xref: /dflybsd-src/sys/net/if.c (revision c6cf4f8f1ebc9e3fe2a8c566f08adfc86122c7bf)
1 /*
2  * Copyright (c) 1980, 1986, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)if.c	8.3 (Berkeley) 1/4/94
34  * $FreeBSD: src/sys/net/if.c,v 1.185 2004/03/13 02:35:03 brooks Exp $
35  * $DragonFly: src/sys/net/if.c,v 1.29 2005/03/04 02:21:48 hsu Exp $
36  */
37 
38 #include "opt_compat.h"
39 #include "opt_inet6.h"
40 #include "opt_inet.h"
41 
42 #include <sys/param.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/systm.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/socketops.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/sockio.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 #include <sys/domain.h>
57 
58 #include <net/if.h>
59 #include <net/if_arp.h>
60 #include <net/if_dl.h>
61 #include <net/if_types.h>
62 #include <net/if_var.h>
63 #include <net/ifq_var.h>
64 #include <net/radix.h>
65 #include <net/route.h>
66 #include <machine/stdarg.h>
67 
68 #if defined(INET) || defined(INET6)
69 /*XXX*/
70 #include <netinet/in.h>
71 #include <netinet/in_var.h>
72 #include <netinet/if_ether.h>
73 #ifdef INET6
74 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
75 #include <netinet6/in6_var.h>
76 #include <netinet6/in6_ifattach.h>
77 #endif
78 #endif
79 
80 #if defined(COMPAT_43)
81 #include <emulation/43bsd/43bsd_socket.h>
82 #endif /* COMPAT_43 */
83 
84 /*
85  * System initialization
86  */
87 
88 static void	if_attachdomain(void *);
89 static void	if_attachdomain1(struct ifnet *);
90 static int ifconf (u_long, caddr_t, struct thread *);
91 static void ifinit (void *);
92 static void if_slowtimo (void *);
93 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
94 static int  if_rtdel (struct radix_node *, void *);
95 
96 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
97 
98 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
99 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
100 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
101 
102 int	ifqmaxlen = IFQ_MAXLEN;
103 struct	ifnethead ifnet;	/* depend on static init XXX */
104 
105 #ifdef INET6
106 /*
107  * XXX: declare here to avoid to include many inet6 related files..
108  * should be more generalized?
109  */
110 extern void	nd6_setmtu (struct ifnet *);
111 #endif
112 
113 struct if_clone *if_clone_lookup (const char *, int *);
114 int if_clone_list (struct if_clonereq *);
115 
116 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
117 int if_cloners_count;
118 
119 struct callout if_slowtimo_timer;
120 
121 /*
122  * Network interface utility routines.
123  *
124  * Routines with ifa_ifwith* names take sockaddr *'s as
125  * parameters.
126  */
127 /* ARGSUSED*/
128 void
129 ifinit(void *dummy)
130 {
131 	struct ifnet *ifp;
132 	int s;
133 
134 	callout_init(&if_slowtimo_timer);
135 
136 	s = splimp();
137 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
138 		if (ifp->if_snd.ifq_maxlen == 0) {
139 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
140 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
141 		}
142 	}
143 	splx(s);
144 
145 	if_slowtimo(0);
146 }
147 
148 int if_index = 0;
149 struct ifaddr **ifnet_addrs;
150 struct ifnet **ifindex2ifnet = NULL;
151 
152 /*
153  * Attach an interface to the
154  * list of "active" interfaces.
155  */
156 void
157 if_attach(struct ifnet *ifp)
158 {
159 	unsigned socksize, ifasize;
160 	int namelen, masklen;
161 	struct sockaddr_dl *sdl;
162 	struct ifaddr *ifa;
163 
164 	static int if_indexlim = 8;
165 	static boolean_t inited;
166 
167 	if (!inited) {
168 		TAILQ_INIT(&ifnet);
169 		inited = TRUE;
170 	}
171 
172 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
173 	ifp->if_index = ++if_index;
174 	/*
175 	 * XXX -
176 	 * The old code would work if the interface passed a pre-existing
177 	 * chain of ifaddrs to this code.  We don't trust our callers to
178 	 * properly initialize the tailq, however, so we no longer allow
179 	 * this unlikely case.
180 	 */
181 	TAILQ_INIT(&ifp->if_addrhead);
182 	TAILQ_INIT(&ifp->if_prefixhead);
183 	LIST_INIT(&ifp->if_multiaddrs);
184 	getmicrotime(&ifp->if_lastchange);
185 	if (ifnet_addrs == NULL || if_index >= if_indexlim) {
186 		unsigned int n;
187 		caddr_t q;
188 
189 		if_indexlim <<= 1;
190 		n = if_indexlim * sizeof(struct ifaddr *);
191 		q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
192 		if (ifnet_addrs != NULL) {
193 			bcopy(ifnet_addrs, q, n/2);
194 			free(ifnet_addrs, M_IFADDR);
195 		}
196 		ifnet_addrs = (struct ifaddr **)q;
197 
198 		/* grow ifindex2ifnet */
199 		n = if_indexlim * sizeof(struct ifnet *);
200 		q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
201 		if (ifindex2ifnet) {
202 			bcopy(ifindex2ifnet, q, n/2);
203 			free(ifindex2ifnet, M_IFADDR);
204 		}
205 		ifindex2ifnet = (struct ifnet **)q;
206 	}
207 
208 	ifindex2ifnet[if_index] = ifp;
209 
210 	/*
211 	 * create a Link Level name for this device
212 	 */
213 	namelen = strlen(ifp->if_xname);
214 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
215 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
216 	socksize = masklen + ifp->if_addrlen;
217 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
218 	if (socksize < sizeof(*sdl))
219 		socksize = sizeof(*sdl);
220 	socksize = ROUNDUP(socksize);
221 	ifasize = sizeof(struct ifaddr) + 2 * socksize;
222 	ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
223 	sdl = (struct sockaddr_dl *)(ifa + 1);
224 	sdl->sdl_len = socksize;
225 	sdl->sdl_family = AF_LINK;
226 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
227 	sdl->sdl_nlen = namelen;
228 	sdl->sdl_index = ifp->if_index;
229 	sdl->sdl_type = ifp->if_type;
230 	ifnet_addrs[if_index - 1] = ifa;
231 	ifa->ifa_ifp = ifp;
232 	ifa->ifa_rtrequest = link_rtrequest;
233 	ifa->ifa_addr = (struct sockaddr *)sdl;
234 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
235 	ifa->ifa_netmask = (struct sockaddr *)sdl;
236 	sdl->sdl_len = masklen;
237 	while (namelen != 0)
238 		sdl->sdl_data[--namelen] = 0xff;
239 	TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
240 
241 	EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
242 
243 	ifp->if_snd.altq_type = 0;
244 	ifp->if_snd.altq_disc = NULL;
245 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
246 	ifp->if_snd.altq_tbr = NULL;
247 	ifp->if_snd.altq_ifp = ifp;
248 
249 	if (!SLIST_EMPTY(&domains))
250 		if_attachdomain1(ifp);
251 
252 	/* Announce the interface. */
253 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
254 }
255 
256 static void
257 if_attachdomain(void *dummy)
258 {
259 	struct ifnet *ifp;
260 	int s;
261 
262 	s = splnet();
263 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
264 		if_attachdomain1(ifp);
265 	splx(s);
266 }
267 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
268 	if_attachdomain, NULL);
269 
270 static void
271 if_attachdomain1(struct ifnet *ifp)
272 {
273 	struct domain *dp;
274 	int s;
275 
276 	s = splnet();
277 
278 	/* address family dependent data region */
279 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
280 	SLIST_FOREACH(dp, &domains, dom_next)
281 		if (dp->dom_ifattach)
282 			ifp->if_afdata[dp->dom_family] =
283 				(*dp->dom_ifattach)(ifp);
284 	splx(s);
285 }
286 
287 /*
288  * Detach an interface, removing it from the
289  * list of "active" interfaces.
290  */
291 void
292 if_detach(struct ifnet *ifp)
293 {
294 	struct ifaddr *ifa;
295 	struct radix_node_head	*rnh;
296 	int s;
297 	int i;
298 	struct domain *dp;
299 
300 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
301 
302 	/*
303 	 * Remove routes and flush queues.
304 	 */
305 	s = splnet();
306 	if_down(ifp);
307 
308 	if (ifq_is_enabled(&ifp->if_snd))
309 		altq_disable(&ifp->if_snd);
310 	if (ifq_is_attached(&ifp->if_snd))
311 		altq_detach(&ifp->if_snd);
312 
313 	/*
314 	 * Remove address from ifnet_addrs[] and maybe decrement if_index.
315 	 * Clean up all addresses.
316 	 */
317 	ifnet_addrs[ifp->if_index - 1] = 0;
318 	while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
319 		if_index--;
320 
321 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
322 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
323 #ifdef INET
324 		/* XXX: Ugly!! ad hoc just for INET */
325 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
326 			struct ifaliasreq ifr;
327 
328 			bzero(&ifr, sizeof ifr);
329 			ifr.ifra_addr = *ifa->ifa_addr;
330 			if (ifa->ifa_dstaddr)
331 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
332 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
333 				       NULL) == 0)
334 				continue;
335 		}
336 #endif /* INET */
337 #ifdef INET6
338 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
339 			in6_purgeaddr(ifa);
340 			/* ifp_addrhead is already updated */
341 			continue;
342 		}
343 #endif /* INET6 */
344 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
345 		IFAFREE(ifa);
346 	}
347 
348 #ifdef INET6
349 	/*
350 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
351 	 * before removing routing entries below, since IPv6 interface direct
352 	 * routes are expected to be removed by the IPv6-specific kernel API.
353 	 * Otherwise, the kernel will detect some inconsistency and bark it.
354 	 */
355 	in6_ifdetach(ifp);
356 #endif
357 
358 	/*
359 	 * Delete all remaining routes using this interface
360 	 * Unfortuneatly the only way to do this is to slog through
361 	 * the entire routing table looking for routes which point
362 	 * to this interface...oh well...
363 	 */
364 	for (i = 1; i <= AF_MAX; i++) {
365 		if ((rnh = rt_tables[i]) == NULL)
366 			continue;
367 		rnh->rnh_walktree(rnh, if_rtdel, ifp);
368 	}
369 
370 	/* Announce that the interface is gone. */
371 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
372 
373 	SLIST_FOREACH(dp, &domains, dom_next)
374 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
375 			(*dp->dom_ifdetach)(ifp,
376 				ifp->if_afdata[dp->dom_family]);
377 
378 	ifindex2ifnet[ifp->if_index] = NULL;
379 
380 	TAILQ_REMOVE(&ifnet, ifp, if_link);
381 	splx(s);
382 }
383 
384 /*
385  * Delete Routes for a Network Interface
386  *
387  * Called for each routing entry via the rnh->rnh_walktree() call above
388  * to delete all route entries referencing a detaching network interface.
389  *
390  * Arguments:
391  *	rn	pointer to node in the routing table
392  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
393  *
394  * Returns:
395  *	0	successful
396  *	errno	failed - reason indicated
397  *
398  */
399 static int
400 if_rtdel(struct radix_node *rn, void *arg)
401 {
402 	struct rtentry	*rt = (struct rtentry *)rn;
403 	struct ifnet	*ifp = arg;
404 	int		err;
405 
406 	if (rt->rt_ifp == ifp) {
407 
408 		/*
409 		 * Protect (sorta) against walktree recursion problems
410 		 * with cloned routes
411 		 */
412 		if (!(rt->rt_flags & RTF_UP))
413 			return (0);
414 
415 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
416 				rt_mask(rt), rt->rt_flags,
417 				(struct rtentry **) NULL);
418 		if (err) {
419 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
420 		}
421 	}
422 
423 	return (0);
424 }
425 
426 /*
427  * Create a clone network interface.
428  */
429 int
430 if_clone_create(char *name, int len)
431 {
432 	struct if_clone *ifc;
433 	char *dp;
434 	int wildcard, bytoff, bitoff;
435 	int unit;
436 	int err;
437 
438 	ifc = if_clone_lookup(name, &unit);
439 	if (ifc == NULL)
440 		return (EINVAL);
441 
442 	if (ifunit(name) != NULL)
443 		return (EEXIST);
444 
445 	bytoff = bitoff = 0;
446 	wildcard = (unit < 0);
447 	/*
448 	 * Find a free unit if none was given.
449 	 */
450 	if (wildcard) {
451 		while (bytoff < ifc->ifc_bmlen &&
452 		    ifc->ifc_units[bytoff] == 0xff)
453 			bytoff++;
454 		if (bytoff >= ifc->ifc_bmlen)
455 			return (ENOSPC);
456 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
457 			bitoff++;
458 		unit = (bytoff << 3) + bitoff;
459 	}
460 
461 	if (unit > ifc->ifc_maxunit)
462 		return (ENXIO);
463 
464 	err = (*ifc->ifc_create)(ifc, unit);
465 	if (err != 0)
466 		return (err);
467 
468 	if (!wildcard) {
469 		bytoff = unit >> 3;
470 		bitoff = unit - (bytoff << 3);
471 	}
472 
473 	/*
474 	 * Allocate the unit in the bitmap.
475 	 */
476 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
477 	    ("%s: bit is already set", __func__));
478 	ifc->ifc_units[bytoff] |= (1 << bitoff);
479 
480 	/* In the wildcard case, we need to update the name. */
481 	if (wildcard) {
482 		for (dp = name; *dp != '\0'; dp++);
483 		if (snprintf(dp, len - (dp-name), "%d", unit) >
484 		    len - (dp-name) - 1) {
485 			/*
486 			 * This can only be a programmer error and
487 			 * there's no straightforward way to recover if
488 			 * it happens.
489 			 */
490 			panic("if_clone_create(): interface name too long");
491 		}
492 
493 	}
494 
495 	EVENTHANDLER_INVOKE(if_clone_event, ifc);
496 
497 	return (0);
498 }
499 
500 /*
501  * Destroy a clone network interface.
502  */
503 int
504 if_clone_destroy(const char *name)
505 {
506 	struct if_clone *ifc;
507 	struct ifnet *ifp;
508 	int bytoff, bitoff;
509 	int unit;
510 
511 	ifc = if_clone_lookup(name, &unit);
512 	if (ifc == NULL)
513 		return (EINVAL);
514 
515 	if (unit < ifc->ifc_minifs)
516 		return (EINVAL);
517 
518 	ifp = ifunit(name);
519 	if (ifp == NULL)
520 		return (ENXIO);
521 
522 	if (ifc->ifc_destroy == NULL)
523 		return (EOPNOTSUPP);
524 
525 	(*ifc->ifc_destroy)(ifp);
526 
527 	/*
528 	 * Compute offset in the bitmap and deallocate the unit.
529 	 */
530 	bytoff = unit >> 3;
531 	bitoff = unit - (bytoff << 3);
532 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
533 	    ("%s: bit is already cleared", __func__));
534 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
535 	return (0);
536 }
537 
538 /*
539  * Look up a network interface cloner.
540  */
541 struct if_clone *
542 if_clone_lookup(const char *name, int *unitp)
543 {
544 	struct if_clone *ifc;
545 	const char *cp;
546 	int i;
547 
548 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
549 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
550 			if (ifc->ifc_name[i] != *cp)
551 				goto next_ifc;
552 		}
553 		goto found_name;
554  next_ifc:
555 		ifc = LIST_NEXT(ifc, ifc_list);
556 	}
557 
558 	/* No match. */
559 	return ((struct if_clone *)NULL);
560 
561  found_name:
562 	if (*cp == '\0') {
563 		i = -1;
564 	} else {
565 		for (i = 0; *cp != '\0'; cp++) {
566 			if (*cp < '0' || *cp > '9') {
567 				/* Bogus unit number. */
568 				return (NULL);
569 			}
570 			i = (i * 10) + (*cp - '0');
571 		}
572 	}
573 
574 	if (unitp != NULL)
575 		*unitp = i;
576 	return (ifc);
577 }
578 
579 /*
580  * Register a network interface cloner.
581  */
582 void
583 if_clone_attach(struct if_clone *ifc)
584 {
585 	int bytoff, bitoff;
586 	int err;
587 	int len, maxclone;
588 	int unit;
589 
590 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
591 	    ("%s: %s requested more units then allowed (%d > %d)",
592 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
593 	    ifc->ifc_maxunit + 1));
594 	/*
595 	 * Compute bitmap size and allocate it.
596 	 */
597 	maxclone = ifc->ifc_maxunit + 1;
598 	len = maxclone >> 3;
599 	if ((len << 3) < maxclone)
600 		len++;
601 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
602 	ifc->ifc_bmlen = len;
603 
604 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
605 	if_cloners_count++;
606 
607 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
608 		err = (*ifc->ifc_create)(ifc, unit);
609 		KASSERT(err == 0,
610 		    ("%s: failed to create required interface %s%d",
611 		    __func__, ifc->ifc_name, unit));
612 
613 		/* Allocate the unit in the bitmap. */
614 		bytoff = unit >> 3;
615 		bitoff = unit - (bytoff << 3);
616 		ifc->ifc_units[bytoff] |= (1 << bitoff);
617 	}
618 }
619 
620 /*
621  * Unregister a network interface cloner.
622  */
623 void
624 if_clone_detach(struct if_clone *ifc)
625 {
626 
627 	LIST_REMOVE(ifc, ifc_list);
628 	free(ifc->ifc_units, M_CLONE);
629 	if_cloners_count--;
630 }
631 
632 /*
633  * Provide list of interface cloners to userspace.
634  */
635 int
636 if_clone_list(struct if_clonereq *ifcr)
637 {
638 	char outbuf[IFNAMSIZ], *dst;
639 	struct if_clone *ifc;
640 	int count, error = 0;
641 
642 	ifcr->ifcr_total = if_cloners_count;
643 	if ((dst = ifcr->ifcr_buffer) == NULL) {
644 		/* Just asking how many there are. */
645 		return (0);
646 	}
647 
648 	if (ifcr->ifcr_count < 0)
649 		return (EINVAL);
650 
651 	count = (if_cloners_count < ifcr->ifcr_count) ?
652 	    if_cloners_count : ifcr->ifcr_count;
653 
654 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
655 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
656 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
657 		error = copyout(outbuf, dst, IFNAMSIZ);
658 		if (error)
659 			break;
660 	}
661 
662 	return (error);
663 }
664 
665 /*
666  * Locate an interface based on a complete address.
667  */
668 struct ifaddr *
669 ifa_ifwithaddr(struct sockaddr *addr)
670 {
671 	struct ifnet *ifp;
672 	struct ifaddr *ifa;
673 
674 	TAILQ_FOREACH(ifp, &ifnet, if_link)
675 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
676 		if (ifa->ifa_addr->sa_family != addr->sa_family)
677 			continue;
678 		if (sa_equal(addr, ifa->ifa_addr))
679 			return (ifa);
680 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
681 		    /* IPv6 doesn't have broadcast */
682 		    ifa->ifa_broadaddr->sa_len != 0 &&
683 		    sa_equal(ifa->ifa_broadaddr, addr))
684 			return (ifa);
685 	}
686 	return ((struct ifaddr *)NULL);
687 }
688 /*
689  * Locate the point to point interface with a given destination address.
690  */
691 struct ifaddr *
692 ifa_ifwithdstaddr(struct sockaddr *addr)
693 {
694 	struct ifnet *ifp;
695 	struct ifaddr *ifa;
696 
697 	TAILQ_FOREACH(ifp, &ifnet, if_link)
698 	    if (ifp->if_flags & IFF_POINTOPOINT)
699 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
700 			if (ifa->ifa_addr->sa_family != addr->sa_family)
701 				continue;
702 			if (ifa->ifa_dstaddr &&
703 			    sa_equal(addr, ifa->ifa_dstaddr))
704 				return (ifa);
705 	}
706 	return ((struct ifaddr *)NULL);
707 }
708 
709 /*
710  * Find an interface on a specific network.  If many, choice
711  * is most specific found.
712  */
713 struct ifaddr *
714 ifa_ifwithnet(struct sockaddr *addr)
715 {
716 	struct ifnet *ifp;
717 	struct ifaddr *ifa;
718 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
719 	u_int af = addr->sa_family;
720 	char *addr_data = addr->sa_data, *cplim;
721 
722 	/*
723 	 * AF_LINK addresses can be looked up directly by their index number,
724 	 * so do that if we can.
725 	 */
726 	if (af == AF_LINK) {
727 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
728 
729 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
730 		return (ifnet_addrs[sdl->sdl_index - 1]);
731 	}
732 
733 	/*
734 	 * Scan though each interface, looking for ones that have
735 	 * addresses in this address family.
736 	 */
737 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
738 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
739 			char *cp, *cp2, *cp3;
740 
741 			if (ifa->ifa_addr->sa_family != af)
742 next:				continue;
743 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
744 				/*
745 				 * This is a bit broken as it doesn't
746 				 * take into account that the remote end may
747 				 * be a single node in the network we are
748 				 * looking for.
749 				 * The trouble is that we don't know the
750 				 * netmask for the remote end.
751 				 */
752 				if (ifa->ifa_dstaddr != NULL &&
753 				    sa_equal(addr, ifa->ifa_dstaddr))
754 					return (ifa);
755 			} else {
756 				/*
757 				 * if we have a special address handler,
758 				 * then use it instead of the generic one.
759 				 */
760 				if (ifa->ifa_claim_addr) {
761 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
762 						return (ifa);
763 					} else {
764 						continue;
765 					}
766 				}
767 
768 				/*
769 				 * Scan all the bits in the ifa's address.
770 				 * If a bit dissagrees with what we are
771 				 * looking for, mask it with the netmask
772 				 * to see if it really matters.
773 				 * (A byte at a time)
774 				 */
775 				if (ifa->ifa_netmask == 0)
776 					continue;
777 				cp = addr_data;
778 				cp2 = ifa->ifa_addr->sa_data;
779 				cp3 = ifa->ifa_netmask->sa_data;
780 				cplim = ifa->ifa_netmask->sa_len +
781 					(char *)ifa->ifa_netmask;
782 				while (cp3 < cplim)
783 					if ((*cp++ ^ *cp2++) & *cp3++)
784 						goto next; /* next address! */
785 				/*
786 				 * If the netmask of what we just found
787 				 * is more specific than what we had before
788 				 * (if we had one) then remember the new one
789 				 * before continuing to search
790 				 * for an even better one.
791 				 */
792 				if (ifa_maybe == 0 ||
793 				    rn_refines((char *)ifa->ifa_netmask,
794 					       (char *)ifa_maybe->ifa_netmask))
795 					ifa_maybe = ifa;
796 			}
797 		}
798 	}
799 	return (ifa_maybe);
800 }
801 
802 /*
803  * Find an interface address specific to an interface best matching
804  * a given address.
805  */
806 struct ifaddr *
807 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
808 {
809 	struct ifaddr *ifa;
810 	char *cp, *cp2, *cp3;
811 	char *cplim;
812 	struct ifaddr *ifa_maybe = 0;
813 	u_int af = addr->sa_family;
814 
815 	if (af >= AF_MAX)
816 		return (0);
817 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
818 		if (ifa->ifa_addr->sa_family != af)
819 			continue;
820 		if (ifa_maybe == 0)
821 			ifa_maybe = ifa;
822 		if (ifa->ifa_netmask == NULL) {
823 			if (sa_equal(addr, ifa->ifa_addr) ||
824 			    (ifa->ifa_dstaddr != NULL &&
825 			     sa_equal(addr, ifa->ifa_dstaddr)))
826 				return (ifa);
827 			continue;
828 		}
829 		if (ifp->if_flags & IFF_POINTOPOINT) {
830 			if (sa_equal(addr, ifa->ifa_dstaddr))
831 				return (ifa);
832 		} else {
833 			cp = addr->sa_data;
834 			cp2 = ifa->ifa_addr->sa_data;
835 			cp3 = ifa->ifa_netmask->sa_data;
836 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
837 			for (; cp3 < cplim; cp3++)
838 				if ((*cp++ ^ *cp2++) & *cp3)
839 					break;
840 			if (cp3 == cplim)
841 				return (ifa);
842 		}
843 	}
844 	return (ifa_maybe);
845 }
846 
847 #include <net/route.h>
848 
849 /*
850  * Default action when installing a route with a Link Level gateway.
851  * Lookup an appropriate real ifa to point to.
852  * This should be moved to /sys/net/link.c eventually.
853  */
854 static void
855 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
856 {
857 	struct ifaddr *ifa;
858 	struct sockaddr *dst;
859 	struct ifnet *ifp;
860 
861 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
862 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
863 		return;
864 	ifa = ifaof_ifpforaddr(dst, ifp);
865 	if (ifa != NULL) {
866 		IFAFREE(rt->rt_ifa);
867 		IFAREF(ifa);
868 		rt->rt_ifa = ifa;
869 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
870 			ifa->ifa_rtrequest(cmd, rt, info);
871 	}
872 }
873 
874 /*
875  * Mark an interface down and notify protocols of
876  * the transition.
877  * NOTE: must be called at splnet or eqivalent.
878  */
879 void
880 if_unroute(struct ifnet *ifp, int flag, int fam)
881 {
882 	struct ifaddr *ifa;
883 
884 	ifp->if_flags &= ~flag;
885 	getmicrotime(&ifp->if_lastchange);
886 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
887 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
888 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
889 	ifq_purge(&ifp->if_snd);
890 	rt_ifmsg(ifp);
891 }
892 
893 /*
894  * Mark an interface up and notify protocols of
895  * the transition.
896  * NOTE: must be called at splnet or eqivalent.
897  */
898 void
899 if_route(struct ifnet *ifp, int flag, int fam)
900 {
901 	struct ifaddr *ifa;
902 
903 	ifp->if_flags |= flag;
904 	getmicrotime(&ifp->if_lastchange);
905 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
906 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
907 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
908 	rt_ifmsg(ifp);
909 #ifdef INET6
910 	in6_if_up(ifp);
911 #endif
912 }
913 
914 /*
915  * Mark an interface down and notify protocols of the transition.  An
916  * interface going down is also considered to be a synchronizing event.
917  * We must ensure that all packet processing related to the interface
918  * has completed before we return so e.g. the caller can free the ifnet
919  * structure that the mbufs may be referencing.
920  *
921  * NOTE: must be called at splnet or eqivalent.
922  */
923 void
924 if_down(struct ifnet *ifp)
925 {
926 
927 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
928 	netmsg_service_sync();
929 }
930 
931 /*
932  * Mark an interface up and notify protocols of
933  * the transition.
934  * NOTE: must be called at splnet or eqivalent.
935  */
936 void
937 if_up(struct ifnet *ifp)
938 {
939 
940 	if_route(ifp, IFF_UP, AF_UNSPEC);
941 }
942 
943 /*
944  * Handle interface watchdog timer routines.  Called
945  * from softclock, we decrement timers (if set) and
946  * call the appropriate interface routine on expiration.
947  */
948 static void
949 if_slowtimo(void *arg)
950 {
951 	struct ifnet *ifp;
952 	int s = splimp();
953 
954 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
955 		if (ifp->if_timer == 0 || --ifp->if_timer)
956 			continue;
957 		if (ifp->if_watchdog)
958 			(*ifp->if_watchdog)(ifp);
959 	}
960 	splx(s);
961 	callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
962 }
963 
964 /*
965  * Map interface name to
966  * interface structure pointer.
967  */
968 struct ifnet *
969 ifunit(const char *name)
970 {
971 	struct ifnet *ifp;
972 
973 	/*
974 	 * Search all the interfaces for this name/number
975 	 */
976 
977 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
978 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
979 			break;
980 	}
981 	return (ifp);
982 }
983 
984 
985 /*
986  * Map interface name in a sockaddr_dl to
987  * interface structure pointer.
988  */
989 struct ifnet *
990 if_withname(struct sockaddr *sa)
991 {
992 	char ifname[IFNAMSIZ+1];
993 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
994 
995 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
996 	     (sdl->sdl_nlen > IFNAMSIZ) )
997 		return NULL;
998 
999 	/*
1000 	 * ifunit wants a null-terminated name.  It may not be null-terminated
1001 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
1002 	 * and there might not be room to put the trailing null anyway, so we
1003 	 * make a local copy that we know we can null terminate safely.
1004 	 */
1005 
1006 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1007 	ifname[sdl->sdl_nlen] = '\0';
1008 	return ifunit(ifname);
1009 }
1010 
1011 
1012 /*
1013  * Interface ioctls.
1014  */
1015 int
1016 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1017 {
1018 	struct ifnet *ifp;
1019 	struct ifreq *ifr;
1020 	struct ifstat *ifs;
1021 	int error;
1022 	short oif_flags;
1023 	int new_flags;
1024 	size_t namelen, onamelen;
1025 	char new_name[IFNAMSIZ];
1026 	struct ifaddr *ifa;
1027 	struct sockaddr_dl *sdl;
1028 
1029 	switch (cmd) {
1030 
1031 	case SIOCGIFCONF:
1032 	case OSIOCGIFCONF:
1033 		return (ifconf(cmd, data, td));
1034 	}
1035 	ifr = (struct ifreq *)data;
1036 
1037 	switch (cmd) {
1038 	case SIOCIFCREATE:
1039 	case SIOCIFDESTROY:
1040 		if ((error = suser(td)) != 0)
1041 			return (error);
1042 		return ((cmd == SIOCIFCREATE) ?
1043 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1044 			if_clone_destroy(ifr->ifr_name));
1045 
1046 	case SIOCIFGCLONERS:
1047 		return (if_clone_list((struct if_clonereq *)data));
1048 	}
1049 
1050 	ifp = ifunit(ifr->ifr_name);
1051 	if (ifp == 0)
1052 		return (ENXIO);
1053 	switch (cmd) {
1054 
1055 	case SIOCGIFFLAGS:
1056 		ifr->ifr_flags = ifp->if_flags;
1057 		ifr->ifr_flagshigh = ifp->if_flags >> 16;
1058 		break;
1059 
1060 	case SIOCGIFCAP:
1061 		ifr->ifr_reqcap = ifp->if_capabilities;
1062 		ifr->ifr_curcap = ifp->if_capenable;
1063 		break;
1064 
1065 	case SIOCGIFMETRIC:
1066 		ifr->ifr_metric = ifp->if_metric;
1067 		break;
1068 
1069 	case SIOCGIFMTU:
1070 		ifr->ifr_mtu = ifp->if_mtu;
1071 		break;
1072 
1073 	case SIOCGIFPHYS:
1074 		ifr->ifr_phys = ifp->if_physical;
1075 		break;
1076 
1077 	case SIOCSIFFLAGS:
1078 		error = suser(td);
1079 		if (error)
1080 			return (error);
1081 		new_flags = (ifr->ifr_flags & 0xffff) |
1082 		    (ifr->ifr_flagshigh << 16);
1083 		if (ifp->if_flags & IFF_SMART) {
1084 			/* Smart drivers twiddle their own routes */
1085 		} else if (ifp->if_flags & IFF_UP &&
1086 		    (new_flags & IFF_UP) == 0) {
1087 			int s = splimp();
1088 			if_down(ifp);
1089 			splx(s);
1090 		} else if (new_flags & IFF_UP &&
1091 		    (ifp->if_flags & IFF_UP) == 0) {
1092 			int s = splimp();
1093 			if_up(ifp);
1094 			splx(s);
1095 		}
1096 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1097 			(new_flags &~ IFF_CANTCHANGE);
1098 		if (new_flags & IFF_PPROMISC) {
1099 			/* Permanently promiscuous mode requested */
1100 			ifp->if_flags |= IFF_PROMISC;
1101 		} else if (ifp->if_pcount == 0) {
1102 			ifp->if_flags &= ~IFF_PROMISC;
1103 		}
1104 		if (ifp->if_ioctl)
1105 			(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1106 		getmicrotime(&ifp->if_lastchange);
1107 		break;
1108 
1109 	case SIOCSIFCAP:
1110 		error = suser(td);
1111 		if (error)
1112 			return (error);
1113 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1114 			return (EINVAL);
1115 		(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1116 		break;
1117 
1118 	case SIOCSIFNAME:
1119 		error = suser(td);
1120 		if (error != 0)
1121 			return (error);
1122 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1123 		if (error != 0)
1124 			return (error);
1125 		if (new_name[0] == '\0')
1126 			return (EINVAL);
1127 		if (ifunit(new_name) != NULL)
1128 			return (EEXIST);
1129 
1130 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1131 
1132 		/* Announce the departure of the interface. */
1133 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1134 
1135 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1136 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1137 		/* XXX IFA_LOCK(ifa); */
1138 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1139 		namelen = strlen(new_name);
1140 		onamelen = sdl->sdl_nlen;
1141 		/*
1142 		 * Move the address if needed.  This is safe because we
1143 		 * allocate space for a name of length IFNAMSIZ when we
1144 		 * create this in if_attach().
1145 		 */
1146 		if (namelen != onamelen) {
1147 			bcopy(sdl->sdl_data + onamelen,
1148 			    sdl->sdl_data + namelen, sdl->sdl_alen);
1149 		}
1150 		bcopy(new_name, sdl->sdl_data, namelen);
1151 		sdl->sdl_nlen = namelen;
1152 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1153 		bzero(sdl->sdl_data, onamelen);
1154 		while (namelen != 0)
1155 			sdl->sdl_data[--namelen] = 0xff;
1156 		/* XXX IFA_UNLOCK(ifa) */
1157 
1158 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1159 
1160 		/* Announce the return of the interface. */
1161 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1162 		break;
1163 
1164 	case SIOCSIFMETRIC:
1165 		error = suser(td);
1166 		if (error)
1167 			return (error);
1168 		ifp->if_metric = ifr->ifr_metric;
1169 		getmicrotime(&ifp->if_lastchange);
1170 		break;
1171 
1172 	case SIOCSIFPHYS:
1173 		error = suser(td);
1174 		if (error)
1175 			return error;
1176 		if (!ifp->if_ioctl)
1177 		        return EOPNOTSUPP;
1178 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1179 		if (error == 0)
1180 			getmicrotime(&ifp->if_lastchange);
1181 		return (error);
1182 
1183 	case SIOCSIFMTU:
1184 	{
1185 		u_long oldmtu = ifp->if_mtu;
1186 
1187 		error = suser(td);
1188 		if (error)
1189 			return (error);
1190 		if (ifp->if_ioctl == NULL)
1191 			return (EOPNOTSUPP);
1192 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1193 			return (EINVAL);
1194 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1195 		if (error == 0) {
1196 			getmicrotime(&ifp->if_lastchange);
1197 			rt_ifmsg(ifp);
1198 		}
1199 		/*
1200 		 * If the link MTU changed, do network layer specific procedure.
1201 		 */
1202 		if (ifp->if_mtu != oldmtu) {
1203 #ifdef INET6
1204 			nd6_setmtu(ifp);
1205 #endif
1206 		}
1207 		return (error);
1208 	}
1209 
1210 	case SIOCADDMULTI:
1211 	case SIOCDELMULTI:
1212 		error = suser(td);
1213 		if (error)
1214 			return (error);
1215 
1216 		/* Don't allow group membership on non-multicast interfaces. */
1217 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1218 			return EOPNOTSUPP;
1219 
1220 		/* Don't let users screw up protocols' entries. */
1221 		if (ifr->ifr_addr.sa_family != AF_LINK)
1222 			return EINVAL;
1223 
1224 		if (cmd == SIOCADDMULTI) {
1225 			struct ifmultiaddr *ifma;
1226 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1227 		} else {
1228 			error = if_delmulti(ifp, &ifr->ifr_addr);
1229 		}
1230 		if (error == 0)
1231 			getmicrotime(&ifp->if_lastchange);
1232 		return error;
1233 
1234 	case SIOCSIFPHYADDR:
1235 	case SIOCDIFPHYADDR:
1236 #ifdef INET6
1237 	case SIOCSIFPHYADDR_IN6:
1238 #endif
1239 	case SIOCSLIFPHYADDR:
1240         case SIOCSIFMEDIA:
1241 	case SIOCSIFGENERIC:
1242 		error = suser(td);
1243 		if (error)
1244 			return (error);
1245 		if (ifp->if_ioctl == 0)
1246 			return (EOPNOTSUPP);
1247 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1248 		if (error == 0)
1249 			getmicrotime(&ifp->if_lastchange);
1250 		return error;
1251 
1252 	case SIOCGIFSTATUS:
1253 		ifs = (struct ifstat *)data;
1254 		ifs->ascii[0] = '\0';
1255 
1256 	case SIOCGIFPSRCADDR:
1257 	case SIOCGIFPDSTADDR:
1258 	case SIOCGLIFPHYADDR:
1259 	case SIOCGIFMEDIA:
1260 	case SIOCGIFGENERIC:
1261 		if (ifp->if_ioctl == 0)
1262 			return (EOPNOTSUPP);
1263 		return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1264 
1265 	case SIOCSIFLLADDR:
1266 		error = suser(td);
1267 		if (error)
1268 			return (error);
1269 		return if_setlladdr(ifp,
1270 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1271 
1272 	default:
1273 		oif_flags = ifp->if_flags;
1274 		if (so->so_proto == 0)
1275 			return (EOPNOTSUPP);
1276 #ifndef COMPAT_43
1277 		error = so_pru_control(so, cmd, data, ifp, td);
1278 #else
1279 	    {
1280 		int ocmd = cmd;
1281 
1282 		switch (cmd) {
1283 
1284 		case SIOCSIFDSTADDR:
1285 		case SIOCSIFADDR:
1286 		case SIOCSIFBRDADDR:
1287 		case SIOCSIFNETMASK:
1288 #if BYTE_ORDER != BIG_ENDIAN
1289 			if (ifr->ifr_addr.sa_family == 0 &&
1290 			    ifr->ifr_addr.sa_len < 16) {
1291 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1292 				ifr->ifr_addr.sa_len = 16;
1293 			}
1294 #else
1295 			if (ifr->ifr_addr.sa_len == 0)
1296 				ifr->ifr_addr.sa_len = 16;
1297 #endif
1298 			break;
1299 
1300 		case OSIOCGIFADDR:
1301 			cmd = SIOCGIFADDR;
1302 			break;
1303 
1304 		case OSIOCGIFDSTADDR:
1305 			cmd = SIOCGIFDSTADDR;
1306 			break;
1307 
1308 		case OSIOCGIFBRDADDR:
1309 			cmd = SIOCGIFBRDADDR;
1310 			break;
1311 
1312 		case OSIOCGIFNETMASK:
1313 			cmd = SIOCGIFNETMASK;
1314 		}
1315 		error =  so_pru_control(so, cmd, data, ifp, td);
1316 		switch (ocmd) {
1317 
1318 		case OSIOCGIFADDR:
1319 		case OSIOCGIFDSTADDR:
1320 		case OSIOCGIFBRDADDR:
1321 		case OSIOCGIFNETMASK:
1322 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1323 
1324 		}
1325 	    }
1326 #endif /* COMPAT_43 */
1327 
1328 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1329 #ifdef INET6
1330 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1331 			if (ifp->if_flags & IFF_UP) {
1332 				int s = splimp();
1333 				in6_if_up(ifp);
1334 				splx(s);
1335 			}
1336 #endif
1337 		}
1338 		return (error);
1339 
1340 	}
1341 	return (0);
1342 }
1343 
1344 /*
1345  * Set/clear promiscuous mode on interface ifp based on the truth value
1346  * of pswitch.  The calls are reference counted so that only the first
1347  * "on" request actually has an effect, as does the final "off" request.
1348  * Results are undefined if the "off" and "on" requests are not matched.
1349  */
1350 int
1351 ifpromisc(struct ifnet *ifp, int pswitch)
1352 {
1353 	struct ifreq ifr;
1354 	int error;
1355 	int oldflags;
1356 
1357 	oldflags = ifp->if_flags;
1358 	if (ifp->if_flags & IFF_PPROMISC) {
1359 		/* Do nothing if device is in permanently promiscuous mode */
1360 		ifp->if_pcount += pswitch ? 1 : -1;
1361 		return (0);
1362 	}
1363 	if (pswitch) {
1364 		/*
1365 		 * If the device is not configured up, we cannot put it in
1366 		 * promiscuous mode.
1367 		 */
1368 		if ((ifp->if_flags & IFF_UP) == 0)
1369 			return (ENETDOWN);
1370 		if (ifp->if_pcount++ != 0)
1371 			return (0);
1372 		ifp->if_flags |= IFF_PROMISC;
1373 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1374 		    ifp->if_xname);
1375 	} else {
1376 		if (--ifp->if_pcount > 0)
1377 			return (0);
1378 		ifp->if_flags &= ~IFF_PROMISC;
1379 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1380 		    ifp->if_xname);
1381 	}
1382 	ifr.ifr_flags = ifp->if_flags;
1383 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
1384 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1385 				 (struct ucred *)NULL);
1386 	if (error == 0)
1387 		rt_ifmsg(ifp);
1388 	else
1389 		ifp->if_flags = oldflags;
1390 	return error;
1391 }
1392 
1393 /*
1394  * Return interface configuration
1395  * of system.  List may be used
1396  * in later ioctl's (above) to get
1397  * other information.
1398  */
1399 static int
1400 ifconf(u_long cmd, caddr_t data, struct thread *td)
1401 {
1402 	struct ifconf *ifc = (struct ifconf *)data;
1403 	struct ifnet *ifp;
1404 	struct ifaddr *ifa;
1405 	struct sockaddr *sa;
1406 	struct ifreq ifr, *ifrp;
1407 	int space = ifc->ifc_len, error = 0;
1408 
1409 	ifrp = ifc->ifc_req;
1410 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1411 		int addrs;
1412 
1413 		if (space <= sizeof ifr)
1414 			break;
1415 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1416 		    >= sizeof(ifr.ifr_name)) {
1417 			error = ENAMETOOLONG;
1418 			break;
1419 		}
1420 
1421 		addrs = 0;
1422 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1423 			if (space <= sizeof ifr)
1424 				break;
1425 			sa = ifa->ifa_addr;
1426 			if (td->td_proc->p_ucred->cr_prison &&
1427 			    prison_if(td, sa))
1428 				continue;
1429 			addrs++;
1430 #ifdef COMPAT_43
1431 			if (cmd == OSIOCGIFCONF) {
1432 				struct osockaddr *osa =
1433 					 (struct osockaddr *)&ifr.ifr_addr;
1434 				ifr.ifr_addr = *sa;
1435 				osa->sa_family = sa->sa_family;
1436 				error = copyout(&ifr, ifrp, sizeof ifr);
1437 				ifrp++;
1438 			} else
1439 #endif
1440 			if (sa->sa_len <= sizeof(*sa)) {
1441 				ifr.ifr_addr = *sa;
1442 				error = copyout(&ifr, ifrp, sizeof ifr);
1443 				ifrp++;
1444 			} else {
1445 				if (space < (sizeof ifr) + sa->sa_len -
1446 					    sizeof(*sa))
1447 					break;
1448 				space -= sa->sa_len - sizeof(*sa);
1449 				error = copyout(&ifr, ifrp,
1450 						sizeof ifr.ifr_name);
1451 				if (error == 0)
1452 					error = copyout(sa, &ifrp->ifr_addr,
1453 							sa->sa_len);
1454 				ifrp = (struct ifreq *)
1455 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1456 			}
1457 			if (error)
1458 				break;
1459 			space -= sizeof ifr;
1460 		}
1461 		if (error)
1462 			break;
1463 		if (!addrs) {
1464 			bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1465 			error = copyout(&ifr, ifrp, sizeof ifr);
1466 			if (error)
1467 				break;
1468 			space -= sizeof ifr;
1469 			ifrp++;
1470 		}
1471 	}
1472 	ifc->ifc_len -= space;
1473 	return (error);
1474 }
1475 
1476 /*
1477  * Just like if_promisc(), but for all-multicast-reception mode.
1478  */
1479 int
1480 if_allmulti(struct ifnet *ifp, int onswitch)
1481 {
1482 	int error = 0;
1483 	int s = splimp();
1484 	struct ifreq ifr;
1485 
1486 	if (onswitch) {
1487 		if (ifp->if_amcount++ == 0) {
1488 			ifp->if_flags |= IFF_ALLMULTI;
1489 			ifr.ifr_flags = ifp->if_flags;
1490 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1491 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1492 					      (struct ucred *)NULL);
1493 		}
1494 	} else {
1495 		if (ifp->if_amcount > 1) {
1496 			ifp->if_amcount--;
1497 		} else {
1498 			ifp->if_amcount = 0;
1499 			ifp->if_flags &= ~IFF_ALLMULTI;
1500 			ifr.ifr_flags = ifp->if_flags;
1501 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1502 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1503 					      (struct ucred *)NULL);
1504 		}
1505 	}
1506 	splx(s);
1507 
1508 	if (error == 0)
1509 		rt_ifmsg(ifp);
1510 	return error;
1511 }
1512 
1513 /*
1514  * Add a multicast listenership to the interface in question.
1515  * The link layer provides a routine which converts
1516  */
1517 int
1518 if_addmulti(
1519 	struct ifnet *ifp,	/* interface to manipulate */
1520 	struct sockaddr *sa,	/* address to add */
1521 	struct ifmultiaddr **retifma)
1522 {
1523 	struct sockaddr *llsa, *dupsa;
1524 	int error, s;
1525 	struct ifmultiaddr *ifma;
1526 
1527 	/*
1528 	 * If the matching multicast address already exists
1529 	 * then don't add a new one, just add a reference
1530 	 */
1531 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1532 		if (sa_equal(sa, ifma->ifma_addr)) {
1533 			ifma->ifma_refcount++;
1534 			if (retifma)
1535 				*retifma = ifma;
1536 			return 0;
1537 		}
1538 	}
1539 
1540 	/*
1541 	 * Give the link layer a chance to accept/reject it, and also
1542 	 * find out which AF_LINK address this maps to, if it isn't one
1543 	 * already.
1544 	 */
1545 	if (ifp->if_resolvemulti) {
1546 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1547 		if (error) return error;
1548 	} else {
1549 		llsa = 0;
1550 	}
1551 
1552 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1553 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1554 	bcopy(sa, dupsa, sa->sa_len);
1555 
1556 	ifma->ifma_addr = dupsa;
1557 	ifma->ifma_lladdr = llsa;
1558 	ifma->ifma_ifp = ifp;
1559 	ifma->ifma_refcount = 1;
1560 	ifma->ifma_protospec = 0;
1561 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1562 
1563 	/*
1564 	 * Some network interfaces can scan the address list at
1565 	 * interrupt time; lock them out.
1566 	 */
1567 	s = splimp();
1568 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1569 	splx(s);
1570 	*retifma = ifma;
1571 
1572 	if (llsa != 0) {
1573 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1574 			if (sa_equal(ifma->ifma_addr, llsa))
1575 				break;
1576 		}
1577 		if (ifma) {
1578 			ifma->ifma_refcount++;
1579 		} else {
1580 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1581 			       M_IFMADDR, M_WAITOK);
1582 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1583 			       M_IFMADDR, M_WAITOK);
1584 			bcopy(llsa, dupsa, llsa->sa_len);
1585 			ifma->ifma_addr = dupsa;
1586 			ifma->ifma_ifp = ifp;
1587 			ifma->ifma_refcount = 1;
1588 			s = splimp();
1589 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1590 			splx(s);
1591 		}
1592 	}
1593 	/*
1594 	 * We are certain we have added something, so call down to the
1595 	 * interface to let them know about it.
1596 	 */
1597 	s = splimp();
1598 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1599 	splx(s);
1600 
1601 	return 0;
1602 }
1603 
1604 /*
1605  * Remove a reference to a multicast address on this interface.  Yell
1606  * if the request does not match an existing membership.
1607  */
1608 int
1609 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1610 {
1611 	struct ifmultiaddr *ifma;
1612 	int s;
1613 
1614 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1615 		if (sa_equal(sa, ifma->ifma_addr))
1616 			break;
1617 	if (ifma == 0)
1618 		return ENOENT;
1619 
1620 	if (ifma->ifma_refcount > 1) {
1621 		ifma->ifma_refcount--;
1622 		return 0;
1623 	}
1624 
1625 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1626 	sa = ifma->ifma_lladdr;
1627 	s = splimp();
1628 	LIST_REMOVE(ifma, ifma_link);
1629 	/*
1630 	 * Make sure the interface driver is notified
1631 	 * in the case of a link layer mcast group being left.
1632 	 */
1633 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1634 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1635 	splx(s);
1636 	free(ifma->ifma_addr, M_IFMADDR);
1637 	free(ifma, M_IFMADDR);
1638 	if (sa == 0)
1639 		return 0;
1640 
1641 	/*
1642 	 * Now look for the link-layer address which corresponds to
1643 	 * this network address.  It had been squirreled away in
1644 	 * ifma->ifma_lladdr for this purpose (so we don't have
1645 	 * to call ifp->if_resolvemulti() again), and we saved that
1646 	 * value in sa above.  If some nasty deleted the
1647 	 * link-layer address out from underneath us, we can deal because
1648 	 * the address we stored was is not the same as the one which was
1649 	 * in the record for the link-layer address.  (So we don't complain
1650 	 * in that case.)
1651 	 */
1652 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1653 		if (sa_equal(sa, ifma->ifma_addr))
1654 			break;
1655 	if (ifma == 0)
1656 		return 0;
1657 
1658 	if (ifma->ifma_refcount > 1) {
1659 		ifma->ifma_refcount--;
1660 		return 0;
1661 	}
1662 
1663 	s = splimp();
1664 	LIST_REMOVE(ifma, ifma_link);
1665 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1666 	splx(s);
1667 	free(ifma->ifma_addr, M_IFMADDR);
1668 	free(sa, M_IFMADDR);
1669 	free(ifma, M_IFMADDR);
1670 
1671 	return 0;
1672 }
1673 
1674 /*
1675  * Set the link layer address on an interface.
1676  *
1677  * At this time we only support certain types of interfaces,
1678  * and we don't allow the length of the address to change.
1679  */
1680 int
1681 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1682 {
1683 	struct sockaddr_dl *sdl;
1684 	struct ifaddr *ifa;
1685 	struct ifreq ifr;
1686 
1687 	ifa = ifnet_addrs[ifp->if_index - 1];
1688 	if (ifa == NULL)
1689 		return (EINVAL);
1690 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1691 	if (sdl == NULL)
1692 		return (EINVAL);
1693 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1694 		return (EINVAL);
1695 	switch (ifp->if_type) {
1696 	case IFT_ETHER:			/* these types use struct arpcom */
1697 	case IFT_FDDI:
1698 	case IFT_XETHER:
1699 	case IFT_ISO88025:
1700 	case IFT_L2VLAN:
1701 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1702 		/* FALLTHROUGH */
1703 	case IFT_ARCNET:
1704 		bcopy(lladdr, LLADDR(sdl), len);
1705 		break;
1706 	default:
1707 		return (ENODEV);
1708 	}
1709 	/*
1710 	 * If the interface is already up, we need
1711 	 * to re-init it in order to reprogram its
1712 	 * address filter.
1713 	 */
1714 	if ((ifp->if_flags & IFF_UP) != 0) {
1715 		ifp->if_flags &= ~IFF_UP;
1716 		ifr.ifr_flags = ifp->if_flags;
1717 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1718 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1719 				 (struct ucred *)NULL);
1720 		ifp->if_flags |= IFF_UP;
1721 		ifr.ifr_flags = ifp->if_flags;
1722 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1723 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1724 				 (struct ucred *)NULL);
1725 #ifdef INET
1726 		/*
1727 		 * Also send gratuitous ARPs to notify other nodes about
1728 		 * the address change.
1729 		 */
1730 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1731 			if (ifa->ifa_addr != NULL &&
1732 			    ifa->ifa_addr->sa_family == AF_INET)
1733 				arp_ifinit(ifp, ifa);
1734 		}
1735 #endif
1736 	}
1737 	return (0);
1738 }
1739 
1740 struct ifmultiaddr *
1741 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1742 {
1743 	struct ifmultiaddr *ifma;
1744 
1745 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1746 		if (sa_equal(ifma->ifma_addr, sa))
1747 			break;
1748 
1749 	return ifma;
1750 }
1751 
1752 /*
1753  * The name argument must be a pointer to storage which will last as
1754  * long as the interface does.  For physical devices, the result of
1755  * device_get_name(dev) is a good choice and for pseudo-devices a
1756  * static string works well.
1757  */
1758 void
1759 if_initname(struct ifnet *ifp, const char *name, int unit)
1760 {
1761 	ifp->if_dname = name;
1762 	ifp->if_dunit = unit;
1763 	if (unit != IF_DUNIT_NONE)
1764 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1765 	else
1766 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1767 }
1768 
1769 int
1770 if_printf(struct ifnet *ifp, const char *fmt, ...)
1771 {
1772 	__va_list ap;
1773 	int retval;
1774 
1775 	retval = printf("%s: ", ifp->if_xname);
1776 	__va_start(ap, fmt);
1777 	retval += vprintf(fmt, ap);
1778 	__va_end(ap);
1779 	return (retval);
1780 }
1781 
1782 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1783 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1784