xref: /dragonfly/sys/net/if.c (revision 27f48495)
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	8.3 (Berkeley) 1/4/94
34  * $FreeBSD: src/sys/net/if.c,v 1.185 2004/03/13 02:35:03 brooks Exp $
35  * $DragonFly: src/sys/net/if.c,v 1.28 2005/02/11 22:25:57 joerg Exp $
36  */
37 
38 #include "opt_compat.h"
39 #include "opt_inet6.h"
40 #include "opt_inet.h"
41 
42 #include <sys/param.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/systm.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/socketops.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/domain.h>
57 
58 #include <net/if.h>
59 #include <net/if_arp.h>
60 #include <net/if_dl.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/ifq_var.h>
64 #include <net/radix.h>
65 #include <net/route.h>
66 #include <machine/stdarg.h>
67 
68 #if defined(INET) || defined(INET6)
69 /*XXX*/
70 #include <netinet/in.h>
71 #include <netinet/in_var.h>
72 #include <netinet/if_ether.h>
73 #ifdef INET6
74 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
75 #include <netinet6/in6_var.h>
76 #include <netinet6/in6_ifattach.h>
77 #endif
78 #endif
79 
80 #if defined(COMPAT_43)
81 #include <emulation/43bsd/43bsd_socket.h>
82 #endif /* COMPAT_43 */
83 
84 /*
85  * System initialization
86  */
87 
88 static void	if_attachdomain(void *);
89 static void	if_attachdomain1(struct ifnet *);
90 static int ifconf (u_long, caddr_t, struct thread *);
91 static void ifinit (void *);
92 static void if_slowtimo (void *);
93 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
94 static int  if_rtdel (struct radix_node *, void *);
95 
96 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
97 
98 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
99 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
100 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
101 
102 int	ifqmaxlen = IFQ_MAXLEN;
103 struct	ifnethead ifnet;	/* depend on static init XXX */
104 
105 #ifdef INET6
106 /*
107  * XXX: declare here to avoid to include many inet6 related files..
108  * should be more generalized?
109  */
110 extern void	nd6_setmtu (struct ifnet *);
111 #endif
112 
113 struct if_clone *if_clone_lookup (const char *, int *);
114 int if_clone_list (struct if_clonereq *);
115 
116 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
117 int if_cloners_count;
118 
119 struct callout if_slowtimo_timer;
120 
121 /*
122  * Network interface utility routines.
123  *
124  * Routines with ifa_ifwith* names take sockaddr *'s as
125  * parameters.
126  */
127 /* ARGSUSED*/
128 void
129 ifinit(void *dummy)
130 {
131 	struct ifnet *ifp;
132 	int s;
133 
134 	callout_init(&if_slowtimo_timer);
135 
136 	s = splimp();
137 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
138 		if (ifp->if_snd.ifq_maxlen == 0) {
139 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
140 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
141 		}
142 	}
143 	splx(s);
144 
145 	if_slowtimo(0);
146 }
147 
148 int if_index = 0;
149 struct ifaddr **ifnet_addrs;
150 struct ifnet **ifindex2ifnet = NULL;
151 
152 /*
153  * Attach an interface to the
154  * list of "active" interfaces.
155  */
156 void
157 if_attach(struct ifnet *ifp)
158 {
159 	unsigned socksize, ifasize;
160 	int namelen, masklen;
161 	struct sockaddr_dl *sdl;
162 	struct ifaddr *ifa;
163 
164 	static int if_indexlim = 8;
165 	static boolean_t inited;
166 
167 	if (!inited) {
168 		TAILQ_INIT(&ifnet);
169 		inited = TRUE;
170 	}
171 
172 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
173 	ifp->if_index = ++if_index;
174 	/*
175 	 * XXX -
176 	 * The old code would work if the interface passed a pre-existing
177 	 * chain of ifaddrs to this code.  We don't trust our callers to
178 	 * properly initialize the tailq, however, so we no longer allow
179 	 * this unlikely case.
180 	 */
181 	TAILQ_INIT(&ifp->if_addrhead);
182 	TAILQ_INIT(&ifp->if_prefixhead);
183 	LIST_INIT(&ifp->if_multiaddrs);
184 	getmicrotime(&ifp->if_lastchange);
185 	if (ifnet_addrs == NULL || if_index >= if_indexlim) {
186 		unsigned int n;
187 		caddr_t q;
188 
189 		if_indexlim <<= 1;
190 		n = if_indexlim * sizeof(struct ifaddr *);
191 		q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
192 		if (ifnet_addrs != NULL) {
193 			bcopy(ifnet_addrs, q, n/2);
194 			free(ifnet_addrs, M_IFADDR);
195 		}
196 		ifnet_addrs = (struct ifaddr **)q;
197 
198 		/* grow ifindex2ifnet */
199 		n = if_indexlim * sizeof(struct ifnet *);
200 		q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
201 		if (ifindex2ifnet) {
202 			bcopy(ifindex2ifnet, q, n/2);
203 			free(ifindex2ifnet, M_IFADDR);
204 		}
205 		ifindex2ifnet = (struct ifnet **)q;
206 	}
207 
208 	ifindex2ifnet[if_index] = ifp;
209 
210 	/*
211 	 * create a Link Level name for this device
212 	 */
213 	namelen = strlen(ifp->if_xname);
214 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
215 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
216 	socksize = masklen + ifp->if_addrlen;
217 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
218 	if (socksize < sizeof(*sdl))
219 		socksize = sizeof(*sdl);
220 	socksize = ROUNDUP(socksize);
221 	ifasize = sizeof(struct ifaddr) + 2 * socksize;
222 	ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
223 	sdl = (struct sockaddr_dl *)(ifa + 1);
224 	sdl->sdl_len = socksize;
225 	sdl->sdl_family = AF_LINK;
226 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
227 	sdl->sdl_nlen = namelen;
228 	sdl->sdl_index = ifp->if_index;
229 	sdl->sdl_type = ifp->if_type;
230 	ifnet_addrs[if_index - 1] = ifa;
231 	ifa->ifa_ifp = ifp;
232 	ifa->ifa_rtrequest = link_rtrequest;
233 	ifa->ifa_addr = (struct sockaddr *)sdl;
234 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
235 	ifa->ifa_netmask = (struct sockaddr *)sdl;
236 	sdl->sdl_len = masklen;
237 	while (namelen != 0)
238 		sdl->sdl_data[--namelen] = 0xff;
239 	TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
240 
241 	EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
242 
243 	ifp->if_snd.altq_type = 0;
244 	ifp->if_snd.altq_disc = NULL;
245 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
246 	ifp->if_snd.altq_tbr = NULL;
247 	ifp->if_snd.altq_ifp = ifp;
248 
249 	if (domains)
250 		if_attachdomain1(ifp);
251 
252 	/* Announce the interface. */
253 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
254 }
255 
256 static void
257 if_attachdomain(void *dummy)
258 {
259 	struct ifnet *ifp;
260 	int s;
261 
262 	s = splnet();
263 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
264 		if_attachdomain1(ifp);
265 	splx(s);
266 }
267 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
268 	if_attachdomain, NULL);
269 
270 static void
271 if_attachdomain1(struct ifnet *ifp)
272 {
273 	struct domain *dp;
274 	int s;
275 
276 	s = splnet();
277 
278 	/* address family dependent data region */
279 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
280 	for (dp = domains; dp; dp = dp->dom_next) {
281 		if (dp->dom_ifattach)
282 			ifp->if_afdata[dp->dom_family] =
283 				(*dp->dom_ifattach)(ifp);
284 	}
285 	splx(s);
286 }
287 
288 /*
289  * Detach an interface, removing it from the
290  * list of "active" interfaces.
291  */
292 void
293 if_detach(struct ifnet *ifp)
294 {
295 	struct ifaddr *ifa;
296 	struct radix_node_head	*rnh;
297 	int s;
298 	int i;
299 	struct domain *dp;
300 
301 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
302 
303 	/*
304 	 * Remove routes and flush queues.
305 	 */
306 	s = splnet();
307 	if_down(ifp);
308 
309 	if (ifq_is_enabled(&ifp->if_snd))
310 		altq_disable(&ifp->if_snd);
311 	if (ifq_is_attached(&ifp->if_snd))
312 		altq_detach(&ifp->if_snd);
313 
314 	/*
315 	 * Remove address from ifnet_addrs[] and maybe decrement if_index.
316 	 * Clean up all addresses.
317 	 */
318 	ifnet_addrs[ifp->if_index - 1] = 0;
319 	while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
320 		if_index--;
321 
322 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
323 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
324 #ifdef INET
325 		/* XXX: Ugly!! ad hoc just for INET */
326 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
327 			struct ifaliasreq ifr;
328 
329 			bzero(&ifr, sizeof ifr);
330 			ifr.ifra_addr = *ifa->ifa_addr;
331 			if (ifa->ifa_dstaddr)
332 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
333 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
334 				       NULL) == 0)
335 				continue;
336 		}
337 #endif /* INET */
338 #ifdef INET6
339 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
340 			in6_purgeaddr(ifa);
341 			/* ifp_addrhead is already updated */
342 			continue;
343 		}
344 #endif /* INET6 */
345 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
346 		IFAFREE(ifa);
347 	}
348 
349 #ifdef INET6
350 	/*
351 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
352 	 * before removing routing entries below, since IPv6 interface direct
353 	 * routes are expected to be removed by the IPv6-specific kernel API.
354 	 * Otherwise, the kernel will detect some inconsistency and bark it.
355 	 */
356 	in6_ifdetach(ifp);
357 #endif
358 
359 	/*
360 	 * Delete all remaining routes using this interface
361 	 * Unfortuneatly the only way to do this is to slog through
362 	 * the entire routing table looking for routes which point
363 	 * to this interface...oh well...
364 	 */
365 	for (i = 1; i <= AF_MAX; i++) {
366 		if ((rnh = rt_tables[i]) == NULL)
367 			continue;
368 		rnh->rnh_walktree(rnh, if_rtdel, ifp);
369 	}
370 
371 	/* Announce that the interface is gone. */
372 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
373 
374 	for (dp = domains; dp; dp = dp->dom_next) {
375 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
376 			(*dp->dom_ifdetach)(ifp,
377 				ifp->if_afdata[dp->dom_family]);
378 	}
379 
380 	ifindex2ifnet[ifp->if_index] = NULL;
381 
382 	TAILQ_REMOVE(&ifnet, ifp, if_link);
383 	splx(s);
384 }
385 
386 /*
387  * Delete Routes for a Network Interface
388  *
389  * Called for each routing entry via the rnh->rnh_walktree() call above
390  * to delete all route entries referencing a detaching network interface.
391  *
392  * Arguments:
393  *	rn	pointer to node in the routing table
394  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
395  *
396  * Returns:
397  *	0	successful
398  *	errno	failed - reason indicated
399  *
400  */
401 static int
402 if_rtdel(struct radix_node *rn, void *arg)
403 {
404 	struct rtentry	*rt = (struct rtentry *)rn;
405 	struct ifnet	*ifp = arg;
406 	int		err;
407 
408 	if (rt->rt_ifp == ifp) {
409 
410 		/*
411 		 * Protect (sorta) against walktree recursion problems
412 		 * with cloned routes
413 		 */
414 		if (!(rt->rt_flags & RTF_UP))
415 			return (0);
416 
417 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
418 				rt_mask(rt), rt->rt_flags,
419 				(struct rtentry **) NULL);
420 		if (err) {
421 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
422 		}
423 	}
424 
425 	return (0);
426 }
427 
428 /*
429  * Create a clone network interface.
430  */
431 int
432 if_clone_create(char *name, int len)
433 {
434 	struct if_clone *ifc;
435 	char *dp;
436 	int wildcard, bytoff, bitoff;
437 	int unit;
438 	int err;
439 
440 	ifc = if_clone_lookup(name, &unit);
441 	if (ifc == NULL)
442 		return (EINVAL);
443 
444 	if (ifunit(name) != NULL)
445 		return (EEXIST);
446 
447 	bytoff = bitoff = 0;
448 	wildcard = (unit < 0);
449 	/*
450 	 * Find a free unit if none was given.
451 	 */
452 	if (wildcard) {
453 		while (bytoff < ifc->ifc_bmlen &&
454 		    ifc->ifc_units[bytoff] == 0xff)
455 			bytoff++;
456 		if (bytoff >= ifc->ifc_bmlen)
457 			return (ENOSPC);
458 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
459 			bitoff++;
460 		unit = (bytoff << 3) + bitoff;
461 	}
462 
463 	if (unit > ifc->ifc_maxunit)
464 		return (ENXIO);
465 
466 	err = (*ifc->ifc_create)(ifc, unit);
467 	if (err != 0)
468 		return (err);
469 
470 	if (!wildcard) {
471 		bytoff = unit >> 3;
472 		bitoff = unit - (bytoff << 3);
473 	}
474 
475 	/*
476 	 * Allocate the unit in the bitmap.
477 	 */
478 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
479 	    ("%s: bit is already set", __func__));
480 	ifc->ifc_units[bytoff] |= (1 << bitoff);
481 
482 	/* In the wildcard case, we need to update the name. */
483 	if (wildcard) {
484 		for (dp = name; *dp != '\0'; dp++);
485 		if (snprintf(dp, len - (dp-name), "%d", unit) >
486 		    len - (dp-name) - 1) {
487 			/*
488 			 * This can only be a programmer error and
489 			 * there's no straightforward way to recover if
490 			 * it happens.
491 			 */
492 			panic("if_clone_create(): interface name too long");
493 		}
494 
495 	}
496 
497 	EVENTHANDLER_INVOKE(if_clone_event, ifc);
498 
499 	return (0);
500 }
501 
502 /*
503  * Destroy a clone network interface.
504  */
505 int
506 if_clone_destroy(const char *name)
507 {
508 	struct if_clone *ifc;
509 	struct ifnet *ifp;
510 	int bytoff, bitoff;
511 	int unit;
512 
513 	ifc = if_clone_lookup(name, &unit);
514 	if (ifc == NULL)
515 		return (EINVAL);
516 
517 	if (unit < ifc->ifc_minifs)
518 		return (EINVAL);
519 
520 	ifp = ifunit(name);
521 	if (ifp == NULL)
522 		return (ENXIO);
523 
524 	if (ifc->ifc_destroy == NULL)
525 		return (EOPNOTSUPP);
526 
527 	(*ifc->ifc_destroy)(ifp);
528 
529 	/*
530 	 * Compute offset in the bitmap and deallocate the unit.
531 	 */
532 	bytoff = unit >> 3;
533 	bitoff = unit - (bytoff << 3);
534 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
535 	    ("%s: bit is already cleared", __func__));
536 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
537 	return (0);
538 }
539 
540 /*
541  * Look up a network interface cloner.
542  */
543 struct if_clone *
544 if_clone_lookup(const char *name, int *unitp)
545 {
546 	struct if_clone *ifc;
547 	const char *cp;
548 	int i;
549 
550 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
551 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
552 			if (ifc->ifc_name[i] != *cp)
553 				goto next_ifc;
554 		}
555 		goto found_name;
556  next_ifc:
557 		ifc = LIST_NEXT(ifc, ifc_list);
558 	}
559 
560 	/* No match. */
561 	return ((struct if_clone *)NULL);
562 
563  found_name:
564 	if (*cp == '\0') {
565 		i = -1;
566 	} else {
567 		for (i = 0; *cp != '\0'; cp++) {
568 			if (*cp < '0' || *cp > '9') {
569 				/* Bogus unit number. */
570 				return (NULL);
571 			}
572 			i = (i * 10) + (*cp - '0');
573 		}
574 	}
575 
576 	if (unitp != NULL)
577 		*unitp = i;
578 	return (ifc);
579 }
580 
581 /*
582  * Register a network interface cloner.
583  */
584 void
585 if_clone_attach(struct if_clone *ifc)
586 {
587 	int bytoff, bitoff;
588 	int err;
589 	int len, maxclone;
590 	int unit;
591 
592 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
593 	    ("%s: %s requested more units then allowed (%d > %d)",
594 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
595 	    ifc->ifc_maxunit + 1));
596 	/*
597 	 * Compute bitmap size and allocate it.
598 	 */
599 	maxclone = ifc->ifc_maxunit + 1;
600 	len = maxclone >> 3;
601 	if ((len << 3) < maxclone)
602 		len++;
603 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
604 	ifc->ifc_bmlen = len;
605 
606 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
607 	if_cloners_count++;
608 
609 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
610 		err = (*ifc->ifc_create)(ifc, unit);
611 		KASSERT(err == 0,
612 		    ("%s: failed to create required interface %s%d",
613 		    __func__, ifc->ifc_name, unit));
614 
615 		/* Allocate the unit in the bitmap. */
616 		bytoff = unit >> 3;
617 		bitoff = unit - (bytoff << 3);
618 		ifc->ifc_units[bytoff] |= (1 << bitoff);
619 	}
620 }
621 
622 /*
623  * Unregister a network interface cloner.
624  */
625 void
626 if_clone_detach(struct if_clone *ifc)
627 {
628 
629 	LIST_REMOVE(ifc, ifc_list);
630 	free(ifc->ifc_units, M_CLONE);
631 	if_cloners_count--;
632 }
633 
634 /*
635  * Provide list of interface cloners to userspace.
636  */
637 int
638 if_clone_list(struct if_clonereq *ifcr)
639 {
640 	char outbuf[IFNAMSIZ], *dst;
641 	struct if_clone *ifc;
642 	int count, error = 0;
643 
644 	ifcr->ifcr_total = if_cloners_count;
645 	if ((dst = ifcr->ifcr_buffer) == NULL) {
646 		/* Just asking how many there are. */
647 		return (0);
648 	}
649 
650 	if (ifcr->ifcr_count < 0)
651 		return (EINVAL);
652 
653 	count = (if_cloners_count < ifcr->ifcr_count) ?
654 	    if_cloners_count : ifcr->ifcr_count;
655 
656 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
657 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
658 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
659 		error = copyout(outbuf, dst, IFNAMSIZ);
660 		if (error)
661 			break;
662 	}
663 
664 	return (error);
665 }
666 
667 /*
668  * Locate an interface based on a complete address.
669  */
670 struct ifaddr *
671 ifa_ifwithaddr(struct sockaddr *addr)
672 {
673 	struct ifnet *ifp;
674 	struct ifaddr *ifa;
675 
676 	TAILQ_FOREACH(ifp, &ifnet, if_link)
677 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
678 		if (ifa->ifa_addr->sa_family != addr->sa_family)
679 			continue;
680 		if (sa_equal(addr, ifa->ifa_addr))
681 			return (ifa);
682 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
683 		    /* IPv6 doesn't have broadcast */
684 		    ifa->ifa_broadaddr->sa_len != 0 &&
685 		    sa_equal(ifa->ifa_broadaddr, addr))
686 			return (ifa);
687 	}
688 	return ((struct ifaddr *)NULL);
689 }
690 /*
691  * Locate the point to point interface with a given destination address.
692  */
693 struct ifaddr *
694 ifa_ifwithdstaddr(struct sockaddr *addr)
695 {
696 	struct ifnet *ifp;
697 	struct ifaddr *ifa;
698 
699 	TAILQ_FOREACH(ifp, &ifnet, if_link)
700 	    if (ifp->if_flags & IFF_POINTOPOINT)
701 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
702 			if (ifa->ifa_addr->sa_family != addr->sa_family)
703 				continue;
704 			if (ifa->ifa_dstaddr &&
705 			    sa_equal(addr, ifa->ifa_dstaddr))
706 				return (ifa);
707 	}
708 	return ((struct ifaddr *)NULL);
709 }
710 
711 /*
712  * Find an interface on a specific network.  If many, choice
713  * is most specific found.
714  */
715 struct ifaddr *
716 ifa_ifwithnet(struct sockaddr *addr)
717 {
718 	struct ifnet *ifp;
719 	struct ifaddr *ifa;
720 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
721 	u_int af = addr->sa_family;
722 	char *addr_data = addr->sa_data, *cplim;
723 
724 	/*
725 	 * AF_LINK addresses can be looked up directly by their index number,
726 	 * so do that if we can.
727 	 */
728 	if (af == AF_LINK) {
729 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
730 
731 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
732 		return (ifnet_addrs[sdl->sdl_index - 1]);
733 	}
734 
735 	/*
736 	 * Scan though each interface, looking for ones that have
737 	 * addresses in this address family.
738 	 */
739 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
740 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
741 			char *cp, *cp2, *cp3;
742 
743 			if (ifa->ifa_addr->sa_family != af)
744 next:				continue;
745 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
746 				/*
747 				 * This is a bit broken as it doesn't
748 				 * take into account that the remote end may
749 				 * be a single node in the network we are
750 				 * looking for.
751 				 * The trouble is that we don't know the
752 				 * netmask for the remote end.
753 				 */
754 				if (ifa->ifa_dstaddr != NULL &&
755 				    sa_equal(addr, ifa->ifa_dstaddr))
756 					return (ifa);
757 			} else {
758 				/*
759 				 * if we have a special address handler,
760 				 * then use it instead of the generic one.
761 				 */
762 				if (ifa->ifa_claim_addr) {
763 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
764 						return (ifa);
765 					} else {
766 						continue;
767 					}
768 				}
769 
770 				/*
771 				 * Scan all the bits in the ifa's address.
772 				 * If a bit dissagrees with what we are
773 				 * looking for, mask it with the netmask
774 				 * to see if it really matters.
775 				 * (A byte at a time)
776 				 */
777 				if (ifa->ifa_netmask == 0)
778 					continue;
779 				cp = addr_data;
780 				cp2 = ifa->ifa_addr->sa_data;
781 				cp3 = ifa->ifa_netmask->sa_data;
782 				cplim = ifa->ifa_netmask->sa_len +
783 					(char *)ifa->ifa_netmask;
784 				while (cp3 < cplim)
785 					if ((*cp++ ^ *cp2++) & *cp3++)
786 						goto next; /* next address! */
787 				/*
788 				 * If the netmask of what we just found
789 				 * is more specific than what we had before
790 				 * (if we had one) then remember the new one
791 				 * before continuing to search
792 				 * for an even better one.
793 				 */
794 				if (ifa_maybe == 0 ||
795 				    rn_refines((char *)ifa->ifa_netmask,
796 					       (char *)ifa_maybe->ifa_netmask))
797 					ifa_maybe = ifa;
798 			}
799 		}
800 	}
801 	return (ifa_maybe);
802 }
803 
804 /*
805  * Find an interface address specific to an interface best matching
806  * a given address.
807  */
808 struct ifaddr *
809 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
810 {
811 	struct ifaddr *ifa;
812 	char *cp, *cp2, *cp3;
813 	char *cplim;
814 	struct ifaddr *ifa_maybe = 0;
815 	u_int af = addr->sa_family;
816 
817 	if (af >= AF_MAX)
818 		return (0);
819 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
820 		if (ifa->ifa_addr->sa_family != af)
821 			continue;
822 		if (ifa_maybe == 0)
823 			ifa_maybe = ifa;
824 		if (ifa->ifa_netmask == NULL) {
825 			if (sa_equal(addr, ifa->ifa_addr) ||
826 			    (ifa->ifa_dstaddr != NULL &&
827 			     sa_equal(addr, ifa->ifa_dstaddr)))
828 				return (ifa);
829 			continue;
830 		}
831 		if (ifp->if_flags & IFF_POINTOPOINT) {
832 			if (sa_equal(addr, ifa->ifa_dstaddr))
833 				return (ifa);
834 		} else {
835 			cp = addr->sa_data;
836 			cp2 = ifa->ifa_addr->sa_data;
837 			cp3 = ifa->ifa_netmask->sa_data;
838 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
839 			for (; cp3 < cplim; cp3++)
840 				if ((*cp++ ^ *cp2++) & *cp3)
841 					break;
842 			if (cp3 == cplim)
843 				return (ifa);
844 		}
845 	}
846 	return (ifa_maybe);
847 }
848 
849 #include <net/route.h>
850 
851 /*
852  * Default action when installing a route with a Link Level gateway.
853  * Lookup an appropriate real ifa to point to.
854  * This should be moved to /sys/net/link.c eventually.
855  */
856 static void
857 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
858 {
859 	struct ifaddr *ifa;
860 	struct sockaddr *dst;
861 	struct ifnet *ifp;
862 
863 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
864 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
865 		return;
866 	ifa = ifaof_ifpforaddr(dst, ifp);
867 	if (ifa != NULL) {
868 		IFAFREE(rt->rt_ifa);
869 		IFAREF(ifa);
870 		rt->rt_ifa = ifa;
871 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
872 			ifa->ifa_rtrequest(cmd, rt, info);
873 	}
874 }
875 
876 /*
877  * Mark an interface down and notify protocols of
878  * the transition.
879  * NOTE: must be called at splnet or eqivalent.
880  */
881 void
882 if_unroute(struct ifnet *ifp, int flag, int fam)
883 {
884 	struct ifaddr *ifa;
885 
886 	ifp->if_flags &= ~flag;
887 	getmicrotime(&ifp->if_lastchange);
888 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
889 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
890 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
891 	ifq_purge(&ifp->if_snd);
892 	rt_ifmsg(ifp);
893 }
894 
895 /*
896  * Mark an interface up and notify protocols of
897  * the transition.
898  * NOTE: must be called at splnet or eqivalent.
899  */
900 void
901 if_route(struct ifnet *ifp, int flag, int fam)
902 {
903 	struct ifaddr *ifa;
904 
905 	ifp->if_flags |= flag;
906 	getmicrotime(&ifp->if_lastchange);
907 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
908 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
909 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
910 	rt_ifmsg(ifp);
911 #ifdef INET6
912 	in6_if_up(ifp);
913 #endif
914 }
915 
916 /*
917  * Mark an interface down and notify protocols of the transition.  An
918  * interface going down is also considered to be a synchronizing event.
919  * We must ensure that all packet processing related to the interface
920  * has completed before we return so e.g. the caller can free the ifnet
921  * structure that the mbufs may be referencing.
922  *
923  * NOTE: must be called at splnet or eqivalent.
924  */
925 void
926 if_down(struct ifnet *ifp)
927 {
928 
929 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
930 	netmsg_service_sync();
931 }
932 
933 /*
934  * Mark an interface up and notify protocols of
935  * the transition.
936  * NOTE: must be called at splnet or eqivalent.
937  */
938 void
939 if_up(struct ifnet *ifp)
940 {
941 
942 	if_route(ifp, IFF_UP, AF_UNSPEC);
943 }
944 
945 /*
946  * Handle interface watchdog timer routines.  Called
947  * from softclock, we decrement timers (if set) and
948  * call the appropriate interface routine on expiration.
949  */
950 static void
951 if_slowtimo(void *arg)
952 {
953 	struct ifnet *ifp;
954 	int s = splimp();
955 
956 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
957 		if (ifp->if_timer == 0 || --ifp->if_timer)
958 			continue;
959 		if (ifp->if_watchdog)
960 			(*ifp->if_watchdog)(ifp);
961 	}
962 	splx(s);
963 	callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
964 }
965 
966 /*
967  * Map interface name to
968  * interface structure pointer.
969  */
970 struct ifnet *
971 ifunit(const char *name)
972 {
973 	struct ifnet *ifp;
974 
975 	/*
976 	 * Search all the interfaces for this name/number
977 	 */
978 
979 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
980 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
981 			break;
982 	}
983 	return (ifp);
984 }
985 
986 
987 /*
988  * Map interface name in a sockaddr_dl to
989  * interface structure pointer.
990  */
991 struct ifnet *
992 if_withname(struct sockaddr *sa)
993 {
994 	char ifname[IFNAMSIZ+1];
995 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
996 
997 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
998 	     (sdl->sdl_nlen > IFNAMSIZ) )
999 		return NULL;
1000 
1001 	/*
1002 	 * ifunit wants a null-terminated name.  It may not be null-terminated
1003 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
1004 	 * and there might not be room to put the trailing null anyway, so we
1005 	 * make a local copy that we know we can null terminate safely.
1006 	 */
1007 
1008 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1009 	ifname[sdl->sdl_nlen] = '\0';
1010 	return ifunit(ifname);
1011 }
1012 
1013 
1014 /*
1015  * Interface ioctls.
1016  */
1017 int
1018 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1019 {
1020 	struct ifnet *ifp;
1021 	struct ifreq *ifr;
1022 	struct ifstat *ifs;
1023 	int error;
1024 	short oif_flags;
1025 	int new_flags;
1026 	size_t namelen, onamelen;
1027 	char new_name[IFNAMSIZ];
1028 	struct ifaddr *ifa;
1029 	struct sockaddr_dl *sdl;
1030 
1031 	switch (cmd) {
1032 
1033 	case SIOCGIFCONF:
1034 	case OSIOCGIFCONF:
1035 		return (ifconf(cmd, data, td));
1036 	}
1037 	ifr = (struct ifreq *)data;
1038 
1039 	switch (cmd) {
1040 	case SIOCIFCREATE:
1041 	case SIOCIFDESTROY:
1042 		if ((error = suser(td)) != 0)
1043 			return (error);
1044 		return ((cmd == SIOCIFCREATE) ?
1045 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1046 			if_clone_destroy(ifr->ifr_name));
1047 
1048 	case SIOCIFGCLONERS:
1049 		return (if_clone_list((struct if_clonereq *)data));
1050 	}
1051 
1052 	ifp = ifunit(ifr->ifr_name);
1053 	if (ifp == 0)
1054 		return (ENXIO);
1055 	switch (cmd) {
1056 
1057 	case SIOCGIFFLAGS:
1058 		ifr->ifr_flags = ifp->if_flags;
1059 		ifr->ifr_flagshigh = ifp->if_flags >> 16;
1060 		break;
1061 
1062 	case SIOCGIFCAP:
1063 		ifr->ifr_reqcap = ifp->if_capabilities;
1064 		ifr->ifr_curcap = ifp->if_capenable;
1065 		break;
1066 
1067 	case SIOCGIFMETRIC:
1068 		ifr->ifr_metric = ifp->if_metric;
1069 		break;
1070 
1071 	case SIOCGIFMTU:
1072 		ifr->ifr_mtu = ifp->if_mtu;
1073 		break;
1074 
1075 	case SIOCGIFPHYS:
1076 		ifr->ifr_phys = ifp->if_physical;
1077 		break;
1078 
1079 	case SIOCSIFFLAGS:
1080 		error = suser(td);
1081 		if (error)
1082 			return (error);
1083 		new_flags = (ifr->ifr_flags & 0xffff) |
1084 		    (ifr->ifr_flagshigh << 16);
1085 		if (ifp->if_flags & IFF_SMART) {
1086 			/* Smart drivers twiddle their own routes */
1087 		} else if (ifp->if_flags & IFF_UP &&
1088 		    (new_flags & IFF_UP) == 0) {
1089 			int s = splimp();
1090 			if_down(ifp);
1091 			splx(s);
1092 		} else if (new_flags & IFF_UP &&
1093 		    (ifp->if_flags & IFF_UP) == 0) {
1094 			int s = splimp();
1095 			if_up(ifp);
1096 			splx(s);
1097 		}
1098 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1099 			(new_flags &~ IFF_CANTCHANGE);
1100 		if (new_flags & IFF_PPROMISC) {
1101 			/* Permanently promiscuous mode requested */
1102 			ifp->if_flags |= IFF_PROMISC;
1103 		} else if (ifp->if_pcount == 0) {
1104 			ifp->if_flags &= ~IFF_PROMISC;
1105 		}
1106 		if (ifp->if_ioctl)
1107 			(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1108 		getmicrotime(&ifp->if_lastchange);
1109 		break;
1110 
1111 	case SIOCSIFCAP:
1112 		error = suser(td);
1113 		if (error)
1114 			return (error);
1115 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1116 			return (EINVAL);
1117 		(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1118 		break;
1119 
1120 	case SIOCSIFNAME:
1121 		error = suser(td);
1122 		if (error != 0)
1123 			return (error);
1124 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1125 		if (error != 0)
1126 			return (error);
1127 		if (new_name[0] == '\0')
1128 			return (EINVAL);
1129 		if (ifunit(new_name) != NULL)
1130 			return (EEXIST);
1131 
1132 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1133 
1134 		/* Announce the departure of the interface. */
1135 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1136 
1137 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1138 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1139 		/* XXX IFA_LOCK(ifa); */
1140 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1141 		namelen = strlen(new_name);
1142 		onamelen = sdl->sdl_nlen;
1143 		/*
1144 		 * Move the address if needed.  This is safe because we
1145 		 * allocate space for a name of length IFNAMSIZ when we
1146 		 * create this in if_attach().
1147 		 */
1148 		if (namelen != onamelen) {
1149 			bcopy(sdl->sdl_data + onamelen,
1150 			    sdl->sdl_data + namelen, sdl->sdl_alen);
1151 		}
1152 		bcopy(new_name, sdl->sdl_data, namelen);
1153 		sdl->sdl_nlen = namelen;
1154 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1155 		bzero(sdl->sdl_data, onamelen);
1156 		while (namelen != 0)
1157 			sdl->sdl_data[--namelen] = 0xff;
1158 		/* XXX IFA_UNLOCK(ifa) */
1159 
1160 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1161 
1162 		/* Announce the return of the interface. */
1163 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1164 		break;
1165 
1166 	case SIOCSIFMETRIC:
1167 		error = suser(td);
1168 		if (error)
1169 			return (error);
1170 		ifp->if_metric = ifr->ifr_metric;
1171 		getmicrotime(&ifp->if_lastchange);
1172 		break;
1173 
1174 	case SIOCSIFPHYS:
1175 		error = suser(td);
1176 		if (error)
1177 			return error;
1178 		if (!ifp->if_ioctl)
1179 		        return EOPNOTSUPP;
1180 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1181 		if (error == 0)
1182 			getmicrotime(&ifp->if_lastchange);
1183 		return (error);
1184 
1185 	case SIOCSIFMTU:
1186 	{
1187 		u_long oldmtu = ifp->if_mtu;
1188 
1189 		error = suser(td);
1190 		if (error)
1191 			return (error);
1192 		if (ifp->if_ioctl == NULL)
1193 			return (EOPNOTSUPP);
1194 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1195 			return (EINVAL);
1196 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1197 		if (error == 0) {
1198 			getmicrotime(&ifp->if_lastchange);
1199 			rt_ifmsg(ifp);
1200 		}
1201 		/*
1202 		 * If the link MTU changed, do network layer specific procedure.
1203 		 */
1204 		if (ifp->if_mtu != oldmtu) {
1205 #ifdef INET6
1206 			nd6_setmtu(ifp);
1207 #endif
1208 		}
1209 		return (error);
1210 	}
1211 
1212 	case SIOCADDMULTI:
1213 	case SIOCDELMULTI:
1214 		error = suser(td);
1215 		if (error)
1216 			return (error);
1217 
1218 		/* Don't allow group membership on non-multicast interfaces. */
1219 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1220 			return EOPNOTSUPP;
1221 
1222 		/* Don't let users screw up protocols' entries. */
1223 		if (ifr->ifr_addr.sa_family != AF_LINK)
1224 			return EINVAL;
1225 
1226 		if (cmd == SIOCADDMULTI) {
1227 			struct ifmultiaddr *ifma;
1228 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1229 		} else {
1230 			error = if_delmulti(ifp, &ifr->ifr_addr);
1231 		}
1232 		if (error == 0)
1233 			getmicrotime(&ifp->if_lastchange);
1234 		return error;
1235 
1236 	case SIOCSIFPHYADDR:
1237 	case SIOCDIFPHYADDR:
1238 #ifdef INET6
1239 	case SIOCSIFPHYADDR_IN6:
1240 #endif
1241 	case SIOCSLIFPHYADDR:
1242         case SIOCSIFMEDIA:
1243 	case SIOCSIFGENERIC:
1244 		error = suser(td);
1245 		if (error)
1246 			return (error);
1247 		if (ifp->if_ioctl == 0)
1248 			return (EOPNOTSUPP);
1249 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1250 		if (error == 0)
1251 			getmicrotime(&ifp->if_lastchange);
1252 		return error;
1253 
1254 	case SIOCGIFSTATUS:
1255 		ifs = (struct ifstat *)data;
1256 		ifs->ascii[0] = '\0';
1257 
1258 	case SIOCGIFPSRCADDR:
1259 	case SIOCGIFPDSTADDR:
1260 	case SIOCGLIFPHYADDR:
1261 	case SIOCGIFMEDIA:
1262 	case SIOCGIFGENERIC:
1263 		if (ifp->if_ioctl == 0)
1264 			return (EOPNOTSUPP);
1265 		return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1266 
1267 	case SIOCSIFLLADDR:
1268 		error = suser(td);
1269 		if (error)
1270 			return (error);
1271 		return if_setlladdr(ifp,
1272 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1273 
1274 	default:
1275 		oif_flags = ifp->if_flags;
1276 		if (so->so_proto == 0)
1277 			return (EOPNOTSUPP);
1278 #ifndef COMPAT_43
1279 		error = so_pru_control(so, cmd, data, ifp, td);
1280 #else
1281 	    {
1282 		int ocmd = cmd;
1283 
1284 		switch (cmd) {
1285 
1286 		case SIOCSIFDSTADDR:
1287 		case SIOCSIFADDR:
1288 		case SIOCSIFBRDADDR:
1289 		case SIOCSIFNETMASK:
1290 #if BYTE_ORDER != BIG_ENDIAN
1291 			if (ifr->ifr_addr.sa_family == 0 &&
1292 			    ifr->ifr_addr.sa_len < 16) {
1293 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1294 				ifr->ifr_addr.sa_len = 16;
1295 			}
1296 #else
1297 			if (ifr->ifr_addr.sa_len == 0)
1298 				ifr->ifr_addr.sa_len = 16;
1299 #endif
1300 			break;
1301 
1302 		case OSIOCGIFADDR:
1303 			cmd = SIOCGIFADDR;
1304 			break;
1305 
1306 		case OSIOCGIFDSTADDR:
1307 			cmd = SIOCGIFDSTADDR;
1308 			break;
1309 
1310 		case OSIOCGIFBRDADDR:
1311 			cmd = SIOCGIFBRDADDR;
1312 			break;
1313 
1314 		case OSIOCGIFNETMASK:
1315 			cmd = SIOCGIFNETMASK;
1316 		}
1317 		error =  so_pru_control(so, cmd, data, ifp, td);
1318 		switch (ocmd) {
1319 
1320 		case OSIOCGIFADDR:
1321 		case OSIOCGIFDSTADDR:
1322 		case OSIOCGIFBRDADDR:
1323 		case OSIOCGIFNETMASK:
1324 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1325 
1326 		}
1327 	    }
1328 #endif /* COMPAT_43 */
1329 
1330 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1331 #ifdef INET6
1332 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1333 			if (ifp->if_flags & IFF_UP) {
1334 				int s = splimp();
1335 				in6_if_up(ifp);
1336 				splx(s);
1337 			}
1338 #endif
1339 		}
1340 		return (error);
1341 
1342 	}
1343 	return (0);
1344 }
1345 
1346 /*
1347  * Set/clear promiscuous mode on interface ifp based on the truth value
1348  * of pswitch.  The calls are reference counted so that only the first
1349  * "on" request actually has an effect, as does the final "off" request.
1350  * Results are undefined if the "off" and "on" requests are not matched.
1351  */
1352 int
1353 ifpromisc(struct ifnet *ifp, int pswitch)
1354 {
1355 	struct ifreq ifr;
1356 	int error;
1357 	int oldflags;
1358 
1359 	oldflags = ifp->if_flags;
1360 	if (ifp->if_flags & IFF_PPROMISC) {
1361 		/* Do nothing if device is in permanently promiscuous mode */
1362 		ifp->if_pcount += pswitch ? 1 : -1;
1363 		return (0);
1364 	}
1365 	if (pswitch) {
1366 		/*
1367 		 * If the device is not configured up, we cannot put it in
1368 		 * promiscuous mode.
1369 		 */
1370 		if ((ifp->if_flags & IFF_UP) == 0)
1371 			return (ENETDOWN);
1372 		if (ifp->if_pcount++ != 0)
1373 			return (0);
1374 		ifp->if_flags |= IFF_PROMISC;
1375 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1376 		    ifp->if_xname);
1377 	} else {
1378 		if (--ifp->if_pcount > 0)
1379 			return (0);
1380 		ifp->if_flags &= ~IFF_PROMISC;
1381 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1382 		    ifp->if_xname);
1383 	}
1384 	ifr.ifr_flags = ifp->if_flags;
1385 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
1386 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1387 				 (struct ucred *)NULL);
1388 	if (error == 0)
1389 		rt_ifmsg(ifp);
1390 	else
1391 		ifp->if_flags = oldflags;
1392 	return error;
1393 }
1394 
1395 /*
1396  * Return interface configuration
1397  * of system.  List may be used
1398  * in later ioctl's (above) to get
1399  * other information.
1400  */
1401 static int
1402 ifconf(u_long cmd, caddr_t data, struct thread *td)
1403 {
1404 	struct ifconf *ifc = (struct ifconf *)data;
1405 	struct ifnet *ifp;
1406 	struct ifaddr *ifa;
1407 	struct sockaddr *sa;
1408 	struct ifreq ifr, *ifrp;
1409 	int space = ifc->ifc_len, error = 0;
1410 
1411 	ifrp = ifc->ifc_req;
1412 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1413 		int addrs;
1414 
1415 		if (space <= sizeof ifr)
1416 			break;
1417 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1418 		    >= sizeof(ifr.ifr_name)) {
1419 			error = ENAMETOOLONG;
1420 			break;
1421 		}
1422 
1423 		addrs = 0;
1424 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1425 			if (space <= sizeof ifr)
1426 				break;
1427 			sa = ifa->ifa_addr;
1428 			if (td->td_proc->p_ucred->cr_prison &&
1429 			    prison_if(td, sa))
1430 				continue;
1431 			addrs++;
1432 #ifdef COMPAT_43
1433 			if (cmd == OSIOCGIFCONF) {
1434 				struct osockaddr *osa =
1435 					 (struct osockaddr *)&ifr.ifr_addr;
1436 				ifr.ifr_addr = *sa;
1437 				osa->sa_family = sa->sa_family;
1438 				error = copyout(&ifr, ifrp, sizeof ifr);
1439 				ifrp++;
1440 			} else
1441 #endif
1442 			if (sa->sa_len <= sizeof(*sa)) {
1443 				ifr.ifr_addr = *sa;
1444 				error = copyout(&ifr, ifrp, sizeof ifr);
1445 				ifrp++;
1446 			} else {
1447 				if (space < (sizeof ifr) + sa->sa_len -
1448 					    sizeof(*sa))
1449 					break;
1450 				space -= sa->sa_len - sizeof(*sa);
1451 				error = copyout(&ifr, ifrp,
1452 						sizeof ifr.ifr_name);
1453 				if (error == 0)
1454 					error = copyout(sa, &ifrp->ifr_addr,
1455 							sa->sa_len);
1456 				ifrp = (struct ifreq *)
1457 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1458 			}
1459 			if (error)
1460 				break;
1461 			space -= sizeof ifr;
1462 		}
1463 		if (error)
1464 			break;
1465 		if (!addrs) {
1466 			bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1467 			error = copyout(&ifr, ifrp, sizeof ifr);
1468 			if (error)
1469 				break;
1470 			space -= sizeof ifr;
1471 			ifrp++;
1472 		}
1473 	}
1474 	ifc->ifc_len -= space;
1475 	return (error);
1476 }
1477 
1478 /*
1479  * Just like if_promisc(), but for all-multicast-reception mode.
1480  */
1481 int
1482 if_allmulti(struct ifnet *ifp, int onswitch)
1483 {
1484 	int error = 0;
1485 	int s = splimp();
1486 	struct ifreq ifr;
1487 
1488 	if (onswitch) {
1489 		if (ifp->if_amcount++ == 0) {
1490 			ifp->if_flags |= IFF_ALLMULTI;
1491 			ifr.ifr_flags = ifp->if_flags;
1492 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1493 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1494 					      (struct ucred *)NULL);
1495 		}
1496 	} else {
1497 		if (ifp->if_amcount > 1) {
1498 			ifp->if_amcount--;
1499 		} else {
1500 			ifp->if_amcount = 0;
1501 			ifp->if_flags &= ~IFF_ALLMULTI;
1502 			ifr.ifr_flags = ifp->if_flags;
1503 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1504 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1505 					      (struct ucred *)NULL);
1506 		}
1507 	}
1508 	splx(s);
1509 
1510 	if (error == 0)
1511 		rt_ifmsg(ifp);
1512 	return error;
1513 }
1514 
1515 /*
1516  * Add a multicast listenership to the interface in question.
1517  * The link layer provides a routine which converts
1518  */
1519 int
1520 if_addmulti(
1521 	struct ifnet *ifp,	/* interface to manipulate */
1522 	struct sockaddr *sa,	/* address to add */
1523 	struct ifmultiaddr **retifma)
1524 {
1525 	struct sockaddr *llsa, *dupsa;
1526 	int error, s;
1527 	struct ifmultiaddr *ifma;
1528 
1529 	/*
1530 	 * If the matching multicast address already exists
1531 	 * then don't add a new one, just add a reference
1532 	 */
1533 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1534 		if (sa_equal(sa, ifma->ifma_addr)) {
1535 			ifma->ifma_refcount++;
1536 			if (retifma)
1537 				*retifma = ifma;
1538 			return 0;
1539 		}
1540 	}
1541 
1542 	/*
1543 	 * Give the link layer a chance to accept/reject it, and also
1544 	 * find out which AF_LINK address this maps to, if it isn't one
1545 	 * already.
1546 	 */
1547 	if (ifp->if_resolvemulti) {
1548 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1549 		if (error) return error;
1550 	} else {
1551 		llsa = 0;
1552 	}
1553 
1554 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1555 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1556 	bcopy(sa, dupsa, sa->sa_len);
1557 
1558 	ifma->ifma_addr = dupsa;
1559 	ifma->ifma_lladdr = llsa;
1560 	ifma->ifma_ifp = ifp;
1561 	ifma->ifma_refcount = 1;
1562 	ifma->ifma_protospec = 0;
1563 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1564 
1565 	/*
1566 	 * Some network interfaces can scan the address list at
1567 	 * interrupt time; lock them out.
1568 	 */
1569 	s = splimp();
1570 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1571 	splx(s);
1572 	*retifma = ifma;
1573 
1574 	if (llsa != 0) {
1575 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1576 			if (sa_equal(ifma->ifma_addr, llsa))
1577 				break;
1578 		}
1579 		if (ifma) {
1580 			ifma->ifma_refcount++;
1581 		} else {
1582 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1583 			       M_IFMADDR, M_WAITOK);
1584 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1585 			       M_IFMADDR, M_WAITOK);
1586 			bcopy(llsa, dupsa, llsa->sa_len);
1587 			ifma->ifma_addr = dupsa;
1588 			ifma->ifma_ifp = ifp;
1589 			ifma->ifma_refcount = 1;
1590 			s = splimp();
1591 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1592 			splx(s);
1593 		}
1594 	}
1595 	/*
1596 	 * We are certain we have added something, so call down to the
1597 	 * interface to let them know about it.
1598 	 */
1599 	s = splimp();
1600 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1601 	splx(s);
1602 
1603 	return 0;
1604 }
1605 
1606 /*
1607  * Remove a reference to a multicast address on this interface.  Yell
1608  * if the request does not match an existing membership.
1609  */
1610 int
1611 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1612 {
1613 	struct ifmultiaddr *ifma;
1614 	int s;
1615 
1616 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1617 		if (sa_equal(sa, ifma->ifma_addr))
1618 			break;
1619 	if (ifma == 0)
1620 		return ENOENT;
1621 
1622 	if (ifma->ifma_refcount > 1) {
1623 		ifma->ifma_refcount--;
1624 		return 0;
1625 	}
1626 
1627 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1628 	sa = ifma->ifma_lladdr;
1629 	s = splimp();
1630 	LIST_REMOVE(ifma, ifma_link);
1631 	/*
1632 	 * Make sure the interface driver is notified
1633 	 * in the case of a link layer mcast group being left.
1634 	 */
1635 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1636 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1637 	splx(s);
1638 	free(ifma->ifma_addr, M_IFMADDR);
1639 	free(ifma, M_IFMADDR);
1640 	if (sa == 0)
1641 		return 0;
1642 
1643 	/*
1644 	 * Now look for the link-layer address which corresponds to
1645 	 * this network address.  It had been squirreled away in
1646 	 * ifma->ifma_lladdr for this purpose (so we don't have
1647 	 * to call ifp->if_resolvemulti() again), and we saved that
1648 	 * value in sa above.  If some nasty deleted the
1649 	 * link-layer address out from underneath us, we can deal because
1650 	 * the address we stored was is not the same as the one which was
1651 	 * in the record for the link-layer address.  (So we don't complain
1652 	 * in that case.)
1653 	 */
1654 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1655 		if (sa_equal(sa, ifma->ifma_addr))
1656 			break;
1657 	if (ifma == 0)
1658 		return 0;
1659 
1660 	if (ifma->ifma_refcount > 1) {
1661 		ifma->ifma_refcount--;
1662 		return 0;
1663 	}
1664 
1665 	s = splimp();
1666 	LIST_REMOVE(ifma, ifma_link);
1667 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1668 	splx(s);
1669 	free(ifma->ifma_addr, M_IFMADDR);
1670 	free(sa, M_IFMADDR);
1671 	free(ifma, M_IFMADDR);
1672 
1673 	return 0;
1674 }
1675 
1676 /*
1677  * Set the link layer address on an interface.
1678  *
1679  * At this time we only support certain types of interfaces,
1680  * and we don't allow the length of the address to change.
1681  */
1682 int
1683 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1684 {
1685 	struct sockaddr_dl *sdl;
1686 	struct ifaddr *ifa;
1687 	struct ifreq ifr;
1688 
1689 	ifa = ifnet_addrs[ifp->if_index - 1];
1690 	if (ifa == NULL)
1691 		return (EINVAL);
1692 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1693 	if (sdl == NULL)
1694 		return (EINVAL);
1695 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1696 		return (EINVAL);
1697 	switch (ifp->if_type) {
1698 	case IFT_ETHER:			/* these types use struct arpcom */
1699 	case IFT_FDDI:
1700 	case IFT_XETHER:
1701 	case IFT_ISO88025:
1702 	case IFT_L2VLAN:
1703 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1704 		/* FALLTHROUGH */
1705 	case IFT_ARCNET:
1706 		bcopy(lladdr, LLADDR(sdl), len);
1707 		break;
1708 	default:
1709 		return (ENODEV);
1710 	}
1711 	/*
1712 	 * If the interface is already up, we need
1713 	 * to re-init it in order to reprogram its
1714 	 * address filter.
1715 	 */
1716 	if ((ifp->if_flags & IFF_UP) != 0) {
1717 		ifp->if_flags &= ~IFF_UP;
1718 		ifr.ifr_flags = ifp->if_flags;
1719 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1720 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1721 				 (struct ucred *)NULL);
1722 		ifp->if_flags |= IFF_UP;
1723 		ifr.ifr_flags = ifp->if_flags;
1724 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1725 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1726 				 (struct ucred *)NULL);
1727 #ifdef INET
1728 		/*
1729 		 * Also send gratuitous ARPs to notify other nodes about
1730 		 * the address change.
1731 		 */
1732 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1733 			if (ifa->ifa_addr != NULL &&
1734 			    ifa->ifa_addr->sa_family == AF_INET)
1735 				arp_ifinit(ifp, ifa);
1736 		}
1737 #endif
1738 	}
1739 	return (0);
1740 }
1741 
1742 struct ifmultiaddr *
1743 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1744 {
1745 	struct ifmultiaddr *ifma;
1746 
1747 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1748 		if (sa_equal(ifma->ifma_addr, sa))
1749 			break;
1750 
1751 	return ifma;
1752 }
1753 
1754 /*
1755  * The name argument must be a pointer to storage which will last as
1756  * long as the interface does.  For physical devices, the result of
1757  * device_get_name(dev) is a good choice and for pseudo-devices a
1758  * static string works well.
1759  */
1760 void
1761 if_initname(struct ifnet *ifp, const char *name, int unit)
1762 {
1763 	ifp->if_dname = name;
1764 	ifp->if_dunit = unit;
1765 	if (unit != IF_DUNIT_NONE)
1766 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1767 	else
1768 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1769 }
1770 
1771 int
1772 if_printf(struct ifnet *ifp, const char *fmt, ...)
1773 {
1774 	__va_list ap;
1775 	int retval;
1776 
1777 	retval = printf("%s: ", ifp->if_xname);
1778 	__va_start(ap, fmt);
1779 	retval += vprintf(fmt, ap);
1780 	__va_end(ap);
1781 	return (retval);
1782 }
1783 
1784 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1785 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1786