		u_long oldmtu = ifp->if_mtu;
1688 
1689 		error = priv_check(td, PRIV_NET_SETIFMTU);
1690 		if (error)
1691 			return (error);
1692 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1693 			return (EINVAL);
1694 		if (ifp->if_ioctl == NULL)
1695 			return (EOPNOTSUPP);
1696 		IFF_LOCKGIANT(ifp);
1697 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1698 		IFF_UNLOCKGIANT(ifp);
1699 		if (error == 0) {
1700 			getmicrotime(&ifp->if_lastchange);
1701 			rt_ifmsg(ifp);
1702 		}
1703 		/*
1704 		 * If the link MTU changed, do network layer specific procedure.
1705 		 */
1706 		if (ifp->if_mtu != oldmtu) {
1707 #ifdef INET6
1708 			nd6_setmtu(ifp);
1709 #endif
1710 		}
1711 		break;
1712 	}
1713 
1714 	case SIOCADDMULTI:
1715 	case SIOCDELMULTI:
1716 		if (cmd == SIOCADDMULTI)
1717 			error = priv_check(td, PRIV_NET_ADDMULTI);
1718 		else
1719 			error = priv_check(td, PRIV_NET_DELMULTI);
1720 		if (error)
1721 			return (error);
1722 
1723 		/* Don't allow group membership on non-multicast interfaces. */
1724 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1725 			return (EOPNOTSUPP);
1726 
1727 		/* Don't let users screw up protocols' entries. */
1728 		if (ifr->ifr_addr.sa_family != AF_LINK)
1729 			return (EINVAL);
1730 
1731 		if (cmd == SIOCADDMULTI) {
1732 			struct ifmultiaddr *ifma;
1733 
1734 			/*
1735 			 * Userland is only permitted to join groups once
1736 			 * via the if_addmulti() KPI, because it cannot hold
1737 			 * struct ifmultiaddr * between calls. It may also
1738 			 * lose a race while we check if the membership
1739 			 * already exists.
1740 			 */
1741 			IF_ADDR_LOCK(ifp);
1742 			ifma = if_findmulti(ifp, &ifr->ifr_addr);
1743 			IF_ADDR_UNLOCK(ifp);
1744 			if (ifma != NULL)
1745 				error = EADDRINUSE;
1746 			else
1747 				error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1748 		} else {
1749 			error = if_delmulti(ifp, &ifr->ifr_addr);
1750 		}
1751 		if (error == 0)
1752 			getmicrotime(&ifp->if_lastchange);
1753 		break;
1754 
1755 	case SIOCSIFPHYADDR:
1756 	case SIOCDIFPHYADDR:
1757 #ifdef INET6
1758 	case SIOCSIFPHYADDR_IN6:
1759 #endif
1760 	case SIOCSLIFPHYADDR:
1761 	case SIOCSIFMEDIA:
1762 	case SIOCSIFGENERIC:
1763 		error = priv_check(td, PRIV_NET_HWIOCTL);
1764 		if (error)
1765 			return (error);
1766 		if (ifp->if_ioctl == NULL)
1767 			return (EOPNOTSUPP);
1768 		IFF_LOCKGIANT(ifp);
1769 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1770 		IFF_UNLOCKGIANT(ifp);
1771 		if (error == 0)
1772 			getmicrotime(&ifp->if_lastchange);
1773 		break;
1774 
1775 	case SIOCGIFSTATUS:
1776 		ifs = (struct ifstat *)data;
1777 		ifs->ascii[0] = '\0';
1778 
1779 	case SIOCGIFPSRCADDR:
1780 	case SIOCGIFPDSTADDR:
1781 	case SIOCGLIFPHYADDR:
1782 	case SIOCGIFMEDIA:
1783 	case SIOCGIFGENERIC:
1784 		if (ifp->if_ioctl == NULL)
1785 			return (EOPNOTSUPP);
1786 		IFF_LOCKGIANT(ifp);
1787 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1788 		IFF_UNLOCKGIANT(ifp);
1789 		break;
1790 
1791 	case SIOCSIFLLADDR:
1792 		error = priv_check(td, PRIV_NET_SETLLADDR);
1793 		if (error)
1794 			return (error);
1795 		error = if_setlladdr(ifp,
1796 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1797 		break;
1798 
1799 	case SIOCAIFGROUP:
1800 	{
1801 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
1802 
1803 		error = priv_check(td, PRIV_NET_ADDIFGROUP);
1804 		if (error)
1805 			return (error);
1806 		if ((error = if_addgroup(ifp, ifgr->ifgr_group)))
1807 			return (error);
1808 		break;
1809 	}
1810 
1811 	case SIOCGIFGROUP:
1812 		if ((error = if_getgroup((struct ifgroupreq *)ifr, ifp)))
1813 			return (error);
1814 		break;
1815 
1816 	case SIOCDIFGROUP:
1817 	{
1818 		struct ifgroupreq *ifgr = (struct ifgroupreq *)ifr;
1819 
1820 		error = priv_check(td, PRIV_NET_DELIFGROUP);
1821 		if (error)
1822 			return (error);
1823 		if ((error = if_delgroup(ifp, ifgr->ifgr_group)))
1824 			return (error);
1825 		break;
1826 	}
1827 
1828 	default:
1829 		error = ENOIOCTL;
1830 		break;
1831 	}
1832 	return (error);
1833 }
1834 
1835 /*
1836  * Interface ioctls.
1837  */
1838 int
1839 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1840 {
1841 	struct ifnet *ifp;
1842 	struct ifreq *ifr;
1843 	int error;
1844 	int oif_flags;
1845 
1846 	switch (cmd) {
1847 	case SIOCGIFCONF:
1848 	case OSIOCGIFCONF:
1849 #ifdef __amd64__
1850 	case SIOCGIFCONF32:
1851 #endif
1852 		return (ifconf(cmd, data));
1853 	}
1854 	ifr = (struct ifreq *)data;
1855 
1856 	switch (cmd) {
1857 	case SIOCIFCREATE:
1858 	case SIOCIFCREATE2:
1859 		error = priv_check(td, PRIV_NET_IFCREATE);
1860 		if (error)
1861 			return (error);
1862 		return (if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name),
1863 			cmd == SIOCIFCREATE2 ? ifr->ifr_data : NULL));
1864 	case SIOCIFDESTROY:
1865 		error = priv_check(td, PRIV_NET_IFDESTROY);
1866 		if (error)
1867 			return (error);
1868 		return if_clone_destroy(ifr->ifr_name);
1869 
1870 	case SIOCIFGCLONERS:
1871 		return (if_clone_list((struct if_clonereq *)data));
1872 	case SIOCGIFGMEMB:
1873 		return (if_getgroupmembers((struct ifgroupreq *)data));
1874 	}
1875 
1876 	ifp = ifunit(ifr->ifr_name);
1877 	if (ifp == 0)
1878 		return (ENXIO);
1879 
1880 	error = ifhwioctl(cmd, ifp, data, td);
1881 	if (error != ENOIOCTL)
1882 		return (error);
1883 
1884 	oif_flags = ifp->if_flags;
1885 	if (so->so_proto == 0)
1886 		return (EOPNOTSUPP);
1887 #ifndef COMPAT_43
1888 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1889 								 data,
1890 								 ifp, td));
1891 #else
1892 	{
1893 		int ocmd = cmd;
1894 
1895 		switch (cmd) {
1896 
1897 		case SIOCSIFDSTADDR:
1898 		case SIOCSIFADDR:
1899 		case SIOCSIFBRDADDR:
1900 		case SIOCSIFNETMASK:
1901 #if BYTE_ORDER != BIG_ENDIAN
1902 			if (ifr->ifr_addr.sa_family == 0 &&
1903 			    ifr->ifr_addr.sa_len < 16) {
1904 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1905 				ifr->ifr_addr.sa_len = 16;
1906 			}
1907 #else
1908 			if (ifr->ifr_addr.sa_len == 0)
1909 				ifr->ifr_addr.sa_len = 16;
1910 #endif
1911 			break;
1912 
1913 		case OSIOCGIFADDR:
1914 			cmd = SIOCGIFADDR;
1915 			break;
1916 
1917 		case OSIOCGIFDSTADDR:
1918 			cmd = SIOCGIFDSTADDR;
1919 			break;
1920 
1921 		case OSIOCGIFBRDADDR:
1922 			cmd = SIOCGIFBRDADDR;
1923 			break;
1924 
1925 		case OSIOCGIFNETMASK:
1926 			cmd = SIOCGIFNETMASK;
1927 		}
1928 		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
1929 								   cmd,
1930 								   data,
1931 								   ifp, td));
1932 		switch (ocmd) {
1933 
1934 		case OSIOCGIFADDR:
1935 		case OSIOCGIFDSTADDR:
1936 		case OSIOCGIFBRDADDR:
1937 		case OSIOCGIFNETMASK:
1938 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1939 
1940 		}
1941 	}
1942 #endif /* COMPAT_43 */
1943 
1944 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1945 #ifdef INET6
1946 		DELAY(100);/* XXX: temporary workaround for fxp issue*/
1947 		if (ifp->if_flags & IFF_UP) {
1948 			int s = splimp();
1949 			in6_if_up(ifp);
1950 			splx(s);
1951 		}
1952 #endif
1953 	}
1954 	return (error);
1955 }
1956 
1957 /*
1958  * The code common to handling reference counted flags,
1959  * e.g., in ifpromisc() and if_allmulti().
1960  * The "pflag" argument can specify a permanent mode flag to check,
1961  * such as IFF_PPROMISC for promiscuous mode; should be 0 if none.
1962  *
1963  * Only to be used on stack-owned flags, not driver-owned flags.
1964  */
1965 static int
1966 if_setflag(struct ifnet *ifp, int flag, int pflag, int *refcount, int onswitch)
1967 {
1968 	struct ifreq ifr;
1969 	int error;
1970 	int oldflags, oldcount;
1971 
1972 	/* Sanity checks to catch programming errors */
1973 	KASSERT((flag & (IFF_DRV_OACTIVE|IFF_DRV_RUNNING)) == 0,
1974 	    ("%s: setting driver-owned flag %d", __func__, flag));
1975 
1976 	if (onswitch)
1977 		KASSERT(*refcount >= 0,
1978 		    ("%s: increment negative refcount %d for flag %d",
1979 		    __func__, *refcount, flag));
1980 	else
1981 		KASSERT(*refcount > 0,
1982 		    ("%s: decrement non-positive refcount %d for flag %d",
1983 		    __func__, *refcount, flag));
1984 
1985 	/* In case this mode is permanent, just touch refcount */
1986 	if (ifp->if_flags & pflag) {
1987 		*refcount += onswitch ? 1 : -1;
1988 		return (0);
1989 	}
1990 
1991 	/* Save ifnet parameters for if_ioctl() may fail */
1992 	oldcount = *refcount;
1993 	oldflags = ifp->if_flags;
1994 
1995 	/*
1996 	 * See if we aren't the only and touching refcount is enough.
1997 	 * Actually toggle interface flag if we are the first or last.
1998 	 */
1999 	if (onswitch) {
2000 		if ((*refcount)++)
2001 			return (0);
2002 		ifp->if_flags |= flag;
2003 	} else {
2004 		if (--(*refcount))
2005 			return (0);
2006 		ifp->if_flags &= ~flag;
2007 	}
2008 
2009 	/* Call down the driver since we've changed interface flags */
2010 	if (ifp->if_ioctl == NULL) {
2011 		error = EOPNOTSUPP;
2012 		goto recover;
2013 	}
2014 	ifr.ifr_flags = ifp->if_flags & 0xffff;
2015 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
2016 	IFF_LOCKGIANT(ifp);
2017 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2018 	IFF_UNLOCKGIANT(ifp);
2019 	if (error)
2020 		goto recover;
2021 	/* Notify userland that interface flags have changed */
2022 	rt_ifmsg(ifp);
2023 	return (0);
2024 
2025 recover:
2026 	/* Recover after driver error */
2027 	*refcount = oldcount;
2028 	ifp->if_flags = oldflags;
2029 	return (error);
2030 }
2031 
2032 /*
2033  * Set/clear promiscuous mode on interface ifp based on the truth value
2034  * of pswitch.  The calls are reference counted so that only the first
2035  * "on" request actually has an effect, as does the final "off" request.
2036  * Results are undefined if the "off" and "on" requests are not matched.
2037  */
2038 int
2039 ifpromisc(struct ifnet *ifp, int pswitch)
2040 {
2041 	int error;
2042 	int oldflags = ifp->if_flags;
2043 
2044 	error = if_setflag(ifp, IFF_PROMISC, IFF_PPROMISC,
2045 			   &ifp->if_pcount, pswitch);
2046 	/* If promiscuous mode status has changed, log a message */
2047 	if (error == 0 && ((ifp->if_flags ^ oldflags) & IFF_PROMISC))
2048 		log(LOG_INFO, "%s: promiscuous mode %s\n",
2049 		    ifp->if_xname,
2050 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
2051 	return (error);
2052 }
2053 
2054 /*
2055  * Return interface configuration
2056  * of system.  List may be used
2057  * in later ioctl's (above) to get
2058  * other information.
2059  */
2060 /*ARGSUSED*/
2061 static int
2062 ifconf(u_long cmd, caddr_t data)
2063 {
2064 	struct ifconf *ifc = (struct ifconf *)data;
2065 #ifdef __amd64__
2066 	struct ifconf32 *ifc32 = (struct ifconf32 *)data;
2067 	struct ifconf ifc_swab;
2068 #endif
2069 	struct ifnet *ifp;
2070 	struct ifaddr *ifa;
2071 	struct ifreq ifr;
2072 	struct sbuf *sb;
2073 	int error, full = 0, valid_len, max_len;
2074 
2075 #ifdef __amd64__
2076 	if (cmd == SIOCGIFCONF32) {
2077 		ifc_swab.ifc_len = ifc32->ifc_len;
2078 		ifc_swab.ifc_buf = (caddr_t)(uintptr_t)ifc32->ifc_buf;
2079 		ifc = &ifc_swab;
2080 	}
2081 #endif
2082 	/* Limit initial buffer size to MAXPHYS to avoid DoS from userspace. */
2083 	max_len = MAXPHYS - 1;
2084 
2085 	/* Prevent hostile input from being able to crash the system */
2086 	if (ifc->ifc_len <= 0)
2087 		return (EINVAL);
2088 
2089 again:
2090 	if (ifc->ifc_len <= max_len) {
2091 		max_len = ifc->ifc_len;
2092 		full = 1;
2093 	}
2094 	sb = sbuf_new(NULL, NULL, max_len + 1, SBUF_FIXEDLEN);
2095 	max_len = 0;
2096 	valid_len = 0;
2097 
2098 	IFNET_RLOCK();		/* could sleep XXX */
2099 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
2100 		int addrs;
2101 
2102 		/*
2103 		 * Zero the ifr_name buffer to make sure we don't
2104 		 * disclose the contents of the stack.
2105 		 */
2106 		memset(ifr.ifr_name, 0, sizeof(ifr.ifr_name));
2107 
2108 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
2109 		    >= sizeof(ifr.ifr_name)) {
2110 			sbuf_delete(sb);
2111 			IFNET_RUNLOCK();
2112 			return (ENAMETOOLONG);
2113 		}
2114 
2115 		addrs = 0;
2116 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2117 			struct sockaddr *sa = ifa->ifa_addr;
2118 
2119 			if (jailed(curthread->td_ucred) &&
2120 			    prison_if(curthread->td_ucred, sa))
2121 				continue;
2122 			addrs++;
2123 #ifdef COMPAT_43
2124 			if (cmd == OSIOCGIFCONF) {
2125 				struct osockaddr *osa =
2126 					 (struct osockaddr *)&ifr.ifr_addr;
2127 				ifr.ifr_addr = *sa;
2128 				osa->sa_family = sa->sa_family;
2129 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2130 				max_len += sizeof(ifr);
2131 			} else
2132 #endif
2133 			if (sa->sa_len <= sizeof(*sa)) {
2134 				ifr.ifr_addr = *sa;
2135 				sbuf_bcat(sb, &ifr, sizeof(ifr));
2136 				max_len += sizeof(ifr);
2137 			} else {
2138 				sbuf_bcat(sb, &ifr,
2139 				    offsetof(struct ifreq, ifr_addr));
2140 				max_len += offsetof(struct ifreq, ifr_addr);
2141 				sbuf_bcat(sb, sa, sa->sa_len);
2142 				max_len += sa->sa_len;
2143 			}
2144 
2145 			if (!sbuf_overflowed(sb))
2146 				valid_len = sbuf_len(sb);
2147 		}
2148 		if (addrs == 0) {
2149 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
2150 			sbuf_bcat(sb, &ifr, sizeof(ifr));
2151 			max_len += sizeof(ifr);
2152 
2153 			if (!sbuf_overflowed(sb))
2154 				valid_len = sbuf_len(sb);
2155 		}
2156 	}
2157 	IFNET_RUNLOCK();
2158 
2159 	/*
2160 	 * If we didn't allocate enough space (uncommon), try again.  If
2161 	 * we have already allocated as much space as we are allowed,
2162 	 * return what we've got.
2163 	 */
2164 	if (valid_len != max_len && !full) {
2165 		sbuf_delete(sb);
2166 		goto again;
2167 	}
2168 
2169 	ifc->ifc_len = valid_len;
2170 #ifdef __amd64__
2171 	if (cmd == SIOCGIFCONF32)
2172 		ifc32->ifc_len = valid_len;
2173 #endif
2174 	sbuf_finish(sb);
2175 	error = copyout(sbuf_data(sb), ifc->ifc_req, ifc->ifc_len);
2176 	sbuf_delete(sb);
2177 	return (error);
2178 }
2179 
2180 /*
2181  * Just like ifpromisc(), but for all-multicast-reception mode.
2182  */
2183 int
2184 if_allmulti(struct ifnet *ifp, int onswitch)
2185 {
2186 
2187 	return (if_setflag(ifp, IFF_ALLMULTI, 0, &ifp->if_amcount, onswitch));
2188 }
2189 
2190 struct ifmultiaddr *
2191 if_findmulti(struct ifnet *ifp, struct sockaddr *sa)
2192 {
2193 	struct ifmultiaddr *ifma;
2194 
2195 	IF_ADDR_LOCK_ASSERT(ifp);
2196 
2197 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
2198 		if (sa->sa_family == AF_LINK) {
2199 			if (sa_dl_equal(ifma->ifma_addr, sa))
2200 				break;
2201 		} else {
2202 			if (sa_equal(ifma->ifma_addr, sa))
2203 				break;
2204 		}
2205 	}
2206 
2207 	return ifma;
2208 }
2209 
2210 /*
2211  * Allocate a new ifmultiaddr and initialize based on passed arguments.  We
2212  * make copies of passed sockaddrs.  The ifmultiaddr will not be added to
2213  * the ifnet multicast address list here, so the caller must do that and
2214  * other setup work (such as notifying the device driver).  The reference
2215  * count is initialized to 1.
2216  */
2217 static struct ifmultiaddr *
2218 if_allocmulti(struct ifnet *ifp, struct sockaddr *sa, struct sockaddr *llsa,
2219     int mflags)
2220 {
2221 	struct ifmultiaddr *ifma;
2222 	struct sockaddr *dupsa;
2223 
2224 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, mflags |
2225 	    M_ZERO);
2226 	if (ifma == NULL)
2227 		return (NULL);
2228 
2229 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, mflags);
2230 	if (dupsa == NULL) {
2231 		FREE(ifma, M_IFMADDR);
2232 		return (NULL);
2233 	}
2234 	bcopy(sa, dupsa, sa->sa_len);
2235 	ifma->ifma_addr = dupsa;
2236 
2237 	ifma->ifma_ifp = ifp;
2238 	ifma->ifma_refcount = 1;
2239 	ifma->ifma_protospec = NULL;
2240 
2241 	if (llsa == NULL) {
2242 		ifma->ifma_lladdr = NULL;
2243 		return (ifma);
2244 	}
2245 
2246 	MALLOC(dupsa, struct sockaddr *, llsa->sa_len, M_IFMADDR, mflags);
2247 	if (dupsa == NULL) {
2248 		FREE(ifma->ifma_addr, M_IFMADDR);
2249 		FREE(ifma, M_IFMADDR);
2250 		return (NULL);
2251 	}
2252 	bcopy(llsa, dupsa, llsa->sa_len);
2253 	ifma->ifma_lladdr = dupsa;
2254 
2255 	return (ifma);
2256 }
2257 
2258 /*
2259  * if_freemulti: free ifmultiaddr structure and possibly attached related
2260  * addresses.  The caller is responsible for implementing reference
2261  * counting, notifying the driver, handling routing messages, and releasing
2262  * any dependent link layer state.
2263  */
2264 static void
2265 if_freemulti(struct ifmultiaddr *ifma)
2266 {
2267 
2268 	KASSERT(ifma->ifma_refcount == 0, ("if_freemulti: refcount %d",
2269 	    ifma->ifma_refcount));
2270 	KASSERT(ifma->ifma_protospec == NULL,
2271 	    ("if_freemulti: protospec not NULL"));
2272 
2273 	if (ifma->ifma_lladdr != NULL)
2274 		FREE(ifma->ifma_lladdr, M_IFMADDR);
2275 	FREE(ifma->ifma_addr, M_IFMADDR);
2276 	FREE(ifma, M_IFMADDR);
2277 }
2278 
2279 /*
2280  * Register an additional multicast address with a network interface.
2281  *
2282  * - If the address is already present, bump the reference count on the
2283  *   address and return.
2284  * - If the address is not link-layer, look up a link layer address.
2285  * - Allocate address structures for one or both addresses, and attach to the
2286  *   multicast address list on the interface.  If automatically adding a link
2287  *   layer address, the protocol address will own a reference to the link
2288  *   layer address, to be freed when it is freed.
2289  * - Notify the network device driver of an addition to the multicast address
2290  *   list.
2291  *
2292  * 'sa' points to caller-owned memory with the desired multicast address.
2293  *
2294  * 'retifma' will be used to return a pointer to the resulting multicast
2295  * address reference, if desired.
2296  */
2297 int
2298 if_addmulti(struct ifnet *ifp, struct sockaddr *sa,
2299     struct ifmultiaddr **retifma)
2300 {
2301 	struct ifmultiaddr *ifma, *ll_ifma;
2302 	struct sockaddr *llsa;
2303 	int error;
2304 
2305 	/*
2306 	 * If the address is already present, return a new reference to it;
2307 	 * otherwise, allocate storage and set up a new address.
2308 	 */
2309 	IF_ADDR_LOCK(ifp);
2310 	ifma = if_findmulti(ifp, sa);
2311 	if (ifma != NULL) {
2312 		ifma->ifma_refcount++;
2313 		if (retifma != NULL)
2314 			*retifma = ifma;
2315 		IF_ADDR_UNLOCK(ifp);
2316 		return (0);
2317 	}
2318 
2319 	/*
2320 	 * The address isn't already present; resolve the protocol address
2321 	 * into a link layer address, and then look that up, bump its
2322 	 * refcount or allocate an ifma for that also.  If 'llsa' was
2323 	 * returned, we will need to free it later.
2324 	 */
2325 	llsa = NULL;
2326 	ll_ifma = NULL;
2327 	if (ifp->if_resolvemulti != NULL) {
2328 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
2329 		if (error)
2330 			goto unlock_out;
2331 	}
2332 
2333 	/*
2334 	 * Allocate the new address.  Don't hook it up yet, as we may also
2335 	 * need to allocate a link layer multicast address.
2336 	 */
2337 	ifma = if_allocmulti(ifp, sa, llsa, M_NOWAIT);
2338 	if (ifma == NULL) {
2339 		error = ENOMEM;
2340 		goto free_llsa_out;
2341 	}
2342 
2343 	/*
2344 	 * If a link layer address is found, we'll need to see if it's
2345 	 * already present in the address list, or allocate is as well.
2346 	 * When this block finishes, the link layer address will be on the
2347 	 * list.
2348 	 */
2349 	if (llsa != NULL) {
2350 		ll_ifma = if_findmulti(ifp, llsa);
2351 		if (ll_ifma == NULL) {
2352 			ll_ifma = if_allocmulti(ifp, llsa, NULL, M_NOWAIT);
2353 			if (ll_ifma == NULL) {
2354 				--ifma->ifma_refcount;
2355 				if_freemulti(ifma);
2356 				error = ENOMEM;
2357 				goto free_llsa_out;
2358 			}
2359 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ll_ifma,
2360 			    ifma_link);
2361 		} else
2362 			ll_ifma->ifma_refcount++;
2363 		ifma->ifma_llifma = ll_ifma;
2364 	}
2365 
2366 	/*
2367 	 * We now have a new multicast address, ifma, and possibly a new or
2368 	 * referenced link layer address.  Add the primary address to the
2369 	 * ifnet address list.
2370 	 */
2371 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
2372 
2373 	if (retifma != NULL)
2374 		*retifma = ifma;
2375 
2376 	/*
2377 	 * Must generate the message while holding the lock so that 'ifma'
2378 	 * pointer is still valid.
2379 	 */
2380 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
2381 	IF_ADDR_UNLOCK(ifp);
2382 
2383 	/*
2384 	 * We are certain we have added something, so call down to the
2385 	 * interface to let them know about it.
2386 	 */
2387 	if (ifp->if_ioctl != NULL) {
2388 		IFF_LOCKGIANT(ifp);
2389 		(void) (*ifp->if_ioctl)(ifp, SIOCADDMULTI, 0);
2390 		IFF_UNLOCKGIANT(ifp);
2391 	}
2392 
2393 	if (llsa != NULL)
2394 		FREE(llsa, M_IFMADDR);
2395 
2396 	return (0);
2397 
2398 free_llsa_out:
2399 	if (llsa != NULL)
2400 		FREE(llsa, M_IFMADDR);
2401 
2402 unlock_out:
2403 	IF_ADDR_UNLOCK(ifp);
2404 	return (error);
2405 }
2406 
2407 /*
2408  * Delete a multicast group membership by network-layer group address.
2409  *
2410  * Returns ENOENT if the entry could not be found. If ifp no longer
2411  * exists, results are undefined. This entry point should only be used
2412  * from subsystems which do appropriate locking to hold ifp for the
2413  * duration of the call.
2414  * Network-layer protocol domains must use if_delmulti_ifma().
2415  */
2416 int
2417 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
2418 {
2419 	struct ifmultiaddr *ifma;
2420 	int lastref;
2421 #ifdef INVARIANTS
2422 	struct ifnet *oifp;
2423 
2424 	IFNET_RLOCK();
2425 	TAILQ_FOREACH(oifp, &ifnet, if_link)
2426 		if (ifp == oifp)
2427 			break;
2428 	if (ifp != oifp)
2429 		ifp = NULL;
2430 	IFNET_RUNLOCK();
2431 
2432 	KASSERT(ifp != NULL, ("%s: ifnet went away", __func__));
2433 #endif
2434 	if (ifp == NULL)
2435 		return (ENOENT);
2436 
2437 	IF_ADDR_LOCK(ifp);
2438 	lastref = 0;
2439 	ifma = if_findmulti(ifp, sa);
2440 	if (ifma != NULL)
2441 		lastref = if_delmulti_locked(ifp, ifma, 0);
2442 	IF_ADDR_UNLOCK(ifp);
2443 
2444 	if (ifma == NULL)
2445 		return (ENOENT);
2446 
2447 	if (lastref && ifp->if_ioctl != NULL) {
2448 		IFF_LOCKGIANT(ifp);
2449 		(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
2450 		IFF_UNLOCKGIANT(ifp);
2451 	}
2452 
2453 	return (0);
2454 }
2455 
2456 /*
2457  * Delete a multicast group membership by group membership pointer.
2458  * Network-layer protocol domains must use this routine.
2459  *
2460  * It is safe to call this routine if the ifp disappeared. Callers should
2461  * hold IFF_LOCKGIANT() to avoid a LOR in case the hardware needs to be
2462  * reconfigured.
2463  */
2464 void
2465 if_delmulti_ifma(struct ifmultiaddr *ifma)
2466 {
2467 	struct ifnet *ifp;
2468 	int lastref;
2469 
2470 	ifp = ifma->ifma_ifp;
2471 #ifdef DIAGNOSTIC
2472 	if (ifp == NULL) {
2473 		printf("%s: ifma_ifp seems to be detached\n", __func__);
2474 	} else {
2475 		struct ifnet *oifp;
2476 
2477 		IFNET_RLOCK();
2478 		TAILQ_FOREACH(oifp, &ifnet, if_link)
2479 			if (ifp == oifp)
2480 				break;
2481 		if (ifp != oifp) {
2482 			printf("%s: ifnet %p disappeared\n", __func__, ifp);
2483 			ifp = NULL;
2484 		}
2485 		IFNET_RUNLOCK();
2486 	}
2487 #endif
2488 	/*
2489 	 * If and only if the ifnet instance exists: Acquire the address lock.
2490 	 */
2491 	if (ifp != NULL)
2492 		IF_ADDR_LOCK(ifp);
2493 
2494 	lastref = if_delmulti_locked(ifp, ifma, 0);
2495 
2496 	if (ifp != NULL) {
2497 		/*
2498 		 * If and only if the ifnet instance exists:
2499 		 *  Release the address lock.
2500 		 *  If the group was left: update the hardware hash filter.
2501 		 */
2502 		IF_ADDR_UNLOCK(ifp);
2503 		if (lastref && ifp->if_ioctl != NULL) {
2504 			IFF_LOCKGIANT(ifp);
2505 			(void)(*ifp->if_ioctl)(ifp, SIOCDELMULTI, 0);
2506 			IFF_UNLOCKGIANT(ifp);
2507 		}
2508 	}
2509 }
2510 
2511 /*
2512  * Perform deletion of network-layer and/or link-layer multicast address.
2513  *
2514  * Return 0 if the reference count was decremented.
2515  * Return 1 if the final reference was released, indicating that the
2516  * hardware hash filter should be reprogrammed.
2517  */
2518 static int
2519 if_delmulti_locked(struct ifnet *ifp, struct ifmultiaddr *ifma, int detaching)
2520 {
2521 	struct ifmultiaddr *ll_ifma;
2522 
2523 	if (ifp != NULL && ifma->ifma_ifp != NULL) {
2524 		KASSERT(ifma->ifma_ifp == ifp,
2525 		    ("%s: inconsistent ifp %p", __func__, ifp));
2526 		IF_ADDR_LOCK_ASSERT(ifp);
2527 	}
2528 
2529 	ifp = ifma->ifma_ifp;
2530 
2531 	/*
2532 	 * If the ifnet is detaching, null out references to ifnet,
2533 	 * so that upper protocol layers will notice, and not attempt
2534 	 * to obtain locks for an ifnet which no longer exists. The
2535 	 * routing socket announcement must happen before the ifnet
2536 	 * instance is detached from the system.
2537 	 */
2538 	if (detaching) {
2539 #ifdef DIAGNOSTIC
2540 		printf("%s: detaching ifnet instance %p\n", __func__, ifp);
2541 #endif
2542 		/*
2543 		 * ifp may already be nulled out if we are being reentered
2544 		 * to delete the ll_ifma.
2545 		 */
2546 		if (ifp != NULL) {
2547 			rt_newmaddrmsg(RTM_DELMADDR, ifma);
2548 			ifma->ifma_ifp = NULL;
2549 		}
2550 	}
2551 
2552 	if (--ifma->ifma_refcount > 0)
2553 		return 0;
2554 
2555 	/*
2556 	 * If this ifma is a network-layer ifma, a link-layer ifma may
2557 	 * have been associated with it. Release it first if so.
2558 	 */
2559 	ll_ifma = ifma->ifma_llifma;
2560 	if (ll_ifma != NULL) {
2561 		KASSERT(ifma->ifma_lladdr != NULL,
2562 		    ("%s: llifma w/o lladdr", __func__));
2563 		if (detaching)
2564 			ll_ifma->ifma_ifp = NULL;	/* XXX */
2565 		if (--ll_ifma->ifma_refcount == 0) {
2566 			if (ifp != NULL) {
2567 				TAILQ_REMOVE(&ifp->if_multiaddrs, ll_ifma,
2568 				    ifma_link);
2569 			}
2570 			if_freemulti(ll_ifma);
2571 		}
2572 	}
2573 
2574 	if (ifp != NULL)
2575 		TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
2576 
2577 	if_freemulti(ifma);
2578 
2579 	/*
2580 	 * The last reference to this instance of struct ifmultiaddr
2581 	 * was released; the hardware should be notified of this change.
2582 	 */
2583 	return 1;
2584 }
2585 
2586 /*
2587  * Set the link layer address on an interface.
2588  *
2589  * At this time we only support certain types of interfaces,
2590  * and we don't allow the length of the address to change.
2591  */
2592 int
2593 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
2594 {
2595 	struct sockaddr_dl *sdl;
2596 	struct ifaddr *ifa;
2597 	struct ifreq ifr;
2598 
2599 	ifa = ifp->if_addr;
2600 	if (ifa == NULL)
2601 		return (EINVAL);
2602 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
2603 	if (sdl == NULL)
2604 		return (EINVAL);
2605 	if (len != sdl->sdl_alen)	/* don't allow length to change */
2606 		return (EINVAL);
2607 	switch (ifp->if_type) {
2608 	case IFT_ETHER:
2609 	case IFT_FDDI:
2610 	case IFT_XETHER:
2611 	case IFT_ISO88025:
2612 	case IFT_L2VLAN:
2613 	case IFT_BRIDGE:
2614 	case IFT_ARCNET:
2615 	case IFT_IEEE8023ADLAG:
2616 		bcopy(lladdr, LLADDR(sdl), len);
2617 		break;
2618 	default:
2619 		return (ENODEV);
2620 	}
2621 	/*
2622 	 * If the interface is already up, we need
2623 	 * to re-init it in order to reprogram its
2624 	 * address filter.
2625 	 */
2626 	if ((ifp->if_flags & IFF_UP) != 0) {
2627 		if (ifp->if_ioctl) {
2628 			IFF_LOCKGIANT(ifp);
2629 			ifp->if_flags &= ~IFF_UP;
2630 			ifr.ifr_flags = ifp->if_flags & 0xffff;
2631 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
2632 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2633 			ifp->if_flags |= IFF_UP;
2634 			ifr.ifr_flags = ifp->if_flags & 0xffff;
2635 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
2636 			(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
2637 			IFF_UNLOCKGIANT(ifp);
2638 		}
2639 #ifdef INET
2640 		/*
2641 		 * Also send gratuitous ARPs to notify other nodes about
2642 		 * the address change.
2643 		 */
2644 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2645 			if (ifa->ifa_addr->sa_family == AF_INET)
2646 				arp_ifinit(ifp, ifa);
2647 		}
2648 #endif
2649 	}
2650 	return (0);
2651 }
2652 
2653 /*
2654  * The name argument must be a pointer to storage which will last as
2655  * long as the interface does.  For physical devices, the result of
2656  * device_get_name(dev) is a good choice and for pseudo-devices a
2657  * static string works well.
2658  */
2659 void
2660 if_initname(struct ifnet *ifp, const char *name, int unit)
2661 {
2662 	ifp->if_dname = name;
2663 	ifp->if_dunit = unit;
2664 	if (unit != IF_DUNIT_NONE)
2665 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
2666 	else
2667 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
2668 }
2669 
2670 int
2671 if_printf(struct ifnet *ifp, const char * fmt, ...)
2672 {
2673 	va_list ap;
2674 	int retval;
2675 
2676 	retval = printf("%s: ", ifp->if_xname);
2677 	va_start(ap, fmt);
2678 	retval += vprintf(fmt, ap);
2679 	va_end(ap);
2680 	return (retval);
2681 }
2682 
2683 /*
2684  * When an interface is marked IFF_NEEDSGIANT, its if_start() routine cannot
2685  * be called without Giant.  However, we often can't acquire the Giant lock
2686  * at those points; instead, we run it via a task queue that holds Giant via
2687  * if_start_deferred.
2688  *
2689  * XXXRW: We need to make sure that the ifnet isn't fully detached until any
2690  * outstanding if_start_deferred() tasks that will run after the free.  This
2691  * probably means waiting in if_detach().
2692  */
2693 void
2694 if_start(struct ifnet *ifp)
2695 {
2696 
2697 	NET_ASSERT_GIANT();
2698 
2699 	if ((ifp->if_flags & IFF_NEEDSGIANT) != 0 && debug_mpsafenet != 0) {
2700 		if (mtx_owned(&Giant))
2701 			(*(ifp)->if_start)(ifp);
2702 		else
2703 			taskqueue_enqueue(taskqueue_swi_giant,
2704 			    &ifp->if_starttask);
2705 	} else
2706 		(*(ifp)->if_start)(ifp);
2707 }
2708 
2709 static void
2710 if_start_deferred(void *context, int pending)
2711 {
2712 	struct ifnet *ifp;
2713 
2714 	/*
2715 	 * This code must be entered with Giant, and should never run if
2716 	 * we're not running with debug.mpsafenet.
2717 	 */
2718 	KASSERT(debug_mpsafenet != 0, ("if_start_deferred: debug.mpsafenet"));
2719 	GIANT_REQUIRED;
2720 
2721 	ifp = context;
2722 	(ifp->if_start)(ifp);
2723 }
2724 
2725 int
2726 if_handoff(struct ifqueue *ifq, struct mbuf *m, struct ifnet *ifp, int adjust)
2727 {
2728 	int active = 0;
2729 
2730 	IF_LOCK(ifq);
2731 	if (_IF_QFULL(ifq)) {
2732 		_IF_DROP(ifq);
2733 		IF_UNLOCK(ifq);
2734 		m_freem(m);
2735 		return (0);
2736 	}
2737 	if (ifp != NULL) {
2738 		ifp->if_obytes += m->m_pkthdr.len + adjust;
2739 		if (m->m_flags & (M_BCAST|M_MCAST))
2740 			ifp->if_omcasts++;
2741 		active = ifp->if_drv_flags & IFF_DRV_OACTIVE;
2742 	}
2743 	_IF_ENQUEUE(ifq, m);
2744 	IF_UNLOCK(ifq);
2745 	if (ifp != NULL && !active)
2746 		if_start(ifp);
2747 	return (1);
2748 }
2749 
2750 void
2751 if_register_com_alloc(u_char type,
2752     if_com_alloc_t *a, if_com_free_t *f)
2753 {
2754 
2755 	KASSERT(if_com_alloc[type] == NULL,
2756 	    ("if_register_com_alloc: %d already registered", type));
2757 	KASSERT(if_com_free[type] == NULL,
2758 	    ("if_register_com_alloc: %d free already registered", type));
2759 
2760 	if_com_alloc[type] = a;
2761 	if_com_free[type] = f;
2762 }
2763 
2764 void
2765 if_deregister_com_alloc(u_char type)
2766 {
2767 
2768 	KASSERT(if_com_alloc[type] != NULL,
2769 	    ("if_deregister_com_alloc: %d not registered", type));
2770 	KASSERT(if_com_free[type] != NULL,
2771 	    ("if_deregister_com_alloc: %d free not registered", type));
2772 	if_com_alloc[type] = NULL;
2773 	if_com_free[type] = NULL;
2774 }
2775