xref: /freebsd/sys/net/if.c (revision 68e7a217f8019b955f87547f218e95ab237597af)
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  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)if.c	8.5 (Berkeley) 1/9/95
34  * $FreeBSD$
35  */
36 
37 #include "opt_compat.h"
38 #include "opt_inet6.h"
39 #include "opt_inet.h"
40 
41 #include <sys/param.h>
42 #include <sys/conf.h>
43 #include <sys/malloc.h>
44 #include <sys/bus.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/proc.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/protosw.h>
51 #include <sys/kernel.h>
52 #include <sys/sockio.h>
53 #include <sys/syslog.h>
54 #include <sys/sysctl.h>
55 #include <sys/jail.h>
56 
57 #include <net/if.h>
58 #include <net/if_arp.h>
59 #include <net/if_dl.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/radix.h>
63 #include <net/route.h>
64 
65 #if defined(INET) || defined(INET6)
66 /*XXX*/
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #ifdef INET6
70 #include <netinet6/in6_var.h>
71 #include <netinet6/in6_ifattach.h>
72 #endif
73 #endif
74 #ifdef INET
75 #include <netinet/if_ether.h>
76 #endif
77 
78 static int	ifconf(u_long, caddr_t);
79 static void	if_grow(void);
80 static void	if_init(void *);
81 static void	if_check(void *);
82 static int	if_findindex(struct ifnet *);
83 static void	if_qflush(struct ifqueue *);
84 static void	if_slowtimo(void *);
85 static void	link_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
86 static int	if_rtdel(struct radix_node *, void *);
87 static struct	if_clone *if_clone_lookup(const char *, int *);
88 static int	if_clone_list(struct if_clonereq *);
89 static int	ifhwioctl(u_long, struct ifnet *, caddr_t, struct thread *);
90 #ifdef INET6
91 /*
92  * XXX: declare here to avoid to include many inet6 related files..
93  * should be more generalized?
94  */
95 extern void	nd6_setmtu(struct ifnet *);
96 #endif
97 
98 int	if_index = 0;
99 struct	ifindex_entry *ifindex_table = NULL;
100 int	ifqmaxlen = IFQ_MAXLEN;
101 struct	ifnethead ifnet;	/* depend on static init XXX */
102 int	if_cloners_count;
103 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
104 
105 static int	if_indexlim = 8;
106 static struct	klist ifklist;
107 
108 static void	filt_netdetach(struct knote *kn);
109 static int	filt_netdev(struct knote *kn, long hint);
110 
111 static struct filterops netdev_filtops =
112     { 1, NULL, filt_netdetach, filt_netdev };
113 
114 /*
115  * System initialization
116  */
117 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL)
118 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL)
119 
120 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
121 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
122 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
123 
124 #define CDEV_MAJOR	165
125 
126 static d_open_t		netopen;
127 static d_close_t	netclose;
128 static d_ioctl_t	netioctl;
129 static d_kqfilter_t	netkqfilter;
130 
131 static struct cdevsw net_cdevsw = {
132 	/* open */	netopen,
133 	/* close */	netclose,
134 	/* read */	noread,
135 	/* write */	nowrite,
136 	/* ioctl */	netioctl,
137 	/* poll */	nopoll,
138 	/* mmap */	nommap,
139 	/* strategy */	nostrategy,
140 	/* name */	"net",
141 	/* maj */	CDEV_MAJOR,
142 	/* dump */	nodump,
143 	/* psize */	nopsize,
144 	/* flags */	D_KQFILTER,
145 	/* kqfilter */	netkqfilter,
146 };
147 
148 static int
149 netopen(dev_t dev, int flag, int mode, struct thread *td)
150 {
151 	return (0);
152 }
153 
154 static int
155 netclose(dev_t dev, int flags, int fmt, struct thread *td)
156 {
157 	return (0);
158 }
159 
160 static int
161 netioctl(dev_t dev, u_long cmd, caddr_t data, int flag, struct thread *td)
162 {
163 	struct ifnet *ifp;
164 	int error, idx;
165 
166 	/* only support interface specific ioctls */
167 	if (IOCGROUP(cmd) != 'i')
168 		return (EOPNOTSUPP);
169 	idx = minor(dev);
170 	if (idx == 0) {
171 		/*
172 		 * special network device, not interface.
173 		 */
174 		if (cmd == SIOCGIFCONF)
175 			return (ifconf(cmd, data));	/* XXX remove cmd */
176 		return (EOPNOTSUPP);
177 	}
178 
179 	ifp = ifnet_byindex(idx);
180 	if (ifp == NULL)
181 		return (ENXIO);
182 
183 	error = ifhwioctl(cmd, ifp, data, td);
184 	if (error == ENOIOCTL)
185 		error = EOPNOTSUPP;
186 	return (error);
187 }
188 
189 static int
190 netkqfilter(dev_t dev, struct knote *kn)
191 {
192 	struct klist *klist;
193 	struct ifnet *ifp;
194 	int idx;
195 
196 	idx = minor(dev);
197 	if (idx == 0) {
198 		klist = &ifklist;
199 	} else {
200 		ifp = ifnet_byindex(idx);
201 		if (ifp == NULL)
202 			return (1);
203 		klist = &ifp->if_klist;
204 	}
205 
206 	switch (kn->kn_filter) {
207 	case EVFILT_NETDEV:
208 		kn->kn_fop = &netdev_filtops;
209 		break;
210 	default:
211 		return (1);
212 	}
213 
214 	kn->kn_hook = (caddr_t)klist;
215 
216 	/* XXX locking? */
217 	SLIST_INSERT_HEAD(klist, kn, kn_selnext);
218 
219 	return (0);
220 }
221 
222 static void
223 filt_netdetach(struct knote *kn)
224 {
225 	struct klist *klist = (struct klist *)kn->kn_hook;
226 
227 	if (kn->kn_status & KN_DETACHED)
228 		return;
229 	SLIST_REMOVE(klist, kn, knote, kn_selnext);
230 }
231 
232 static int
233 filt_netdev(struct knote *kn, long hint)
234 {
235 
236 	/*
237 	 * Currently NOTE_EXIT is abused to indicate device detach.
238 	 */
239 	if (hint == NOTE_EXIT) {
240 		kn->kn_data = NOTE_LINKINV;
241                 kn->kn_status |= KN_DETACHED;
242                 kn->kn_flags |= (EV_EOF | EV_ONESHOT);
243                 return (1);
244         }
245 	kn->kn_data = hint;			/* current status */
246 	if (kn->kn_sfflags & hint)
247 		kn->kn_fflags |= hint;
248 	return (kn->kn_fflags != 0);
249 }
250 
251 /*
252  * Network interface utility routines.
253  *
254  * Routines with ifa_ifwith* names take sockaddr *'s as
255  * parameters.
256  */
257 /* ARGSUSED*/
258 static void
259 if_init(dummy)
260 	void *dummy;
261 {
262 
263 	TAILQ_INIT(&ifnet);
264 	SLIST_INIT(&ifklist);
265 	if_grow();				/* create initial table */
266 	ifdev_byindex(0) = make_dev(&net_cdevsw, 0,
267 	    UID_ROOT, GID_WHEEL, 0600, "network");
268 }
269 
270 static void
271 if_grow(void)
272 {
273 	u_int n;
274 	struct ifindex_entry *e;
275 
276 	if_indexlim <<= 1;
277 	n = if_indexlim * sizeof(*e);
278 	e = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
279 	if (ifindex_table != NULL) {
280 		memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2);
281 		free((caddr_t)ifindex_table, M_IFADDR);
282 	}
283 	ifindex_table = e;
284 }
285 
286 /* ARGSUSED*/
287 static void
288 if_check(dummy)
289 	void *dummy;
290 {
291 	struct ifnet *ifp;
292 	int s;
293 
294 	s = splimp();
295 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
296 		if (ifp->if_snd.ifq_maxlen == 0) {
297 			printf("%s%d XXX: driver didn't set ifq_maxlen\n",
298 			    ifp->if_name, ifp->if_unit);
299 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
300 		}
301 		if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) {
302 			printf("%s%d XXX: driver didn't initialize queue mtx\n",
303 			    ifp->if_name, ifp->if_unit);
304 			mtx_init(&ifp->if_snd.ifq_mtx, "unknown",
305 			    MTX_NETWORK_LOCK, MTX_DEF);
306 		}
307 	}
308 	splx(s);
309 	if_slowtimo(0);
310 }
311 
312 static int
313 if_findindex(struct ifnet *ifp)
314 {
315 	int i, unit;
316 	char eaddr[18], devname[32];
317 	const char *name, *p;
318 
319 	switch (ifp->if_type) {
320 	case IFT_ETHER:			/* these types use struct arpcom */
321 	case IFT_FDDI:
322 	case IFT_XETHER:
323 	case IFT_ISO88025:
324 	case IFT_L2VLAN:
325 		snprintf(eaddr, 18, "%6D",
326 		    ((struct arpcom *)ifp->if_softc)->ac_enaddr, ":");
327 		break;
328 	default:
329 		eaddr[0] = '\0';
330 		break;
331 	}
332 	snprintf(devname, 32, "%s%d", ifp->if_name, ifp->if_unit);
333 	name = net_cdevsw.d_name;
334 	i = 0;
335 	while ((resource_find_dev(&i, name, &unit, NULL, NULL)) == 0) {
336 		if (resource_string_value(name, unit, "ether", &p) == 0)
337 			if (strcmp(p, eaddr) == 0)
338 				goto found;
339 		if (resource_string_value(name, unit, "dev", &p) == 0)
340 			if (strcmp(p, devname) == 0)
341 				goto found;
342 	}
343 	unit = 0;
344 found:
345 	if (unit != 0) {
346 		if (ifaddr_byindex(unit) == NULL)
347 			return (unit);
348 		printf("%s%d in use, cannot hardwire it to %s.\n",
349 		    name, unit, devname);
350 	}
351 	for (unit = 1; ; unit++) {
352 		if (unit <= if_index && ifaddr_byindex(unit) != NULL)
353 			continue;
354 		if (resource_string_value(name, unit, "ether", &p) == 0 ||
355 		    resource_string_value(name, unit, "dev", &p) == 0)
356 			continue;
357 		break;
358 	}
359 	return (unit);
360 }
361 
362 /*
363  * Attach an interface to the
364  * list of "active" interfaces.
365  */
366 void
367 if_attach(ifp)
368 	struct ifnet *ifp;
369 {
370 	unsigned socksize, ifasize;
371 	int namelen, masklen;
372 	char workbuf[64];
373 	register struct sockaddr_dl *sdl;
374 	register struct ifaddr *ifa;
375 
376 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
377 	/*
378 	 * XXX -
379 	 * The old code would work if the interface passed a pre-existing
380 	 * chain of ifaddrs to this code.  We don't trust our callers to
381 	 * properly initialize the tailq, however, so we no longer allow
382 	 * this unlikely case.
383 	 */
384 	TAILQ_INIT(&ifp->if_addrhead);
385 	TAILQ_INIT(&ifp->if_prefixhead);
386 	TAILQ_INIT(&ifp->if_multiaddrs);
387 	SLIST_INIT(&ifp->if_klist);
388 	getmicrotime(&ifp->if_lastchange);
389 	ifp->if_index = if_findindex(ifp);
390 	if (ifp->if_index > if_index)
391 		if_index = ifp->if_index;
392 	if (if_index >= if_indexlim)
393 		if_grow();
394 
395 	ifnet_byindex(ifp->if_index) = ifp;
396 	ifdev_byindex(ifp->if_index) = make_dev(&net_cdevsw, ifp->if_index,
397 	    UID_ROOT, GID_WHEEL, 0600, "%s/%s%d",
398 	    net_cdevsw.d_name, ifp->if_name, ifp->if_unit);
399 	make_dev_alias(ifdev_byindex(ifp->if_index), "%s%d",
400 	    net_cdevsw.d_name, ifp->if_index);
401 
402 	mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_name, "if send queue", MTX_DEF);
403 
404 	/*
405 	 * create a Link Level name for this device
406 	 */
407 	namelen = snprintf(workbuf, sizeof(workbuf),
408 	    "%s%d", ifp->if_name, ifp->if_unit);
409 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
410 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
411 	socksize = masklen + ifp->if_addrlen;
412 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
413 	if (socksize < sizeof(*sdl))
414 		socksize = sizeof(*sdl);
415 	socksize = ROUNDUP(socksize);
416 	ifasize = sizeof(*ifa) + 2 * socksize;
417 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
418 	if (ifa) {
419 		sdl = (struct sockaddr_dl *)(ifa + 1);
420 		sdl->sdl_len = socksize;
421 		sdl->sdl_family = AF_LINK;
422 		bcopy(workbuf, sdl->sdl_data, namelen);
423 		sdl->sdl_nlen = namelen;
424 		sdl->sdl_index = ifp->if_index;
425 		sdl->sdl_type = ifp->if_type;
426 		ifaddr_byindex(ifp->if_index) = ifa;
427 		ifa->ifa_ifp = ifp;
428 		ifa->ifa_rtrequest = link_rtrequest;
429 		ifa->ifa_addr = (struct sockaddr *)sdl;
430 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
431 		ifa->ifa_netmask = (struct sockaddr *)sdl;
432 		sdl->sdl_len = masklen;
433 		while (namelen != 0)
434 			sdl->sdl_data[--namelen] = 0xff;
435 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
436 	}
437 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
438 
439 	/* Announce the interface. */
440 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
441 }
442 
443 /*
444  * Detach an interface, removing it from the
445  * list of "active" interfaces.
446  */
447 void
448 if_detach(ifp)
449 	struct ifnet *ifp;
450 {
451 	struct ifaddr *ifa;
452 	struct radix_node_head	*rnh;
453 	int s;
454 	int i;
455 
456 	/*
457 	 * Remove routes and flush queues.
458 	 */
459 	s = splnet();
460 	if_down(ifp);
461 
462 	/*
463 	 * Remove address from ifindex_table[] and maybe decrement if_index.
464 	 * Clean up all addresses.
465 	 */
466 	ifaddr_byindex(ifp->if_index) = NULL;
467 	revoke_and_destroy_dev(ifdev_byindex(ifp->if_index));
468 	ifdev_byindex(ifp->if_index) = NULL;
469 
470 	while (if_index > 0 && ifaddr_byindex(if_index) == NULL)
471 		if_index--;
472 
473 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
474 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
475 #ifdef INET
476 		/* XXX: Ugly!! ad hoc just for INET */
477 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
478 			struct ifaliasreq ifr;
479 
480 			bzero(&ifr, sizeof(ifr));
481 			ifr.ifra_addr = *ifa->ifa_addr;
482 			if (ifa->ifa_dstaddr)
483 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
484 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
485 			    NULL) == 0)
486 				continue;
487 		}
488 #endif /* INET */
489 #ifdef INET6
490 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
491 			in6_purgeaddr(ifa);
492 			/* ifp_addrhead is already updated */
493 			continue;
494 		}
495 #endif /* INET6 */
496 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
497 		IFAFREE(ifa);
498 	}
499 
500 #ifdef INET6
501 	/*
502 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
503 	 * before removing routing entries below, since IPv6 interface direct
504 	 * routes are expected to be removed by the IPv6-specific kernel API.
505 	 * Otherwise, the kernel will detect some inconsistency and bark it.
506 	 */
507 	in6_ifdetach(ifp);
508 #endif
509 
510 	/*
511 	 * Delete all remaining routes using this interface
512 	 * Unfortuneatly the only way to do this is to slog through
513 	 * the entire routing table looking for routes which point
514 	 * to this interface...oh well...
515 	 */
516 	for (i = 1; i <= AF_MAX; i++) {
517 		if ((rnh = rt_tables[i]) == NULL)
518 			continue;
519 		(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
520 	}
521 
522 	/* Announce that the interface is gone. */
523 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
524 
525 	KNOTE(&ifp->if_klist, NOTE_EXIT);
526 	TAILQ_REMOVE(&ifnet, ifp, if_link);
527 	mtx_destroy(&ifp->if_snd.ifq_mtx);
528 	splx(s);
529 }
530 
531 /*
532  * Delete Routes for a Network Interface
533  *
534  * Called for each routing entry via the rnh->rnh_walktree() call above
535  * to delete all route entries referencing a detaching network interface.
536  *
537  * Arguments:
538  *	rn	pointer to node in the routing table
539  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
540  *
541  * Returns:
542  *	0	successful
543  *	errno	failed - reason indicated
544  *
545  */
546 static int
547 if_rtdel(rn, arg)
548 	struct radix_node	*rn;
549 	void			*arg;
550 {
551 	struct rtentry	*rt = (struct rtentry *)rn;
552 	struct ifnet	*ifp = arg;
553 	int		err;
554 
555 	if (rt->rt_ifp == ifp) {
556 
557 		/*
558 		 * Protect (sorta) against walktree recursion problems
559 		 * with cloned routes
560 		 */
561 		if ((rt->rt_flags & RTF_UP) == 0)
562 			return (0);
563 
564 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
565 				rt_mask(rt), rt->rt_flags,
566 				(struct rtentry **) NULL);
567 		if (err) {
568 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
569 		}
570 	}
571 
572 	return (0);
573 }
574 
575 /*
576  * Create a clone network interface.
577  */
578 int
579 if_clone_create(name, len)
580 	char *name;
581 	int len;
582 {
583 	struct if_clone *ifc;
584 	char *dp;
585 	int wildcard, bytoff, bitoff;
586 	int unit;
587 	int err;
588 
589 	ifc = if_clone_lookup(name, &unit);
590 	if (ifc == NULL)
591 		return (EINVAL);
592 
593 	if (ifunit(name) != NULL)
594 		return (EEXIST);
595 
596 	bytoff = bitoff = 0;
597 	wildcard = (unit < 0);
598 	/*
599 	 * Find a free unit if none was given.
600 	 */
601 	if (wildcard) {
602 		while ((bytoff < ifc->ifc_bmlen)
603 		    && (ifc->ifc_units[bytoff] == 0xff))
604 			bytoff++;
605 		if (bytoff >= ifc->ifc_bmlen)
606 			return (ENOSPC);
607 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
608 			bitoff++;
609 		unit = (bytoff << 3) + bitoff;
610 	}
611 
612 	if (unit > ifc->ifc_maxunit)
613 		return (ENXIO);
614 
615 	err = (*ifc->ifc_create)(ifc, unit);
616 	if (err != 0)
617 		return (err);
618 
619 	if (!wildcard) {
620 		bytoff = unit >> 3;
621 		bitoff = unit - (bytoff << 3);
622 	}
623 
624 	/*
625 	 * Allocate the unit in the bitmap.
626 	 */
627 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
628 	    ("%s: bit is already set", __func__));
629 	ifc->ifc_units[bytoff] |= (1 << bitoff);
630 
631 	/* In the wildcard case, we need to update the name. */
632 	if (wildcard) {
633 		for (dp = name; *dp != '\0'; dp++);
634 		if (snprintf(dp, len - (dp-name), "%d", unit) >
635 		    len - (dp-name) - 1) {
636 			/*
637 			 * This can only be a programmer error and
638 			 * there's no straightforward way to recover if
639 			 * it happens.
640 			 */
641 			panic("if_clone_create(): interface name too long");
642 		}
643 
644 	}
645 
646 	return (0);
647 }
648 
649 /*
650  * Destroy a clone network interface.
651  */
652 int
653 if_clone_destroy(name)
654 	const char *name;
655 {
656 	struct if_clone *ifc;
657 	struct ifnet *ifp;
658 	int bytoff, bitoff;
659 	int err, unit;
660 
661 	ifc = if_clone_lookup(name, &unit);
662 	if (ifc == NULL)
663 		return (EINVAL);
664 
665 	ifp = ifunit(name);
666 	if (ifp == NULL)
667 		return (ENXIO);
668 
669 	if (ifc->ifc_destroy == NULL)
670 		return (EOPNOTSUPP);
671 
672 	err = (*ifc->ifc_destroy)(ifp);
673 	if (err != 0)
674 		return (err);
675 
676 	/*
677 	 * Compute offset in the bitmap and deallocate the unit.
678 	 */
679 	bytoff = unit >> 3;
680 	bitoff = unit - (bytoff << 3);
681 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
682 	    ("%s: bit is already cleared", __func__));
683 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
684 	return (0);
685 }
686 
687 /*
688  * Look up a network interface cloner.
689  */
690 static struct if_clone *
691 if_clone_lookup(name, unitp)
692 	const char *name;
693 	int *unitp;
694 {
695 	struct if_clone *ifc;
696 	const char *cp;
697 	int i;
698 
699 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
700 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
701 			if (ifc->ifc_name[i] != *cp)
702 				goto next_ifc;
703 		}
704 		goto found_name;
705  next_ifc:
706 		ifc = LIST_NEXT(ifc, ifc_list);
707 	}
708 
709 	/* No match. */
710 	return ((struct if_clone *)NULL);
711 
712  found_name:
713 	if (*cp == '\0') {
714 		i = -1;
715 	} else {
716 		for (i = 0; *cp != '\0'; cp++) {
717 			if (*cp < '0' || *cp > '9') {
718 				/* Bogus unit number. */
719 				return (NULL);
720 			}
721 			i = (i * 10) + (*cp - '0');
722 		}
723 	}
724 
725 	if (unitp != NULL)
726 		*unitp = i;
727 	return (ifc);
728 }
729 
730 /*
731  * Register a network interface cloner.
732  */
733 void
734 if_clone_attach(ifc)
735 	struct if_clone *ifc;
736 {
737 	int len, maxclone;
738 
739 	/*
740 	 * Compute bitmap size and allocate it.
741 	 */
742 	maxclone = ifc->ifc_maxunit + 1;
743 	len = maxclone >> 3;
744 	if ((len << 3) < maxclone)
745 		len++;
746 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
747 	ifc->ifc_bmlen = len;
748 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
749 	if_cloners_count++;
750 }
751 
752 /*
753  * Unregister a network interface cloner.
754  */
755 void
756 if_clone_detach(ifc)
757 	struct if_clone *ifc;
758 {
759 
760 	LIST_REMOVE(ifc, ifc_list);
761 	free(ifc->ifc_units, M_CLONE);
762 	if_cloners_count--;
763 }
764 
765 /*
766  * Provide list of interface cloners to userspace.
767  */
768 static int
769 if_clone_list(ifcr)
770 	struct if_clonereq *ifcr;
771 {
772 	char outbuf[IFNAMSIZ], *dst;
773 	struct if_clone *ifc;
774 	int count, error = 0;
775 
776 	ifcr->ifcr_total = if_cloners_count;
777 	if ((dst = ifcr->ifcr_buffer) == NULL) {
778 		/* Just asking how many there are. */
779 		return (0);
780 	}
781 
782 	if (ifcr->ifcr_count < 0)
783 		return (EINVAL);
784 
785 	count = (if_cloners_count < ifcr->ifcr_count) ?
786 	    if_cloners_count : ifcr->ifcr_count;
787 
788 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
789 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
790 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
791 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
792 		error = copyout(outbuf, dst, IFNAMSIZ);
793 		if (error)
794 			break;
795 	}
796 
797 	return (error);
798 }
799 
800 /*
801  * Locate an interface based on a complete address.
802  */
803 /*ARGSUSED*/
804 struct ifaddr *
805 ifa_ifwithaddr(addr)
806 	struct sockaddr *addr;
807 {
808 	struct ifnet *ifp;
809 	struct ifaddr *ifa;
810 
811 #define	equal(a1, a2) \
812   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
813 	TAILQ_FOREACH(ifp, &ifnet, if_link)
814 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
815 			if (ifa->ifa_addr->sa_family != addr->sa_family)
816 				continue;
817 			if (equal(addr, ifa->ifa_addr))
818 				goto done;
819 			/* IP6 doesn't have broadcast */
820 			if ((ifp->if_flags & IFF_BROADCAST) &&
821 			    ifa->ifa_broadaddr &&
822 			    ifa->ifa_broadaddr->sa_len != 0 &&
823 			    equal(ifa->ifa_broadaddr, addr))
824 				goto done;
825 		}
826 	ifa = NULL;
827 done:
828 	return (ifa);
829 }
830 
831 /*
832  * Locate the point to point interface with a given destination address.
833  */
834 /*ARGSUSED*/
835 struct ifaddr *
836 ifa_ifwithdstaddr(addr)
837 	struct sockaddr *addr;
838 {
839 	struct ifnet *ifp;
840 	struct ifaddr *ifa;
841 
842 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
843 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
844 			continue;
845 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
846 			if (ifa->ifa_addr->sa_family != addr->sa_family)
847 				continue;
848 			if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
849 				goto done;
850 		}
851 	}
852 	ifa = NULL;
853 done:
854 	return (ifa);
855 }
856 
857 /*
858  * Find an interface on a specific network.  If many, choice
859  * is most specific found.
860  */
861 struct ifaddr *
862 ifa_ifwithnet(addr)
863 	struct sockaddr *addr;
864 {
865 	register struct ifnet *ifp;
866 	register struct ifaddr *ifa;
867 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
868 	u_int af = addr->sa_family;
869 	char *addr_data = addr->sa_data, *cplim;
870 
871 	/*
872 	 * AF_LINK addresses can be looked up directly by their index number,
873 	 * so do that if we can.
874 	 */
875 	if (af == AF_LINK) {
876 	    register struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
877 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
878 		return (ifaddr_byindex(sdl->sdl_index));
879 	}
880 
881 	/*
882 	 * Scan though each interface, looking for ones that have
883 	 * addresses in this address family.
884 	 */
885 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
886 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
887 			register char *cp, *cp2, *cp3;
888 
889 			if (ifa->ifa_addr->sa_family != af)
890 next:				continue;
891 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
892 				/*
893 				 * This is a bit broken as it doesn't
894 				 * take into account that the remote end may
895 				 * be a single node in the network we are
896 				 * looking for.
897 				 * The trouble is that we don't know the
898 				 * netmask for the remote end.
899 				 */
900 				if (ifa->ifa_dstaddr != 0
901 				    && equal(addr, ifa->ifa_dstaddr))
902 					goto done;
903 			} else {
904 				/*
905 				 * if we have a special address handler,
906 				 * then use it instead of the generic one.
907 				 */
908 	          		if (ifa->ifa_claim_addr) {
909 					if ((*ifa->ifa_claim_addr)(ifa, addr))
910 						goto done;
911 					continue;
912 				}
913 
914 				/*
915 				 * Scan all the bits in the ifa's address.
916 				 * If a bit dissagrees with what we are
917 				 * looking for, mask it with the netmask
918 				 * to see if it really matters.
919 				 * (A byte at a time)
920 				 */
921 				if (ifa->ifa_netmask == 0)
922 					continue;
923 				cp = addr_data;
924 				cp2 = ifa->ifa_addr->sa_data;
925 				cp3 = ifa->ifa_netmask->sa_data;
926 				cplim = ifa->ifa_netmask->sa_len
927 					+ (char *)ifa->ifa_netmask;
928 				while (cp3 < cplim)
929 					if ((*cp++ ^ *cp2++) & *cp3++)
930 						goto next; /* next address! */
931 				/*
932 				 * If the netmask of what we just found
933 				 * is more specific than what we had before
934 				 * (if we had one) then remember the new one
935 				 * before continuing to search
936 				 * for an even better one.
937 				 */
938 				if (ifa_maybe == 0 ||
939 				    rn_refines((caddr_t)ifa->ifa_netmask,
940 				    (caddr_t)ifa_maybe->ifa_netmask))
941 					ifa_maybe = ifa;
942 			}
943 		}
944 	}
945 	ifa = ifa_maybe;
946 done:
947 	return (ifa);
948 }
949 
950 /*
951  * Find an interface address specific to an interface best matching
952  * a given address.
953  */
954 struct ifaddr *
955 ifaof_ifpforaddr(addr, ifp)
956 	struct sockaddr *addr;
957 	register struct ifnet *ifp;
958 {
959 	register struct ifaddr *ifa;
960 	register char *cp, *cp2, *cp3;
961 	register char *cplim;
962 	struct ifaddr *ifa_maybe = 0;
963 	u_int af = addr->sa_family;
964 
965 	if (af >= AF_MAX)
966 		return (0);
967 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
968 		if (ifa->ifa_addr->sa_family != af)
969 			continue;
970 		if (ifa_maybe == 0)
971 			ifa_maybe = ifa;
972 		if (ifa->ifa_netmask == 0) {
973 			if (equal(addr, ifa->ifa_addr) ||
974 			    (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
975 				goto done;
976 			continue;
977 		}
978 		if (ifp->if_flags & IFF_POINTOPOINT) {
979 			if (equal(addr, ifa->ifa_dstaddr))
980 				goto done;
981 		} else {
982 			cp = addr->sa_data;
983 			cp2 = ifa->ifa_addr->sa_data;
984 			cp3 = ifa->ifa_netmask->sa_data;
985 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
986 			for (; cp3 < cplim; cp3++)
987 				if ((*cp++ ^ *cp2++) & *cp3)
988 					break;
989 			if (cp3 == cplim)
990 				goto done;
991 		}
992 	}
993 	ifa = ifa_maybe;
994 done:
995 	return (ifa);
996 }
997 
998 #include <net/route.h>
999 
1000 /*
1001  * Default action when installing a route with a Link Level gateway.
1002  * Lookup an appropriate real ifa to point to.
1003  * This should be moved to /sys/net/link.c eventually.
1004  */
1005 static void
1006 link_rtrequest(cmd, rt, info)
1007 	int cmd;
1008 	register struct rtentry *rt;
1009 	struct rt_addrinfo *info;
1010 {
1011 	register struct ifaddr *ifa;
1012 	struct sockaddr *dst;
1013 	struct ifnet *ifp;
1014 
1015 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
1016 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
1017 		return;
1018 	ifa = ifaof_ifpforaddr(dst, ifp);
1019 	if (ifa) {
1020 		IFAFREE(rt->rt_ifa);
1021 		rt->rt_ifa = ifa;
1022 		ifa->ifa_refcnt++;
1023 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1024 			ifa->ifa_rtrequest(cmd, rt, info);
1025 	}
1026 }
1027 
1028 /*
1029  * Mark an interface down and notify protocols of
1030  * the transition.
1031  * NOTE: must be called at splnet or eqivalent.
1032  */
1033 void
1034 if_unroute(ifp, flag, fam)
1035 	register struct ifnet *ifp;
1036 	int flag, fam;
1037 {
1038 	register struct ifaddr *ifa;
1039 
1040 	ifp->if_flags &= ~flag;
1041 	getmicrotime(&ifp->if_lastchange);
1042 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1043 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1044 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1045 	if_qflush(&ifp->if_snd);
1046 	rt_ifmsg(ifp);
1047 }
1048 
1049 /*
1050  * Mark an interface up and notify protocols of
1051  * the transition.
1052  * NOTE: must be called at splnet or eqivalent.
1053  */
1054 void
1055 if_route(ifp, flag, fam)
1056 	register struct ifnet *ifp;
1057 	int flag, fam;
1058 {
1059 	register struct ifaddr *ifa;
1060 
1061 	ifp->if_flags |= flag;
1062 	getmicrotime(&ifp->if_lastchange);
1063 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
1064 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
1065 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
1066 	rt_ifmsg(ifp);
1067 #ifdef INET6
1068 	in6_if_up(ifp);
1069 #endif
1070 }
1071 
1072 /*
1073  * Mark an interface down and notify protocols of
1074  * the transition.
1075  * NOTE: must be called at splnet or eqivalent.
1076  */
1077 void
1078 if_down(ifp)
1079 	register struct ifnet *ifp;
1080 {
1081 
1082 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
1083 }
1084 
1085 /*
1086  * Mark an interface up and notify protocols of
1087  * the transition.
1088  * NOTE: must be called at splnet or eqivalent.
1089  */
1090 void
1091 if_up(ifp)
1092 	register struct ifnet *ifp;
1093 {
1094 
1095 	if_route(ifp, IFF_UP, AF_UNSPEC);
1096 }
1097 
1098 /*
1099  * Flush an interface queue.
1100  */
1101 static void
1102 if_qflush(ifq)
1103 	register struct ifqueue *ifq;
1104 {
1105 	register struct mbuf *m, *n;
1106 
1107 	n = ifq->ifq_head;
1108 	while ((m = n) != 0) {
1109 		n = m->m_act;
1110 		m_freem(m);
1111 	}
1112 	ifq->ifq_head = 0;
1113 	ifq->ifq_tail = 0;
1114 	ifq->ifq_len = 0;
1115 }
1116 
1117 /*
1118  * Handle interface watchdog timer routines.  Called
1119  * from softclock, we decrement timers (if set) and
1120  * call the appropriate interface routine on expiration.
1121  */
1122 static void
1123 if_slowtimo(arg)
1124 	void *arg;
1125 {
1126 	register struct ifnet *ifp;
1127 	int s = splimp();
1128 
1129 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1130 		if (ifp->if_timer == 0 || --ifp->if_timer)
1131 			continue;
1132 		if (ifp->if_watchdog)
1133 			(*ifp->if_watchdog)(ifp);
1134 	}
1135 	splx(s);
1136 	timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
1137 }
1138 
1139 /*
1140  * Map interface name to
1141  * interface structure pointer.
1142  */
1143 struct ifnet *
1144 ifunit(const char *name)
1145 {
1146 	char namebuf[IFNAMSIZ + 1];
1147 	struct ifnet *ifp;
1148 	dev_t dev;
1149 
1150 	/*
1151 	 * Now search all the interfaces for this name/number
1152 	 */
1153 
1154 	/*
1155 	 * XXX
1156 	 * Devices should really be known as /dev/fooN, not /dev/net/fooN.
1157 	 */
1158 	snprintf(namebuf, IFNAMSIZ, "%s/%s", net_cdevsw.d_name, name);
1159 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1160 		dev = ifdev_byindex(ifp->if_index);
1161 		if (strcmp(devtoname(dev), namebuf) == 0)
1162 			break;
1163 		if (dev_named(dev, name))
1164 			break;
1165 	}
1166 	return (ifp);
1167 }
1168 
1169 /*
1170  * Map interface name in a sockaddr_dl to
1171  * interface structure pointer.
1172  */
1173 struct ifnet *
1174 if_withname(sa)
1175 	struct sockaddr *sa;
1176 {
1177 	char ifname[IFNAMSIZ+1];
1178 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1179 
1180 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1181 	     (sdl->sdl_nlen > IFNAMSIZ) )
1182 		return NULL;
1183 
1184 	/*
1185 	 * ifunit wants a null-terminated name.  It may not be null-terminated
1186 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
1187 	 * and there might not be room to put the trailing null anyway, so we
1188 	 * make a local copy that we know we can null terminate safely.
1189 	 */
1190 
1191 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1192 	ifname[sdl->sdl_nlen] = '\0';
1193 	return ifunit(ifname);
1194 }
1195 
1196 /*
1197  * Hardware specific interface ioctls.
1198  */
1199 static int
1200 ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
1201 {
1202 	struct ifreq *ifr;
1203 	struct ifstat *ifs;
1204 	int error = 0;
1205 
1206 	ifr = (struct ifreq *)data;
1207 	switch (cmd) {
1208 	case SIOCGIFINDEX:
1209 		ifr->ifr_index = ifp->if_index;
1210 		break;
1211 
1212 	case SIOCGIFFLAGS:
1213 		ifr->ifr_flags = ifp->if_flags;
1214 		break;
1215 
1216 	case SIOCGIFCAP:
1217 		ifr->ifr_reqcap = ifp->if_capabilities;
1218 		ifr->ifr_curcap = ifp->if_capenable;
1219 		break;
1220 
1221 	case SIOCGIFMETRIC:
1222 		ifr->ifr_metric = ifp->if_metric;
1223 		break;
1224 
1225 	case SIOCGIFMTU:
1226 		ifr->ifr_mtu = ifp->if_mtu;
1227 		break;
1228 
1229 	case SIOCGIFPHYS:
1230 		ifr->ifr_phys = ifp->if_physical;
1231 		break;
1232 
1233 	case SIOCSIFFLAGS:
1234 		error = suser(td);
1235 		if (error)
1236 			return (error);
1237 		ifr->ifr_prevflags = ifp->if_flags;
1238 		if (ifp->if_flags & IFF_SMART) {
1239 			/* Smart drivers twiddle their own routes */
1240 		} else if (ifp->if_flags & IFF_UP &&
1241 		    (ifr->ifr_flags & IFF_UP) == 0) {
1242 			int s = splimp();
1243 			if_down(ifp);
1244 			splx(s);
1245 		} else if (ifr->ifr_flags & IFF_UP &&
1246 		    (ifp->if_flags & IFF_UP) == 0) {
1247 			int s = splimp();
1248 			if_up(ifp);
1249 			splx(s);
1250 		}
1251 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1252 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1253 		if (ifp->if_ioctl)
1254 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1255 		getmicrotime(&ifp->if_lastchange);
1256 		break;
1257 
1258 	case SIOCSIFCAP:
1259 		error = suser(td);
1260 		if (error)
1261 			return (error);
1262 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1263 			return (EINVAL);
1264 		(void) (*ifp->if_ioctl)(ifp, cmd, data);
1265 		break;
1266 
1267 	case SIOCSIFMETRIC:
1268 		error = suser(td);
1269 		if (error)
1270 			return (error);
1271 		ifp->if_metric = ifr->ifr_metric;
1272 		getmicrotime(&ifp->if_lastchange);
1273 		break;
1274 
1275 	case SIOCSIFPHYS:
1276 		error = suser(td);
1277 		if (error)
1278 			return error;
1279 		if (!ifp->if_ioctl)
1280 		        return EOPNOTSUPP;
1281 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1282 		if (error == 0)
1283 			getmicrotime(&ifp->if_lastchange);
1284 		return(error);
1285 
1286 	case SIOCSIFMTU:
1287 	{
1288 		u_long oldmtu = ifp->if_mtu;
1289 
1290 		error = suser(td);
1291 		if (error)
1292 			return (error);
1293 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1294 			return (EINVAL);
1295 		if (ifp->if_ioctl == NULL)
1296 			return (EOPNOTSUPP);
1297 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1298 		if (error == 0) {
1299 			getmicrotime(&ifp->if_lastchange);
1300 			rt_ifmsg(ifp);
1301 		}
1302 		/*
1303 		 * If the link MTU changed, do network layer specific procedure.
1304 		 */
1305 		if (ifp->if_mtu != oldmtu) {
1306 #ifdef INET6
1307 			nd6_setmtu(ifp);
1308 #endif
1309 		}
1310 		break;
1311 	}
1312 
1313 	case SIOCADDMULTI:
1314 	case SIOCDELMULTI:
1315 		error = suser(td);
1316 		if (error)
1317 			return (error);
1318 
1319 		/* Don't allow group membership on non-multicast interfaces. */
1320 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1321 			return (EOPNOTSUPP);
1322 
1323 		/* Don't let users screw up protocols' entries. */
1324 		if (ifr->ifr_addr.sa_family != AF_LINK)
1325 			return (EINVAL);
1326 
1327 		if (cmd == SIOCADDMULTI) {
1328 			struct ifmultiaddr *ifma;
1329 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1330 		} else {
1331 			error = if_delmulti(ifp, &ifr->ifr_addr);
1332 		}
1333 		if (error == 0)
1334 			getmicrotime(&ifp->if_lastchange);
1335 		break;
1336 
1337 	case SIOCSIFPHYADDR:
1338 	case SIOCDIFPHYADDR:
1339 #ifdef INET6
1340 	case SIOCSIFPHYADDR_IN6:
1341 #endif
1342 	case SIOCSLIFPHYADDR:
1343         case SIOCSIFMEDIA:
1344 	case SIOCSIFGENERIC:
1345 		error = suser(td);
1346 		if (error)
1347 			return (error);
1348 		if (ifp->if_ioctl == NULL)
1349 			return (EOPNOTSUPP);
1350 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1351 		if (error == 0)
1352 			getmicrotime(&ifp->if_lastchange);
1353 		break;
1354 
1355 	case SIOCGIFSTATUS:
1356 		ifs = (struct ifstat *)data;
1357 		ifs->ascii[0] = '\0';
1358 
1359 	case SIOCGIFPSRCADDR:
1360 	case SIOCGIFPDSTADDR:
1361 	case SIOCGLIFPHYADDR:
1362 	case SIOCGIFMEDIA:
1363 	case SIOCGIFGENERIC:
1364 		if (ifp->if_ioctl == 0)
1365 			return (EOPNOTSUPP);
1366 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1367 		break;
1368 
1369 	case SIOCSIFLLADDR:
1370 		error = suser(td);
1371 		if (error)
1372 			return (error);
1373 		error = if_setlladdr(ifp,
1374 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1375 		break;
1376 
1377 	default:
1378 		error = ENOIOCTL;
1379 		break;
1380 	}
1381 	return (error);
1382 }
1383 
1384 /*
1385  * Interface ioctls.
1386  */
1387 int
1388 ifioctl(so, cmd, data, td)
1389 	struct socket *so;
1390 	u_long cmd;
1391 	caddr_t data;
1392 	struct thread *td;
1393 {
1394 	struct ifnet *ifp;
1395 	struct ifreq *ifr;
1396 	int error;
1397 	short oif_flags;
1398 
1399 	switch (cmd) {
1400 	case SIOCGIFCONF:
1401 	case OSIOCGIFCONF:
1402 		return (ifconf(cmd, data));
1403 	}
1404 	ifr = (struct ifreq *)data;
1405 
1406 	switch (cmd) {
1407 	case SIOCIFCREATE:
1408 	case SIOCIFDESTROY:
1409 		if ((error = suser(td)) != 0)
1410 			return (error);
1411 		return ((cmd == SIOCIFCREATE) ?
1412 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1413 			if_clone_destroy(ifr->ifr_name));
1414 
1415 	case SIOCIFGCLONERS:
1416 		return (if_clone_list((struct if_clonereq *)data));
1417 	}
1418 
1419 	ifp = ifunit(ifr->ifr_name);
1420 	if (ifp == 0)
1421 		return (ENXIO);
1422 
1423 	error = ifhwioctl(cmd, ifp, data, td);
1424 	if (error != ENOIOCTL)
1425 		return (error);
1426 
1427 	oif_flags = ifp->if_flags;
1428 	if (so->so_proto == 0)
1429 		return (EOPNOTSUPP);
1430 #ifndef COMPAT_43
1431 	error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1432 								 data,
1433 								 ifp, td));
1434 #else
1435 	{
1436 		int ocmd = cmd;
1437 
1438 		switch (cmd) {
1439 
1440 		case SIOCSIFDSTADDR:
1441 		case SIOCSIFADDR:
1442 		case SIOCSIFBRDADDR:
1443 		case SIOCSIFNETMASK:
1444 #if BYTE_ORDER != BIG_ENDIAN
1445 			if (ifr->ifr_addr.sa_family == 0 &&
1446 			    ifr->ifr_addr.sa_len < 16) {
1447 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1448 				ifr->ifr_addr.sa_len = 16;
1449 			}
1450 #else
1451 			if (ifr->ifr_addr.sa_len == 0)
1452 				ifr->ifr_addr.sa_len = 16;
1453 #endif
1454 			break;
1455 
1456 		case OSIOCGIFADDR:
1457 			cmd = SIOCGIFADDR;
1458 			break;
1459 
1460 		case OSIOCGIFDSTADDR:
1461 			cmd = SIOCGIFDSTADDR;
1462 			break;
1463 
1464 		case OSIOCGIFBRDADDR:
1465 			cmd = SIOCGIFBRDADDR;
1466 			break;
1467 
1468 		case OSIOCGIFNETMASK:
1469 			cmd = SIOCGIFNETMASK;
1470 		}
1471 		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
1472 								   cmd,
1473 								   data,
1474 								   ifp, td));
1475 		switch (ocmd) {
1476 
1477 		case OSIOCGIFADDR:
1478 		case OSIOCGIFDSTADDR:
1479 		case OSIOCGIFBRDADDR:
1480 		case OSIOCGIFNETMASK:
1481 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1482 
1483 		}
1484 	}
1485 #endif /* COMPAT_43 */
1486 
1487 	if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1488 #ifdef INET6
1489 		DELAY(100);/* XXX: temporary workaround for fxp issue*/
1490 		if (ifp->if_flags & IFF_UP) {
1491 			int s = splimp();
1492 			in6_if_up(ifp);
1493 			splx(s);
1494 		}
1495 #endif
1496 	}
1497 	return (error);
1498 }
1499 
1500 /*
1501  * Set/clear promiscuous mode on interface ifp based on the truth value
1502  * of pswitch.  The calls are reference counted so that only the first
1503  * "on" request actually has an effect, as does the final "off" request.
1504  * Results are undefined if the "off" and "on" requests are not matched.
1505  */
1506 int
1507 ifpromisc(ifp, pswitch)
1508 	struct ifnet *ifp;
1509 	int pswitch;
1510 {
1511 	struct ifreq ifr;
1512 	int error;
1513 	int oldflags, oldpcount;
1514 
1515 	oldpcount = ifp->if_pcount;
1516 	oldflags = ifp->if_flags;
1517 	if (pswitch) {
1518 		/*
1519 		 * If the device is not configured up, we cannot put it in
1520 		 * promiscuous mode.
1521 		 */
1522 		if ((ifp->if_flags & IFF_UP) == 0)
1523 			return (ENETDOWN);
1524 		if (ifp->if_pcount++ != 0)
1525 			return (0);
1526 		ifp->if_flags |= IFF_PROMISC;
1527 	} else {
1528 		if (--ifp->if_pcount > 0)
1529 			return (0);
1530 		ifp->if_flags &= ~IFF_PROMISC;
1531 	}
1532 	ifr.ifr_flags = ifp->if_flags;
1533 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1534 	if (error == 0) {
1535 		log(LOG_INFO, "%s%d: promiscuous mode %s\n",
1536 		    ifp->if_name, ifp->if_unit,
1537 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
1538 		rt_ifmsg(ifp);
1539 	} else {
1540 		ifp->if_pcount = oldpcount;
1541 		ifp->if_flags = oldflags;
1542 	}
1543 	return error;
1544 }
1545 
1546 /*
1547  * Return interface configuration
1548  * of system.  List may be used
1549  * in later ioctl's (above) to get
1550  * other information.
1551  */
1552 /*ARGSUSED*/
1553 static int
1554 ifconf(cmd, data)
1555 	u_long cmd;
1556 	caddr_t data;
1557 {
1558 	struct ifconf *ifc = (struct ifconf *)data;
1559 	struct ifnet *ifp;
1560 	struct ifaddr *ifa;
1561 	struct ifreq ifr, *ifrp;
1562 	int space = ifc->ifc_len, error = 0;
1563 
1564 	ifrp = ifc->ifc_req;
1565 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1566 		char workbuf[64];
1567 		int ifnlen, addrs;
1568 
1569 		if (space < sizeof(ifr))
1570 			break;
1571 		ifnlen = snprintf(workbuf, sizeof(workbuf),
1572 		    "%s%d", ifp->if_name, ifp->if_unit);
1573 		if(ifnlen + 1 > sizeof ifr.ifr_name) {
1574 			error = ENAMETOOLONG;
1575 			break;
1576 		} else {
1577 			strcpy(ifr.ifr_name, workbuf);
1578 		}
1579 
1580 		addrs = 0;
1581 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1582 			struct sockaddr *sa = ifa->ifa_addr;
1583 
1584 			if (space < sizeof(ifr))
1585 				break;
1586 			if (jailed(curthread->td_ucred) &&
1587 			    prison_if(curthread->td_ucred, sa))
1588 				continue;
1589 			addrs++;
1590 #ifdef COMPAT_43
1591 			if (cmd == OSIOCGIFCONF) {
1592 				struct osockaddr *osa =
1593 					 (struct osockaddr *)&ifr.ifr_addr;
1594 				ifr.ifr_addr = *sa;
1595 				osa->sa_family = sa->sa_family;
1596 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1597 						sizeof (ifr));
1598 				ifrp++;
1599 			} else
1600 #endif
1601 			if (sa->sa_len <= sizeof(*sa)) {
1602 				ifr.ifr_addr = *sa;
1603 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1604 						sizeof (ifr));
1605 				ifrp++;
1606 			} else {
1607 				if (space < sizeof (ifr) + sa->sa_len -
1608 					    sizeof(*sa))
1609 					break;
1610 				space -= sa->sa_len - sizeof(*sa);
1611 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1612 						sizeof (ifr.ifr_name));
1613 				if (error == 0)
1614 				    error = copyout((caddr_t)sa,
1615 				      (caddr_t)&ifrp->ifr_addr, sa->sa_len);
1616 				ifrp = (struct ifreq *)
1617 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1618 			}
1619 			if (error)
1620 				break;
1621 			space -= sizeof (ifr);
1622 		}
1623 		if (error)
1624 			break;
1625 		if (!addrs) {
1626 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1627 			error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1628 			    sizeof (ifr));
1629 			if (error)
1630 				break;
1631 			space -= sizeof (ifr);
1632 			ifrp++;
1633 		}
1634 	}
1635 	ifc->ifc_len -= space;
1636 	return (error);
1637 }
1638 
1639 /*
1640  * Just like if_promisc(), but for all-multicast-reception mode.
1641  */
1642 int
1643 if_allmulti(ifp, onswitch)
1644 	struct ifnet *ifp;
1645 	int onswitch;
1646 {
1647 	int error = 0;
1648 	int s = splimp();
1649 	struct ifreq ifr;
1650 
1651 	if (onswitch) {
1652 		if (ifp->if_amcount++ == 0) {
1653 			ifp->if_flags |= IFF_ALLMULTI;
1654 			ifr.ifr_flags = ifp->if_flags;
1655 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1656 		}
1657 	} else {
1658 		if (ifp->if_amcount > 1) {
1659 			ifp->if_amcount--;
1660 		} else {
1661 			ifp->if_amcount = 0;
1662 			ifp->if_flags &= ~IFF_ALLMULTI;
1663 			ifr.ifr_flags = ifp->if_flags;
1664 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1665 		}
1666 	}
1667 	splx(s);
1668 
1669 	if (error == 0)
1670 		rt_ifmsg(ifp);
1671 	return error;
1672 }
1673 
1674 /*
1675  * Add a multicast listenership to the interface in question.
1676  * The link layer provides a routine which converts
1677  */
1678 int
1679 if_addmulti(ifp, sa, retifma)
1680 	struct ifnet *ifp;	/* interface to manipulate */
1681 	struct sockaddr *sa;	/* address to add */
1682 	struct ifmultiaddr **retifma;
1683 {
1684 	struct sockaddr *llsa, *dupsa;
1685 	int error, s;
1686 	struct ifmultiaddr *ifma;
1687 
1688 	/*
1689 	 * If the matching multicast address already exists
1690 	 * then don't add a new one, just add a reference
1691 	 */
1692 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1693 		if (equal(sa, ifma->ifma_addr)) {
1694 			ifma->ifma_refcount++;
1695 			if (retifma)
1696 				*retifma = ifma;
1697 			return 0;
1698 		}
1699 	}
1700 
1701 	/*
1702 	 * Give the link layer a chance to accept/reject it, and also
1703 	 * find out which AF_LINK address this maps to, if it isn't one
1704 	 * already.
1705 	 */
1706 	if (ifp->if_resolvemulti) {
1707 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1708 		if (error) return error;
1709 	} else {
1710 		llsa = 0;
1711 	}
1712 
1713 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1714 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1715 	bcopy(sa, dupsa, sa->sa_len);
1716 
1717 	ifma->ifma_addr = dupsa;
1718 	ifma->ifma_lladdr = llsa;
1719 	ifma->ifma_ifp = ifp;
1720 	ifma->ifma_refcount = 1;
1721 	ifma->ifma_protospec = 0;
1722 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1723 
1724 	/*
1725 	 * Some network interfaces can scan the address list at
1726 	 * interrupt time; lock them out.
1727 	 */
1728 	s = splimp();
1729 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1730 	splx(s);
1731 	*retifma = ifma;
1732 
1733 	if (llsa != 0) {
1734 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1735 			if (equal(ifma->ifma_addr, llsa))
1736 				break;
1737 		}
1738 		if (ifma) {
1739 			ifma->ifma_refcount++;
1740 		} else {
1741 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1742 			       M_IFMADDR, M_WAITOK);
1743 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1744 			       M_IFMADDR, M_WAITOK);
1745 			bcopy(llsa, dupsa, llsa->sa_len);
1746 			ifma->ifma_addr = dupsa;
1747 			ifma->ifma_ifp = ifp;
1748 			ifma->ifma_refcount = 1;
1749 			s = splimp();
1750 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1751 			splx(s);
1752 		}
1753 	}
1754 	/*
1755 	 * We are certain we have added something, so call down to the
1756 	 * interface to let them know about it.
1757 	 */
1758 	s = splimp();
1759 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0);
1760 	splx(s);
1761 
1762 	return 0;
1763 }
1764 
1765 /*
1766  * Remove a reference to a multicast address on this interface.  Yell
1767  * if the request does not match an existing membership.
1768  */
1769 int
1770 if_delmulti(ifp, sa)
1771 	struct ifnet *ifp;
1772 	struct sockaddr *sa;
1773 {
1774 	struct ifmultiaddr *ifma;
1775 	int s;
1776 
1777 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1778 		if (equal(sa, ifma->ifma_addr))
1779 			break;
1780 	if (ifma == 0)
1781 		return ENOENT;
1782 
1783 	if (ifma->ifma_refcount > 1) {
1784 		ifma->ifma_refcount--;
1785 		return 0;
1786 	}
1787 
1788 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1789 	sa = ifma->ifma_lladdr;
1790 	s = splimp();
1791 	TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
1792 	/*
1793 	 * Make sure the interface driver is notified
1794 	 * in the case of a link layer mcast group being left.
1795 	 */
1796 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1797 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1798 	splx(s);
1799 	free(ifma->ifma_addr, M_IFMADDR);
1800 	free(ifma, M_IFMADDR);
1801 	if (sa == 0)
1802 		return 0;
1803 
1804 	/*
1805 	 * Now look for the link-layer address which corresponds to
1806 	 * this network address.  It had been squirreled away in
1807 	 * ifma->ifma_lladdr for this purpose (so we don't have
1808 	 * to call ifp->if_resolvemulti() again), and we saved that
1809 	 * value in sa above.  If some nasty deleted the
1810 	 * link-layer address out from underneath us, we can deal because
1811 	 * the address we stored was is not the same as the one which was
1812 	 * in the record for the link-layer address.  (So we don't complain
1813 	 * in that case.)
1814 	 */
1815 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1816 		if (equal(sa, ifma->ifma_addr))
1817 			break;
1818 	if (ifma == 0)
1819 		return 0;
1820 
1821 	if (ifma->ifma_refcount > 1) {
1822 		ifma->ifma_refcount--;
1823 		return 0;
1824 	}
1825 
1826 	s = splimp();
1827 	TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
1828 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1829 	splx(s);
1830 	free(ifma->ifma_addr, M_IFMADDR);
1831 	free(sa, M_IFMADDR);
1832 	free(ifma, M_IFMADDR);
1833 
1834 	return 0;
1835 }
1836 
1837 /*
1838  * Set the link layer address on an interface.
1839  *
1840  * At this time we only support certain types of interfaces,
1841  * and we don't allow the length of the address to change.
1842  */
1843 int
1844 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1845 {
1846 	struct sockaddr_dl *sdl;
1847 	struct ifaddr *ifa;
1848 	struct ifreq ifr;
1849 
1850 	ifa = ifaddr_byindex(ifp->if_index);
1851 	if (ifa == NULL)
1852 		return (EINVAL);
1853 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1854 	if (sdl == NULL)
1855 		return (EINVAL);
1856 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1857 		return (EINVAL);
1858 	switch (ifp->if_type) {
1859 	case IFT_ETHER:			/* these types use struct arpcom */
1860 	case IFT_FDDI:
1861 	case IFT_XETHER:
1862 	case IFT_ISO88025:
1863 	case IFT_L2VLAN:
1864 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1865 		bcopy(lladdr, LLADDR(sdl), len);
1866 		break;
1867 	default:
1868 		return (ENODEV);
1869 	}
1870 	/*
1871 	 * If the interface is already up, we need
1872 	 * to re-init it in order to reprogram its
1873 	 * address filter.
1874 	 */
1875 	if ((ifp->if_flags & IFF_UP) != 0) {
1876 		ifp->if_flags &= ~IFF_UP;
1877 		ifr.ifr_flags = ifp->if_flags;
1878 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1879 		ifp->if_flags |= IFF_UP;
1880 		ifr.ifr_flags = ifp->if_flags;
1881 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1882 #ifdef INET
1883 		/*
1884 		 * Also send gratuitous ARPs to notify other nodes about
1885 		 * the address change.
1886 		 */
1887 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1888 			if (ifa->ifa_addr != NULL &&
1889 			    ifa->ifa_addr->sa_family == AF_INET)
1890 				arp_ifinit(ifp, ifa);
1891 		}
1892 #endif
1893 	}
1894 	return (0);
1895 }
1896 
1897 struct ifmultiaddr *
1898 ifmaof_ifpforaddr(sa, ifp)
1899 	struct sockaddr *sa;
1900 	struct ifnet *ifp;
1901 {
1902 	struct ifmultiaddr *ifma;
1903 
1904 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1905 		if (equal(ifma->ifma_addr, sa))
1906 			break;
1907 
1908 	return ifma;
1909 }
1910 
1911 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1912 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1913