xref: /dflybsd-src/sys/net/if.c (revision 53e987cee557d989dbf172d8a3c2ade9ea6fc46f)
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 1.33 2005/04/15 18:39:13 dillon 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  * Support for non-ALTQ interfaces.
86  */
87 static int	ifq_classic_enqueue(struct ifaltq *, struct mbuf *,
88 				    struct altq_pktattr *);
89 static struct mbuf *
90 		ifq_classic_dequeue(struct ifaltq *, int);
91 static int	ifq_classic_request(struct ifaltq *, int, void *);
92 
93 /*
94  * System initialization
95  */
96 
97 static void	if_attachdomain(void *);
98 static void	if_attachdomain1(struct ifnet *);
99 static int ifconf (u_long, caddr_t, struct thread *);
100 static void ifinit (void *);
101 static void if_slowtimo (void *);
102 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
103 static int  if_rtdel (struct radix_node *, void *);
104 
105 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
106 
107 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
108 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
109 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
110 
111 int	ifqmaxlen = IFQ_MAXLEN;
112 struct	ifnethead ifnet;	/* depend on static init XXX */
113 
114 #ifdef INET6
115 /*
116  * XXX: declare here to avoid to include many inet6 related files..
117  * should be more generalized?
118  */
119 extern void	nd6_setmtu (struct ifnet *);
120 #endif
121 
122 struct if_clone *if_clone_lookup (const char *, int *);
123 int if_clone_list (struct if_clonereq *);
124 
125 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
126 int if_cloners_count;
127 
128 struct callout if_slowtimo_timer;
129 
130 /*
131  * Network interface utility routines.
132  *
133  * Routines with ifa_ifwith* names take sockaddr *'s as
134  * parameters.
135  */
136 /* ARGSUSED*/
137 void
138 ifinit(void *dummy)
139 {
140 	struct ifnet *ifp;
141 	int s;
142 
143 	callout_init(&if_slowtimo_timer);
144 
145 	s = splimp();
146 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
147 		if (ifp->if_snd.ifq_maxlen == 0) {
148 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
149 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
150 		}
151 	}
152 	splx(s);
153 
154 	if_slowtimo(0);
155 }
156 
157 int if_index = 0;
158 struct ifaddr **ifnet_addrs;
159 struct ifnet **ifindex2ifnet = NULL;
160 
161 /*
162  * Attach an interface to the
163  * list of "active" interfaces.
164  */
165 void
166 if_attach(struct ifnet *ifp)
167 {
168 	unsigned socksize, ifasize;
169 	int namelen, masklen;
170 	struct sockaddr_dl *sdl;
171 	struct ifaddr *ifa;
172 	struct ifaltq *ifq;
173 
174 	static int if_indexlim = 8;
175 	static boolean_t inited;
176 
177 	if (!inited) {
178 		TAILQ_INIT(&ifnet);
179 		inited = TRUE;
180 	}
181 
182 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
183 	ifp->if_index = ++if_index;
184 	/*
185 	 * XXX -
186 	 * The old code would work if the interface passed a pre-existing
187 	 * chain of ifaddrs to this code.  We don't trust our callers to
188 	 * properly initialize the tailq, however, so we no longer allow
189 	 * this unlikely case.
190 	 */
191 	TAILQ_INIT(&ifp->if_addrhead);
192 	TAILQ_INIT(&ifp->if_prefixhead);
193 	LIST_INIT(&ifp->if_multiaddrs);
194 	getmicrotime(&ifp->if_lastchange);
195 	if (ifnet_addrs == NULL || if_index >= if_indexlim) {
196 		unsigned int n;
197 		caddr_t q;
198 
199 		if_indexlim <<= 1;
200 		n = if_indexlim * sizeof(struct ifaddr *);
201 		q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
202 		if (ifnet_addrs != NULL) {
203 			bcopy(ifnet_addrs, q, n/2);
204 			free(ifnet_addrs, M_IFADDR);
205 		}
206 		ifnet_addrs = (struct ifaddr **)q;
207 
208 		/* grow ifindex2ifnet */
209 		n = if_indexlim * sizeof(struct ifnet *);
210 		q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
211 		if (ifindex2ifnet) {
212 			bcopy(ifindex2ifnet, q, n/2);
213 			free(ifindex2ifnet, M_IFADDR);
214 		}
215 		ifindex2ifnet = (struct ifnet **)q;
216 	}
217 
218 	ifindex2ifnet[if_index] = ifp;
219 
220 	/*
221 	 * create a Link Level name for this device
222 	 */
223 	namelen = strlen(ifp->if_xname);
224 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
225 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
226 	socksize = masklen + ifp->if_addrlen;
227 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
228 	if (socksize < sizeof(*sdl))
229 		socksize = sizeof(*sdl);
230 	socksize = ROUNDUP(socksize);
231 	ifasize = sizeof(struct ifaddr) + 2 * socksize;
232 	ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
233 	sdl = (struct sockaddr_dl *)(ifa + 1);
234 	sdl->sdl_len = socksize;
235 	sdl->sdl_family = AF_LINK;
236 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
237 	sdl->sdl_nlen = namelen;
238 	sdl->sdl_index = ifp->if_index;
239 	sdl->sdl_type = ifp->if_type;
240 	ifnet_addrs[if_index - 1] = ifa;
241 	ifa->ifa_ifp = ifp;
242 	ifa->ifa_rtrequest = link_rtrequest;
243 	ifa->ifa_addr = (struct sockaddr *)sdl;
244 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
245 	ifa->ifa_netmask = (struct sockaddr *)sdl;
246 	sdl->sdl_len = masklen;
247 	while (namelen != 0)
248 		sdl->sdl_data[--namelen] = 0xff;
249 	TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
250 
251 	EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
252 
253 	ifq = &ifp->if_snd;
254 	ifq->altq_type = 0;
255 	ifq->altq_disc = NULL;
256 	ifq->altq_flags &= ALTQF_CANTCHANGE;
257 	ifq->altq_tbr = NULL;
258 	ifq->altq_ifp = ifp;
259 	ifq_set_classic(ifq);
260 
261 	if (!SLIST_EMPTY(&domains))
262 		if_attachdomain1(ifp);
263 
264 	/* Announce the interface. */
265 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
266 }
267 
268 static void
269 if_attachdomain(void *dummy)
270 {
271 	struct ifnet *ifp;
272 	int s;
273 
274 	s = splnet();
275 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
276 		if_attachdomain1(ifp);
277 	splx(s);
278 }
279 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
280 	if_attachdomain, NULL);
281 
282 static void
283 if_attachdomain1(struct ifnet *ifp)
284 {
285 	struct domain *dp;
286 	int s;
287 
288 	s = splnet();
289 
290 	/* address family dependent data region */
291 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
292 	SLIST_FOREACH(dp, &domains, dom_next)
293 		if (dp->dom_ifattach)
294 			ifp->if_afdata[dp->dom_family] =
295 				(*dp->dom_ifattach)(ifp);
296 	splx(s);
297 }
298 
299 /*
300  * Detach an interface, removing it from the
301  * list of "active" interfaces.
302  */
303 void
304 if_detach(struct ifnet *ifp)
305 {
306 	struct ifaddr *ifa;
307 	struct radix_node_head	*rnh;
308 	int s;
309 	int i;
310 	struct domain *dp;
311 
312 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
313 
314 	/*
315 	 * Remove routes and flush queues.
316 	 */
317 	s = splnet();
318 	if_down(ifp);
319 
320 	if (ifq_is_enabled(&ifp->if_snd))
321 		altq_disable(&ifp->if_snd);
322 	if (ifq_is_attached(&ifp->if_snd))
323 		altq_detach(&ifp->if_snd);
324 
325 	/*
326 	 * Remove address from ifnet_addrs[] and maybe decrement if_index.
327 	 * Clean up all addresses.
328 	 */
329 	ifnet_addrs[ifp->if_index - 1] = 0;
330 	while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
331 		if_index--;
332 
333 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
334 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
335 #ifdef INET
336 		/* XXX: Ugly!! ad hoc just for INET */
337 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
338 			struct ifaliasreq ifr;
339 
340 			bzero(&ifr, sizeof ifr);
341 			ifr.ifra_addr = *ifa->ifa_addr;
342 			if (ifa->ifa_dstaddr)
343 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
344 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
345 				       NULL) == 0)
346 				continue;
347 		}
348 #endif /* INET */
349 #ifdef INET6
350 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
351 			in6_purgeaddr(ifa);
352 			/* ifp_addrhead is already updated */
353 			continue;
354 		}
355 #endif /* INET6 */
356 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
357 		IFAFREE(ifa);
358 	}
359 
360 #ifdef INET6
361 	/*
362 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
363 	 * before removing routing entries below, since IPv6 interface direct
364 	 * routes are expected to be removed by the IPv6-specific kernel API.
365 	 * Otherwise, the kernel will detect some inconsistency and bark it.
366 	 */
367 	in6_ifdetach(ifp);
368 #endif
369 
370 	/*
371 	 * Delete all remaining routes using this interface
372 	 * Unfortuneatly the only way to do this is to slog through
373 	 * the entire routing table looking for routes which point
374 	 * to this interface...oh well...
375 	 */
376 	for (i = 1; i <= AF_MAX; i++) {
377 		if ((rnh = rt_tables[i]) == NULL)
378 			continue;
379 		rnh->rnh_walktree(rnh, if_rtdel, ifp);
380 	}
381 
382 	/* Announce that the interface is gone. */
383 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
384 
385 	SLIST_FOREACH(dp, &domains, dom_next)
386 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
387 			(*dp->dom_ifdetach)(ifp,
388 				ifp->if_afdata[dp->dom_family]);
389 
390 	ifindex2ifnet[ifp->if_index] = NULL;
391 
392 	TAILQ_REMOVE(&ifnet, ifp, if_link);
393 	splx(s);
394 }
395 
396 /*
397  * Delete Routes for a Network Interface
398  *
399  * Called for each routing entry via the rnh->rnh_walktree() call above
400  * to delete all route entries referencing a detaching network interface.
401  *
402  * Arguments:
403  *	rn	pointer to node in the routing table
404  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
405  *
406  * Returns:
407  *	0	successful
408  *	errno	failed - reason indicated
409  *
410  */
411 static int
412 if_rtdel(struct radix_node *rn, void *arg)
413 {
414 	struct rtentry	*rt = (struct rtentry *)rn;
415 	struct ifnet	*ifp = arg;
416 	int		err;
417 
418 	if (rt->rt_ifp == ifp) {
419 
420 		/*
421 		 * Protect (sorta) against walktree recursion problems
422 		 * with cloned routes
423 		 */
424 		if (!(rt->rt_flags & RTF_UP))
425 			return (0);
426 
427 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
428 				rt_mask(rt), rt->rt_flags,
429 				(struct rtentry **) NULL);
430 		if (err) {
431 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
432 		}
433 	}
434 
435 	return (0);
436 }
437 
438 /*
439  * Create a clone network interface.
440  */
441 int
442 if_clone_create(char *name, int len)
443 {
444 	struct if_clone *ifc;
445 	char *dp;
446 	int wildcard, bytoff, bitoff;
447 	int unit;
448 	int err;
449 
450 	ifc = if_clone_lookup(name, &unit);
451 	if (ifc == NULL)
452 		return (EINVAL);
453 
454 	if (ifunit(name) != NULL)
455 		return (EEXIST);
456 
457 	bytoff = bitoff = 0;
458 	wildcard = (unit < 0);
459 	/*
460 	 * Find a free unit if none was given.
461 	 */
462 	if (wildcard) {
463 		while (bytoff < ifc->ifc_bmlen &&
464 		    ifc->ifc_units[bytoff] == 0xff)
465 			bytoff++;
466 		if (bytoff >= ifc->ifc_bmlen)
467 			return (ENOSPC);
468 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
469 			bitoff++;
470 		unit = (bytoff << 3) + bitoff;
471 	}
472 
473 	if (unit > ifc->ifc_maxunit)
474 		return (ENXIO);
475 
476 	err = (*ifc->ifc_create)(ifc, unit);
477 	if (err != 0)
478 		return (err);
479 
480 	if (!wildcard) {
481 		bytoff = unit >> 3;
482 		bitoff = unit - (bytoff << 3);
483 	}
484 
485 	/*
486 	 * Allocate the unit in the bitmap.
487 	 */
488 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
489 	    ("%s: bit is already set", __func__));
490 	ifc->ifc_units[bytoff] |= (1 << bitoff);
491 
492 	/* In the wildcard case, we need to update the name. */
493 	if (wildcard) {
494 		for (dp = name; *dp != '\0'; dp++);
495 		if (snprintf(dp, len - (dp-name), "%d", unit) >
496 		    len - (dp-name) - 1) {
497 			/*
498 			 * This can only be a programmer error and
499 			 * there's no straightforward way to recover if
500 			 * it happens.
501 			 */
502 			panic("if_clone_create(): interface name too long");
503 		}
504 
505 	}
506 
507 	EVENTHANDLER_INVOKE(if_clone_event, ifc);
508 
509 	return (0);
510 }
511 
512 /*
513  * Destroy a clone network interface.
514  */
515 int
516 if_clone_destroy(const char *name)
517 {
518 	struct if_clone *ifc;
519 	struct ifnet *ifp;
520 	int bytoff, bitoff;
521 	int unit;
522 
523 	ifc = if_clone_lookup(name, &unit);
524 	if (ifc == NULL)
525 		return (EINVAL);
526 
527 	if (unit < ifc->ifc_minifs)
528 		return (EINVAL);
529 
530 	ifp = ifunit(name);
531 	if (ifp == NULL)
532 		return (ENXIO);
533 
534 	if (ifc->ifc_destroy == NULL)
535 		return (EOPNOTSUPP);
536 
537 	(*ifc->ifc_destroy)(ifp);
538 
539 	/*
540 	 * Compute offset in the bitmap and deallocate the unit.
541 	 */
542 	bytoff = unit >> 3;
543 	bitoff = unit - (bytoff << 3);
544 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
545 	    ("%s: bit is already cleared", __func__));
546 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
547 	return (0);
548 }
549 
550 /*
551  * Look up a network interface cloner.
552  */
553 struct if_clone *
554 if_clone_lookup(const char *name, int *unitp)
555 {
556 	struct if_clone *ifc;
557 	const char *cp;
558 	int i;
559 
560 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
561 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
562 			if (ifc->ifc_name[i] != *cp)
563 				goto next_ifc;
564 		}
565 		goto found_name;
566  next_ifc:
567 		ifc = LIST_NEXT(ifc, ifc_list);
568 	}
569 
570 	/* No match. */
571 	return ((struct if_clone *)NULL);
572 
573  found_name:
574 	if (*cp == '\0') {
575 		i = -1;
576 	} else {
577 		for (i = 0; *cp != '\0'; cp++) {
578 			if (*cp < '0' || *cp > '9') {
579 				/* Bogus unit number. */
580 				return (NULL);
581 			}
582 			i = (i * 10) + (*cp - '0');
583 		}
584 	}
585 
586 	if (unitp != NULL)
587 		*unitp = i;
588 	return (ifc);
589 }
590 
591 /*
592  * Register a network interface cloner.
593  */
594 void
595 if_clone_attach(struct if_clone *ifc)
596 {
597 	int bytoff, bitoff;
598 	int err;
599 	int len, maxclone;
600 	int unit;
601 
602 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
603 	    ("%s: %s requested more units then allowed (%d > %d)",
604 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
605 	    ifc->ifc_maxunit + 1));
606 	/*
607 	 * Compute bitmap size and allocate it.
608 	 */
609 	maxclone = ifc->ifc_maxunit + 1;
610 	len = maxclone >> 3;
611 	if ((len << 3) < maxclone)
612 		len++;
613 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
614 	ifc->ifc_bmlen = len;
615 
616 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
617 	if_cloners_count++;
618 
619 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
620 		err = (*ifc->ifc_create)(ifc, unit);
621 		KASSERT(err == 0,
622 		    ("%s: failed to create required interface %s%d",
623 		    __func__, ifc->ifc_name, unit));
624 
625 		/* Allocate the unit in the bitmap. */
626 		bytoff = unit >> 3;
627 		bitoff = unit - (bytoff << 3);
628 		ifc->ifc_units[bytoff] |= (1 << bitoff);
629 	}
630 }
631 
632 /*
633  * Unregister a network interface cloner.
634  */
635 void
636 if_clone_detach(struct if_clone *ifc)
637 {
638 
639 	LIST_REMOVE(ifc, ifc_list);
640 	free(ifc->ifc_units, M_CLONE);
641 	if_cloners_count--;
642 }
643 
644 /*
645  * Provide list of interface cloners to userspace.
646  */
647 int
648 if_clone_list(struct if_clonereq *ifcr)
649 {
650 	char outbuf[IFNAMSIZ], *dst;
651 	struct if_clone *ifc;
652 	int count, error = 0;
653 
654 	ifcr->ifcr_total = if_cloners_count;
655 	if ((dst = ifcr->ifcr_buffer) == NULL) {
656 		/* Just asking how many there are. */
657 		return (0);
658 	}
659 
660 	if (ifcr->ifcr_count < 0)
661 		return (EINVAL);
662 
663 	count = (if_cloners_count < ifcr->ifcr_count) ?
664 	    if_cloners_count : ifcr->ifcr_count;
665 
666 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
667 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
668 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
669 		error = copyout(outbuf, dst, IFNAMSIZ);
670 		if (error)
671 			break;
672 	}
673 
674 	return (error);
675 }
676 
677 /*
678  * Locate an interface based on a complete address.
679  */
680 struct ifaddr *
681 ifa_ifwithaddr(struct sockaddr *addr)
682 {
683 	struct ifnet *ifp;
684 	struct ifaddr *ifa;
685 
686 	TAILQ_FOREACH(ifp, &ifnet, if_link)
687 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
688 		if (ifa->ifa_addr->sa_family != addr->sa_family)
689 			continue;
690 		if (sa_equal(addr, ifa->ifa_addr))
691 			return (ifa);
692 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
693 		    /* IPv6 doesn't have broadcast */
694 		    ifa->ifa_broadaddr->sa_len != 0 &&
695 		    sa_equal(ifa->ifa_broadaddr, addr))
696 			return (ifa);
697 	}
698 	return ((struct ifaddr *)NULL);
699 }
700 /*
701  * Locate the point to point interface with a given destination address.
702  */
703 struct ifaddr *
704 ifa_ifwithdstaddr(struct sockaddr *addr)
705 {
706 	struct ifnet *ifp;
707 	struct ifaddr *ifa;
708 
709 	TAILQ_FOREACH(ifp, &ifnet, if_link)
710 	    if (ifp->if_flags & IFF_POINTOPOINT)
711 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
712 			if (ifa->ifa_addr->sa_family != addr->sa_family)
713 				continue;
714 			if (ifa->ifa_dstaddr &&
715 			    sa_equal(addr, ifa->ifa_dstaddr))
716 				return (ifa);
717 	}
718 	return ((struct ifaddr *)NULL);
719 }
720 
721 /*
722  * Find an interface on a specific network.  If many, choice
723  * is most specific found.
724  */
725 struct ifaddr *
726 ifa_ifwithnet(struct sockaddr *addr)
727 {
728 	struct ifnet *ifp;
729 	struct ifaddr *ifa;
730 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
731 	u_int af = addr->sa_family;
732 	char *addr_data = addr->sa_data, *cplim;
733 
734 	/*
735 	 * AF_LINK addresses can be looked up directly by their index number,
736 	 * so do that if we can.
737 	 */
738 	if (af == AF_LINK) {
739 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
740 
741 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
742 		return (ifnet_addrs[sdl->sdl_index - 1]);
743 	}
744 
745 	/*
746 	 * Scan though each interface, looking for ones that have
747 	 * addresses in this address family.
748 	 */
749 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
750 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
751 			char *cp, *cp2, *cp3;
752 
753 			if (ifa->ifa_addr->sa_family != af)
754 next:				continue;
755 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
756 				/*
757 				 * This is a bit broken as it doesn't
758 				 * take into account that the remote end may
759 				 * be a single node in the network we are
760 				 * looking for.
761 				 * The trouble is that we don't know the
762 				 * netmask for the remote end.
763 				 */
764 				if (ifa->ifa_dstaddr != NULL &&
765 				    sa_equal(addr, ifa->ifa_dstaddr))
766 					return (ifa);
767 			} else {
768 				/*
769 				 * if we have a special address handler,
770 				 * then use it instead of the generic one.
771 				 */
772 				if (ifa->ifa_claim_addr) {
773 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
774 						return (ifa);
775 					} else {
776 						continue;
777 					}
778 				}
779 
780 				/*
781 				 * Scan all the bits in the ifa's address.
782 				 * If a bit dissagrees with what we are
783 				 * looking for, mask it with the netmask
784 				 * to see if it really matters.
785 				 * (A byte at a time)
786 				 */
787 				if (ifa->ifa_netmask == 0)
788 					continue;
789 				cp = addr_data;
790 				cp2 = ifa->ifa_addr->sa_data;
791 				cp3 = ifa->ifa_netmask->sa_data;
792 				cplim = ifa->ifa_netmask->sa_len +
793 					(char *)ifa->ifa_netmask;
794 				while (cp3 < cplim)
795 					if ((*cp++ ^ *cp2++) & *cp3++)
796 						goto next; /* next address! */
797 				/*
798 				 * If the netmask of what we just found
799 				 * is more specific than what we had before
800 				 * (if we had one) then remember the new one
801 				 * before continuing to search
802 				 * for an even better one.
803 				 */
804 				if (ifa_maybe == 0 ||
805 				    rn_refines((char *)ifa->ifa_netmask,
806 					       (char *)ifa_maybe->ifa_netmask))
807 					ifa_maybe = ifa;
808 			}
809 		}
810 	}
811 	return (ifa_maybe);
812 }
813 
814 /*
815  * Find an interface address specific to an interface best matching
816  * a given address.
817  */
818 struct ifaddr *
819 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
820 {
821 	struct ifaddr *ifa;
822 	char *cp, *cp2, *cp3;
823 	char *cplim;
824 	struct ifaddr *ifa_maybe = 0;
825 	u_int af = addr->sa_family;
826 
827 	if (af >= AF_MAX)
828 		return (0);
829 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
830 		if (ifa->ifa_addr->sa_family != af)
831 			continue;
832 		if (ifa_maybe == 0)
833 			ifa_maybe = ifa;
834 		if (ifa->ifa_netmask == NULL) {
835 			if (sa_equal(addr, ifa->ifa_addr) ||
836 			    (ifa->ifa_dstaddr != NULL &&
837 			     sa_equal(addr, ifa->ifa_dstaddr)))
838 				return (ifa);
839 			continue;
840 		}
841 		if (ifp->if_flags & IFF_POINTOPOINT) {
842 			if (sa_equal(addr, ifa->ifa_dstaddr))
843 				return (ifa);
844 		} else {
845 			cp = addr->sa_data;
846 			cp2 = ifa->ifa_addr->sa_data;
847 			cp3 = ifa->ifa_netmask->sa_data;
848 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
849 			for (; cp3 < cplim; cp3++)
850 				if ((*cp++ ^ *cp2++) & *cp3)
851 					break;
852 			if (cp3 == cplim)
853 				return (ifa);
854 		}
855 	}
856 	return (ifa_maybe);
857 }
858 
859 #include <net/route.h>
860 
861 /*
862  * Default action when installing a route with a Link Level gateway.
863  * Lookup an appropriate real ifa to point to.
864  * This should be moved to /sys/net/link.c eventually.
865  */
866 static void
867 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
868 {
869 	struct ifaddr *ifa;
870 	struct sockaddr *dst;
871 	struct ifnet *ifp;
872 
873 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
874 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
875 		return;
876 	ifa = ifaof_ifpforaddr(dst, ifp);
877 	if (ifa != NULL) {
878 		IFAFREE(rt->rt_ifa);
879 		IFAREF(ifa);
880 		rt->rt_ifa = ifa;
881 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
882 			ifa->ifa_rtrequest(cmd, rt, info);
883 	}
884 }
885 
886 /*
887  * Mark an interface down and notify protocols of
888  * the transition.
889  * NOTE: must be called at splnet or eqivalent.
890  */
891 void
892 if_unroute(struct ifnet *ifp, int flag, int fam)
893 {
894 	struct ifaddr *ifa;
895 
896 	ifp->if_flags &= ~flag;
897 	getmicrotime(&ifp->if_lastchange);
898 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
899 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
900 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
901 	ifq_purge(&ifp->if_snd);
902 	rt_ifmsg(ifp);
903 }
904 
905 /*
906  * Mark an interface up and notify protocols of
907  * the transition.
908  * NOTE: must be called at splnet or eqivalent.
909  */
910 void
911 if_route(struct ifnet *ifp, int flag, int fam)
912 {
913 	struct ifaddr *ifa;
914 
915 	ifp->if_flags |= flag;
916 	getmicrotime(&ifp->if_lastchange);
917 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
918 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
919 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
920 	rt_ifmsg(ifp);
921 #ifdef INET6
922 	in6_if_up(ifp);
923 #endif
924 }
925 
926 /*
927  * Mark an interface down and notify protocols of the transition.  An
928  * interface going down is also considered to be a synchronizing event.
929  * We must ensure that all packet processing related to the interface
930  * has completed before we return so e.g. the caller can free the ifnet
931  * structure that the mbufs may be referencing.
932  *
933  * NOTE: must be called at splnet or eqivalent.
934  */
935 void
936 if_down(struct ifnet *ifp)
937 {
938 
939 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
940 	netmsg_service_sync();
941 }
942 
943 /*
944  * Mark an interface up and notify protocols of
945  * the transition.
946  * NOTE: must be called at splnet or eqivalent.
947  */
948 void
949 if_up(struct ifnet *ifp)
950 {
951 
952 	if_route(ifp, IFF_UP, AF_UNSPEC);
953 }
954 
955 /*
956  * Handle interface watchdog timer routines.  Called
957  * from softclock, we decrement timers (if set) and
958  * call the appropriate interface routine on expiration.
959  */
960 static void
961 if_slowtimo(void *arg)
962 {
963 	struct ifnet *ifp;
964 	int s = splimp();
965 
966 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
967 		if (ifp->if_timer == 0 || --ifp->if_timer)
968 			continue;
969 		if (ifp->if_watchdog)
970 			(*ifp->if_watchdog)(ifp);
971 	}
972 	splx(s);
973 	callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
974 }
975 
976 /*
977  * Map interface name to
978  * interface structure pointer.
979  */
980 struct ifnet *
981 ifunit(const char *name)
982 {
983 	struct ifnet *ifp;
984 
985 	/*
986 	 * Search all the interfaces for this name/number
987 	 */
988 
989 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
990 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
991 			break;
992 	}
993 	return (ifp);
994 }
995 
996 
997 /*
998  * Map interface name in a sockaddr_dl to
999  * interface structure pointer.
1000  */
1001 struct ifnet *
1002 if_withname(struct sockaddr *sa)
1003 {
1004 	char ifname[IFNAMSIZ+1];
1005 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1006 
1007 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1008 	     (sdl->sdl_nlen > IFNAMSIZ) )
1009 		return NULL;
1010 
1011 	/*
1012 	 * ifunit wants a null-terminated name.  It may not be null-terminated
1013 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
1014 	 * and there might not be room to put the trailing null anyway, so we
1015 	 * make a local copy that we know we can null terminate safely.
1016 	 */
1017 
1018 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1019 	ifname[sdl->sdl_nlen] = '\0';
1020 	return ifunit(ifname);
1021 }
1022 
1023 
1024 /*
1025  * Interface ioctls.
1026  */
1027 int
1028 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1029 {
1030 	struct ifnet *ifp;
1031 	struct ifreq *ifr;
1032 	struct ifstat *ifs;
1033 	int error;
1034 	short oif_flags;
1035 	int new_flags;
1036 	size_t namelen, onamelen;
1037 	char new_name[IFNAMSIZ];
1038 	struct ifaddr *ifa;
1039 	struct sockaddr_dl *sdl;
1040 
1041 	switch (cmd) {
1042 
1043 	case SIOCGIFCONF:
1044 	case OSIOCGIFCONF:
1045 		return (ifconf(cmd, data, td));
1046 	}
1047 	ifr = (struct ifreq *)data;
1048 
1049 	switch (cmd) {
1050 	case SIOCIFCREATE:
1051 	case SIOCIFDESTROY:
1052 		if ((error = suser(td)) != 0)
1053 			return (error);
1054 		return ((cmd == SIOCIFCREATE) ?
1055 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1056 			if_clone_destroy(ifr->ifr_name));
1057 
1058 	case SIOCIFGCLONERS:
1059 		return (if_clone_list((struct if_clonereq *)data));
1060 	}
1061 
1062 	ifp = ifunit(ifr->ifr_name);
1063 	if (ifp == 0)
1064 		return (ENXIO);
1065 	switch (cmd) {
1066 
1067 	case SIOCGIFFLAGS:
1068 		ifr->ifr_flags = ifp->if_flags;
1069 		ifr->ifr_flagshigh = ifp->if_flags >> 16;
1070 		break;
1071 
1072 	case SIOCGIFCAP:
1073 		ifr->ifr_reqcap = ifp->if_capabilities;
1074 		ifr->ifr_curcap = ifp->if_capenable;
1075 		break;
1076 
1077 	case SIOCGIFMETRIC:
1078 		ifr->ifr_metric = ifp->if_metric;
1079 		break;
1080 
1081 	case SIOCGIFMTU:
1082 		ifr->ifr_mtu = ifp->if_mtu;
1083 		break;
1084 
1085 	case SIOCGIFPHYS:
1086 		ifr->ifr_phys = ifp->if_physical;
1087 		break;
1088 
1089 	case SIOCSIFFLAGS:
1090 		error = suser(td);
1091 		if (error)
1092 			return (error);
1093 		new_flags = (ifr->ifr_flags & 0xffff) |
1094 		    (ifr->ifr_flagshigh << 16);
1095 		if (ifp->if_flags & IFF_SMART) {
1096 			/* Smart drivers twiddle their own routes */
1097 		} else if (ifp->if_flags & IFF_UP &&
1098 		    (new_flags & IFF_UP) == 0) {
1099 			int s = splimp();
1100 			if_down(ifp);
1101 			splx(s);
1102 		} else if (new_flags & IFF_UP &&
1103 		    (ifp->if_flags & IFF_UP) == 0) {
1104 			int s = splimp();
1105 			if_up(ifp);
1106 			splx(s);
1107 		}
1108 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1109 			(new_flags &~ IFF_CANTCHANGE);
1110 		if (new_flags & IFF_PPROMISC) {
1111 			/* Permanently promiscuous mode requested */
1112 			ifp->if_flags |= IFF_PROMISC;
1113 		} else if (ifp->if_pcount == 0) {
1114 			ifp->if_flags &= ~IFF_PROMISC;
1115 		}
1116 		if (ifp->if_ioctl)
1117 			(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1118 		getmicrotime(&ifp->if_lastchange);
1119 		break;
1120 
1121 	case SIOCSIFCAP:
1122 		error = suser(td);
1123 		if (error)
1124 			return (error);
1125 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1126 			return (EINVAL);
1127 		(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1128 		break;
1129 
1130 	case SIOCSIFNAME:
1131 		error = suser(td);
1132 		if (error != 0)
1133 			return (error);
1134 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1135 		if (error != 0)
1136 			return (error);
1137 		if (new_name[0] == '\0')
1138 			return (EINVAL);
1139 		if (ifunit(new_name) != NULL)
1140 			return (EEXIST);
1141 
1142 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1143 
1144 		/* Announce the departure of the interface. */
1145 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1146 
1147 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1148 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1149 		/* XXX IFA_LOCK(ifa); */
1150 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1151 		namelen = strlen(new_name);
1152 		onamelen = sdl->sdl_nlen;
1153 		/*
1154 		 * Move the address if needed.  This is safe because we
1155 		 * allocate space for a name of length IFNAMSIZ when we
1156 		 * create this in if_attach().
1157 		 */
1158 		if (namelen != onamelen) {
1159 			bcopy(sdl->sdl_data + onamelen,
1160 			    sdl->sdl_data + namelen, sdl->sdl_alen);
1161 		}
1162 		bcopy(new_name, sdl->sdl_data, namelen);
1163 		sdl->sdl_nlen = namelen;
1164 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1165 		bzero(sdl->sdl_data, onamelen);
1166 		while (namelen != 0)
1167 			sdl->sdl_data[--namelen] = 0xff;
1168 		/* XXX IFA_UNLOCK(ifa) */
1169 
1170 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1171 
1172 		/* Announce the return of the interface. */
1173 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1174 		break;
1175 
1176 	case SIOCSIFMETRIC:
1177 		error = suser(td);
1178 		if (error)
1179 			return (error);
1180 		ifp->if_metric = ifr->ifr_metric;
1181 		getmicrotime(&ifp->if_lastchange);
1182 		break;
1183 
1184 	case SIOCSIFPHYS:
1185 		error = suser(td);
1186 		if (error)
1187 			return error;
1188 		if (!ifp->if_ioctl)
1189 		        return EOPNOTSUPP;
1190 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1191 		if (error == 0)
1192 			getmicrotime(&ifp->if_lastchange);
1193 		return (error);
1194 
1195 	case SIOCSIFMTU:
1196 	{
1197 		u_long oldmtu = ifp->if_mtu;
1198 
1199 		error = suser(td);
1200 		if (error)
1201 			return (error);
1202 		if (ifp->if_ioctl == NULL)
1203 			return (EOPNOTSUPP);
1204 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1205 			return (EINVAL);
1206 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1207 		if (error == 0) {
1208 			getmicrotime(&ifp->if_lastchange);
1209 			rt_ifmsg(ifp);
1210 		}
1211 		/*
1212 		 * If the link MTU changed, do network layer specific procedure.
1213 		 */
1214 		if (ifp->if_mtu != oldmtu) {
1215 #ifdef INET6
1216 			nd6_setmtu(ifp);
1217 #endif
1218 		}
1219 		return (error);
1220 	}
1221 
1222 	case SIOCADDMULTI:
1223 	case SIOCDELMULTI:
1224 		error = suser(td);
1225 		if (error)
1226 			return (error);
1227 
1228 		/* Don't allow group membership on non-multicast interfaces. */
1229 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1230 			return EOPNOTSUPP;
1231 
1232 		/* Don't let users screw up protocols' entries. */
1233 		if (ifr->ifr_addr.sa_family != AF_LINK)
1234 			return EINVAL;
1235 
1236 		if (cmd == SIOCADDMULTI) {
1237 			struct ifmultiaddr *ifma;
1238 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1239 		} else {
1240 			error = if_delmulti(ifp, &ifr->ifr_addr);
1241 		}
1242 		if (error == 0)
1243 			getmicrotime(&ifp->if_lastchange);
1244 		return error;
1245 
1246 	case SIOCSIFPHYADDR:
1247 	case SIOCDIFPHYADDR:
1248 #ifdef INET6
1249 	case SIOCSIFPHYADDR_IN6:
1250 #endif
1251 	case SIOCSLIFPHYADDR:
1252         case SIOCSIFMEDIA:
1253 	case SIOCSIFGENERIC:
1254 		error = suser(td);
1255 		if (error)
1256 			return (error);
1257 		if (ifp->if_ioctl == 0)
1258 			return (EOPNOTSUPP);
1259 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1260 		if (error == 0)
1261 			getmicrotime(&ifp->if_lastchange);
1262 		return error;
1263 
1264 	case SIOCGIFSTATUS:
1265 		ifs = (struct ifstat *)data;
1266 		ifs->ascii[0] = '\0';
1267 
1268 	case SIOCGIFPSRCADDR:
1269 	case SIOCGIFPDSTADDR:
1270 	case SIOCGLIFPHYADDR:
1271 	case SIOCGIFMEDIA:
1272 	case SIOCGIFGENERIC:
1273 		if (ifp->if_ioctl == 0)
1274 			return (EOPNOTSUPP);
1275 		return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1276 
1277 	case SIOCSIFLLADDR:
1278 		error = suser(td);
1279 		if (error)
1280 			return (error);
1281 		return if_setlladdr(ifp,
1282 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1283 
1284 	default:
1285 		oif_flags = ifp->if_flags;
1286 		if (so->so_proto == 0)
1287 			return (EOPNOTSUPP);
1288 #ifndef COMPAT_43
1289 		error = so_pru_control(so, cmd, data, ifp, td);
1290 #else
1291 	    {
1292 		int ocmd = cmd;
1293 
1294 		switch (cmd) {
1295 
1296 		case SIOCSIFDSTADDR:
1297 		case SIOCSIFADDR:
1298 		case SIOCSIFBRDADDR:
1299 		case SIOCSIFNETMASK:
1300 #if BYTE_ORDER != BIG_ENDIAN
1301 			if (ifr->ifr_addr.sa_family == 0 &&
1302 			    ifr->ifr_addr.sa_len < 16) {
1303 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1304 				ifr->ifr_addr.sa_len = 16;
1305 			}
1306 #else
1307 			if (ifr->ifr_addr.sa_len == 0)
1308 				ifr->ifr_addr.sa_len = 16;
1309 #endif
1310 			break;
1311 
1312 		case OSIOCGIFADDR:
1313 			cmd = SIOCGIFADDR;
1314 			break;
1315 
1316 		case OSIOCGIFDSTADDR:
1317 			cmd = SIOCGIFDSTADDR;
1318 			break;
1319 
1320 		case OSIOCGIFBRDADDR:
1321 			cmd = SIOCGIFBRDADDR;
1322 			break;
1323 
1324 		case OSIOCGIFNETMASK:
1325 			cmd = SIOCGIFNETMASK;
1326 		}
1327 		error =  so_pru_control(so, cmd, data, ifp, td);
1328 		switch (ocmd) {
1329 
1330 		case OSIOCGIFADDR:
1331 		case OSIOCGIFDSTADDR:
1332 		case OSIOCGIFBRDADDR:
1333 		case OSIOCGIFNETMASK:
1334 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1335 
1336 		}
1337 	    }
1338 #endif /* COMPAT_43 */
1339 
1340 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1341 #ifdef INET6
1342 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1343 			if (ifp->if_flags & IFF_UP) {
1344 				int s = splimp();
1345 				in6_if_up(ifp);
1346 				splx(s);
1347 			}
1348 #endif
1349 		}
1350 		return (error);
1351 
1352 	}
1353 	return (0);
1354 }
1355 
1356 /*
1357  * Set/clear promiscuous mode on interface ifp based on the truth value
1358  * of pswitch.  The calls are reference counted so that only the first
1359  * "on" request actually has an effect, as does the final "off" request.
1360  * Results are undefined if the "off" and "on" requests are not matched.
1361  */
1362 int
1363 ifpromisc(struct ifnet *ifp, int pswitch)
1364 {
1365 	struct ifreq ifr;
1366 	int error;
1367 	int oldflags;
1368 
1369 	oldflags = ifp->if_flags;
1370 	if (ifp->if_flags & IFF_PPROMISC) {
1371 		/* Do nothing if device is in permanently promiscuous mode */
1372 		ifp->if_pcount += pswitch ? 1 : -1;
1373 		return (0);
1374 	}
1375 	if (pswitch) {
1376 		/*
1377 		 * If the device is not configured up, we cannot put it in
1378 		 * promiscuous mode.
1379 		 */
1380 		if ((ifp->if_flags & IFF_UP) == 0)
1381 			return (ENETDOWN);
1382 		if (ifp->if_pcount++ != 0)
1383 			return (0);
1384 		ifp->if_flags |= IFF_PROMISC;
1385 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1386 		    ifp->if_xname);
1387 	} else {
1388 		if (--ifp->if_pcount > 0)
1389 			return (0);
1390 		ifp->if_flags &= ~IFF_PROMISC;
1391 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1392 		    ifp->if_xname);
1393 	}
1394 	ifr.ifr_flags = ifp->if_flags;
1395 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
1396 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1397 				 (struct ucred *)NULL);
1398 	if (error == 0)
1399 		rt_ifmsg(ifp);
1400 	else
1401 		ifp->if_flags = oldflags;
1402 	return error;
1403 }
1404 
1405 /*
1406  * Return interface configuration
1407  * of system.  List may be used
1408  * in later ioctl's (above) to get
1409  * other information.
1410  */
1411 static int
1412 ifconf(u_long cmd, caddr_t data, struct thread *td)
1413 {
1414 	struct ifconf *ifc = (struct ifconf *)data;
1415 	struct ifnet *ifp;
1416 	struct ifaddr *ifa;
1417 	struct sockaddr *sa;
1418 	struct ifreq ifr, *ifrp;
1419 	int space = ifc->ifc_len, error = 0;
1420 
1421 	ifrp = ifc->ifc_req;
1422 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1423 		int addrs;
1424 
1425 		if (space <= sizeof ifr)
1426 			break;
1427 
1428 		/*
1429 		 * Zero the stack declared structure first to prevent
1430 		 * memory disclosure.
1431 		 */
1432 		bzero(&ifr, sizeof(ifr));
1433 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1434 		    >= sizeof(ifr.ifr_name)) {
1435 			error = ENAMETOOLONG;
1436 			break;
1437 		}
1438 
1439 		addrs = 0;
1440 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1441 			if (space <= sizeof ifr)
1442 				break;
1443 			sa = ifa->ifa_addr;
1444 			if (td->td_proc->p_ucred->cr_prison &&
1445 			    prison_if(td, sa))
1446 				continue;
1447 			addrs++;
1448 #ifdef COMPAT_43
1449 			if (cmd == OSIOCGIFCONF) {
1450 				struct osockaddr *osa =
1451 					 (struct osockaddr *)&ifr.ifr_addr;
1452 				ifr.ifr_addr = *sa;
1453 				osa->sa_family = sa->sa_family;
1454 				error = copyout(&ifr, ifrp, sizeof ifr);
1455 				ifrp++;
1456 			} else
1457 #endif
1458 			if (sa->sa_len <= sizeof(*sa)) {
1459 				ifr.ifr_addr = *sa;
1460 				error = copyout(&ifr, ifrp, sizeof ifr);
1461 				ifrp++;
1462 			} else {
1463 				if (space < (sizeof ifr) + sa->sa_len -
1464 					    sizeof(*sa))
1465 					break;
1466 				space -= sa->sa_len - sizeof(*sa);
1467 				error = copyout(&ifr, ifrp,
1468 						sizeof ifr.ifr_name);
1469 				if (error == 0)
1470 					error = copyout(sa, &ifrp->ifr_addr,
1471 							sa->sa_len);
1472 				ifrp = (struct ifreq *)
1473 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1474 			}
1475 			if (error)
1476 				break;
1477 			space -= sizeof ifr;
1478 		}
1479 		if (error)
1480 			break;
1481 		if (!addrs) {
1482 			bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1483 			error = copyout(&ifr, ifrp, sizeof ifr);
1484 			if (error)
1485 				break;
1486 			space -= sizeof ifr;
1487 			ifrp++;
1488 		}
1489 	}
1490 	ifc->ifc_len -= space;
1491 	return (error);
1492 }
1493 
1494 /*
1495  * Just like if_promisc(), but for all-multicast-reception mode.
1496  */
1497 int
1498 if_allmulti(struct ifnet *ifp, int onswitch)
1499 {
1500 	int error = 0;
1501 	int s = splimp();
1502 	struct ifreq ifr;
1503 
1504 	if (onswitch) {
1505 		if (ifp->if_amcount++ == 0) {
1506 			ifp->if_flags |= IFF_ALLMULTI;
1507 			ifr.ifr_flags = ifp->if_flags;
1508 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1509 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1510 					      (struct ucred *)NULL);
1511 		}
1512 	} else {
1513 		if (ifp->if_amcount > 1) {
1514 			ifp->if_amcount--;
1515 		} else {
1516 			ifp->if_amcount = 0;
1517 			ifp->if_flags &= ~IFF_ALLMULTI;
1518 			ifr.ifr_flags = ifp->if_flags;
1519 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1520 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1521 					      (struct ucred *)NULL);
1522 		}
1523 	}
1524 	splx(s);
1525 
1526 	if (error == 0)
1527 		rt_ifmsg(ifp);
1528 	return error;
1529 }
1530 
1531 /*
1532  * Add a multicast listenership to the interface in question.
1533  * The link layer provides a routine which converts
1534  */
1535 int
1536 if_addmulti(
1537 	struct ifnet *ifp,	/* interface to manipulate */
1538 	struct sockaddr *sa,	/* address to add */
1539 	struct ifmultiaddr **retifma)
1540 {
1541 	struct sockaddr *llsa, *dupsa;
1542 	int error, s;
1543 	struct ifmultiaddr *ifma;
1544 
1545 	/*
1546 	 * If the matching multicast address already exists
1547 	 * then don't add a new one, just add a reference
1548 	 */
1549 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1550 		if (sa_equal(sa, ifma->ifma_addr)) {
1551 			ifma->ifma_refcount++;
1552 			if (retifma)
1553 				*retifma = ifma;
1554 			return 0;
1555 		}
1556 	}
1557 
1558 	/*
1559 	 * Give the link layer a chance to accept/reject it, and also
1560 	 * find out which AF_LINK address this maps to, if it isn't one
1561 	 * already.
1562 	 */
1563 	if (ifp->if_resolvemulti) {
1564 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1565 		if (error) return error;
1566 	} else {
1567 		llsa = 0;
1568 	}
1569 
1570 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1571 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1572 	bcopy(sa, dupsa, sa->sa_len);
1573 
1574 	ifma->ifma_addr = dupsa;
1575 	ifma->ifma_lladdr = llsa;
1576 	ifma->ifma_ifp = ifp;
1577 	ifma->ifma_refcount = 1;
1578 	ifma->ifma_protospec = 0;
1579 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1580 
1581 	/*
1582 	 * Some network interfaces can scan the address list at
1583 	 * interrupt time; lock them out.
1584 	 */
1585 	s = splimp();
1586 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1587 	splx(s);
1588 	*retifma = ifma;
1589 
1590 	if (llsa != 0) {
1591 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1592 			if (sa_equal(ifma->ifma_addr, llsa))
1593 				break;
1594 		}
1595 		if (ifma) {
1596 			ifma->ifma_refcount++;
1597 		} else {
1598 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1599 			       M_IFMADDR, M_WAITOK);
1600 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1601 			       M_IFMADDR, M_WAITOK);
1602 			bcopy(llsa, dupsa, llsa->sa_len);
1603 			ifma->ifma_addr = dupsa;
1604 			ifma->ifma_ifp = ifp;
1605 			ifma->ifma_refcount = 1;
1606 			s = splimp();
1607 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1608 			splx(s);
1609 		}
1610 	}
1611 	/*
1612 	 * We are certain we have added something, so call down to the
1613 	 * interface to let them know about it.
1614 	 */
1615 	s = splimp();
1616 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1617 	splx(s);
1618 
1619 	return 0;
1620 }
1621 
1622 /*
1623  * Remove a reference to a multicast address on this interface.  Yell
1624  * if the request does not match an existing membership.
1625  */
1626 int
1627 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1628 {
1629 	struct ifmultiaddr *ifma;
1630 	int s;
1631 
1632 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1633 		if (sa_equal(sa, ifma->ifma_addr))
1634 			break;
1635 	if (ifma == 0)
1636 		return ENOENT;
1637 
1638 	if (ifma->ifma_refcount > 1) {
1639 		ifma->ifma_refcount--;
1640 		return 0;
1641 	}
1642 
1643 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1644 	sa = ifma->ifma_lladdr;
1645 	s = splimp();
1646 	LIST_REMOVE(ifma, ifma_link);
1647 	/*
1648 	 * Make sure the interface driver is notified
1649 	 * in the case of a link layer mcast group being left.
1650 	 */
1651 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1652 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1653 	splx(s);
1654 	free(ifma->ifma_addr, M_IFMADDR);
1655 	free(ifma, M_IFMADDR);
1656 	if (sa == 0)
1657 		return 0;
1658 
1659 	/*
1660 	 * Now look for the link-layer address which corresponds to
1661 	 * this network address.  It had been squirreled away in
1662 	 * ifma->ifma_lladdr for this purpose (so we don't have
1663 	 * to call ifp->if_resolvemulti() again), and we saved that
1664 	 * value in sa above.  If some nasty deleted the
1665 	 * link-layer address out from underneath us, we can deal because
1666 	 * the address we stored was is not the same as the one which was
1667 	 * in the record for the link-layer address.  (So we don't complain
1668 	 * in that case.)
1669 	 */
1670 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1671 		if (sa_equal(sa, ifma->ifma_addr))
1672 			break;
1673 	if (ifma == 0)
1674 		return 0;
1675 
1676 	if (ifma->ifma_refcount > 1) {
1677 		ifma->ifma_refcount--;
1678 		return 0;
1679 	}
1680 
1681 	s = splimp();
1682 	LIST_REMOVE(ifma, ifma_link);
1683 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1684 	splx(s);
1685 	free(ifma->ifma_addr, M_IFMADDR);
1686 	free(sa, M_IFMADDR);
1687 	free(ifma, M_IFMADDR);
1688 
1689 	return 0;
1690 }
1691 
1692 /*
1693  * Set the link layer address on an interface.
1694  *
1695  * At this time we only support certain types of interfaces,
1696  * and we don't allow the length of the address to change.
1697  */
1698 int
1699 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1700 {
1701 	struct sockaddr_dl *sdl;
1702 	struct ifaddr *ifa;
1703 	struct ifreq ifr;
1704 
1705 	ifa = ifnet_addrs[ifp->if_index - 1];
1706 	if (ifa == NULL)
1707 		return (EINVAL);
1708 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1709 	if (sdl == NULL)
1710 		return (EINVAL);
1711 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1712 		return (EINVAL);
1713 	switch (ifp->if_type) {
1714 	case IFT_ETHER:			/* these types use struct arpcom */
1715 	case IFT_FDDI:
1716 	case IFT_XETHER:
1717 	case IFT_ISO88025:
1718 	case IFT_L2VLAN:
1719 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1720 		/* FALLTHROUGH */
1721 	case IFT_ARCNET:
1722 		bcopy(lladdr, LLADDR(sdl), len);
1723 		break;
1724 	default:
1725 		return (ENODEV);
1726 	}
1727 	/*
1728 	 * If the interface is already up, we need
1729 	 * to re-init it in order to reprogram its
1730 	 * address filter.
1731 	 */
1732 	if ((ifp->if_flags & IFF_UP) != 0) {
1733 		ifp->if_flags &= ~IFF_UP;
1734 		ifr.ifr_flags = ifp->if_flags;
1735 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1736 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1737 				 (struct ucred *)NULL);
1738 		ifp->if_flags |= IFF_UP;
1739 		ifr.ifr_flags = ifp->if_flags;
1740 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1741 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1742 				 (struct ucred *)NULL);
1743 #ifdef INET
1744 		/*
1745 		 * Also send gratuitous ARPs to notify other nodes about
1746 		 * the address change.
1747 		 */
1748 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1749 			if (ifa->ifa_addr != NULL &&
1750 			    ifa->ifa_addr->sa_family == AF_INET)
1751 				arp_ifinit(ifp, ifa);
1752 		}
1753 #endif
1754 	}
1755 	return (0);
1756 }
1757 
1758 struct ifmultiaddr *
1759 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1760 {
1761 	struct ifmultiaddr *ifma;
1762 
1763 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1764 		if (sa_equal(ifma->ifma_addr, sa))
1765 			break;
1766 
1767 	return ifma;
1768 }
1769 
1770 /*
1771  * The name argument must be a pointer to storage which will last as
1772  * long as the interface does.  For physical devices, the result of
1773  * device_get_name(dev) is a good choice and for pseudo-devices a
1774  * static string works well.
1775  */
1776 void
1777 if_initname(struct ifnet *ifp, const char *name, int unit)
1778 {
1779 	ifp->if_dname = name;
1780 	ifp->if_dunit = unit;
1781 	if (unit != IF_DUNIT_NONE)
1782 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1783 	else
1784 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1785 }
1786 
1787 int
1788 if_printf(struct ifnet *ifp, const char *fmt, ...)
1789 {
1790 	__va_list ap;
1791 	int retval;
1792 
1793 	retval = printf("%s: ", ifp->if_xname);
1794 	__va_start(ap, fmt);
1795 	retval += vprintf(fmt, ap);
1796 	__va_end(ap);
1797 	return (retval);
1798 }
1799 
1800 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1801 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1802 
1803 void
1804 ifq_set_classic(struct ifaltq *ifq)
1805 {
1806 	ifq->altq_enqueue = ifq_classic_enqueue;
1807 	ifq->altq_dequeue = ifq_classic_dequeue;
1808 	ifq->altq_request = ifq_classic_request;
1809 }
1810 
1811 static int
1812 ifq_classic_enqueue(struct ifaltq *ifq, struct mbuf *m,
1813 		    struct altq_pktattr *pa __unused)
1814 {
1815 	if (IF_QFULL(ifq)) {
1816 		m_freem(m);
1817 		return(ENOBUFS);
1818 	} else {
1819 		IF_ENQUEUE(ifq, m);
1820 		return(0);
1821 	}
1822 }
1823 
1824 static struct mbuf *
1825 ifq_classic_dequeue(struct ifaltq *ifq, int op)
1826 {
1827 	struct mbuf *m;
1828 
1829 	switch (op) {
1830 	case ALTDQ_POLL:
1831 		IF_POLL(ifq, m);
1832 		break;
1833 	case ALTDQ_REMOVE:
1834 		IF_DEQUEUE(ifq, m);
1835 		break;
1836 	default:
1837 		panic("unsupported ALTQ dequeue op: %d", op);
1838 	}
1839 
1840 	return(m);
1841 }
1842 
1843 static int
1844 ifq_classic_request(struct ifaltq *ifq, int req, void *arg)
1845 {
1846 	switch (req) {
1847 	case ALTRQ_PURGE:
1848 		IF_DRAIN(ifq);
1849 		break;
1850 	default:
1851 		panic("unspported ALTQ request: %d", req);
1852 	}
1853 
1854 	return(0);
1855 }
1856