xref: /dragonfly/sys/net/if.c (revision 984263bc)
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 1.85.2.23 2003/04/15 18:11:19 fjoe Exp $
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/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/systm.h>
45 #include <sys/proc.h>
46 #include <sys/socket.h>
47 #include <sys/socketvar.h>
48 #include <sys/protosw.h>
49 #include <sys/kernel.h>
50 #include <sys/sockio.h>
51 #include <sys/syslog.h>
52 #include <sys/sysctl.h>
53 
54 #include <net/if.h>
55 #include <net/if_arp.h>
56 #include <net/if_dl.h>
57 #include <net/if_types.h>
58 #include <net/if_var.h>
59 #include <net/radix.h>
60 #include <net/route.h>
61 #include <machine/stdarg.h>
62 
63 #if defined(INET) || defined(INET6)
64 /*XXX*/
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #include <netinet/if_ether.h>
68 #ifdef INET6
69 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
70 #include <netinet6/in6_var.h>
71 #include <netinet6/in6_ifattach.h>
72 #endif
73 #endif
74 
75 /*
76  * System initialization
77  */
78 
79 static int ifconf __P((u_long, caddr_t));
80 static void ifinit __P((void *));
81 static void if_qflush __P((struct ifqueue *));
82 static void if_slowtimo __P((void *));
83 static void link_rtrequest __P((int, struct rtentry *, struct rt_addrinfo *));
84 static int  if_rtdel __P((struct radix_node *, void *));
85 
86 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
87 
88 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
89 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
90 
91 int	ifqmaxlen = IFQ_MAXLEN;
92 struct	ifnethead ifnet;	/* depend on static init XXX */
93 
94 #ifdef INET6
95 /*
96  * XXX: declare here to avoid to include many inet6 related files..
97  * should be more generalized?
98  */
99 extern void	nd6_setmtu __P((struct ifnet *));
100 #endif
101 
102 struct if_clone *if_clone_lookup __P((const char *, int *));
103 int if_clone_list __P((struct if_clonereq *));
104 
105 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
106 int if_cloners_count;
107 
108 /*
109  * Network interface utility routines.
110  *
111  * Routines with ifa_ifwith* names take sockaddr *'s as
112  * parameters.
113  */
114 /* ARGSUSED*/
115 void
116 ifinit(dummy)
117 	void *dummy;
118 {
119 	struct ifnet *ifp;
120 	int s;
121 
122 	s = splimp();
123 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
124 		if (ifp->if_snd.ifq_maxlen == 0) {
125 			printf("%s%d XXX: driver didn't set ifq_maxlen\n",
126 			    ifp->if_name, ifp->if_unit);
127 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
128 		}
129 	}
130 	splx(s);
131 	if_slowtimo(0);
132 }
133 
134 int if_index = 0;
135 struct ifaddr **ifnet_addrs;
136 struct ifnet **ifindex2ifnet = NULL;
137 
138 
139 /*
140  * Attach an interface to the
141  * list of "active" interfaces.
142  */
143 void
144 if_attach(ifp)
145 	struct ifnet *ifp;
146 {
147 	unsigned socksize, ifasize;
148 	int namelen, masklen;
149 	char workbuf[64];
150 	register struct sockaddr_dl *sdl;
151 	register struct ifaddr *ifa;
152 	static int if_indexlim = 8;
153 	static int inited;
154 
155 	if (!inited) {
156 		TAILQ_INIT(&ifnet);
157 		inited = 1;
158 	}
159 
160 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
161 	ifp->if_index = ++if_index;
162 	/*
163 	 * XXX -
164 	 * The old code would work if the interface passed a pre-existing
165 	 * chain of ifaddrs to this code.  We don't trust our callers to
166 	 * properly initialize the tailq, however, so we no longer allow
167 	 * this unlikely case.
168 	 */
169 	TAILQ_INIT(&ifp->if_addrhead);
170 	TAILQ_INIT(&ifp->if_prefixhead);
171 	LIST_INIT(&ifp->if_multiaddrs);
172 	getmicrotime(&ifp->if_lastchange);
173 	if (ifnet_addrs == 0 || if_index >= if_indexlim) {
174 		unsigned n = (if_indexlim <<= 1) * sizeof(ifa);
175 		caddr_t q = malloc(n, M_IFADDR, M_WAITOK);
176 		bzero(q, n);
177 		if (ifnet_addrs) {
178 			bcopy((caddr_t)ifnet_addrs, (caddr_t)q, n/2);
179 			free((caddr_t)ifnet_addrs, M_IFADDR);
180 		}
181 		ifnet_addrs = (struct ifaddr **)q;
182 
183 		/* grow ifindex2ifnet */
184 		n = if_indexlim * sizeof(struct ifnet *);
185 		q = malloc(n, M_IFADDR, M_WAITOK);
186 		bzero(q, n);
187 		if (ifindex2ifnet) {
188 			bcopy((caddr_t)ifindex2ifnet, q, n/2);
189 			free((caddr_t)ifindex2ifnet, M_IFADDR);
190 		}
191 		ifindex2ifnet = (struct ifnet **)q;
192 	}
193 
194 	ifindex2ifnet[if_index] = ifp;
195 
196 	/*
197 	 * create a Link Level name for this device
198 	 */
199 	namelen = snprintf(workbuf, sizeof(workbuf),
200 	    "%s%d", ifp->if_name, ifp->if_unit);
201 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
202 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
203 	socksize = masklen + ifp->if_addrlen;
204 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
205 	if (socksize < sizeof(*sdl))
206 		socksize = sizeof(*sdl);
207 	socksize = ROUNDUP(socksize);
208 	ifasize = sizeof(*ifa) + 2 * socksize;
209 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
210 	if (ifa) {
211 		bzero((caddr_t)ifa, ifasize);
212 		sdl = (struct sockaddr_dl *)(ifa + 1);
213 		sdl->sdl_len = socksize;
214 		sdl->sdl_family = AF_LINK;
215 		bcopy(workbuf, sdl->sdl_data, namelen);
216 		sdl->sdl_nlen = namelen;
217 		sdl->sdl_index = ifp->if_index;
218 		sdl->sdl_type = ifp->if_type;
219 		ifnet_addrs[if_index - 1] = ifa;
220 		ifa->ifa_ifp = ifp;
221 		ifa->ifa_rtrequest = link_rtrequest;
222 		ifa->ifa_addr = (struct sockaddr *)sdl;
223 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
224 		ifa->ifa_netmask = (struct sockaddr *)sdl;
225 		sdl->sdl_len = masklen;
226 		while (namelen != 0)
227 			sdl->sdl_data[--namelen] = 0xff;
228 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
229 	}
230 
231 	/* Announce the interface. */
232 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
233 }
234 
235 /*
236  * Detach an interface, removing it from the
237  * list of "active" interfaces.
238  */
239 void
240 if_detach(ifp)
241 	struct ifnet *ifp;
242 {
243 	struct ifaddr *ifa;
244 	struct radix_node_head	*rnh;
245 	int s;
246 	int i;
247 
248 	/*
249 	 * Remove routes and flush queues.
250 	 */
251 	s = splnet();
252 	if_down(ifp);
253 
254 	/*
255 	 * Remove address from ifnet_addrs[] and maybe decrement if_index.
256 	 * Clean up all addresses.
257 	 */
258 	ifnet_addrs[ifp->if_index - 1] = 0;
259 	while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
260 		if_index--;
261 
262 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
263 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
264 #ifdef INET
265 		/* XXX: Ugly!! ad hoc just for INET */
266 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
267 			struct ifaliasreq ifr;
268 
269 			bzero(&ifr, sizeof(ifr));
270 			ifr.ifra_addr = *ifa->ifa_addr;
271 			if (ifa->ifa_dstaddr)
272 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
273 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
274 			    NULL) == 0)
275 				continue;
276 		}
277 #endif /* INET */
278 #ifdef INET6
279 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
280 			in6_purgeaddr(ifa);
281 			/* ifp_addrhead is already updated */
282 			continue;
283 		}
284 #endif /* INET6 */
285 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
286 		IFAFREE(ifa);
287 	}
288 
289 #ifdef INET6
290 	/*
291 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
292 	 * before removing routing entries below, since IPv6 interface direct
293 	 * routes are expected to be removed by the IPv6-specific kernel API.
294 	 * Otherwise, the kernel will detect some inconsistency and bark it.
295 	 */
296 	in6_ifdetach(ifp);
297 #endif
298 
299 	/*
300 	 * Delete all remaining routes using this interface
301 	 * Unfortuneatly the only way to do this is to slog through
302 	 * the entire routing table looking for routes which point
303 	 * to this interface...oh well...
304 	 */
305 	for (i = 1; i <= AF_MAX; i++) {
306 		if ((rnh = rt_tables[i]) == NULL)
307 			continue;
308 		(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
309 	}
310 
311 	/* Announce that the interface is gone. */
312 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
313 
314 	TAILQ_REMOVE(&ifnet, ifp, if_link);
315 	splx(s);
316 }
317 
318 /*
319  * Delete Routes for a Network Interface
320  *
321  * Called for each routing entry via the rnh->rnh_walktree() call above
322  * to delete all route entries referencing a detaching network interface.
323  *
324  * Arguments:
325  *	rn	pointer to node in the routing table
326  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
327  *
328  * Returns:
329  *	0	successful
330  *	errno	failed - reason indicated
331  *
332  */
333 static int
334 if_rtdel(rn, arg)
335 	struct radix_node	*rn;
336 	void			*arg;
337 {
338 	struct rtentry	*rt = (struct rtentry *)rn;
339 	struct ifnet	*ifp = arg;
340 	int		err;
341 
342 	if (rt->rt_ifp == ifp) {
343 
344 		/*
345 		 * Protect (sorta) against walktree recursion problems
346 		 * with cloned routes
347 		 */
348 		if ((rt->rt_flags & RTF_UP) == 0)
349 			return (0);
350 
351 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
352 				rt_mask(rt), rt->rt_flags,
353 				(struct rtentry **) NULL);
354 		if (err) {
355 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
356 		}
357 	}
358 
359 	return (0);
360 }
361 
362 /*
363  * Create a clone network interface.
364  */
365 int
366 if_clone_create(name, len)
367 	char *name;
368 	int len;
369 {
370 	struct if_clone *ifc;
371 	char *dp;
372 	int wildcard;
373 	int unit;
374 	int err;
375 
376 	ifc = if_clone_lookup(name, &unit);
377 	if (ifc == NULL)
378 		return (EINVAL);
379 
380 	if (ifunit(name) != NULL)
381 		return (EEXIST);
382 
383 	wildcard = (unit < 0);
384 
385 	err = (*ifc->ifc_create)(ifc, &unit);
386 	if (err != 0)
387 		return (err);
388 
389 	/* In the wildcard case, we need to update the name. */
390 	if (wildcard) {
391 		for (dp = name; *dp != '\0'; dp++);
392 		if (snprintf(dp, len - (dp-name), "%d", unit) >
393 		    len - (dp-name) - 1) {
394 			/*
395 			 * This can only be a programmer error and
396 			 * there's no straightforward way to recover if
397 			 * it happens.
398 			 */
399 			panic("if_clone_create(): interface name too long");
400 		}
401 
402 	}
403 
404 	return (0);
405 }
406 
407 /*
408  * Destroy a clone network interface.
409  */
410 int
411 if_clone_destroy(name)
412 	const char *name;
413 {
414 	struct if_clone *ifc;
415 	struct ifnet *ifp;
416 
417 	ifc = if_clone_lookup(name, NULL);
418 	if (ifc == NULL)
419 		return (EINVAL);
420 
421 	ifp = ifunit(name);
422 	if (ifp == NULL)
423 		return (ENXIO);
424 
425 	if (ifc->ifc_destroy == NULL)
426 		return (EOPNOTSUPP);
427 
428 	(*ifc->ifc_destroy)(ifp);
429 	return (0);
430 }
431 
432 /*
433  * Look up a network interface cloner.
434  */
435 struct if_clone *
436 if_clone_lookup(name, unitp)
437 	const char *name;
438 	int *unitp;
439 {
440 	struct if_clone *ifc;
441 	const char *cp;
442 	int i;
443 
444 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
445 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
446 			if (ifc->ifc_name[i] != *cp)
447 				goto next_ifc;
448 		}
449 		goto found_name;
450  next_ifc:
451 		ifc = LIST_NEXT(ifc, ifc_list);
452 	}
453 
454 	/* No match. */
455 	return ((struct if_clone *)NULL);
456 
457  found_name:
458 	if (*cp == '\0') {
459 		i = -1;
460 	} else {
461 		for (i = 0; *cp != '\0'; cp++) {
462 			if (*cp < '0' || *cp > '9') {
463 				/* Bogus unit number. */
464 				return (NULL);
465 			}
466 			i = (i * 10) + (*cp - '0');
467 		}
468 	}
469 
470 	if (unitp != NULL)
471 		*unitp = i;
472 	return (ifc);
473 }
474 
475 /*
476  * Register a network interface cloner.
477  */
478 void
479 if_clone_attach(ifc)
480 	struct if_clone *ifc;
481 {
482 
483 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
484 	if_cloners_count++;
485 }
486 
487 /*
488  * Unregister a network interface cloner.
489  */
490 void
491 if_clone_detach(ifc)
492 	struct if_clone *ifc;
493 {
494 
495 	LIST_REMOVE(ifc, ifc_list);
496 	if_cloners_count--;
497 }
498 
499 /*
500  * Provide list of interface cloners to userspace.
501  */
502 int
503 if_clone_list(ifcr)
504 	struct if_clonereq *ifcr;
505 {
506 	char outbuf[IFNAMSIZ], *dst;
507 	struct if_clone *ifc;
508 	int count, error = 0;
509 
510 	ifcr->ifcr_total = if_cloners_count;
511 	if ((dst = ifcr->ifcr_buffer) == NULL) {
512 		/* Just asking how many there are. */
513 		return (0);
514 	}
515 
516 	if (ifcr->ifcr_count < 0)
517 		return (EINVAL);
518 
519 	count = (if_cloners_count < ifcr->ifcr_count) ?
520 	    if_cloners_count : ifcr->ifcr_count;
521 
522 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
523 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
524 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
525 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
526 		error = copyout(outbuf, dst, IFNAMSIZ);
527 		if (error)
528 			break;
529 	}
530 
531 	return (error);
532 }
533 
534 /*
535  * Locate an interface based on a complete address.
536  */
537 /*ARGSUSED*/
538 struct ifaddr *
539 ifa_ifwithaddr(addr)
540 	register struct sockaddr *addr;
541 {
542 	register struct ifnet *ifp;
543 	register struct ifaddr *ifa;
544 
545 #define	equal(a1, a2) \
546   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
547 	TAILQ_FOREACH(ifp, &ifnet, if_link)
548 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
549 		if (ifa->ifa_addr->sa_family != addr->sa_family)
550 			continue;
551 		if (equal(addr, ifa->ifa_addr))
552 			return (ifa);
553 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
554 		    /* IP6 doesn't have broadcast */
555 		    ifa->ifa_broadaddr->sa_len != 0 &&
556 		    equal(ifa->ifa_broadaddr, addr))
557 			return (ifa);
558 	}
559 	return ((struct ifaddr *)0);
560 }
561 /*
562  * Locate the point to point interface with a given destination address.
563  */
564 /*ARGSUSED*/
565 struct ifaddr *
566 ifa_ifwithdstaddr(addr)
567 	register struct sockaddr *addr;
568 {
569 	register struct ifnet *ifp;
570 	register struct ifaddr *ifa;
571 
572 	TAILQ_FOREACH(ifp, &ifnet, if_link)
573 	    if (ifp->if_flags & IFF_POINTOPOINT)
574 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
575 			if (ifa->ifa_addr->sa_family != addr->sa_family)
576 				continue;
577 			if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
578 				return (ifa);
579 	}
580 	return ((struct ifaddr *)0);
581 }
582 
583 /*
584  * Find an interface on a specific network.  If many, choice
585  * is most specific found.
586  */
587 struct ifaddr *
588 ifa_ifwithnet(addr)
589 	struct sockaddr *addr;
590 {
591 	register struct ifnet *ifp;
592 	register struct ifaddr *ifa;
593 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
594 	u_int af = addr->sa_family;
595 	char *addr_data = addr->sa_data, *cplim;
596 
597 	/*
598 	 * AF_LINK addresses can be looked up directly by their index number,
599 	 * so do that if we can.
600 	 */
601 	if (af == AF_LINK) {
602 	    register struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
603 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
604 		return (ifnet_addrs[sdl->sdl_index - 1]);
605 	}
606 
607 	/*
608 	 * Scan though each interface, looking for ones that have
609 	 * addresses in this address family.
610 	 */
611 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
612 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
613 			register char *cp, *cp2, *cp3;
614 
615 			if (ifa->ifa_addr->sa_family != af)
616 next:				continue;
617 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
618 				/*
619 				 * This is a bit broken as it doesn't
620 				 * take into account that the remote end may
621 				 * be a single node in the network we are
622 				 * looking for.
623 				 * The trouble is that we don't know the
624 				 * netmask for the remote end.
625 				 */
626 				if (ifa->ifa_dstaddr != 0
627 				    && equal(addr, ifa->ifa_dstaddr))
628  					return (ifa);
629 			} else {
630 				/*
631 				 * if we have a special address handler,
632 				 * then use it instead of the generic one.
633 				 */
634 	          		if (ifa->ifa_claim_addr) {
635 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
636 						return (ifa);
637 					} else {
638 						continue;
639 					}
640 				}
641 
642 				/*
643 				 * Scan all the bits in the ifa's address.
644 				 * If a bit dissagrees with what we are
645 				 * looking for, mask it with the netmask
646 				 * to see if it really matters.
647 				 * (A byte at a time)
648 				 */
649 				if (ifa->ifa_netmask == 0)
650 					continue;
651 				cp = addr_data;
652 				cp2 = ifa->ifa_addr->sa_data;
653 				cp3 = ifa->ifa_netmask->sa_data;
654 				cplim = ifa->ifa_netmask->sa_len
655 					+ (char *)ifa->ifa_netmask;
656 				while (cp3 < cplim)
657 					if ((*cp++ ^ *cp2++) & *cp3++)
658 						goto next; /* next address! */
659 				/*
660 				 * If the netmask of what we just found
661 				 * is more specific than what we had before
662 				 * (if we had one) then remember the new one
663 				 * before continuing to search
664 				 * for an even better one.
665 				 */
666 				if (ifa_maybe == 0 ||
667 				    rn_refines((caddr_t)ifa->ifa_netmask,
668 				    (caddr_t)ifa_maybe->ifa_netmask))
669 					ifa_maybe = ifa;
670 			}
671 		}
672 	}
673 	return (ifa_maybe);
674 }
675 
676 /*
677  * Find an interface address specific to an interface best matching
678  * a given address.
679  */
680 struct ifaddr *
681 ifaof_ifpforaddr(addr, ifp)
682 	struct sockaddr *addr;
683 	register struct ifnet *ifp;
684 {
685 	register struct ifaddr *ifa;
686 	register char *cp, *cp2, *cp3;
687 	register char *cplim;
688 	struct ifaddr *ifa_maybe = 0;
689 	u_int af = addr->sa_family;
690 
691 	if (af >= AF_MAX)
692 		return (0);
693 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
694 		if (ifa->ifa_addr->sa_family != af)
695 			continue;
696 		if (ifa_maybe == 0)
697 			ifa_maybe = ifa;
698 		if (ifa->ifa_netmask == 0) {
699 			if (equal(addr, ifa->ifa_addr) ||
700 			    (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
701 				return (ifa);
702 			continue;
703 		}
704 		if (ifp->if_flags & IFF_POINTOPOINT) {
705 			if (equal(addr, ifa->ifa_dstaddr))
706 				return (ifa);
707 		} else {
708 			cp = addr->sa_data;
709 			cp2 = ifa->ifa_addr->sa_data;
710 			cp3 = ifa->ifa_netmask->sa_data;
711 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
712 			for (; cp3 < cplim; cp3++)
713 				if ((*cp++ ^ *cp2++) & *cp3)
714 					break;
715 			if (cp3 == cplim)
716 				return (ifa);
717 		}
718 	}
719 	return (ifa_maybe);
720 }
721 
722 #include <net/route.h>
723 
724 /*
725  * Default action when installing a route with a Link Level gateway.
726  * Lookup an appropriate real ifa to point to.
727  * This should be moved to /sys/net/link.c eventually.
728  */
729 static void
730 link_rtrequest(cmd, rt, info)
731 	int cmd;
732 	register struct rtentry *rt;
733 	struct rt_addrinfo *info;
734 {
735 	register struct ifaddr *ifa;
736 	struct sockaddr *dst;
737 	struct ifnet *ifp;
738 
739 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
740 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
741 		return;
742 	ifa = ifaof_ifpforaddr(dst, ifp);
743 	if (ifa) {
744 		IFAFREE(rt->rt_ifa);
745 		rt->rt_ifa = ifa;
746 		ifa->ifa_refcnt++;
747 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
748 			ifa->ifa_rtrequest(cmd, rt, info);
749 	}
750 }
751 
752 /*
753  * Mark an interface down and notify protocols of
754  * the transition.
755  * NOTE: must be called at splnet or eqivalent.
756  */
757 void
758 if_unroute(ifp, flag, fam)
759 	register struct ifnet *ifp;
760 	int flag, fam;
761 {
762 	register struct ifaddr *ifa;
763 
764 	ifp->if_flags &= ~flag;
765 	getmicrotime(&ifp->if_lastchange);
766 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
767 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
768 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
769 	if_qflush(&ifp->if_snd);
770 	rt_ifmsg(ifp);
771 }
772 
773 /*
774  * Mark an interface up and notify protocols of
775  * the transition.
776  * NOTE: must be called at splnet or eqivalent.
777  */
778 void
779 if_route(ifp, flag, fam)
780 	register struct ifnet *ifp;
781 	int flag, fam;
782 {
783 	register struct ifaddr *ifa;
784 
785 	ifp->if_flags |= flag;
786 	getmicrotime(&ifp->if_lastchange);
787 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
788 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
789 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
790 	rt_ifmsg(ifp);
791 #ifdef INET6
792 	in6_if_up(ifp);
793 #endif
794 }
795 
796 /*
797  * Mark an interface down and notify protocols of
798  * the transition.
799  * NOTE: must be called at splnet or eqivalent.
800  */
801 void
802 if_down(ifp)
803 	register struct ifnet *ifp;
804 {
805 
806 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
807 }
808 
809 /*
810  * Mark an interface up and notify protocols of
811  * the transition.
812  * NOTE: must be called at splnet or eqivalent.
813  */
814 void
815 if_up(ifp)
816 	register struct ifnet *ifp;
817 {
818 
819 	if_route(ifp, IFF_UP, AF_UNSPEC);
820 }
821 
822 /*
823  * Flush an interface queue.
824  */
825 static void
826 if_qflush(ifq)
827 	register struct ifqueue *ifq;
828 {
829 	register struct mbuf *m, *n;
830 
831 	n = ifq->ifq_head;
832 	while ((m = n) != 0) {
833 		n = m->m_act;
834 		m_freem(m);
835 	}
836 	ifq->ifq_head = 0;
837 	ifq->ifq_tail = 0;
838 	ifq->ifq_len = 0;
839 }
840 
841 /*
842  * Handle interface watchdog timer routines.  Called
843  * from softclock, we decrement timers (if set) and
844  * call the appropriate interface routine on expiration.
845  */
846 static void
847 if_slowtimo(arg)
848 	void *arg;
849 {
850 	register struct ifnet *ifp;
851 	int s = splimp();
852 
853 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
854 		if (ifp->if_timer == 0 || --ifp->if_timer)
855 			continue;
856 		if (ifp->if_watchdog)
857 			(*ifp->if_watchdog)(ifp);
858 	}
859 	splx(s);
860 	timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
861 }
862 
863 /*
864  * Map interface name to
865  * interface structure pointer.
866  */
867 struct ifnet *
868 ifunit(const char *name)
869 {
870 	char namebuf[IFNAMSIZ + 1];
871 	const char *cp;
872 	struct ifnet *ifp;
873 	int unit;
874 	unsigned len, m;
875 	char c;
876 
877 	len = strlen(name);
878 	if (len < 2 || len > IFNAMSIZ)
879 		return NULL;
880 	cp = name + len - 1;
881 	c = *cp;
882 	if (c < '0' || c > '9')
883 		return NULL;		/* trailing garbage */
884 	unit = 0;
885 	m = 1;
886 	do {
887 		if (cp == name)
888 			return NULL;	/* no interface name */
889 		unit += (c - '0') * m;
890 		if (unit > 1000000)
891 			return NULL;	/* number is unreasonable */
892 		m *= 10;
893 		c = *--cp;
894 	} while (c >= '0' && c <= '9');
895 	len = cp - name + 1;
896 	bcopy(name, namebuf, len);
897 	namebuf[len] = '\0';
898 	/*
899 	 * Now search all the interfaces for this name/number
900 	 */
901 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
902 		if (strcmp(ifp->if_name, namebuf))
903 			continue;
904 		if (unit == ifp->if_unit)
905 			break;
906 	}
907 	return (ifp);
908 }
909 
910 
911 /*
912  * Map interface name in a sockaddr_dl to
913  * interface structure pointer.
914  */
915 struct ifnet *
916 if_withname(sa)
917 	struct sockaddr *sa;
918 {
919 	char ifname[IFNAMSIZ+1];
920 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
921 
922 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
923 	     (sdl->sdl_nlen > IFNAMSIZ) )
924 		return NULL;
925 
926 	/*
927 	 * ifunit wants a null-terminated name.  It may not be null-terminated
928 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
929 	 * and there might not be room to put the trailing null anyway, so we
930 	 * make a local copy that we know we can null terminate safely.
931 	 */
932 
933 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
934 	ifname[sdl->sdl_nlen] = '\0';
935 	return ifunit(ifname);
936 }
937 
938 
939 /*
940  * Interface ioctls.
941  */
942 int
943 ifioctl(so, cmd, data, p)
944 	struct socket *so;
945 	u_long cmd;
946 	caddr_t data;
947 	struct proc *p;
948 {
949 	register struct ifnet *ifp;
950 	register struct ifreq *ifr;
951 	struct ifstat *ifs;
952 	int error;
953 	short oif_flags;
954 	int new_flags;
955 
956 	switch (cmd) {
957 
958 	case SIOCGIFCONF:
959 	case OSIOCGIFCONF:
960 		return (ifconf(cmd, data));
961 	}
962 	ifr = (struct ifreq *)data;
963 
964 	switch (cmd) {
965 	case SIOCIFCREATE:
966 	case SIOCIFDESTROY:
967 		if ((error = suser(p)) != 0)
968 			return (error);
969 		return ((cmd == SIOCIFCREATE) ?
970 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
971 			if_clone_destroy(ifr->ifr_name));
972 
973 	case SIOCIFGCLONERS:
974 		return (if_clone_list((struct if_clonereq *)data));
975 	}
976 
977 	ifp = ifunit(ifr->ifr_name);
978 	if (ifp == 0)
979 		return (ENXIO);
980 	switch (cmd) {
981 
982 	case SIOCGIFFLAGS:
983 		ifr->ifr_flags = ifp->if_flags;
984 		ifr->ifr_flagshigh = ifp->if_ipending >> 16;
985 		break;
986 
987 	case SIOCGIFCAP:
988 		ifr->ifr_reqcap = ifp->if_capabilities;
989 		ifr->ifr_curcap = ifp->if_capenable;
990 		break;
991 
992 	case SIOCGIFMETRIC:
993 		ifr->ifr_metric = ifp->if_metric;
994 		break;
995 
996 	case SIOCGIFMTU:
997 		ifr->ifr_mtu = ifp->if_mtu;
998 		break;
999 
1000 	case SIOCGIFPHYS:
1001 		ifr->ifr_phys = ifp->if_physical;
1002 		break;
1003 
1004 	case SIOCSIFFLAGS:
1005 		error = suser(p);
1006 		if (error)
1007 			return (error);
1008 		new_flags = (ifr->ifr_flags & 0xffff) |
1009 		    (ifr->ifr_flagshigh << 16);
1010 		if (ifp->if_flags & IFF_SMART) {
1011 			/* Smart drivers twiddle their own routes */
1012 		} else if (ifp->if_flags & IFF_UP &&
1013 		    (new_flags & IFF_UP) == 0) {
1014 			int s = splimp();
1015 			if_down(ifp);
1016 			splx(s);
1017 		} else if (new_flags & IFF_UP &&
1018 		    (ifp->if_flags & IFF_UP) == 0) {
1019 			int s = splimp();
1020 			if_up(ifp);
1021 			splx(s);
1022 		}
1023 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1024 			(new_flags &~ IFF_CANTCHANGE);
1025 		ifp->if_ipending = (ifp->if_ipending & IFF_CANTCHANGE) |
1026 			(new_flags &~ IFF_CANTCHANGE);
1027 		if (new_flags & IFF_PPROMISC) {
1028 			/* Permanently promiscuous mode requested */
1029 			ifp->if_flags |= IFF_PROMISC;
1030 		} else if (ifp->if_pcount == 0) {
1031 			ifp->if_flags &= ~IFF_PROMISC;
1032 		}
1033 		if (ifp->if_ioctl)
1034 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1035 		getmicrotime(&ifp->if_lastchange);
1036 		break;
1037 
1038 	case SIOCSIFCAP:
1039 		error = suser(p);
1040 		if (error)
1041 			return (error);
1042 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1043 			return (EINVAL);
1044 		(void) (*ifp->if_ioctl)(ifp, cmd, data);
1045 		break;
1046 
1047 	case SIOCSIFMETRIC:
1048 		error = suser(p);
1049 		if (error)
1050 			return (error);
1051 		ifp->if_metric = ifr->ifr_metric;
1052 		getmicrotime(&ifp->if_lastchange);
1053 		break;
1054 
1055 	case SIOCSIFPHYS:
1056 		error = suser(p);
1057 		if (error)
1058 			return error;
1059 		if (!ifp->if_ioctl)
1060 		        return EOPNOTSUPP;
1061 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1062 		if (error == 0)
1063 			getmicrotime(&ifp->if_lastchange);
1064 		return(error);
1065 
1066 	case SIOCSIFMTU:
1067 	{
1068 		u_long oldmtu = ifp->if_mtu;
1069 
1070 		error = suser(p);
1071 		if (error)
1072 			return (error);
1073 		if (ifp->if_ioctl == NULL)
1074 			return (EOPNOTSUPP);
1075 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1076 			return (EINVAL);
1077 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1078 		if (error == 0) {
1079 			getmicrotime(&ifp->if_lastchange);
1080 			rt_ifmsg(ifp);
1081 		}
1082 		/*
1083 		 * If the link MTU changed, do network layer specific procedure.
1084 		 */
1085 		if (ifp->if_mtu != oldmtu) {
1086 #ifdef INET6
1087 			nd6_setmtu(ifp);
1088 #endif
1089 		}
1090 		return (error);
1091 	}
1092 
1093 	case SIOCADDMULTI:
1094 	case SIOCDELMULTI:
1095 		error = suser(p);
1096 		if (error)
1097 			return (error);
1098 
1099 		/* Don't allow group membership on non-multicast interfaces. */
1100 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1101 			return EOPNOTSUPP;
1102 
1103 		/* Don't let users screw up protocols' entries. */
1104 		if (ifr->ifr_addr.sa_family != AF_LINK)
1105 			return EINVAL;
1106 
1107 		if (cmd == SIOCADDMULTI) {
1108 			struct ifmultiaddr *ifma;
1109 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1110 		} else {
1111 			error = if_delmulti(ifp, &ifr->ifr_addr);
1112 		}
1113 		if (error == 0)
1114 			getmicrotime(&ifp->if_lastchange);
1115 		return error;
1116 
1117 	case SIOCSIFPHYADDR:
1118 	case SIOCDIFPHYADDR:
1119 #ifdef INET6
1120 	case SIOCSIFPHYADDR_IN6:
1121 #endif
1122 	case SIOCSLIFPHYADDR:
1123         case SIOCSIFMEDIA:
1124 	case SIOCSIFGENERIC:
1125 		error = suser(p);
1126 		if (error)
1127 			return (error);
1128 		if (ifp->if_ioctl == 0)
1129 			return (EOPNOTSUPP);
1130 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1131 		if (error == 0)
1132 			getmicrotime(&ifp->if_lastchange);
1133 		return error;
1134 
1135 	case SIOCGIFSTATUS:
1136 		ifs = (struct ifstat *)data;
1137 		ifs->ascii[0] = '\0';
1138 
1139 	case SIOCGIFPSRCADDR:
1140 	case SIOCGIFPDSTADDR:
1141 	case SIOCGLIFPHYADDR:
1142 	case SIOCGIFMEDIA:
1143 	case SIOCGIFGENERIC:
1144 		if (ifp->if_ioctl == 0)
1145 			return (EOPNOTSUPP);
1146 		return ((*ifp->if_ioctl)(ifp, cmd, data));
1147 
1148 	case SIOCSIFLLADDR:
1149 		error = suser(p);
1150 		if (error)
1151 			return (error);
1152 		return if_setlladdr(ifp,
1153 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1154 
1155 	default:
1156 		oif_flags = ifp->if_flags;
1157 		if (so->so_proto == 0)
1158 			return (EOPNOTSUPP);
1159 #ifndef COMPAT_43
1160 		error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1161 								 data,
1162 								 ifp, p));
1163 #else
1164 	    {
1165 		int ocmd = cmd;
1166 
1167 		switch (cmd) {
1168 
1169 		case SIOCSIFDSTADDR:
1170 		case SIOCSIFADDR:
1171 		case SIOCSIFBRDADDR:
1172 		case SIOCSIFNETMASK:
1173 #if BYTE_ORDER != BIG_ENDIAN
1174 			if (ifr->ifr_addr.sa_family == 0 &&
1175 			    ifr->ifr_addr.sa_len < 16) {
1176 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1177 				ifr->ifr_addr.sa_len = 16;
1178 			}
1179 #else
1180 			if (ifr->ifr_addr.sa_len == 0)
1181 				ifr->ifr_addr.sa_len = 16;
1182 #endif
1183 			break;
1184 
1185 		case OSIOCGIFADDR:
1186 			cmd = SIOCGIFADDR;
1187 			break;
1188 
1189 		case OSIOCGIFDSTADDR:
1190 			cmd = SIOCGIFDSTADDR;
1191 			break;
1192 
1193 		case OSIOCGIFBRDADDR:
1194 			cmd = SIOCGIFBRDADDR;
1195 			break;
1196 
1197 		case OSIOCGIFNETMASK:
1198 			cmd = SIOCGIFNETMASK;
1199 		}
1200 		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
1201 								   cmd,
1202 								   data,
1203 								   ifp, p));
1204 		switch (ocmd) {
1205 
1206 		case OSIOCGIFADDR:
1207 		case OSIOCGIFDSTADDR:
1208 		case OSIOCGIFBRDADDR:
1209 		case OSIOCGIFNETMASK:
1210 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1211 
1212 		}
1213 	    }
1214 #endif /* COMPAT_43 */
1215 
1216 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1217 #ifdef INET6
1218 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1219 			if (ifp->if_flags & IFF_UP) {
1220 				int s = splimp();
1221 				in6_if_up(ifp);
1222 				splx(s);
1223 			}
1224 #endif
1225 		}
1226 		return (error);
1227 
1228 	}
1229 	return (0);
1230 }
1231 
1232 /*
1233  * Set/clear promiscuous mode on interface ifp based on the truth value
1234  * of pswitch.  The calls are reference counted so that only the first
1235  * "on" request actually has an effect, as does the final "off" request.
1236  * Results are undefined if the "off" and "on" requests are not matched.
1237  */
1238 int
1239 ifpromisc(ifp, pswitch)
1240 	struct ifnet *ifp;
1241 	int pswitch;
1242 {
1243 	struct ifreq ifr;
1244 	int error;
1245 	int oldflags;
1246 
1247 	oldflags = ifp->if_flags;
1248 	if (ifp->if_ipending & IFF_PPROMISC) {
1249 		/* Do nothing if device is in permanently promiscuous mode */
1250 		ifp->if_pcount += pswitch ? 1 : -1;
1251 		return (0);
1252 	}
1253 	if (pswitch) {
1254 		/*
1255 		 * If the device is not configured up, we cannot put it in
1256 		 * promiscuous mode.
1257 		 */
1258 		if ((ifp->if_flags & IFF_UP) == 0)
1259 			return (ENETDOWN);
1260 		if (ifp->if_pcount++ != 0)
1261 			return (0);
1262 		ifp->if_flags |= IFF_PROMISC;
1263 		log(LOG_INFO, "%s%d: promiscuous mode enabled\n",
1264 		    ifp->if_name, ifp->if_unit);
1265 	} else {
1266 		if (--ifp->if_pcount > 0)
1267 			return (0);
1268 		ifp->if_flags &= ~IFF_PROMISC;
1269 		log(LOG_INFO, "%s%d: promiscuous mode disabled\n",
1270 		    ifp->if_name, ifp->if_unit);
1271 	}
1272 	ifr.ifr_flags = ifp->if_flags;
1273 	ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1274 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1275 	if (error == 0)
1276 		rt_ifmsg(ifp);
1277 	else
1278 		ifp->if_flags = oldflags;
1279 	return error;
1280 }
1281 
1282 /*
1283  * Return interface configuration
1284  * of system.  List may be used
1285  * in later ioctl's (above) to get
1286  * other information.
1287  */
1288 /*ARGSUSED*/
1289 static int
1290 ifconf(cmd, data)
1291 	u_long cmd;
1292 	caddr_t data;
1293 {
1294 	register struct ifconf *ifc = (struct ifconf *)data;
1295 	register struct ifnet *ifp;
1296 	register struct ifaddr *ifa;
1297 	struct sockaddr *sa;
1298 	struct ifreq ifr, *ifrp;
1299 	int space = ifc->ifc_len, error = 0;
1300 
1301 	ifrp = ifc->ifc_req;
1302 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1303 		char workbuf[64];
1304 		int ifnlen, addrs;
1305 
1306 		if (space <= sizeof (ifr))
1307 			break;
1308 		ifnlen = snprintf(workbuf, sizeof(workbuf),
1309 		    "%s%d", ifp->if_name, ifp->if_unit);
1310 		if(ifnlen + 1 > sizeof ifr.ifr_name) {
1311 			error = ENAMETOOLONG;
1312 			break;
1313 		} else {
1314 			strcpy(ifr.ifr_name, workbuf);
1315 		}
1316 
1317 		addrs = 0;
1318 		ifa = ifp->if_addrhead.tqh_first;
1319 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1320 			if (space <= sizeof(ifr))
1321 				break;
1322 			sa = ifa->ifa_addr;
1323 			if (curproc->p_prison && prison_if(curproc, sa))
1324 				continue;
1325 			addrs++;
1326 #ifdef COMPAT_43
1327 			if (cmd == OSIOCGIFCONF) {
1328 				struct osockaddr *osa =
1329 					 (struct osockaddr *)&ifr.ifr_addr;
1330 				ifr.ifr_addr = *sa;
1331 				osa->sa_family = sa->sa_family;
1332 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1333 						sizeof (ifr));
1334 				ifrp++;
1335 			} else
1336 #endif
1337 			if (sa->sa_len <= sizeof(*sa)) {
1338 				ifr.ifr_addr = *sa;
1339 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1340 						sizeof (ifr));
1341 				ifrp++;
1342 			} else {
1343 				if (space < sizeof (ifr) + sa->sa_len -
1344 					    sizeof(*sa))
1345 					break;
1346 				space -= sa->sa_len - sizeof(*sa);
1347 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1348 						sizeof (ifr.ifr_name));
1349 				if (error == 0)
1350 				    error = copyout((caddr_t)sa,
1351 				      (caddr_t)&ifrp->ifr_addr, sa->sa_len);
1352 				ifrp = (struct ifreq *)
1353 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1354 			}
1355 			if (error)
1356 				break;
1357 			space -= sizeof (ifr);
1358 		}
1359 		if (error)
1360 			break;
1361 		if (!addrs) {
1362 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1363 			error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1364 			    sizeof (ifr));
1365 			if (error)
1366 				break;
1367 			space -= sizeof (ifr);
1368 			ifrp++;
1369 		}
1370 	}
1371 	ifc->ifc_len -= space;
1372 	return (error);
1373 }
1374 
1375 /*
1376  * Just like if_promisc(), but for all-multicast-reception mode.
1377  */
1378 int
1379 if_allmulti(ifp, onswitch)
1380 	struct ifnet *ifp;
1381 	int onswitch;
1382 {
1383 	int error = 0;
1384 	int s = splimp();
1385 	struct ifreq ifr;
1386 
1387 	if (onswitch) {
1388 		if (ifp->if_amcount++ == 0) {
1389 			ifp->if_flags |= IFF_ALLMULTI;
1390 			ifr.ifr_flags = ifp->if_flags;
1391 			ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1392 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1393 		}
1394 	} else {
1395 		if (ifp->if_amcount > 1) {
1396 			ifp->if_amcount--;
1397 		} else {
1398 			ifp->if_amcount = 0;
1399 			ifp->if_flags &= ~IFF_ALLMULTI;
1400 			ifr.ifr_flags = ifp->if_flags;
1401 			ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1402 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1403 		}
1404 	}
1405 	splx(s);
1406 
1407 	if (error == 0)
1408 		rt_ifmsg(ifp);
1409 	return error;
1410 }
1411 
1412 /*
1413  * Add a multicast listenership to the interface in question.
1414  * The link layer provides a routine which converts
1415  */
1416 int
1417 if_addmulti(ifp, sa, retifma)
1418 	struct ifnet *ifp;	/* interface to manipulate */
1419 	struct sockaddr *sa;	/* address to add */
1420 	struct ifmultiaddr **retifma;
1421 {
1422 	struct sockaddr *llsa, *dupsa;
1423 	int error, s;
1424 	struct ifmultiaddr *ifma;
1425 
1426 	/*
1427 	 * If the matching multicast address already exists
1428 	 * then don't add a new one, just add a reference
1429 	 */
1430 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1431 		if (equal(sa, ifma->ifma_addr)) {
1432 			ifma->ifma_refcount++;
1433 			if (retifma)
1434 				*retifma = ifma;
1435 			return 0;
1436 		}
1437 	}
1438 
1439 	/*
1440 	 * Give the link layer a chance to accept/reject it, and also
1441 	 * find out which AF_LINK address this maps to, if it isn't one
1442 	 * already.
1443 	 */
1444 	if (ifp->if_resolvemulti) {
1445 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1446 		if (error) return error;
1447 	} else {
1448 		llsa = 0;
1449 	}
1450 
1451 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1452 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1453 	bcopy(sa, dupsa, sa->sa_len);
1454 
1455 	ifma->ifma_addr = dupsa;
1456 	ifma->ifma_lladdr = llsa;
1457 	ifma->ifma_ifp = ifp;
1458 	ifma->ifma_refcount = 1;
1459 	ifma->ifma_protospec = 0;
1460 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1461 
1462 	/*
1463 	 * Some network interfaces can scan the address list at
1464 	 * interrupt time; lock them out.
1465 	 */
1466 	s = splimp();
1467 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1468 	splx(s);
1469 	*retifma = ifma;
1470 
1471 	if (llsa != 0) {
1472 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1473 			if (equal(ifma->ifma_addr, llsa))
1474 				break;
1475 		}
1476 		if (ifma) {
1477 			ifma->ifma_refcount++;
1478 		} else {
1479 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1480 			       M_IFMADDR, M_WAITOK);
1481 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1482 			       M_IFMADDR, M_WAITOK);
1483 			bcopy(llsa, dupsa, llsa->sa_len);
1484 			ifma->ifma_addr = dupsa;
1485 			ifma->ifma_ifp = ifp;
1486 			ifma->ifma_refcount = 1;
1487 			s = splimp();
1488 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1489 			splx(s);
1490 		}
1491 	}
1492 	/*
1493 	 * We are certain we have added something, so call down to the
1494 	 * interface to let them know about it.
1495 	 */
1496 	s = splimp();
1497 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0);
1498 	splx(s);
1499 
1500 	return 0;
1501 }
1502 
1503 /*
1504  * Remove a reference to a multicast address on this interface.  Yell
1505  * if the request does not match an existing membership.
1506  */
1507 int
1508 if_delmulti(ifp, sa)
1509 	struct ifnet *ifp;
1510 	struct sockaddr *sa;
1511 {
1512 	struct ifmultiaddr *ifma;
1513 	int s;
1514 
1515 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1516 		if (equal(sa, ifma->ifma_addr))
1517 			break;
1518 	if (ifma == 0)
1519 		return ENOENT;
1520 
1521 	if (ifma->ifma_refcount > 1) {
1522 		ifma->ifma_refcount--;
1523 		return 0;
1524 	}
1525 
1526 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1527 	sa = ifma->ifma_lladdr;
1528 	s = splimp();
1529 	LIST_REMOVE(ifma, ifma_link);
1530 	/*
1531 	 * Make sure the interface driver is notified
1532 	 * in the case of a link layer mcast group being left.
1533 	 */
1534 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1535 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1536 	splx(s);
1537 	free(ifma->ifma_addr, M_IFMADDR);
1538 	free(ifma, M_IFMADDR);
1539 	if (sa == 0)
1540 		return 0;
1541 
1542 	/*
1543 	 * Now look for the link-layer address which corresponds to
1544 	 * this network address.  It had been squirreled away in
1545 	 * ifma->ifma_lladdr for this purpose (so we don't have
1546 	 * to call ifp->if_resolvemulti() again), and we saved that
1547 	 * value in sa above.  If some nasty deleted the
1548 	 * link-layer address out from underneath us, we can deal because
1549 	 * the address we stored was is not the same as the one which was
1550 	 * in the record for the link-layer address.  (So we don't complain
1551 	 * in that case.)
1552 	 */
1553 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1554 		if (equal(sa, ifma->ifma_addr))
1555 			break;
1556 	if (ifma == 0)
1557 		return 0;
1558 
1559 	if (ifma->ifma_refcount > 1) {
1560 		ifma->ifma_refcount--;
1561 		return 0;
1562 	}
1563 
1564 	s = splimp();
1565 	LIST_REMOVE(ifma, ifma_link);
1566 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1567 	splx(s);
1568 	free(ifma->ifma_addr, M_IFMADDR);
1569 	free(sa, M_IFMADDR);
1570 	free(ifma, M_IFMADDR);
1571 
1572 	return 0;
1573 }
1574 
1575 /*
1576  * Set the link layer address on an interface.
1577  *
1578  * At this time we only support certain types of interfaces,
1579  * and we don't allow the length of the address to change.
1580  */
1581 int
1582 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1583 {
1584 	struct sockaddr_dl *sdl;
1585 	struct ifaddr *ifa;
1586 	struct ifreq ifr;
1587 
1588 	ifa = ifnet_addrs[ifp->if_index - 1];
1589 	if (ifa == NULL)
1590 		return (EINVAL);
1591 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1592 	if (sdl == NULL)
1593 		return (EINVAL);
1594 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1595 		return (EINVAL);
1596 	switch (ifp->if_type) {
1597 	case IFT_ETHER:			/* these types use struct arpcom */
1598 	case IFT_FDDI:
1599 	case IFT_XETHER:
1600 	case IFT_ISO88025:
1601 	case IFT_L2VLAN:
1602 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1603 		/* FALLTHROUGH */
1604 	case IFT_ARCNET:
1605 		bcopy(lladdr, LLADDR(sdl), len);
1606 		break;
1607 	default:
1608 		return (ENODEV);
1609 	}
1610 	/*
1611 	 * If the interface is already up, we need
1612 	 * to re-init it in order to reprogram its
1613 	 * address filter.
1614 	 */
1615 	if ((ifp->if_flags & IFF_UP) != 0) {
1616 		ifp->if_flags &= ~IFF_UP;
1617 		ifr.ifr_flags = ifp->if_flags;
1618 		ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1619 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1620 		ifp->if_flags |= IFF_UP;
1621 		ifr.ifr_flags = ifp->if_flags;
1622 		ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1623 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1624 #ifdef INET
1625 		/*
1626 		 * Also send gratuitous ARPs to notify other nodes about
1627 		 * the address change.
1628 		 */
1629 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1630 			if (ifa->ifa_addr != NULL &&
1631 			    ifa->ifa_addr->sa_family == AF_INET)
1632 				arp_ifinit(ifp, ifa);
1633 		}
1634 #endif
1635 	}
1636 	return (0);
1637 }
1638 
1639 struct ifmultiaddr *
1640 ifmaof_ifpforaddr(sa, ifp)
1641 	struct sockaddr *sa;
1642 	struct ifnet *ifp;
1643 {
1644 	struct ifmultiaddr *ifma;
1645 
1646 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1647 		if (equal(ifma->ifma_addr, sa))
1648 			break;
1649 
1650 	return ifma;
1651 }
1652 
1653 int
1654 if_printf(struct ifnet *ifp, const char *fmt, ...)
1655 {
1656 	va_list ap;
1657 	int retval;
1658 
1659 	retval = printf("%s%d: ", ifp->if_name, ifp->if_unit);
1660 	va_start(ap, fmt);
1661 	retval += vprintf(fmt, ap);
1662 	va_end(ap);
1663 	return (retval);
1664 }
1665 
1666 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1667 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1668