xref: /dflybsd-src/sys/net/if.c (revision ac2e3f5effc58aa364c7e5c199f35ebbae7cda81)
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  * $DragonFly: src/sys/net/if.c,v 1.5 2003/07/26 20:19:33 rob 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/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/kernel.h>
51 #include <sys/sockio.h>
52 #include <sys/syslog.h>
53 #include <sys/sysctl.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 #include <machine/stdarg.h>
63 
64 #if defined(INET) || defined(INET6)
65 /*XXX*/
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/if_ether.h>
69 #ifdef INET6
70 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
71 #include <netinet6/in6_var.h>
72 #include <netinet6/in6_ifattach.h>
73 #endif
74 #endif
75 
76 /*
77  * System initialization
78  */
79 
80 static int ifconf __P((u_long, caddr_t));
81 static void ifinit __P((void *));
82 static void if_qflush __P((struct ifqueue *));
83 static void if_slowtimo __P((void *));
84 static void link_rtrequest __P((int, struct rtentry *, struct rt_addrinfo *));
85 static int  if_rtdel __P((struct radix_node *, void *));
86 
87 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
88 
89 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
90 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
91 
92 int	ifqmaxlen = IFQ_MAXLEN;
93 struct	ifnethead ifnet;	/* depend on static init XXX */
94 
95 #ifdef INET6
96 /*
97  * XXX: declare here to avoid to include many inet6 related files..
98  * should be more generalized?
99  */
100 extern void	nd6_setmtu __P((struct ifnet *));
101 #endif
102 
103 struct if_clone *if_clone_lookup __P((const char *, int *));
104 int if_clone_list __P((struct if_clonereq *));
105 
106 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
107 int if_cloners_count;
108 
109 /*
110  * Network interface utility routines.
111  *
112  * Routines with ifa_ifwith* names take sockaddr *'s as
113  * parameters.
114  */
115 /* ARGSUSED*/
116 void
117 ifinit(dummy)
118 	void *dummy;
119 {
120 	struct ifnet *ifp;
121 	int s;
122 
123 	s = splimp();
124 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
125 		if (ifp->if_snd.ifq_maxlen == 0) {
126 			printf("%s%d XXX: driver didn't set ifq_maxlen\n",
127 			    ifp->if_name, ifp->if_unit);
128 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
129 		}
130 	}
131 	splx(s);
132 	if_slowtimo(0);
133 }
134 
135 int if_index = 0;
136 struct ifaddr **ifnet_addrs;
137 struct ifnet **ifindex2ifnet = NULL;
138 
139 
140 /*
141  * Attach an interface to the
142  * list of "active" interfaces.
143  */
144 void
145 if_attach(ifp)
146 	struct ifnet *ifp;
147 {
148 	unsigned socksize, ifasize;
149 	int namelen, masklen;
150 	char workbuf[64];
151 	struct sockaddr_dl *sdl;
152 	struct ifaddr *ifa;
153 	static int if_indexlim = 8;
154 	static int inited;
155 
156 	if (!inited) {
157 		TAILQ_INIT(&ifnet);
158 		inited = 1;
159 	}
160 
161 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
162 	ifp->if_index = ++if_index;
163 	/*
164 	 * XXX -
165 	 * The old code would work if the interface passed a pre-existing
166 	 * chain of ifaddrs to this code.  We don't trust our callers to
167 	 * properly initialize the tailq, however, so we no longer allow
168 	 * this unlikely case.
169 	 */
170 	TAILQ_INIT(&ifp->if_addrhead);
171 	TAILQ_INIT(&ifp->if_prefixhead);
172 	LIST_INIT(&ifp->if_multiaddrs);
173 	getmicrotime(&ifp->if_lastchange);
174 	if (ifnet_addrs == 0 || if_index >= if_indexlim) {
175 		unsigned n = (if_indexlim <<= 1) * sizeof(ifa);
176 		caddr_t q = malloc(n, M_IFADDR, M_WAITOK);
177 		bzero(q, n);
178 		if (ifnet_addrs) {
179 			bcopy((caddr_t)ifnet_addrs, (caddr_t)q, n/2);
180 			free((caddr_t)ifnet_addrs, M_IFADDR);
181 		}
182 		ifnet_addrs = (struct ifaddr **)q;
183 
184 		/* grow ifindex2ifnet */
185 		n = if_indexlim * sizeof(struct ifnet *);
186 		q = malloc(n, M_IFADDR, M_WAITOK);
187 		bzero(q, n);
188 		if (ifindex2ifnet) {
189 			bcopy((caddr_t)ifindex2ifnet, q, n/2);
190 			free((caddr_t)ifindex2ifnet, M_IFADDR);
191 		}
192 		ifindex2ifnet = (struct ifnet **)q;
193 	}
194 
195 	ifindex2ifnet[if_index] = ifp;
196 
197 	/*
198 	 * create a Link Level name for this device
199 	 */
200 	namelen = snprintf(workbuf, sizeof(workbuf),
201 	    "%s%d", ifp->if_name, ifp->if_unit);
202 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
203 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
204 	socksize = masklen + ifp->if_addrlen;
205 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
206 	if (socksize < sizeof(*sdl))
207 		socksize = sizeof(*sdl);
208 	socksize = ROUNDUP(socksize);
209 	ifasize = sizeof(*ifa) + 2 * socksize;
210 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
211 	if (ifa) {
212 		bzero((caddr_t)ifa, ifasize);
213 		sdl = (struct sockaddr_dl *)(ifa + 1);
214 		sdl->sdl_len = socksize;
215 		sdl->sdl_family = AF_LINK;
216 		bcopy(workbuf, sdl->sdl_data, namelen);
217 		sdl->sdl_nlen = namelen;
218 		sdl->sdl_index = ifp->if_index;
219 		sdl->sdl_type = ifp->if_type;
220 		ifnet_addrs[if_index - 1] = ifa;
221 		ifa->ifa_ifp = ifp;
222 		ifa->ifa_rtrequest = link_rtrequest;
223 		ifa->ifa_addr = (struct sockaddr *)sdl;
224 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
225 		ifa->ifa_netmask = (struct sockaddr *)sdl;
226 		sdl->sdl_len = masklen;
227 		while (namelen != 0)
228 			sdl->sdl_data[--namelen] = 0xff;
229 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
230 	}
231 
232 	/* Announce the interface. */
233 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
234 }
235 
236 /*
237  * Detach an interface, removing it from the
238  * list of "active" interfaces.
239  */
240 void
241 if_detach(ifp)
242 	struct ifnet *ifp;
243 {
244 	struct ifaddr *ifa;
245 	struct radix_node_head	*rnh;
246 	int s;
247 	int i;
248 
249 	/*
250 	 * Remove routes and flush queues.
251 	 */
252 	s = splnet();
253 	if_down(ifp);
254 
255 	/*
256 	 * Remove address from ifnet_addrs[] and maybe decrement if_index.
257 	 * Clean up all addresses.
258 	 */
259 	ifnet_addrs[ifp->if_index - 1] = 0;
260 	while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
261 		if_index--;
262 
263 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
264 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
265 #ifdef INET
266 		/* XXX: Ugly!! ad hoc just for INET */
267 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
268 			struct ifaliasreq ifr;
269 
270 			bzero(&ifr, sizeof(ifr));
271 			ifr.ifra_addr = *ifa->ifa_addr;
272 			if (ifa->ifa_dstaddr)
273 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
274 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
275 			    NULL) == 0)
276 				continue;
277 		}
278 #endif /* INET */
279 #ifdef INET6
280 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
281 			in6_purgeaddr(ifa);
282 			/* ifp_addrhead is already updated */
283 			continue;
284 		}
285 #endif /* INET6 */
286 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
287 		IFAFREE(ifa);
288 	}
289 
290 #ifdef INET6
291 	/*
292 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
293 	 * before removing routing entries below, since IPv6 interface direct
294 	 * routes are expected to be removed by the IPv6-specific kernel API.
295 	 * Otherwise, the kernel will detect some inconsistency and bark it.
296 	 */
297 	in6_ifdetach(ifp);
298 #endif
299 
300 	/*
301 	 * Delete all remaining routes using this interface
302 	 * Unfortuneatly the only way to do this is to slog through
303 	 * the entire routing table looking for routes which point
304 	 * to this interface...oh well...
305 	 */
306 	for (i = 1; i <= AF_MAX; i++) {
307 		if ((rnh = rt_tables[i]) == NULL)
308 			continue;
309 		(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
310 	}
311 
312 	/* Announce that the interface is gone. */
313 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
314 
315 	TAILQ_REMOVE(&ifnet, ifp, if_link);
316 	splx(s);
317 }
318 
319 /*
320  * Delete Routes for a Network Interface
321  *
322  * Called for each routing entry via the rnh->rnh_walktree() call above
323  * to delete all route entries referencing a detaching network interface.
324  *
325  * Arguments:
326  *	rn	pointer to node in the routing table
327  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
328  *
329  * Returns:
330  *	0	successful
331  *	errno	failed - reason indicated
332  *
333  */
334 static int
335 if_rtdel(rn, arg)
336 	struct radix_node	*rn;
337 	void			*arg;
338 {
339 	struct rtentry	*rt = (struct rtentry *)rn;
340 	struct ifnet	*ifp = arg;
341 	int		err;
342 
343 	if (rt->rt_ifp == ifp) {
344 
345 		/*
346 		 * Protect (sorta) against walktree recursion problems
347 		 * with cloned routes
348 		 */
349 		if ((rt->rt_flags & RTF_UP) == 0)
350 			return (0);
351 
352 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
353 				rt_mask(rt), rt->rt_flags,
354 				(struct rtentry **) NULL);
355 		if (err) {
356 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
357 		}
358 	}
359 
360 	return (0);
361 }
362 
363 /*
364  * Create a clone network interface.
365  */
366 int
367 if_clone_create(name, len)
368 	char *name;
369 	int len;
370 {
371 	struct if_clone *ifc;
372 	char *dp;
373 	int wildcard;
374 	int unit;
375 	int err;
376 
377 	ifc = if_clone_lookup(name, &unit);
378 	if (ifc == NULL)
379 		return (EINVAL);
380 
381 	if (ifunit(name) != NULL)
382 		return (EEXIST);
383 
384 	wildcard = (unit < 0);
385 
386 	err = (*ifc->ifc_create)(ifc, &unit);
387 	if (err != 0)
388 		return (err);
389 
390 	/* In the wildcard case, we need to update the name. */
391 	if (wildcard) {
392 		for (dp = name; *dp != '\0'; dp++);
393 		if (snprintf(dp, len - (dp-name), "%d", unit) >
394 		    len - (dp-name) - 1) {
395 			/*
396 			 * This can only be a programmer error and
397 			 * there's no straightforward way to recover if
398 			 * it happens.
399 			 */
400 			panic("if_clone_create(): interface name too long");
401 		}
402 
403 	}
404 
405 	return (0);
406 }
407 
408 /*
409  * Destroy a clone network interface.
410  */
411 int
412 if_clone_destroy(name)
413 	const char *name;
414 {
415 	struct if_clone *ifc;
416 	struct ifnet *ifp;
417 
418 	ifc = if_clone_lookup(name, NULL);
419 	if (ifc == NULL)
420 		return (EINVAL);
421 
422 	ifp = ifunit(name);
423 	if (ifp == NULL)
424 		return (ENXIO);
425 
426 	if (ifc->ifc_destroy == NULL)
427 		return (EOPNOTSUPP);
428 
429 	(*ifc->ifc_destroy)(ifp);
430 	return (0);
431 }
432 
433 /*
434  * Look up a network interface cloner.
435  */
436 struct if_clone *
437 if_clone_lookup(name, unitp)
438 	const char *name;
439 	int *unitp;
440 {
441 	struct if_clone *ifc;
442 	const char *cp;
443 	int i;
444 
445 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
446 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
447 			if (ifc->ifc_name[i] != *cp)
448 				goto next_ifc;
449 		}
450 		goto found_name;
451  next_ifc:
452 		ifc = LIST_NEXT(ifc, ifc_list);
453 	}
454 
455 	/* No match. */
456 	return ((struct if_clone *)NULL);
457 
458  found_name:
459 	if (*cp == '\0') {
460 		i = -1;
461 	} else {
462 		for (i = 0; *cp != '\0'; cp++) {
463 			if (*cp < '0' || *cp > '9') {
464 				/* Bogus unit number. */
465 				return (NULL);
466 			}
467 			i = (i * 10) + (*cp - '0');
468 		}
469 	}
470 
471 	if (unitp != NULL)
472 		*unitp = i;
473 	return (ifc);
474 }
475 
476 /*
477  * Register a network interface cloner.
478  */
479 void
480 if_clone_attach(ifc)
481 	struct if_clone *ifc;
482 {
483 
484 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
485 	if_cloners_count++;
486 }
487 
488 /*
489  * Unregister a network interface cloner.
490  */
491 void
492 if_clone_detach(ifc)
493 	struct if_clone *ifc;
494 {
495 
496 	LIST_REMOVE(ifc, ifc_list);
497 	if_cloners_count--;
498 }
499 
500 /*
501  * Provide list of interface cloners to userspace.
502  */
503 int
504 if_clone_list(ifcr)
505 	struct if_clonereq *ifcr;
506 {
507 	char outbuf[IFNAMSIZ], *dst;
508 	struct if_clone *ifc;
509 	int count, error = 0;
510 
511 	ifcr->ifcr_total = if_cloners_count;
512 	if ((dst = ifcr->ifcr_buffer) == NULL) {
513 		/* Just asking how many there are. */
514 		return (0);
515 	}
516 
517 	if (ifcr->ifcr_count < 0)
518 		return (EINVAL);
519 
520 	count = (if_cloners_count < ifcr->ifcr_count) ?
521 	    if_cloners_count : ifcr->ifcr_count;
522 
523 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
524 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
525 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
526 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
527 		error = copyout(outbuf, dst, IFNAMSIZ);
528 		if (error)
529 			break;
530 	}
531 
532 	return (error);
533 }
534 
535 /*
536  * Locate an interface based on a complete address.
537  */
538 /*ARGSUSED*/
539 struct ifaddr *
540 ifa_ifwithaddr(addr)
541 	struct sockaddr *addr;
542 {
543 	struct ifnet *ifp;
544 	struct ifaddr *ifa;
545 
546 #define	equal(a1, a2) \
547   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
548 	TAILQ_FOREACH(ifp, &ifnet, if_link)
549 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
550 		if (ifa->ifa_addr->sa_family != addr->sa_family)
551 			continue;
552 		if (equal(addr, ifa->ifa_addr))
553 			return (ifa);
554 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
555 		    /* IP6 doesn't have broadcast */
556 		    ifa->ifa_broadaddr->sa_len != 0 &&
557 		    equal(ifa->ifa_broadaddr, addr))
558 			return (ifa);
559 	}
560 	return ((struct ifaddr *)0);
561 }
562 /*
563  * Locate the point to point interface with a given destination address.
564  */
565 /*ARGSUSED*/
566 struct ifaddr *
567 ifa_ifwithdstaddr(addr)
568 	struct sockaddr *addr;
569 {
570 	struct ifnet *ifp;
571 	struct ifaddr *ifa;
572 
573 	TAILQ_FOREACH(ifp, &ifnet, if_link)
574 	    if (ifp->if_flags & IFF_POINTOPOINT)
575 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
576 			if (ifa->ifa_addr->sa_family != addr->sa_family)
577 				continue;
578 			if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
579 				return (ifa);
580 	}
581 	return ((struct ifaddr *)0);
582 }
583 
584 /*
585  * Find an interface on a specific network.  If many, choice
586  * is most specific found.
587  */
588 struct ifaddr *
589 ifa_ifwithnet(addr)
590 	struct sockaddr *addr;
591 {
592 	struct ifnet *ifp;
593 	struct ifaddr *ifa;
594 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
595 	u_int af = addr->sa_family;
596 	char *addr_data = addr->sa_data, *cplim;
597 
598 	/*
599 	 * AF_LINK addresses can be looked up directly by their index number,
600 	 * so do that if we can.
601 	 */
602 	if (af == AF_LINK) {
603 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
604 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
605 		return (ifnet_addrs[sdl->sdl_index - 1]);
606 	}
607 
608 	/*
609 	 * Scan though each interface, looking for ones that have
610 	 * addresses in this address family.
611 	 */
612 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
613 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
614 			char *cp, *cp2, *cp3;
615 
616 			if (ifa->ifa_addr->sa_family != af)
617 next:				continue;
618 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
619 				/*
620 				 * This is a bit broken as it doesn't
621 				 * take into account that the remote end may
622 				 * be a single node in the network we are
623 				 * looking for.
624 				 * The trouble is that we don't know the
625 				 * netmask for the remote end.
626 				 */
627 				if (ifa->ifa_dstaddr != 0
628 				    && equal(addr, ifa->ifa_dstaddr))
629  					return (ifa);
630 			} else {
631 				/*
632 				 * if we have a special address handler,
633 				 * then use it instead of the generic one.
634 				 */
635 	          		if (ifa->ifa_claim_addr) {
636 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
637 						return (ifa);
638 					} else {
639 						continue;
640 					}
641 				}
642 
643 				/*
644 				 * Scan all the bits in the ifa's address.
645 				 * If a bit dissagrees with what we are
646 				 * looking for, mask it with the netmask
647 				 * to see if it really matters.
648 				 * (A byte at a time)
649 				 */
650 				if (ifa->ifa_netmask == 0)
651 					continue;
652 				cp = addr_data;
653 				cp2 = ifa->ifa_addr->sa_data;
654 				cp3 = ifa->ifa_netmask->sa_data;
655 				cplim = ifa->ifa_netmask->sa_len
656 					+ (char *)ifa->ifa_netmask;
657 				while (cp3 < cplim)
658 					if ((*cp++ ^ *cp2++) & *cp3++)
659 						goto next; /* next address! */
660 				/*
661 				 * If the netmask of what we just found
662 				 * is more specific than what we had before
663 				 * (if we had one) then remember the new one
664 				 * before continuing to search
665 				 * for an even better one.
666 				 */
667 				if (ifa_maybe == 0 ||
668 				    rn_refines((caddr_t)ifa->ifa_netmask,
669 				    (caddr_t)ifa_maybe->ifa_netmask))
670 					ifa_maybe = ifa;
671 			}
672 		}
673 	}
674 	return (ifa_maybe);
675 }
676 
677 /*
678  * Find an interface address specific to an interface best matching
679  * a given address.
680  */
681 struct ifaddr *
682 ifaof_ifpforaddr(addr, ifp)
683 	struct sockaddr *addr;
684 	struct ifnet *ifp;
685 {
686 	struct ifaddr *ifa;
687 	char *cp, *cp2, *cp3;
688 	char *cplim;
689 	struct ifaddr *ifa_maybe = 0;
690 	u_int af = addr->sa_family;
691 
692 	if (af >= AF_MAX)
693 		return (0);
694 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
695 		if (ifa->ifa_addr->sa_family != af)
696 			continue;
697 		if (ifa_maybe == 0)
698 			ifa_maybe = ifa;
699 		if (ifa->ifa_netmask == 0) {
700 			if (equal(addr, ifa->ifa_addr) ||
701 			    (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
702 				return (ifa);
703 			continue;
704 		}
705 		if (ifp->if_flags & IFF_POINTOPOINT) {
706 			if (equal(addr, ifa->ifa_dstaddr))
707 				return (ifa);
708 		} else {
709 			cp = addr->sa_data;
710 			cp2 = ifa->ifa_addr->sa_data;
711 			cp3 = ifa->ifa_netmask->sa_data;
712 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
713 			for (; cp3 < cplim; cp3++)
714 				if ((*cp++ ^ *cp2++) & *cp3)
715 					break;
716 			if (cp3 == cplim)
717 				return (ifa);
718 		}
719 	}
720 	return (ifa_maybe);
721 }
722 
723 #include <net/route.h>
724 
725 /*
726  * Default action when installing a route with a Link Level gateway.
727  * Lookup an appropriate real ifa to point to.
728  * This should be moved to /sys/net/link.c eventually.
729  */
730 static void
731 link_rtrequest(cmd, rt, info)
732 	int cmd;
733 	struct rtentry *rt;
734 	struct rt_addrinfo *info;
735 {
736 	struct ifaddr *ifa;
737 	struct sockaddr *dst;
738 	struct ifnet *ifp;
739 
740 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
741 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
742 		return;
743 	ifa = ifaof_ifpforaddr(dst, ifp);
744 	if (ifa) {
745 		IFAFREE(rt->rt_ifa);
746 		rt->rt_ifa = ifa;
747 		ifa->ifa_refcnt++;
748 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
749 			ifa->ifa_rtrequest(cmd, rt, info);
750 	}
751 }
752 
753 /*
754  * Mark an interface down and notify protocols of
755  * the transition.
756  * NOTE: must be called at splnet or eqivalent.
757  */
758 void
759 if_unroute(ifp, flag, fam)
760 	struct ifnet *ifp;
761 	int flag, fam;
762 {
763 	struct ifaddr *ifa;
764 
765 	ifp->if_flags &= ~flag;
766 	getmicrotime(&ifp->if_lastchange);
767 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
768 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
769 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
770 	if_qflush(&ifp->if_snd);
771 	rt_ifmsg(ifp);
772 }
773 
774 /*
775  * Mark an interface up and notify protocols of
776  * the transition.
777  * NOTE: must be called at splnet or eqivalent.
778  */
779 void
780 if_route(ifp, flag, fam)
781 	struct ifnet *ifp;
782 	int flag, fam;
783 {
784 	struct ifaddr *ifa;
785 
786 	ifp->if_flags |= flag;
787 	getmicrotime(&ifp->if_lastchange);
788 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
789 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
790 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
791 	rt_ifmsg(ifp);
792 #ifdef INET6
793 	in6_if_up(ifp);
794 #endif
795 }
796 
797 /*
798  * Mark an interface down and notify protocols of
799  * the transition.
800  * NOTE: must be called at splnet or eqivalent.
801  */
802 void
803 if_down(ifp)
804 	struct ifnet *ifp;
805 {
806 
807 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
808 }
809 
810 /*
811  * Mark an interface up and notify protocols of
812  * the transition.
813  * NOTE: must be called at splnet or eqivalent.
814  */
815 void
816 if_up(ifp)
817 	struct ifnet *ifp;
818 {
819 
820 	if_route(ifp, IFF_UP, AF_UNSPEC);
821 }
822 
823 /*
824  * Flush an interface queue.
825  */
826 static void
827 if_qflush(ifq)
828 	struct ifqueue *ifq;
829 {
830 	struct mbuf *m, *n;
831 
832 	n = ifq->ifq_head;
833 	while ((m = n) != 0) {
834 		n = m->m_act;
835 		m_freem(m);
836 	}
837 	ifq->ifq_head = 0;
838 	ifq->ifq_tail = 0;
839 	ifq->ifq_len = 0;
840 }
841 
842 /*
843  * Handle interface watchdog timer routines.  Called
844  * from softclock, we decrement timers (if set) and
845  * call the appropriate interface routine on expiration.
846  */
847 static void
848 if_slowtimo(arg)
849 	void *arg;
850 {
851 	struct ifnet *ifp;
852 	int s = splimp();
853 
854 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
855 		if (ifp->if_timer == 0 || --ifp->if_timer)
856 			continue;
857 		if (ifp->if_watchdog)
858 			(*ifp->if_watchdog)(ifp);
859 	}
860 	splx(s);
861 	timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
862 }
863 
864 /*
865  * Map interface name to
866  * interface structure pointer.
867  */
868 struct ifnet *
869 ifunit(const char *name)
870 {
871 	char namebuf[IFNAMSIZ + 1];
872 	const char *cp;
873 	struct ifnet *ifp;
874 	int unit;
875 	unsigned len, m;
876 	char c;
877 
878 	len = strlen(name);
879 	if (len < 2 || len > IFNAMSIZ)
880 		return NULL;
881 	cp = name + len - 1;
882 	c = *cp;
883 	if (c < '0' || c > '9')
884 		return NULL;		/* trailing garbage */
885 	unit = 0;
886 	m = 1;
887 	do {
888 		if (cp == name)
889 			return NULL;	/* no interface name */
890 		unit += (c - '0') * m;
891 		if (unit > 1000000)
892 			return NULL;	/* number is unreasonable */
893 		m *= 10;
894 		c = *--cp;
895 	} while (c >= '0' && c <= '9');
896 	len = cp - name + 1;
897 	bcopy(name, namebuf, len);
898 	namebuf[len] = '\0';
899 	/*
900 	 * Now search all the interfaces for this name/number
901 	 */
902 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
903 		if (strcmp(ifp->if_name, namebuf))
904 			continue;
905 		if (unit == ifp->if_unit)
906 			break;
907 	}
908 	return (ifp);
909 }
910 
911 
912 /*
913  * Map interface name in a sockaddr_dl to
914  * interface structure pointer.
915  */
916 struct ifnet *
917 if_withname(sa)
918 	struct sockaddr *sa;
919 {
920 	char ifname[IFNAMSIZ+1];
921 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
922 
923 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
924 	     (sdl->sdl_nlen > IFNAMSIZ) )
925 		return NULL;
926 
927 	/*
928 	 * ifunit wants a null-terminated name.  It may not be null-terminated
929 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
930 	 * and there might not be room to put the trailing null anyway, so we
931 	 * make a local copy that we know we can null terminate safely.
932 	 */
933 
934 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
935 	ifname[sdl->sdl_nlen] = '\0';
936 	return ifunit(ifname);
937 }
938 
939 
940 /*
941  * Interface ioctls.
942  */
943 int
944 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
945 {
946 	struct ifnet *ifp;
947 	struct ifreq *ifr;
948 	struct ifstat *ifs;
949 	int error;
950 	short oif_flags;
951 	int new_flags;
952 
953 	switch (cmd) {
954 
955 	case SIOCGIFCONF:
956 	case OSIOCGIFCONF:
957 		return (ifconf(cmd, data));
958 	}
959 	ifr = (struct ifreq *)data;
960 
961 	switch (cmd) {
962 	case SIOCIFCREATE:
963 	case SIOCIFDESTROY:
964 		if ((error = suser(td)) != 0)
965 			return (error);
966 		return ((cmd == SIOCIFCREATE) ?
967 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
968 			if_clone_destroy(ifr->ifr_name));
969 
970 	case SIOCIFGCLONERS:
971 		return (if_clone_list((struct if_clonereq *)data));
972 	}
973 
974 	ifp = ifunit(ifr->ifr_name);
975 	if (ifp == 0)
976 		return (ENXIO);
977 	switch (cmd) {
978 
979 	case SIOCGIFFLAGS:
980 		ifr->ifr_flags = ifp->if_flags;
981 		ifr->ifr_flagshigh = ifp->if_ipending >> 16;
982 		break;
983 
984 	case SIOCGIFCAP:
985 		ifr->ifr_reqcap = ifp->if_capabilities;
986 		ifr->ifr_curcap = ifp->if_capenable;
987 		break;
988 
989 	case SIOCGIFMETRIC:
990 		ifr->ifr_metric = ifp->if_metric;
991 		break;
992 
993 	case SIOCGIFMTU:
994 		ifr->ifr_mtu = ifp->if_mtu;
995 		break;
996 
997 	case SIOCGIFPHYS:
998 		ifr->ifr_phys = ifp->if_physical;
999 		break;
1000 
1001 	case SIOCSIFFLAGS:
1002 		error = suser(td);
1003 		if (error)
1004 			return (error);
1005 		new_flags = (ifr->ifr_flags & 0xffff) |
1006 		    (ifr->ifr_flagshigh << 16);
1007 		if (ifp->if_flags & IFF_SMART) {
1008 			/* Smart drivers twiddle their own routes */
1009 		} else if (ifp->if_flags & IFF_UP &&
1010 		    (new_flags & IFF_UP) == 0) {
1011 			int s = splimp();
1012 			if_down(ifp);
1013 			splx(s);
1014 		} else if (new_flags & IFF_UP &&
1015 		    (ifp->if_flags & IFF_UP) == 0) {
1016 			int s = splimp();
1017 			if_up(ifp);
1018 			splx(s);
1019 		}
1020 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1021 			(new_flags &~ IFF_CANTCHANGE);
1022 		ifp->if_ipending = (ifp->if_ipending & IFF_CANTCHANGE) |
1023 			(new_flags &~ IFF_CANTCHANGE);
1024 		if (new_flags & IFF_PPROMISC) {
1025 			/* Permanently promiscuous mode requested */
1026 			ifp->if_flags |= IFF_PROMISC;
1027 		} else if (ifp->if_pcount == 0) {
1028 			ifp->if_flags &= ~IFF_PROMISC;
1029 		}
1030 		if (ifp->if_ioctl)
1031 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1032 		getmicrotime(&ifp->if_lastchange);
1033 		break;
1034 
1035 	case SIOCSIFCAP:
1036 		error = suser(td);
1037 		if (error)
1038 			return (error);
1039 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1040 			return (EINVAL);
1041 		(void) (*ifp->if_ioctl)(ifp, cmd, data);
1042 		break;
1043 
1044 	case SIOCSIFMETRIC:
1045 		error = suser(td);
1046 		if (error)
1047 			return (error);
1048 		ifp->if_metric = ifr->ifr_metric;
1049 		getmicrotime(&ifp->if_lastchange);
1050 		break;
1051 
1052 	case SIOCSIFPHYS:
1053 		error = suser(td);
1054 		if (error)
1055 			return error;
1056 		if (!ifp->if_ioctl)
1057 		        return EOPNOTSUPP;
1058 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1059 		if (error == 0)
1060 			getmicrotime(&ifp->if_lastchange);
1061 		return(error);
1062 
1063 	case SIOCSIFMTU:
1064 	{
1065 		u_long oldmtu = ifp->if_mtu;
1066 
1067 		error = suser(td);
1068 		if (error)
1069 			return (error);
1070 		if (ifp->if_ioctl == NULL)
1071 			return (EOPNOTSUPP);
1072 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1073 			return (EINVAL);
1074 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1075 		if (error == 0) {
1076 			getmicrotime(&ifp->if_lastchange);
1077 			rt_ifmsg(ifp);
1078 		}
1079 		/*
1080 		 * If the link MTU changed, do network layer specific procedure.
1081 		 */
1082 		if (ifp->if_mtu != oldmtu) {
1083 #ifdef INET6
1084 			nd6_setmtu(ifp);
1085 #endif
1086 		}
1087 		return (error);
1088 	}
1089 
1090 	case SIOCADDMULTI:
1091 	case SIOCDELMULTI:
1092 		error = suser(td);
1093 		if (error)
1094 			return (error);
1095 
1096 		/* Don't allow group membership on non-multicast interfaces. */
1097 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1098 			return EOPNOTSUPP;
1099 
1100 		/* Don't let users screw up protocols' entries. */
1101 		if (ifr->ifr_addr.sa_family != AF_LINK)
1102 			return EINVAL;
1103 
1104 		if (cmd == SIOCADDMULTI) {
1105 			struct ifmultiaddr *ifma;
1106 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1107 		} else {
1108 			error = if_delmulti(ifp, &ifr->ifr_addr);
1109 		}
1110 		if (error == 0)
1111 			getmicrotime(&ifp->if_lastchange);
1112 		return error;
1113 
1114 	case SIOCSIFPHYADDR:
1115 	case SIOCDIFPHYADDR:
1116 #ifdef INET6
1117 	case SIOCSIFPHYADDR_IN6:
1118 #endif
1119 	case SIOCSLIFPHYADDR:
1120         case SIOCSIFMEDIA:
1121 	case SIOCSIFGENERIC:
1122 		error = suser(td);
1123 		if (error)
1124 			return (error);
1125 		if (ifp->if_ioctl == 0)
1126 			return (EOPNOTSUPP);
1127 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1128 		if (error == 0)
1129 			getmicrotime(&ifp->if_lastchange);
1130 		return error;
1131 
1132 	case SIOCGIFSTATUS:
1133 		ifs = (struct ifstat *)data;
1134 		ifs->ascii[0] = '\0';
1135 
1136 	case SIOCGIFPSRCADDR:
1137 	case SIOCGIFPDSTADDR:
1138 	case SIOCGLIFPHYADDR:
1139 	case SIOCGIFMEDIA:
1140 	case SIOCGIFGENERIC:
1141 		if (ifp->if_ioctl == 0)
1142 			return (EOPNOTSUPP);
1143 		return ((*ifp->if_ioctl)(ifp, cmd, data));
1144 
1145 	case SIOCSIFLLADDR:
1146 		error = suser(td);
1147 		if (error)
1148 			return (error);
1149 		return if_setlladdr(ifp,
1150 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1151 
1152 	default:
1153 		oif_flags = ifp->if_flags;
1154 		if (so->so_proto == 0)
1155 			return (EOPNOTSUPP);
1156 #ifndef COMPAT_43
1157 		error = ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
1158 								 data,
1159 								 ifp, p));
1160 #else
1161 	    {
1162 		int ocmd = cmd;
1163 
1164 		switch (cmd) {
1165 
1166 		case SIOCSIFDSTADDR:
1167 		case SIOCSIFADDR:
1168 		case SIOCSIFBRDADDR:
1169 		case SIOCSIFNETMASK:
1170 #if BYTE_ORDER != BIG_ENDIAN
1171 			if (ifr->ifr_addr.sa_family == 0 &&
1172 			    ifr->ifr_addr.sa_len < 16) {
1173 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1174 				ifr->ifr_addr.sa_len = 16;
1175 			}
1176 #else
1177 			if (ifr->ifr_addr.sa_len == 0)
1178 				ifr->ifr_addr.sa_len = 16;
1179 #endif
1180 			break;
1181 
1182 		case OSIOCGIFADDR:
1183 			cmd = SIOCGIFADDR;
1184 			break;
1185 
1186 		case OSIOCGIFDSTADDR:
1187 			cmd = SIOCGIFDSTADDR;
1188 			break;
1189 
1190 		case OSIOCGIFBRDADDR:
1191 			cmd = SIOCGIFBRDADDR;
1192 			break;
1193 
1194 		case OSIOCGIFNETMASK:
1195 			cmd = SIOCGIFNETMASK;
1196 		}
1197 		error =  ((*so->so_proto->pr_usrreqs->pru_control)
1198 				(so, cmd, data, ifp, td));
1199 		switch (ocmd) {
1200 
1201 		case OSIOCGIFADDR:
1202 		case OSIOCGIFDSTADDR:
1203 		case OSIOCGIFBRDADDR:
1204 		case OSIOCGIFNETMASK:
1205 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1206 
1207 		}
1208 	    }
1209 #endif /* COMPAT_43 */
1210 
1211 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1212 #ifdef INET6
1213 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1214 			if (ifp->if_flags & IFF_UP) {
1215 				int s = splimp();
1216 				in6_if_up(ifp);
1217 				splx(s);
1218 			}
1219 #endif
1220 		}
1221 		return (error);
1222 
1223 	}
1224 	return (0);
1225 }
1226 
1227 /*
1228  * Set/clear promiscuous mode on interface ifp based on the truth value
1229  * of pswitch.  The calls are reference counted so that only the first
1230  * "on" request actually has an effect, as does the final "off" request.
1231  * Results are undefined if the "off" and "on" requests are not matched.
1232  */
1233 int
1234 ifpromisc(ifp, pswitch)
1235 	struct ifnet *ifp;
1236 	int pswitch;
1237 {
1238 	struct ifreq ifr;
1239 	int error;
1240 	int oldflags;
1241 
1242 	oldflags = ifp->if_flags;
1243 	if (ifp->if_ipending & IFF_PPROMISC) {
1244 		/* Do nothing if device is in permanently promiscuous mode */
1245 		ifp->if_pcount += pswitch ? 1 : -1;
1246 		return (0);
1247 	}
1248 	if (pswitch) {
1249 		/*
1250 		 * If the device is not configured up, we cannot put it in
1251 		 * promiscuous mode.
1252 		 */
1253 		if ((ifp->if_flags & IFF_UP) == 0)
1254 			return (ENETDOWN);
1255 		if (ifp->if_pcount++ != 0)
1256 			return (0);
1257 		ifp->if_flags |= IFF_PROMISC;
1258 		log(LOG_INFO, "%s%d: promiscuous mode enabled\n",
1259 		    ifp->if_name, ifp->if_unit);
1260 	} else {
1261 		if (--ifp->if_pcount > 0)
1262 			return (0);
1263 		ifp->if_flags &= ~IFF_PROMISC;
1264 		log(LOG_INFO, "%s%d: promiscuous mode disabled\n",
1265 		    ifp->if_name, ifp->if_unit);
1266 	}
1267 	ifr.ifr_flags = ifp->if_flags;
1268 	ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1269 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1270 	if (error == 0)
1271 		rt_ifmsg(ifp);
1272 	else
1273 		ifp->if_flags = oldflags;
1274 	return error;
1275 }
1276 
1277 /*
1278  * Return interface configuration
1279  * of system.  List may be used
1280  * in later ioctl's (above) to get
1281  * other information.
1282  */
1283 /*ARGSUSED*/
1284 static int
1285 ifconf(u_long cmd, caddr_t data)
1286 {
1287 	struct ifconf *ifc = (struct ifconf *)data;
1288 	struct ifnet *ifp;
1289 	struct ifaddr *ifa;
1290 	struct sockaddr *sa;
1291 	struct ifreq ifr, *ifrp;
1292 	int space = ifc->ifc_len, error = 0;
1293 
1294 	ifrp = ifc->ifc_req;
1295 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1296 		char workbuf[64];
1297 		int ifnlen, addrs;
1298 
1299 		if (space <= sizeof (ifr))
1300 			break;
1301 		ifnlen = snprintf(workbuf, sizeof(workbuf),
1302 		    "%s%d", ifp->if_name, ifp->if_unit);
1303 		if(ifnlen + 1 > sizeof ifr.ifr_name) {
1304 			error = ENAMETOOLONG;
1305 			break;
1306 		} else {
1307 			strcpy(ifr.ifr_name, workbuf);
1308 		}
1309 
1310 		addrs = 0;
1311 		ifa = ifp->if_addrhead.tqh_first;
1312 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1313 			if (space <= sizeof(ifr))
1314 				break;
1315 			sa = ifa->ifa_addr;
1316 			if (curproc->p_ucred->cr_prison && prison_if(curthread, sa))
1317 				continue;
1318 			addrs++;
1319 #ifdef COMPAT_43
1320 			if (cmd == OSIOCGIFCONF) {
1321 				struct osockaddr *osa =
1322 					 (struct osockaddr *)&ifr.ifr_addr;
1323 				ifr.ifr_addr = *sa;
1324 				osa->sa_family = sa->sa_family;
1325 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1326 						sizeof (ifr));
1327 				ifrp++;
1328 			} else
1329 #endif
1330 			if (sa->sa_len <= sizeof(*sa)) {
1331 				ifr.ifr_addr = *sa;
1332 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1333 						sizeof (ifr));
1334 				ifrp++;
1335 			} else {
1336 				if (space < sizeof (ifr) + sa->sa_len -
1337 					    sizeof(*sa))
1338 					break;
1339 				space -= sa->sa_len - sizeof(*sa);
1340 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1341 						sizeof (ifr.ifr_name));
1342 				if (error == 0)
1343 				    error = copyout((caddr_t)sa,
1344 				      (caddr_t)&ifrp->ifr_addr, sa->sa_len);
1345 				ifrp = (struct ifreq *)
1346 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1347 			}
1348 			if (error)
1349 				break;
1350 			space -= sizeof (ifr);
1351 		}
1352 		if (error)
1353 			break;
1354 		if (!addrs) {
1355 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1356 			error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1357 			    sizeof (ifr));
1358 			if (error)
1359 				break;
1360 			space -= sizeof (ifr);
1361 			ifrp++;
1362 		}
1363 	}
1364 	ifc->ifc_len -= space;
1365 	return (error);
1366 }
1367 
1368 /*
1369  * Just like if_promisc(), but for all-multicast-reception mode.
1370  */
1371 int
1372 if_allmulti(ifp, onswitch)
1373 	struct ifnet *ifp;
1374 	int onswitch;
1375 {
1376 	int error = 0;
1377 	int s = splimp();
1378 	struct ifreq ifr;
1379 
1380 	if (onswitch) {
1381 		if (ifp->if_amcount++ == 0) {
1382 			ifp->if_flags |= IFF_ALLMULTI;
1383 			ifr.ifr_flags = ifp->if_flags;
1384 			ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1385 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1386 		}
1387 	} else {
1388 		if (ifp->if_amcount > 1) {
1389 			ifp->if_amcount--;
1390 		} else {
1391 			ifp->if_amcount = 0;
1392 			ifp->if_flags &= ~IFF_ALLMULTI;
1393 			ifr.ifr_flags = ifp->if_flags;
1394 			ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1395 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1396 		}
1397 	}
1398 	splx(s);
1399 
1400 	if (error == 0)
1401 		rt_ifmsg(ifp);
1402 	return error;
1403 }
1404 
1405 /*
1406  * Add a multicast listenership to the interface in question.
1407  * The link layer provides a routine which converts
1408  */
1409 int
1410 if_addmulti(ifp, sa, retifma)
1411 	struct ifnet *ifp;	/* interface to manipulate */
1412 	struct sockaddr *sa;	/* address to add */
1413 	struct ifmultiaddr **retifma;
1414 {
1415 	struct sockaddr *llsa, *dupsa;
1416 	int error, s;
1417 	struct ifmultiaddr *ifma;
1418 
1419 	/*
1420 	 * If the matching multicast address already exists
1421 	 * then don't add a new one, just add a reference
1422 	 */
1423 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1424 		if (equal(sa, ifma->ifma_addr)) {
1425 			ifma->ifma_refcount++;
1426 			if (retifma)
1427 				*retifma = ifma;
1428 			return 0;
1429 		}
1430 	}
1431 
1432 	/*
1433 	 * Give the link layer a chance to accept/reject it, and also
1434 	 * find out which AF_LINK address this maps to, if it isn't one
1435 	 * already.
1436 	 */
1437 	if (ifp->if_resolvemulti) {
1438 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1439 		if (error) return error;
1440 	} else {
1441 		llsa = 0;
1442 	}
1443 
1444 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1445 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1446 	bcopy(sa, dupsa, sa->sa_len);
1447 
1448 	ifma->ifma_addr = dupsa;
1449 	ifma->ifma_lladdr = llsa;
1450 	ifma->ifma_ifp = ifp;
1451 	ifma->ifma_refcount = 1;
1452 	ifma->ifma_protospec = 0;
1453 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1454 
1455 	/*
1456 	 * Some network interfaces can scan the address list at
1457 	 * interrupt time; lock them out.
1458 	 */
1459 	s = splimp();
1460 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1461 	splx(s);
1462 	*retifma = ifma;
1463 
1464 	if (llsa != 0) {
1465 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1466 			if (equal(ifma->ifma_addr, llsa))
1467 				break;
1468 		}
1469 		if (ifma) {
1470 			ifma->ifma_refcount++;
1471 		} else {
1472 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1473 			       M_IFMADDR, M_WAITOK);
1474 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1475 			       M_IFMADDR, M_WAITOK);
1476 			bcopy(llsa, dupsa, llsa->sa_len);
1477 			ifma->ifma_addr = dupsa;
1478 			ifma->ifma_ifp = ifp;
1479 			ifma->ifma_refcount = 1;
1480 			s = splimp();
1481 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1482 			splx(s);
1483 		}
1484 	}
1485 	/*
1486 	 * We are certain we have added something, so call down to the
1487 	 * interface to let them know about it.
1488 	 */
1489 	s = splimp();
1490 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0);
1491 	splx(s);
1492 
1493 	return 0;
1494 }
1495 
1496 /*
1497  * Remove a reference to a multicast address on this interface.  Yell
1498  * if the request does not match an existing membership.
1499  */
1500 int
1501 if_delmulti(ifp, sa)
1502 	struct ifnet *ifp;
1503 	struct sockaddr *sa;
1504 {
1505 	struct ifmultiaddr *ifma;
1506 	int s;
1507 
1508 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1509 		if (equal(sa, ifma->ifma_addr))
1510 			break;
1511 	if (ifma == 0)
1512 		return ENOENT;
1513 
1514 	if (ifma->ifma_refcount > 1) {
1515 		ifma->ifma_refcount--;
1516 		return 0;
1517 	}
1518 
1519 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1520 	sa = ifma->ifma_lladdr;
1521 	s = splimp();
1522 	LIST_REMOVE(ifma, ifma_link);
1523 	/*
1524 	 * Make sure the interface driver is notified
1525 	 * in the case of a link layer mcast group being left.
1526 	 */
1527 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1528 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1529 	splx(s);
1530 	free(ifma->ifma_addr, M_IFMADDR);
1531 	free(ifma, M_IFMADDR);
1532 	if (sa == 0)
1533 		return 0;
1534 
1535 	/*
1536 	 * Now look for the link-layer address which corresponds to
1537 	 * this network address.  It had been squirreled away in
1538 	 * ifma->ifma_lladdr for this purpose (so we don't have
1539 	 * to call ifp->if_resolvemulti() again), and we saved that
1540 	 * value in sa above.  If some nasty deleted the
1541 	 * link-layer address out from underneath us, we can deal because
1542 	 * the address we stored was is not the same as the one which was
1543 	 * in the record for the link-layer address.  (So we don't complain
1544 	 * in that case.)
1545 	 */
1546 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1547 		if (equal(sa, ifma->ifma_addr))
1548 			break;
1549 	if (ifma == 0)
1550 		return 0;
1551 
1552 	if (ifma->ifma_refcount > 1) {
1553 		ifma->ifma_refcount--;
1554 		return 0;
1555 	}
1556 
1557 	s = splimp();
1558 	LIST_REMOVE(ifma, ifma_link);
1559 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1560 	splx(s);
1561 	free(ifma->ifma_addr, M_IFMADDR);
1562 	free(sa, M_IFMADDR);
1563 	free(ifma, M_IFMADDR);
1564 
1565 	return 0;
1566 }
1567 
1568 /*
1569  * Set the link layer address on an interface.
1570  *
1571  * At this time we only support certain types of interfaces,
1572  * and we don't allow the length of the address to change.
1573  */
1574 int
1575 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1576 {
1577 	struct sockaddr_dl *sdl;
1578 	struct ifaddr *ifa;
1579 	struct ifreq ifr;
1580 
1581 	ifa = ifnet_addrs[ifp->if_index - 1];
1582 	if (ifa == NULL)
1583 		return (EINVAL);
1584 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1585 	if (sdl == NULL)
1586 		return (EINVAL);
1587 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1588 		return (EINVAL);
1589 	switch (ifp->if_type) {
1590 	case IFT_ETHER:			/* these types use struct arpcom */
1591 	case IFT_FDDI:
1592 	case IFT_XETHER:
1593 	case IFT_ISO88025:
1594 	case IFT_L2VLAN:
1595 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1596 		/* FALLTHROUGH */
1597 	case IFT_ARCNET:
1598 		bcopy(lladdr, LLADDR(sdl), len);
1599 		break;
1600 	default:
1601 		return (ENODEV);
1602 	}
1603 	/*
1604 	 * If the interface is already up, we need
1605 	 * to re-init it in order to reprogram its
1606 	 * address filter.
1607 	 */
1608 	if ((ifp->if_flags & IFF_UP) != 0) {
1609 		ifp->if_flags &= ~IFF_UP;
1610 		ifr.ifr_flags = ifp->if_flags;
1611 		ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1612 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1613 		ifp->if_flags |= IFF_UP;
1614 		ifr.ifr_flags = ifp->if_flags;
1615 		ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1616 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1617 #ifdef INET
1618 		/*
1619 		 * Also send gratuitous ARPs to notify other nodes about
1620 		 * the address change.
1621 		 */
1622 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1623 			if (ifa->ifa_addr != NULL &&
1624 			    ifa->ifa_addr->sa_family == AF_INET)
1625 				arp_ifinit(ifp, ifa);
1626 		}
1627 #endif
1628 	}
1629 	return (0);
1630 }
1631 
1632 struct ifmultiaddr *
1633 ifmaof_ifpforaddr(sa, ifp)
1634 	struct sockaddr *sa;
1635 	struct ifnet *ifp;
1636 {
1637 	struct ifmultiaddr *ifma;
1638 
1639 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1640 		if (equal(ifma->ifma_addr, sa))
1641 			break;
1642 
1643 	return ifma;
1644 }
1645 
1646 int
1647 if_printf(struct ifnet *ifp, const char *fmt, ...)
1648 {
1649 	va_list ap;
1650 	int retval;
1651 
1652 	retval = printf("%s%d: ", ifp->if_name, ifp->if_unit);
1653 	va_start(ap, fmt);
1654 	retval += vprintf(fmt, ap);
1655 	va_end(ap);
1656 	return (retval);
1657 }
1658 
1659 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1660 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1661