xref: /freebsd/sys/net/if.c (revision 6990ffd8a95caaba6858ad44ff1b3157d1efba8f)
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  * $FreeBSD$
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 #include <sys/jail.h>
54 
55 #include <net/if.h>
56 #include <net/if_arp.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <net/if_var.h>
60 #include <net/radix.h>
61 #include <net/route.h>
62 
63 #if defined(INET) || defined(INET6)
64 /*XXX*/
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #ifdef INET6
68 #include <netinet6/in6_var.h>
69 #include <netinet6/in6_ifattach.h>
70 #endif
71 #endif
72 
73 static int	ifconf(u_long, caddr_t);
74 static void	if_grow(void);
75 static void	if_init(void *);
76 static void	if_check(void *);
77 static void	if_qflush(struct ifqueue *);
78 static void	if_slowtimo(void *);
79 static void	link_rtrequest(int, struct rtentry *, struct sockaddr *);
80 static int	if_rtdel(struct radix_node *, void *);
81 static struct	if_clone *if_clone_lookup(const char *, int *);
82 static int	if_clone_list(struct if_clonereq *);
83 #ifdef INET6
84 /*
85  * XXX: declare here to avoid to include many inet6 related files..
86  * should be more generalized?
87  */
88 extern void	nd6_setmtu __P((struct ifnet *));
89 #endif
90 
91 int	if_index = 0;
92 struct	ifindex_entry *ifindex_table = NULL;
93 int	ifqmaxlen = IFQ_MAXLEN;
94 struct	ifnethead ifnet;	/* depend on static init XXX */
95 int	if_cloners_count;
96 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
97 
98 static int	if_indexlim = 8;
99 
100 /*
101  * System initialization
102  */
103 SYSINIT(interfaces, SI_SUB_INIT_IF, SI_ORDER_FIRST, if_init, NULL)
104 SYSINIT(interface_check, SI_SUB_PROTO_IF, SI_ORDER_FIRST, if_check, NULL)
105 
106 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
107 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
108 
109 /*
110  * Network interface utility routines.
111  *
112  * Routines with ifa_ifwith* names take sockaddr *'s as
113  * parameters.
114  */
115 /* ARGSUSED*/
116 static void
117 if_init(dummy)
118 	void *dummy;
119 {
120 
121 	TAILQ_INIT(&ifnet);
122 	if_grow();				/* create initial table */
123 }
124 
125 static void
126 if_grow(void)
127 {
128 	u_int n;
129 	struct ifindex_entry *e;
130 
131 	if_indexlim <<= 1;
132 	n = if_indexlim * sizeof(*e);
133 	e = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
134 	if (ifindex_table != NULL) {
135 		memcpy((caddr_t)e, (caddr_t)ifindex_table, n/2);
136 		free((caddr_t)ifindex_table, M_IFADDR);
137 	}
138 	ifindex_table = e;
139 }
140 
141 /* ARGSUSED*/
142 static void
143 if_check(dummy)
144 	void *dummy;
145 {
146 	struct ifnet *ifp;
147 	int s;
148 
149 	s = splimp();
150 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
151 		if (ifp->if_snd.ifq_maxlen == 0) {
152 			printf("%s%d XXX: driver didn't set ifq_maxlen\n",
153 			    ifp->if_name, ifp->if_unit);
154 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
155 		}
156 		if (!mtx_initialized(&ifp->if_snd.ifq_mtx)) {
157 			printf("%s%d XXX: driver didn't initialize queue mtx\n",
158 			    ifp->if_name, ifp->if_unit);
159 			mtx_init(&ifp->if_snd.ifq_mtx, "unknown", MTX_DEF);
160 		}
161 	}
162 	splx(s);
163 	if_slowtimo(0);
164 }
165 
166 /*
167  * Attach an interface to the
168  * list of "active" interfaces.
169  */
170 void
171 if_attach(ifp)
172 	struct ifnet *ifp;
173 {
174 	unsigned socksize, ifasize;
175 	int namelen, masklen;
176 	char workbuf[64];
177 	register struct sockaddr_dl *sdl;
178 	register struct ifaddr *ifa;
179 
180 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
181 	ifp->if_index = ++if_index;
182 	/*
183 	 * XXX -
184 	 * The old code would work if the interface passed a pre-existing
185 	 * chain of ifaddrs to this code.  We don't trust our callers to
186 	 * properly initialize the tailq, however, so we no longer allow
187 	 * this unlikely case.
188 	 */
189 	TAILQ_INIT(&ifp->if_addrhead);
190 	TAILQ_INIT(&ifp->if_prefixhead);
191 	TAILQ_INIT(&ifp->if_multiaddrs);
192 	getmicrotime(&ifp->if_lastchange);
193 	if (if_index >= if_indexlim)
194 		if_grow();
195 
196 	ifnet_byindex(if_index) = ifp;
197 	mtx_init(&ifp->if_snd.ifq_mtx, ifp->if_name, MTX_DEF);
198 
199 	/*
200 	 * create a Link Level name for this device
201 	 */
202 	namelen = snprintf(workbuf, sizeof(workbuf),
203 	    "%s%d", ifp->if_name, ifp->if_unit);
204 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
205 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
206 	socksize = masklen + ifp->if_addrlen;
207 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
208 	if (socksize < sizeof(*sdl))
209 		socksize = sizeof(*sdl);
210 	socksize = ROUNDUP(socksize);
211 	ifasize = sizeof(*ifa) + 2 * socksize;
212 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
213 	if (ifa) {
214 		sdl = (struct sockaddr_dl *)(ifa + 1);
215 		sdl->sdl_len = socksize;
216 		sdl->sdl_family = AF_LINK;
217 		bcopy(workbuf, sdl->sdl_data, namelen);
218 		sdl->sdl_nlen = namelen;
219 		sdl->sdl_index = ifp->if_index;
220 		sdl->sdl_type = ifp->if_type;
221 		ifaddr_byindex(if_index) = ifa;
222 		ifa->ifa_ifp = ifp;
223 		ifa->ifa_rtrequest = link_rtrequest;
224 		ifa->ifa_addr = (struct sockaddr *)sdl;
225 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
226 		ifa->ifa_netmask = (struct sockaddr *)sdl;
227 		sdl->sdl_len = masklen;
228 		while (namelen != 0)
229 			sdl->sdl_data[--namelen] = 0xff;
230 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
231 	}
232 }
233 
234 /*
235  * Detach an interface, removing it from the
236  * list of "active" interfaces.
237  */
238 void
239 if_detach(ifp)
240 	struct ifnet *ifp;
241 {
242 	struct ifaddr *ifa;
243 	struct radix_node_head	*rnh;
244 	int s;
245 	int i;
246 
247 	/*
248 	 * Remove routes and flush queues.
249 	 */
250 	s = splnet();
251 	if_down(ifp);
252 
253 	/*
254 	 * Remove address from ifindex_table[] and maybe decrement if_index.
255 	 * Clean up all addresses.
256 	 */
257 	ifaddr_byindex(ifp->if_index) = NULL;
258 
259 	while (if_index > 0 && ifaddr_byindex(if_index) == NULL)
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 	TAILQ_REMOVE(&ifnet, ifp, if_link);
312 	mtx_destroy(&ifp->if_snd.ifq_mtx);
313 	splx(s);
314 }
315 
316 /*
317  * Delete Routes for a Network Interface
318  *
319  * Called for each routing entry via the rnh->rnh_walktree() call above
320  * to delete all route entries referencing a detaching network interface.
321  *
322  * Arguments:
323  *	rn	pointer to node in the routing table
324  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
325  *
326  * Returns:
327  *	0	successful
328  *	errno	failed - reason indicated
329  *
330  */
331 static int
332 if_rtdel(rn, arg)
333 	struct radix_node	*rn;
334 	void			*arg;
335 {
336 	struct rtentry	*rt = (struct rtentry *)rn;
337 	struct ifnet	*ifp = arg;
338 	int		err;
339 
340 	if (rt->rt_ifp == ifp) {
341 
342 		/*
343 		 * Protect (sorta) against walktree recursion problems
344 		 * with cloned routes
345 		 */
346 		if ((rt->rt_flags & RTF_UP) == 0)
347 			return (0);
348 
349 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
350 				rt_mask(rt), rt->rt_flags,
351 				(struct rtentry **) NULL);
352 		if (err) {
353 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
354 		}
355 	}
356 
357 	return (0);
358 }
359 
360 /*
361  * Create a clone network interface.
362  */
363 int
364 if_clone_create(name, len)
365 	char *name;
366 	int len;
367 {
368 	struct if_clone *ifc;
369 	char *dp;
370 	int wildcard;
371 	int unit;
372 	int err;
373 
374 	ifc = if_clone_lookup(name, &unit);
375 	if (ifc == NULL)
376 		return (EINVAL);
377 
378 	if (ifunit(name) != NULL)
379 		return (EEXIST);
380 
381 	wildcard = (unit < 0);
382 
383 	err = (*ifc->ifc_create)(ifc, &unit);
384 	if (err != 0)
385 		return (err);
386 
387 	/* In the wildcard case, we need to update the name. */
388 	if (wildcard) {
389 		for (dp = name; *dp != '\0'; dp++);
390 		if (snprintf(dp, len - (dp-name), "%d", unit) >
391 		    len - (dp-name) - 1) {
392 			/*
393 			 * This can only be a programmer error and
394 			 * there's no straightforward way to recover if
395 			 * it happens.
396 			 */
397 			panic("if_clone_create(): interface name too long");
398 		}
399 
400 	}
401 
402 	return (0);
403 }
404 
405 /*
406  * Destroy a clone network interface.
407  */
408 int
409 if_clone_destroy(name)
410 	const char *name;
411 {
412 	struct if_clone *ifc;
413 	struct ifnet *ifp;
414 
415 	ifc = if_clone_lookup(name, NULL);
416 	if (ifc == NULL)
417 		return (EINVAL);
418 
419 	ifp = ifunit(name);
420 	if (ifp == NULL)
421 		return (ENXIO);
422 
423 	if (ifc->ifc_destroy == NULL)
424 		return (EOPNOTSUPP);
425 
426 	(*ifc->ifc_destroy)(ifp);
427 	return (0);
428 }
429 
430 /*
431  * Look up a network interface cloner.
432  */
433 static struct if_clone *
434 if_clone_lookup(name, unitp)
435 	const char *name;
436 	int *unitp;
437 {
438 	struct if_clone *ifc;
439 	const char *cp;
440 	int i;
441 
442 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
443 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
444 			if (ifc->ifc_name[i] != *cp)
445 				goto next_ifc;
446 		}
447 		goto found_name;
448  next_ifc:
449 		ifc = LIST_NEXT(ifc, ifc_list);
450 	}
451 
452 	/* No match. */
453 	return ((struct if_clone *)NULL);
454 
455  found_name:
456 	if (*cp == '\0') {
457 		i = -1;
458 	} else {
459 		for (i = 0; *cp != '\0'; cp++) {
460 			if (*cp < '0' || *cp > '9') {
461 				/* Bogus unit number. */
462 				return (NULL);
463 			}
464 			i = (i * 10) + (*cp - '0');
465 		}
466 	}
467 
468 	if (unitp != NULL)
469 		*unitp = i;
470 	return (ifc);
471 }
472 
473 /*
474  * Register a network interface cloner.
475  */
476 void
477 if_clone_attach(ifc)
478 	struct if_clone *ifc;
479 {
480 
481 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
482 	if_cloners_count++;
483 }
484 
485 /*
486  * Unregister a network interface cloner.
487  */
488 void
489 if_clone_detach(ifc)
490 	struct if_clone *ifc;
491 {
492 
493 	LIST_REMOVE(ifc, ifc_list);
494 	if_cloners_count--;
495 }
496 
497 /*
498  * Provide list of interface cloners to userspace.
499  */
500 static int
501 if_clone_list(ifcr)
502 	struct if_clonereq *ifcr;
503 {
504 	char outbuf[IFNAMSIZ], *dst;
505 	struct if_clone *ifc;
506 	int count, error = 0;
507 
508 	ifcr->ifcr_total = if_cloners_count;
509 	if ((dst = ifcr->ifcr_buffer) == NULL) {
510 		/* Just asking how many there are. */
511 		return (0);
512 	}
513 
514 	if (ifcr->ifcr_count < 0)
515 		return (EINVAL);
516 
517 	count = (if_cloners_count < ifcr->ifcr_count) ?
518 	    if_cloners_count : ifcr->ifcr_count;
519 
520 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
521 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
522 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
523 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
524 		error = copyout(outbuf, dst, IFNAMSIZ);
525 		if (error)
526 			break;
527 	}
528 
529 	return (error);
530 }
531 
532 /*
533  * Locate an interface based on a complete address.
534  */
535 /*ARGSUSED*/
536 struct ifaddr *
537 ifa_ifwithaddr(addr)
538 	struct sockaddr *addr;
539 {
540 	struct ifnet *ifp;
541 	struct ifaddr *ifa;
542 
543 #define	equal(a1, a2) \
544   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
545 	TAILQ_FOREACH(ifp, &ifnet, if_link)
546 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
547 			if (ifa->ifa_addr->sa_family != addr->sa_family)
548 				continue;
549 			if (equal(addr, ifa->ifa_addr))
550 				goto done;
551 			/* IP6 doesn't have broadcast */
552 			if ((ifp->if_flags & IFF_BROADCAST) &&
553 			    ifa->ifa_broadaddr &&
554 			    ifa->ifa_broadaddr->sa_len != 0 &&
555 			    equal(ifa->ifa_broadaddr, addr))
556 				goto done;
557 		}
558 	ifa = NULL;
559 done:
560 	return (ifa);
561 }
562 
563 /*
564  * Locate the point to point interface with a given destination address.
565  */
566 /*ARGSUSED*/
567 struct ifaddr *
568 ifa_ifwithdstaddr(addr)
569 	struct sockaddr *addr;
570 {
571 	struct ifnet *ifp;
572 	struct ifaddr *ifa;
573 
574 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
575 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
576 			continue;
577 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
578 			if (ifa->ifa_addr->sa_family != addr->sa_family)
579 				continue;
580 			if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
581 				goto done;
582 		}
583 	}
584 	ifa = NULL;
585 done:
586 	return (ifa);
587 }
588 
589 /*
590  * Find an interface on a specific network.  If many, choice
591  * is most specific found.
592  */
593 struct ifaddr *
594 ifa_ifwithnet(addr)
595 	struct sockaddr *addr;
596 {
597 	register struct ifnet *ifp;
598 	register struct ifaddr *ifa;
599 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
600 	u_int af = addr->sa_family;
601 	char *addr_data = addr->sa_data, *cplim;
602 
603 	/*
604 	 * AF_LINK addresses can be looked up directly by their index number,
605 	 * so do that if we can.
606 	 */
607 	if (af == AF_LINK) {
608 	    register struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
609 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
610 		return (ifaddr_byindex(sdl->sdl_index));
611 	}
612 
613 	/*
614 	 * Scan though each interface, looking for ones that have
615 	 * addresses in this address family.
616 	 */
617 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
618 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
619 			register char *cp, *cp2, *cp3;
620 
621 			if (ifa->ifa_addr->sa_family != af)
622 next:				continue;
623 			if (
624 #ifdef INET6 /* XXX: for maching gif tunnel dst as routing entry gateway */
625 			    addr->sa_family != AF_INET6 &&
626 #endif
627 			    ifp->if_flags & IFF_POINTOPOINT) {
628 				/*
629 				 * This is a bit broken as it doesn't
630 				 * take into account that the remote end may
631 				 * be a single node in the network we are
632 				 * looking for.
633 				 * The trouble is that we don't know the
634 				 * netmask for the remote end.
635 				 */
636 				if (ifa->ifa_dstaddr != 0
637 				    && equal(addr, ifa->ifa_dstaddr))
638 					goto done;
639 			} else {
640 				/*
641 				 * if we have a special address handler,
642 				 * then use it instead of the generic one.
643 				 */
644 	          		if (ifa->ifa_claim_addr) {
645 					if ((*ifa->ifa_claim_addr)(ifa, addr))
646 						goto done;
647 					continue;
648 				}
649 
650 				/*
651 				 * Scan all the bits in the ifa's address.
652 				 * If a bit dissagrees with what we are
653 				 * looking for, mask it with the netmask
654 				 * to see if it really matters.
655 				 * (A byte at a time)
656 				 */
657 				if (ifa->ifa_netmask == 0)
658 					continue;
659 				cp = addr_data;
660 				cp2 = ifa->ifa_addr->sa_data;
661 				cp3 = ifa->ifa_netmask->sa_data;
662 				cplim = ifa->ifa_netmask->sa_len
663 					+ (char *)ifa->ifa_netmask;
664 				while (cp3 < cplim)
665 					if ((*cp++ ^ *cp2++) & *cp3++)
666 						goto next; /* next address! */
667 				/*
668 				 * If the netmask of what we just found
669 				 * is more specific than what we had before
670 				 * (if we had one) then remember the new one
671 				 * before continuing to search
672 				 * for an even better one.
673 				 */
674 				if (ifa_maybe == 0 ||
675 				    rn_refines((caddr_t)ifa->ifa_netmask,
676 				    (caddr_t)ifa_maybe->ifa_netmask))
677 					ifa_maybe = ifa;
678 			}
679 		}
680 	}
681 	ifa = ifa_maybe;
682 done:
683 	return (ifa);
684 }
685 
686 /*
687  * Find an interface address specific to an interface best matching
688  * a given address.
689  */
690 struct ifaddr *
691 ifaof_ifpforaddr(addr, ifp)
692 	struct sockaddr *addr;
693 	register struct ifnet *ifp;
694 {
695 	register struct ifaddr *ifa;
696 	register char *cp, *cp2, *cp3;
697 	register char *cplim;
698 	struct ifaddr *ifa_maybe = 0;
699 	u_int af = addr->sa_family;
700 
701 	if (af >= AF_MAX)
702 		return (0);
703 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
704 		if (ifa->ifa_addr->sa_family != af)
705 			continue;
706 		if (ifa_maybe == 0)
707 			ifa_maybe = ifa;
708 		if (ifa->ifa_netmask == 0) {
709 			if (equal(addr, ifa->ifa_addr) ||
710 			    (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
711 				goto done;
712 			continue;
713 		}
714 		if (ifp->if_flags & IFF_POINTOPOINT) {
715 			if (equal(addr, ifa->ifa_dstaddr))
716 				goto done;
717 		} else {
718 			cp = addr->sa_data;
719 			cp2 = ifa->ifa_addr->sa_data;
720 			cp3 = ifa->ifa_netmask->sa_data;
721 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
722 			for (; cp3 < cplim; cp3++)
723 				if ((*cp++ ^ *cp2++) & *cp3)
724 					break;
725 			if (cp3 == cplim)
726 				goto done;
727 		}
728 	}
729 	ifa = ifa_maybe;
730 done:
731 	return (ifa);
732 }
733 
734 #include <net/route.h>
735 
736 /*
737  * Default action when installing a route with a Link Level gateway.
738  * Lookup an appropriate real ifa to point to.
739  * This should be moved to /sys/net/link.c eventually.
740  */
741 static void
742 link_rtrequest(cmd, rt, sa)
743 	int cmd;
744 	register struct rtentry *rt;
745 	struct sockaddr *sa;
746 {
747 	register struct ifaddr *ifa;
748 	struct sockaddr *dst;
749 	struct ifnet *ifp;
750 
751 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
752 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
753 		return;
754 	ifa = ifaof_ifpforaddr(dst, ifp);
755 	if (ifa) {
756 		IFAFREE(rt->rt_ifa);
757 		rt->rt_ifa = ifa;
758 		ifa->ifa_refcnt++;
759 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
760 			ifa->ifa_rtrequest(cmd, rt, sa);
761 	}
762 }
763 
764 /*
765  * Mark an interface down and notify protocols of
766  * the transition.
767  * NOTE: must be called at splnet or eqivalent.
768  */
769 void
770 if_unroute(ifp, flag, fam)
771 	register struct ifnet *ifp;
772 	int flag, fam;
773 {
774 	register struct ifaddr *ifa;
775 
776 	ifp->if_flags &= ~flag;
777 	getmicrotime(&ifp->if_lastchange);
778 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
779 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
780 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
781 	if_qflush(&ifp->if_snd);
782 	rt_ifmsg(ifp);
783 }
784 
785 /*
786  * Mark an interface up and notify protocols of
787  * the transition.
788  * NOTE: must be called at splnet or eqivalent.
789  */
790 void
791 if_route(ifp, flag, fam)
792 	register struct ifnet *ifp;
793 	int flag, fam;
794 {
795 	register struct ifaddr *ifa;
796 
797 	ifp->if_flags |= flag;
798 	getmicrotime(&ifp->if_lastchange);
799 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
800 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
801 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
802 	rt_ifmsg(ifp);
803 #ifdef INET6
804 	in6_if_up(ifp);
805 #endif
806 }
807 
808 /*
809  * Mark an interface down and notify protocols of
810  * the transition.
811  * NOTE: must be called at splnet or eqivalent.
812  */
813 void
814 if_down(ifp)
815 	register struct ifnet *ifp;
816 {
817 
818 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
819 }
820 
821 /*
822  * Mark an interface up and notify protocols of
823  * the transition.
824  * NOTE: must be called at splnet or eqivalent.
825  */
826 void
827 if_up(ifp)
828 	register struct ifnet *ifp;
829 {
830 
831 	if_route(ifp, IFF_UP, AF_UNSPEC);
832 }
833 
834 /*
835  * Flush an interface queue.
836  */
837 static void
838 if_qflush(ifq)
839 	register struct ifqueue *ifq;
840 {
841 	register struct mbuf *m, *n;
842 
843 	n = ifq->ifq_head;
844 	while ((m = n) != 0) {
845 		n = m->m_act;
846 		m_freem(m);
847 	}
848 	ifq->ifq_head = 0;
849 	ifq->ifq_tail = 0;
850 	ifq->ifq_len = 0;
851 }
852 
853 /*
854  * Handle interface watchdog timer routines.  Called
855  * from softclock, we decrement timers (if set) and
856  * call the appropriate interface routine on expiration.
857  */
858 static void
859 if_slowtimo(arg)
860 	void *arg;
861 {
862 	register struct ifnet *ifp;
863 	int s = splimp();
864 
865 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
866 		if (ifp->if_timer == 0 || --ifp->if_timer)
867 			continue;
868 		if (ifp->if_watchdog)
869 			(*ifp->if_watchdog)(ifp);
870 	}
871 	splx(s);
872 	timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
873 }
874 
875 /*
876  * Map interface name to
877  * interface structure pointer.
878  */
879 struct ifnet *
880 ifunit(const char *name)
881 {
882 	char namebuf[IFNAMSIZ + 1];
883 	const char *cp;
884 	struct ifnet *ifp;
885 	int unit;
886 	unsigned len, m;
887 	char c;
888 
889 	len = strlen(name);
890 	if (len < 2 || len > IFNAMSIZ)
891 		return NULL;
892 	cp = name + len - 1;
893 	c = *cp;
894 	if (c < '0' || c > '9')
895 		return NULL;		/* trailing garbage */
896 	unit = 0;
897 	m = 1;
898 	do {
899 		if (cp == name)
900 			return NULL;	/* no interface name */
901 		unit += (c - '0') * m;
902 		if (unit > 1000000)
903 			return NULL;	/* number is unreasonable */
904 		m *= 10;
905 		c = *--cp;
906 	} while (c >= '0' && c <= '9');
907 	len = cp - name + 1;
908 	bcopy(name, namebuf, len);
909 	namebuf[len] = '\0';
910 	/*
911 	 * Now search all the interfaces for this name/number
912 	 */
913 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
914 		if (strcmp(ifp->if_name, namebuf))
915 			continue;
916 		if (unit == ifp->if_unit)
917 			break;
918 	}
919 	return (ifp);
920 }
921 
922 
923 /*
924  * Map interface name in a sockaddr_dl to
925  * interface structure pointer.
926  */
927 struct ifnet *
928 if_withname(sa)
929 	struct sockaddr *sa;
930 {
931 	char ifname[IFNAMSIZ+1];
932 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
933 
934 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
935 	     (sdl->sdl_nlen > IFNAMSIZ) )
936 		return NULL;
937 
938 	/*
939 	 * ifunit wants a null-terminated name.  It may not be null-terminated
940 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
941 	 * and there might not be room to put the trailing null anyway, so we
942 	 * make a local copy that we know we can null terminate safely.
943 	 */
944 
945 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
946 	ifname[sdl->sdl_nlen] = '\0';
947 	return ifunit(ifname);
948 }
949 
950 
951 /*
952  * Interface ioctls.
953  */
954 int
955 ifioctl(so, cmd, data, td)
956 	struct socket *so;
957 	u_long cmd;
958 	caddr_t data;
959 	struct thread *td;
960 {
961 	register struct ifnet *ifp;
962 	register struct ifreq *ifr;
963 	struct ifstat *ifs;
964 	int error;
965 	short oif_flags;
966 
967 	switch (cmd) {
968 
969 	case SIOCGIFCONF:
970 	case OSIOCGIFCONF:
971 		return (ifconf(cmd, data));
972 	}
973 	ifr = (struct ifreq *)data;
974 
975 	switch (cmd) {
976 	case SIOCIFCREATE:
977 	case SIOCIFDESTROY:
978 		if ((error = suser_td(td)) != 0)
979 			return (error);
980 		return ((cmd == SIOCIFCREATE) ?
981 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
982 			if_clone_destroy(ifr->ifr_name));
983 
984 	case SIOCIFGCLONERS:
985 		return (if_clone_list((struct if_clonereq *)data));
986 	}
987 
988 	ifp = ifunit(ifr->ifr_name);
989 	if (ifp == 0)
990 		return (ENXIO);
991 	switch (cmd) {
992 
993 	case SIOCGIFFLAGS:
994 		ifr->ifr_flags = ifp->if_flags;
995 		break;
996 
997 	case SIOCGIFMETRIC:
998 		ifr->ifr_metric = ifp->if_metric;
999 		break;
1000 
1001 	case SIOCGIFMTU:
1002 		ifr->ifr_mtu = ifp->if_mtu;
1003 		break;
1004 
1005 	case SIOCGIFPHYS:
1006 		ifr->ifr_phys = ifp->if_physical;
1007 		break;
1008 
1009 	case SIOCSIFFLAGS:
1010 		error = suser_td(td);
1011 		if (error)
1012 			return (error);
1013 		ifr->ifr_prevflags = ifp->if_flags;
1014 		if (ifp->if_flags & IFF_SMART) {
1015 			/* Smart drivers twiddle their own routes */
1016 		} else if (ifp->if_flags & IFF_UP &&
1017 		    (ifr->ifr_flags & IFF_UP) == 0) {
1018 			int s = splimp();
1019 			if_down(ifp);
1020 			splx(s);
1021 		} else if (ifr->ifr_flags & IFF_UP &&
1022 		    (ifp->if_flags & IFF_UP) == 0) {
1023 			int s = splimp();
1024 			if_up(ifp);
1025 			splx(s);
1026 		}
1027 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1028 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
1029 		if (ifp->if_ioctl)
1030 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1031 		getmicrotime(&ifp->if_lastchange);
1032 		break;
1033 
1034 	case SIOCSIFMETRIC:
1035 		error = suser_td(td);
1036 		if (error)
1037 			return (error);
1038 		ifp->if_metric = ifr->ifr_metric;
1039 		getmicrotime(&ifp->if_lastchange);
1040 		break;
1041 
1042 	case SIOCSIFPHYS:
1043 		error = suser_td(td);
1044 		if (error)
1045 			return error;
1046 		if (!ifp->if_ioctl)
1047 		        return EOPNOTSUPP;
1048 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1049 		if (error == 0)
1050 			getmicrotime(&ifp->if_lastchange);
1051 		return(error);
1052 
1053 	case SIOCSIFMTU:
1054 	{
1055 		u_long oldmtu = ifp->if_mtu;
1056 
1057 		error = suser_td(td);
1058 		if (error)
1059 			return (error);
1060 		if (ifp->if_ioctl == NULL)
1061 			return (EOPNOTSUPP);
1062 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1063 			return (EINVAL);
1064 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1065 		if (error == 0) {
1066 			getmicrotime(&ifp->if_lastchange);
1067 			rt_ifmsg(ifp);
1068 		}
1069 		/*
1070 		 * If the link MTU changed, do network layer specific procedure.
1071 		 */
1072 		if (ifp->if_mtu != oldmtu) {
1073 #ifdef INET6
1074 			nd6_setmtu(ifp);
1075 #endif
1076 		}
1077 		return (error);
1078 	}
1079 
1080 	case SIOCADDMULTI:
1081 	case SIOCDELMULTI:
1082 		error = suser_td(td);
1083 		if (error)
1084 			return (error);
1085 
1086 		/* Don't allow group membership on non-multicast interfaces. */
1087 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1088 			return EOPNOTSUPP;
1089 
1090 		/* Don't let users screw up protocols' entries. */
1091 		if (ifr->ifr_addr.sa_family != AF_LINK)
1092 			return EINVAL;
1093 
1094 		if (cmd == SIOCADDMULTI) {
1095 			struct ifmultiaddr *ifma;
1096 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1097 		} else {
1098 			error = if_delmulti(ifp, &ifr->ifr_addr);
1099 		}
1100 		if (error == 0)
1101 			getmicrotime(&ifp->if_lastchange);
1102 		return error;
1103 
1104 	case SIOCSIFPHYADDR:
1105 	case SIOCDIFPHYADDR:
1106 #ifdef INET6
1107 	case SIOCSIFPHYADDR_IN6:
1108 #endif
1109 	case SIOCSLIFPHYADDR:
1110         case SIOCSIFMEDIA:
1111 	case SIOCSIFGENERIC:
1112 		error = suser_td(td);
1113 		if (error)
1114 			return (error);
1115 		if (ifp->if_ioctl == 0)
1116 			return (EOPNOTSUPP);
1117 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1118 		if (error == 0)
1119 			getmicrotime(&ifp->if_lastchange);
1120 		return error;
1121 
1122 	case SIOCGIFSTATUS:
1123 		ifs = (struct ifstat *)data;
1124 		ifs->ascii[0] = '\0';
1125 
1126 	case SIOCGIFPSRCADDR:
1127 	case SIOCGIFPDSTADDR:
1128 	case SIOCGLIFPHYADDR:
1129 	case SIOCGIFMEDIA:
1130 	case SIOCGIFGENERIC:
1131 		if (ifp->if_ioctl == 0)
1132 			return (EOPNOTSUPP);
1133 		return ((*ifp->if_ioctl)(ifp, cmd, data));
1134 
1135 	case SIOCSIFLLADDR:
1136 		error = suser_td(td);
1137 		if (error)
1138 			return (error);
1139 		return if_setlladdr(ifp,
1140 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1141 
1142 	default:
1143 		oif_flags = ifp->if_flags;
1144 		if (so->so_proto == 0)
1145 			return (EOPNOTSUPP);
1146 #ifndef COMPAT_43
1147 		error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1148 								 data,
1149 								 ifp, td));
1150 #else
1151 	    {
1152 		int ocmd = cmd;
1153 
1154 		switch (cmd) {
1155 
1156 		case SIOCSIFDSTADDR:
1157 		case SIOCSIFADDR:
1158 		case SIOCSIFBRDADDR:
1159 		case SIOCSIFNETMASK:
1160 #if BYTE_ORDER != BIG_ENDIAN
1161 			if (ifr->ifr_addr.sa_family == 0 &&
1162 			    ifr->ifr_addr.sa_len < 16) {
1163 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1164 				ifr->ifr_addr.sa_len = 16;
1165 			}
1166 #else
1167 			if (ifr->ifr_addr.sa_len == 0)
1168 				ifr->ifr_addr.sa_len = 16;
1169 #endif
1170 			break;
1171 
1172 		case OSIOCGIFADDR:
1173 			cmd = SIOCGIFADDR;
1174 			break;
1175 
1176 		case OSIOCGIFDSTADDR:
1177 			cmd = SIOCGIFDSTADDR;
1178 			break;
1179 
1180 		case OSIOCGIFBRDADDR:
1181 			cmd = SIOCGIFBRDADDR;
1182 			break;
1183 
1184 		case OSIOCGIFNETMASK:
1185 			cmd = SIOCGIFNETMASK;
1186 		}
1187 		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
1188 								   cmd,
1189 								   data,
1190 								   ifp, td));
1191 		switch (ocmd) {
1192 
1193 		case OSIOCGIFADDR:
1194 		case OSIOCGIFDSTADDR:
1195 		case OSIOCGIFBRDADDR:
1196 		case OSIOCGIFNETMASK:
1197 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1198 
1199 		}
1200 	    }
1201 #endif /* COMPAT_43 */
1202 
1203 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1204 #ifdef INET6
1205 			DELAY(100);/* XXX: temporal workaround for fxp issue*/
1206 			if (ifp->if_flags & IFF_UP) {
1207 				int s = splimp();
1208 				in6_if_up(ifp);
1209 				splx(s);
1210 			}
1211 #endif
1212 		}
1213 		return (error);
1214 
1215 	}
1216 	return (0);
1217 }
1218 
1219 /*
1220  * Set/clear promiscuous mode on interface ifp based on the truth value
1221  * of pswitch.  The calls are reference counted so that only the first
1222  * "on" request actually has an effect, as does the final "off" request.
1223  * Results are undefined if the "off" and "on" requests are not matched.
1224  */
1225 int
1226 ifpromisc(ifp, pswitch)
1227 	struct ifnet *ifp;
1228 	int pswitch;
1229 {
1230 	struct ifreq ifr;
1231 	int error;
1232 	int oldflags, oldpcount;
1233 
1234 	oldpcount = ifp->if_pcount;
1235 	oldflags = ifp->if_flags;
1236 	if (pswitch) {
1237 		/*
1238 		 * If the device is not configured up, we cannot put it in
1239 		 * promiscuous mode.
1240 		 */
1241 		if ((ifp->if_flags & IFF_UP) == 0)
1242 			return (ENETDOWN);
1243 		if (ifp->if_pcount++ != 0)
1244 			return (0);
1245 		ifp->if_flags |= IFF_PROMISC;
1246 	} else {
1247 		if (--ifp->if_pcount > 0)
1248 			return (0);
1249 		ifp->if_flags &= ~IFF_PROMISC;
1250 	}
1251 	ifr.ifr_flags = ifp->if_flags;
1252 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1253 	if (error == 0) {
1254 		log(LOG_INFO, "%s%d: promiscuous mode %s\n",
1255 		    ifp->if_name, ifp->if_unit,
1256 		    (ifp->if_flags & IFF_PROMISC) ? "enabled" : "disabled");
1257 		rt_ifmsg(ifp);
1258 	} else {
1259 		ifp->if_pcount = oldpcount;
1260 		ifp->if_flags = oldflags;
1261 	}
1262 	return error;
1263 }
1264 
1265 /*
1266  * Return interface configuration
1267  * of system.  List may be used
1268  * in later ioctl's (above) to get
1269  * other information.
1270  */
1271 /*ARGSUSED*/
1272 static int
1273 ifconf(cmd, data)
1274 	u_long cmd;
1275 	caddr_t data;
1276 {
1277 	struct ifconf *ifc = (struct ifconf *)data;
1278 	struct ifnet *ifp;
1279 	struct ifaddr *ifa;
1280 	struct ifreq ifr, *ifrp;
1281 	int space = ifc->ifc_len, error = 0;
1282 
1283 	ifrp = ifc->ifc_req;
1284 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1285 		char workbuf[64];
1286 		int ifnlen, addrs;
1287 
1288 		if (space < sizeof(ifr))
1289 			break;
1290 		ifnlen = snprintf(workbuf, sizeof(workbuf),
1291 		    "%s%d", ifp->if_name, ifp->if_unit);
1292 		if(ifnlen + 1 > sizeof ifr.ifr_name) {
1293 			error = ENAMETOOLONG;
1294 			break;
1295 		} else {
1296 			strcpy(ifr.ifr_name, workbuf);
1297 		}
1298 
1299 		addrs = 0;
1300 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1301 			struct sockaddr *sa = ifa->ifa_addr;
1302 
1303 			if (space < sizeof(ifr))
1304 				break;
1305 			if (jailed(curproc->p_ucred) &&
1306 			    prison_if(curproc->p_ucred, sa))
1307 				continue;
1308 			addrs++;
1309 #ifdef COMPAT_43
1310 			if (cmd == OSIOCGIFCONF) {
1311 				struct osockaddr *osa =
1312 					 (struct osockaddr *)&ifr.ifr_addr;
1313 				ifr.ifr_addr = *sa;
1314 				osa->sa_family = sa->sa_family;
1315 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1316 						sizeof (ifr));
1317 				ifrp++;
1318 			} else
1319 #endif
1320 			if (sa->sa_len <= sizeof(*sa)) {
1321 				ifr.ifr_addr = *sa;
1322 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1323 						sizeof (ifr));
1324 				ifrp++;
1325 			} else {
1326 				if (space < sizeof (ifr) + sa->sa_len -
1327 					    sizeof(*sa))
1328 					break;
1329 				space -= sa->sa_len - sizeof(*sa);
1330 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1331 						sizeof (ifr.ifr_name));
1332 				if (error == 0)
1333 				    error = copyout((caddr_t)sa,
1334 				      (caddr_t)&ifrp->ifr_addr, sa->sa_len);
1335 				ifrp = (struct ifreq *)
1336 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1337 			}
1338 			if (error)
1339 				break;
1340 			space -= sizeof (ifr);
1341 		}
1342 		if (error)
1343 			break;
1344 		if (!addrs) {
1345 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1346 			error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1347 			    sizeof (ifr));
1348 			if (error)
1349 				break;
1350 			space -= sizeof (ifr);
1351 			ifrp++;
1352 		}
1353 	}
1354 	ifc->ifc_len -= space;
1355 	return (error);
1356 }
1357 
1358 /*
1359  * Just like if_promisc(), but for all-multicast-reception mode.
1360  */
1361 int
1362 if_allmulti(ifp, onswitch)
1363 	struct ifnet *ifp;
1364 	int onswitch;
1365 {
1366 	int error = 0;
1367 	int s = splimp();
1368 
1369 	if (onswitch) {
1370 		if (ifp->if_amcount++ == 0) {
1371 			ifp->if_flags |= IFF_ALLMULTI;
1372 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, 0);
1373 		}
1374 	} else {
1375 		if (ifp->if_amcount > 1) {
1376 			ifp->if_amcount--;
1377 		} else {
1378 			ifp->if_amcount = 0;
1379 			ifp->if_flags &= ~IFF_ALLMULTI;
1380 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, 0);
1381 		}
1382 	}
1383 	splx(s);
1384 
1385 	if (error == 0)
1386 		rt_ifmsg(ifp);
1387 	return error;
1388 }
1389 
1390 /*
1391  * Add a multicast listenership to the interface in question.
1392  * The link layer provides a routine which converts
1393  */
1394 int
1395 if_addmulti(ifp, sa, retifma)
1396 	struct ifnet *ifp;	/* interface to manipulate */
1397 	struct sockaddr *sa;	/* address to add */
1398 	struct ifmultiaddr **retifma;
1399 {
1400 	struct sockaddr *llsa, *dupsa;
1401 	int error, s;
1402 	struct ifmultiaddr *ifma;
1403 
1404 	/*
1405 	 * If the matching multicast address already exists
1406 	 * then don't add a new one, just add a reference
1407 	 */
1408 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1409 		if (equal(sa, ifma->ifma_addr)) {
1410 			ifma->ifma_refcount++;
1411 			if (retifma)
1412 				*retifma = ifma;
1413 			return 0;
1414 		}
1415 	}
1416 
1417 	/*
1418 	 * Give the link layer a chance to accept/reject it, and also
1419 	 * find out which AF_LINK address this maps to, if it isn't one
1420 	 * already.
1421 	 */
1422 	if (ifp->if_resolvemulti) {
1423 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1424 		if (error) return error;
1425 	} else {
1426 		llsa = 0;
1427 	}
1428 
1429 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1430 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1431 	bcopy(sa, dupsa, sa->sa_len);
1432 
1433 	ifma->ifma_addr = dupsa;
1434 	ifma->ifma_lladdr = llsa;
1435 	ifma->ifma_ifp = ifp;
1436 	ifma->ifma_refcount = 1;
1437 	ifma->ifma_protospec = 0;
1438 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1439 
1440 	/*
1441 	 * Some network interfaces can scan the address list at
1442 	 * interrupt time; lock them out.
1443 	 */
1444 	s = splimp();
1445 	TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1446 	splx(s);
1447 	*retifma = ifma;
1448 
1449 	if (llsa != 0) {
1450 		TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1451 			if (equal(ifma->ifma_addr, llsa))
1452 				break;
1453 		}
1454 		if (ifma) {
1455 			ifma->ifma_refcount++;
1456 		} else {
1457 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1458 			       M_IFMADDR, M_WAITOK);
1459 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1460 			       M_IFMADDR, M_WAITOK);
1461 			bcopy(llsa, dupsa, llsa->sa_len);
1462 			ifma->ifma_addr = dupsa;
1463 			ifma->ifma_ifp = ifp;
1464 			ifma->ifma_refcount = 1;
1465 			s = splimp();
1466 			TAILQ_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1467 			splx(s);
1468 		}
1469 	}
1470 	/*
1471 	 * We are certain we have added something, so call down to the
1472 	 * interface to let them know about it.
1473 	 */
1474 	s = splimp();
1475 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0);
1476 	splx(s);
1477 
1478 	return 0;
1479 }
1480 
1481 /*
1482  * Remove a reference to a multicast address on this interface.  Yell
1483  * if the request does not match an existing membership.
1484  */
1485 int
1486 if_delmulti(ifp, sa)
1487 	struct ifnet *ifp;
1488 	struct sockaddr *sa;
1489 {
1490 	struct ifmultiaddr *ifma;
1491 	int s;
1492 
1493 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1494 		if (equal(sa, ifma->ifma_addr))
1495 			break;
1496 	if (ifma == 0)
1497 		return ENOENT;
1498 
1499 	if (ifma->ifma_refcount > 1) {
1500 		ifma->ifma_refcount--;
1501 		return 0;
1502 	}
1503 
1504 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1505 	sa = ifma->ifma_lladdr;
1506 	s = splimp();
1507 	TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
1508 	/*
1509 	 * Make sure the interface driver is notified
1510 	 * in the case of a link layer mcast group being left.
1511 	 */
1512 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1513 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1514 	splx(s);
1515 	free(ifma->ifma_addr, M_IFMADDR);
1516 	free(ifma, M_IFMADDR);
1517 	if (sa == 0)
1518 		return 0;
1519 
1520 	/*
1521 	 * Now look for the link-layer address which corresponds to
1522 	 * this network address.  It had been squirreled away in
1523 	 * ifma->ifma_lladdr for this purpose (so we don't have
1524 	 * to call ifp->if_resolvemulti() again), and we saved that
1525 	 * value in sa above.  If some nasty deleted the
1526 	 * link-layer address out from underneath us, we can deal because
1527 	 * the address we stored was is not the same as the one which was
1528 	 * in the record for the link-layer address.  (So we don't complain
1529 	 * in that case.)
1530 	 */
1531 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1532 		if (equal(sa, ifma->ifma_addr))
1533 			break;
1534 	if (ifma == 0)
1535 		return 0;
1536 
1537 	if (ifma->ifma_refcount > 1) {
1538 		ifma->ifma_refcount--;
1539 		return 0;
1540 	}
1541 
1542 	s = splimp();
1543 	TAILQ_REMOVE(&ifp->if_multiaddrs, ifma, ifma_link);
1544 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1545 	splx(s);
1546 	free(ifma->ifma_addr, M_IFMADDR);
1547 	free(sa, M_IFMADDR);
1548 	free(ifma, M_IFMADDR);
1549 
1550 	return 0;
1551 }
1552 
1553 /*
1554  * Set the link layer address on an interface.
1555  *
1556  * At this time we only support certain types of interfaces,
1557  * and we don't allow the length of the address to change.
1558  */
1559 int
1560 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1561 {
1562 	struct sockaddr_dl *sdl;
1563 	struct ifaddr *ifa;
1564 
1565 	ifa = ifaddr_byindex(ifp->if_index);
1566 	if (ifa == NULL)
1567 		return (EINVAL);
1568 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1569 	if (sdl == NULL)
1570 		return (EINVAL);
1571 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1572 		return (EINVAL);
1573 	switch (ifp->if_type) {
1574 	case IFT_ETHER:			/* these types use struct arpcom */
1575 	case IFT_FDDI:
1576 	case IFT_XETHER:
1577 	case IFT_ISO88025:
1578 	case IFT_L2VLAN:
1579 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1580 		bcopy(lladdr, LLADDR(sdl), len);
1581 		break;
1582 	default:
1583 		return (ENODEV);
1584 	}
1585 	/*
1586 	 * If the interface is already up, we need
1587 	 * to re-init it in order to reprogram its
1588 	 * address filter.
1589 	 */
1590 	if ((ifp->if_flags & IFF_UP) != 0) {
1591 		ifp->if_flags &= ~IFF_UP;
1592 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, NULL);
1593 		ifp->if_flags |= IFF_UP;
1594 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, NULL);
1595 	}
1596 	return (0);
1597 }
1598 
1599 struct ifmultiaddr *
1600 ifmaof_ifpforaddr(sa, ifp)
1601 	struct sockaddr *sa;
1602 	struct ifnet *ifp;
1603 {
1604 	struct ifmultiaddr *ifma;
1605 
1606 	TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1607 		if (equal(ifma->ifma_addr, sa))
1608 			break;
1609 
1610 	return ifma;
1611 }
1612 
1613 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1614 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1615