xref: /dflybsd-src/sys/net/if.c (revision f7e25d559127833cbbb89b74fcf0cc036406459b)
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 1.23 2004/12/21 02:54:14 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 
57 #include <net/if.h>
58 #include <net/if_arp.h>
59 #include <net/if_dl.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/radix.h>
63 #include <net/route.h>
64 #include <machine/stdarg.h>
65 
66 #if defined(INET) || defined(INET6)
67 /*XXX*/
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <netinet/if_ether.h>
71 #ifdef INET6
72 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
73 #include <netinet6/in6_var.h>
74 #include <netinet6/in6_ifattach.h>
75 #endif
76 #endif
77 
78 #if defined(COMPAT_43)
79 #include <emulation/43bsd/43bsd_socket.h>
80 #endif /* COMPAT_43 */
81 
82 /*
83  * System initialization
84  */
85 
86 static int ifconf (u_long, caddr_t, struct thread *);
87 static void ifinit (void *);
88 static void if_qflush (struct ifqueue *);
89 static void if_slowtimo (void *);
90 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
91 static int  if_rtdel (struct radix_node *, void *);
92 
93 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
94 
95 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
96 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
97 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
98 
99 int	ifqmaxlen = IFQ_MAXLEN;
100 struct	ifnethead ifnet;	/* depend on static init XXX */
101 
102 #ifdef INET6
103 /*
104  * XXX: declare here to avoid to include many inet6 related files..
105  * should be more generalized?
106  */
107 extern void	nd6_setmtu (struct ifnet *);
108 #endif
109 
110 struct if_clone *if_clone_lookup (const char *, int *);
111 int if_clone_list (struct if_clonereq *);
112 
113 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
114 int if_cloners_count;
115 
116 struct callout if_slowtimo_timer;
117 
118 /*
119  * Network interface utility routines.
120  *
121  * Routines with ifa_ifwith* names take sockaddr *'s as
122  * parameters.
123  */
124 /* ARGSUSED*/
125 void
126 ifinit(void *dummy)
127 {
128 	struct ifnet *ifp;
129 	int s;
130 
131 	callout_init(&if_slowtimo_timer);
132 
133 	s = splimp();
134 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
135 		if (ifp->if_snd.ifq_maxlen == 0) {
136 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
137 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
138 		}
139 	}
140 	splx(s);
141 
142 	if_slowtimo(0);
143 }
144 
145 int if_index = 0;
146 struct ifaddr **ifnet_addrs;
147 struct ifnet **ifindex2ifnet = NULL;
148 
149 /*
150  * Attach an interface to the
151  * list of "active" interfaces.
152  */
153 void
154 if_attach(struct ifnet *ifp)
155 {
156 	unsigned socksize, ifasize;
157 	int namelen, masklen;
158 	struct sockaddr_dl *sdl;
159 	struct ifaddr *ifa;
160 	static int if_indexlim = 8;
161 	static int inited;
162 
163 	if (!inited) {
164 		TAILQ_INIT(&ifnet);
165 		inited = 1;
166 	}
167 
168 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
169 	ifp->if_index = ++if_index;
170 	/*
171 	 * XXX -
172 	 * The old code would work if the interface passed a pre-existing
173 	 * chain of ifaddrs to this code.  We don't trust our callers to
174 	 * properly initialize the tailq, however, so we no longer allow
175 	 * this unlikely case.
176 	 */
177 	TAILQ_INIT(&ifp->if_addrhead);
178 	TAILQ_INIT(&ifp->if_prefixhead);
179 	LIST_INIT(&ifp->if_multiaddrs);
180 	getmicrotime(&ifp->if_lastchange);
181 	if (ifnet_addrs == NULL || if_index >= if_indexlim) {
182 		unsigned n = (if_indexlim <<= 1) * sizeof(ifa);
183 		caddr_t q = malloc(n, M_IFADDR, M_WAITOK);
184 		bzero(q, n);
185 		if (ifnet_addrs != NULL) {
186 			bcopy(ifnet_addrs, q, n/2);
187 			free(ifnet_addrs, M_IFADDR);
188 		}
189 		ifnet_addrs = (struct ifaddr **)q;
190 
191 		/* grow ifindex2ifnet */
192 		n = if_indexlim * sizeof(struct ifnet *);
193 		q = malloc(n, M_IFADDR, M_WAITOK);
194 		bzero(q, n);
195 		if (ifindex2ifnet) {
196 			bcopy(ifindex2ifnet, q, n/2);
197 			free(ifindex2ifnet, M_IFADDR);
198 		}
199 		ifindex2ifnet = (struct ifnet **)q;
200 	}
201 
202 	ifindex2ifnet[if_index] = ifp;
203 
204 	/*
205 	 * create a Link Level name for this device
206 	 */
207 	namelen = strlen(ifp->if_xname);
208 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
209 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
210 	socksize = masklen + ifp->if_addrlen;
211 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
212 	if (socksize < sizeof(*sdl))
213 		socksize = sizeof(*sdl);
214 	socksize = ROUNDUP(socksize);
215 	ifasize = sizeof(*ifa) + 2 * socksize;
216 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
217 	if (ifa) {
218 		bzero(ifa, ifasize);
219 		sdl = (struct sockaddr_dl *)(ifa + 1);
220 		sdl->sdl_len = socksize;
221 		sdl->sdl_family = AF_LINK;
222 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
223 		sdl->sdl_nlen = namelen;
224 		sdl->sdl_index = ifp->if_index;
225 		sdl->sdl_type = ifp->if_type;
226 		ifnet_addrs[if_index - 1] = ifa;
227 		ifa->ifa_ifp = ifp;
228 		ifa->ifa_rtrequest = link_rtrequest;
229 		ifa->ifa_addr = (struct sockaddr *)sdl;
230 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
231 		ifa->ifa_netmask = (struct sockaddr *)sdl;
232 		sdl->sdl_len = masklen;
233 		while (namelen != 0)
234 			sdl->sdl_data[--namelen] = 0xff;
235 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
236 	}
237 
238 	EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
239 
240 	/* Announce the interface. */
241 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
242 }
243 
244 /*
245  * Detach an interface, removing it from the
246  * list of "active" interfaces.
247  */
248 void
249 if_detach(struct ifnet *ifp)
250 {
251 	struct ifaddr *ifa;
252 	struct radix_node_head	*rnh;
253 	int s;
254 	int i;
255 
256 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
257 
258 	/*
259 	 * Remove routes and flush queues.
260 	 */
261 	s = splnet();
262 	if_down(ifp);
263 
264 	/*
265 	 * Remove address from ifnet_addrs[] and maybe decrement if_index.
266 	 * Clean up all addresses.
267 	 */
268 	ifnet_addrs[ifp->if_index - 1] = 0;
269 	while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
270 		if_index--;
271 
272 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
273 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
274 #ifdef INET
275 		/* XXX: Ugly!! ad hoc just for INET */
276 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
277 			struct ifaliasreq ifr;
278 
279 			bzero(&ifr, sizeof ifr);
280 			ifr.ifra_addr = *ifa->ifa_addr;
281 			if (ifa->ifa_dstaddr)
282 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
283 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
284 			    NULL) == 0)
285 				continue;
286 		}
287 #endif /* INET */
288 #ifdef INET6
289 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
290 			in6_purgeaddr(ifa);
291 			/* ifp_addrhead is already updated */
292 			continue;
293 		}
294 #endif /* INET6 */
295 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
296 		IFAFREE(ifa);
297 	}
298 
299 #ifdef INET6
300 	/*
301 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
302 	 * before removing routing entries below, since IPv6 interface direct
303 	 * routes are expected to be removed by the IPv6-specific kernel API.
304 	 * Otherwise, the kernel will detect some inconsistency and bark it.
305 	 */
306 	in6_ifdetach(ifp);
307 #endif
308 
309 	/*
310 	 * Delete all remaining routes using this interface
311 	 * Unfortuneatly the only way to do this is to slog through
312 	 * the entire routing table looking for routes which point
313 	 * to this interface...oh well...
314 	 */
315 	for (i = 1; i <= AF_MAX; i++) {
316 		if ((rnh = rt_tables[i]) == NULL)
317 			continue;
318 		rnh->rnh_walktree(rnh, if_rtdel, ifp);
319 	}
320 
321 	/* Announce that the interface is gone. */
322 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
323 
324 	ifindex2ifnet[ifp->if_index] = NULL;
325 
326 	TAILQ_REMOVE(&ifnet, ifp, if_link);
327 	splx(s);
328 }
329 
330 /*
331  * Delete Routes for a Network Interface
332  *
333  * Called for each routing entry via the rnh->rnh_walktree() call above
334  * to delete all route entries referencing a detaching network interface.
335  *
336  * Arguments:
337  *	rn	pointer to node in the routing table
338  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
339  *
340  * Returns:
341  *	0	successful
342  *	errno	failed - reason indicated
343  *
344  */
345 static int
346 if_rtdel(struct radix_node *rn, void *arg)
347 {
348 	struct rtentry	*rt = (struct rtentry *)rn;
349 	struct ifnet	*ifp = arg;
350 	int		err;
351 
352 	if (rt->rt_ifp == ifp) {
353 
354 		/*
355 		 * Protect (sorta) against walktree recursion problems
356 		 * with cloned routes
357 		 */
358 		if (!(rt->rt_flags & RTF_UP))
359 			return (0);
360 
361 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
362 				rt_mask(rt), rt->rt_flags,
363 				(struct rtentry **) NULL);
364 		if (err) {
365 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
366 		}
367 	}
368 
369 	return (0);
370 }
371 
372 /*
373  * Create a clone network interface.
374  */
375 int
376 if_clone_create(char *name, int len)
377 {
378 	struct if_clone *ifc;
379 	char *dp;
380 	int wildcard, bytoff, bitoff;
381 	int unit;
382 	int err;
383 
384 	ifc = if_clone_lookup(name, &unit);
385 	if (ifc == NULL)
386 		return (EINVAL);
387 
388 	if (ifunit(name) != NULL)
389 		return (EEXIST);
390 
391 	bytoff = bitoff = 0;
392 	wildcard = (unit < 0);
393 	/*
394 	 * Find a free unit if none was given.
395 	 */
396 	if (wildcard) {
397 		while (bytoff < ifc->ifc_bmlen &&
398 		    ifc->ifc_units[bytoff] == 0xff)
399 			bytoff++;
400 		if (bytoff >= ifc->ifc_bmlen)
401 			return (ENOSPC);
402 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
403 			bitoff++;
404 		unit = (bytoff << 3) + bitoff;
405 	}
406 
407 	if (unit > ifc->ifc_maxunit)
408 		return (ENXIO);
409 
410 	err = (*ifc->ifc_create)(ifc, unit);
411 	if (err != 0)
412 		return (err);
413 
414 	if (!wildcard) {
415 		bytoff = unit >> 3;
416 		bitoff = unit - (bytoff << 3);
417 	}
418 
419 	/*
420 	 * Allocate the unit in the bitmap.
421 	 */
422 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
423 	    ("%s: bit is already set", __func__));
424 	ifc->ifc_units[bytoff] |= (1 << bitoff);
425 
426 	/* In the wildcard case, we need to update the name. */
427 	if (wildcard) {
428 		for (dp = name; *dp != '\0'; dp++);
429 		if (snprintf(dp, len - (dp-name), "%d", unit) >
430 		    len - (dp-name) - 1) {
431 			/*
432 			 * This can only be a programmer error and
433 			 * there's no straightforward way to recover if
434 			 * it happens.
435 			 */
436 			panic("if_clone_create(): interface name too long");
437 		}
438 
439 	}
440 
441 	EVENTHANDLER_INVOKE(if_clone_event, ifc);
442 
443 	return (0);
444 }
445 
446 /*
447  * Destroy a clone network interface.
448  */
449 int
450 if_clone_destroy(const char *name)
451 {
452 	struct if_clone *ifc;
453 	struct ifnet *ifp;
454 	int bytoff, bitoff;
455 	int unit;
456 
457 	ifc = if_clone_lookup(name, &unit);
458 	if (ifc == NULL)
459 		return (EINVAL);
460 
461 	if (unit < ifc->ifc_minifs)
462 		return (EINVAL);
463 
464 	ifp = ifunit(name);
465 	if (ifp == NULL)
466 		return (ENXIO);
467 
468 	if (ifc->ifc_destroy == NULL)
469 		return (EOPNOTSUPP);
470 
471 	(*ifc->ifc_destroy)(ifp);
472 
473 	/*
474 	 * Compute offset in the bitmap and deallocate the unit.
475 	 */
476 	bytoff = unit >> 3;
477 	bitoff = unit - (bytoff << 3);
478 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
479 	    ("%s: bit is already cleared", __func__));
480 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
481 	return (0);
482 }
483 
484 /*
485  * Look up a network interface cloner.
486  */
487 struct if_clone *
488 if_clone_lookup(const char *name, int *unitp)
489 {
490 	struct if_clone *ifc;
491 	const char *cp;
492 	int i;
493 
494 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
495 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
496 			if (ifc->ifc_name[i] != *cp)
497 				goto next_ifc;
498 		}
499 		goto found_name;
500  next_ifc:
501 		ifc = LIST_NEXT(ifc, ifc_list);
502 	}
503 
504 	/* No match. */
505 	return ((struct if_clone *)NULL);
506 
507  found_name:
508 	if (*cp == '\0') {
509 		i = -1;
510 	} else {
511 		for (i = 0; *cp != '\0'; cp++) {
512 			if (*cp < '0' || *cp > '9') {
513 				/* Bogus unit number. */
514 				return (NULL);
515 			}
516 			i = (i * 10) + (*cp - '0');
517 		}
518 	}
519 
520 	if (unitp != NULL)
521 		*unitp = i;
522 	return (ifc);
523 }
524 
525 /*
526  * Register a network interface cloner.
527  */
528 void
529 if_clone_attach(struct if_clone *ifc)
530 {
531 	int bytoff, bitoff;
532 	int err;
533 	int len, maxclone;
534 	int unit;
535 
536 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
537 	    ("%s: %s requested more units then allowed (%d > %d)",
538 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
539 	    ifc->ifc_maxunit + 1));
540 	/*
541 	 * Compute bitmap size and allocate it.
542 	 */
543 	maxclone = ifc->ifc_maxunit + 1;
544 	len = maxclone >> 3;
545 	if ((len << 3) < maxclone)
546 		len++;
547 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
548 	ifc->ifc_bmlen = len;
549 
550 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
551 	if_cloners_count++;
552 
553 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
554 		err = (*ifc->ifc_create)(ifc, unit);
555 		KASSERT(err == 0,
556 		    ("%s: failed to create required interface %s%d",
557 		    __func__, ifc->ifc_name, unit));
558 
559 		/* Allocate the unit in the bitmap. */
560 		bytoff = unit >> 3;
561 		bitoff = unit - (bytoff << 3);
562 		ifc->ifc_units[bytoff] |= (1 << bitoff);
563 	}
564 }
565 
566 /*
567  * Unregister a network interface cloner.
568  */
569 void
570 if_clone_detach(struct if_clone *ifc)
571 {
572 
573 	LIST_REMOVE(ifc, ifc_list);
574 	free(ifc->ifc_units, M_CLONE);
575 	if_cloners_count--;
576 }
577 
578 /*
579  * Provide list of interface cloners to userspace.
580  */
581 int
582 if_clone_list(struct if_clonereq *ifcr)
583 {
584 	char outbuf[IFNAMSIZ], *dst;
585 	struct if_clone *ifc;
586 	int count, error = 0;
587 
588 	ifcr->ifcr_total = if_cloners_count;
589 	if ((dst = ifcr->ifcr_buffer) == NULL) {
590 		/* Just asking how many there are. */
591 		return (0);
592 	}
593 
594 	if (ifcr->ifcr_count < 0)
595 		return (EINVAL);
596 
597 	count = (if_cloners_count < ifcr->ifcr_count) ?
598 	    if_cloners_count : ifcr->ifcr_count;
599 
600 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
601 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
602 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
603 		error = copyout(outbuf, dst, IFNAMSIZ);
604 		if (error)
605 			break;
606 	}
607 
608 	return (error);
609 }
610 
611 /*
612  * Locate an interface based on a complete address.
613  */
614 struct ifaddr *
615 ifa_ifwithaddr(struct sockaddr *addr)
616 {
617 	struct ifnet *ifp;
618 	struct ifaddr *ifa;
619 
620 #define	equal(a1, a2) \
621   (bcmp((a1), (a2), ((struct sockaddr *)(a1))->sa_len) == 0)
622 
623 	TAILQ_FOREACH(ifp, &ifnet, if_link)
624 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
625 		if (ifa->ifa_addr->sa_family != addr->sa_family)
626 			continue;
627 		if (equal(addr, ifa->ifa_addr))
628 			return (ifa);
629 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
630 		    /* IP6 doesn't have broadcast */
631 		    ifa->ifa_broadaddr->sa_len != 0 &&
632 		    equal(ifa->ifa_broadaddr, addr))
633 			return (ifa);
634 	}
635 	return ((struct ifaddr *)0);
636 }
637 /*
638  * Locate the point to point interface with a given destination address.
639  */
640 struct ifaddr *
641 ifa_ifwithdstaddr(struct sockaddr *addr)
642 {
643 	struct ifnet *ifp;
644 	struct ifaddr *ifa;
645 
646 	TAILQ_FOREACH(ifp, &ifnet, if_link)
647 	    if (ifp->if_flags & IFF_POINTOPOINT)
648 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
649 			if (ifa->ifa_addr->sa_family != addr->sa_family)
650 				continue;
651 			if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
652 				return (ifa);
653 	}
654 	return ((struct ifaddr *)0);
655 }
656 
657 /*
658  * Find an interface on a specific network.  If many, choice
659  * is most specific found.
660  */
661 struct ifaddr *
662 ifa_ifwithnet(struct sockaddr *addr)
663 {
664 	struct ifnet *ifp;
665 	struct ifaddr *ifa;
666 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
667 	u_int af = addr->sa_family;
668 	char *addr_data = addr->sa_data, *cplim;
669 
670 	/*
671 	 * AF_LINK addresses can be looked up directly by their index number,
672 	 * so do that if we can.
673 	 */
674 	if (af == AF_LINK) {
675 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
676 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
677 		return (ifnet_addrs[sdl->sdl_index - 1]);
678 	}
679 
680 	/*
681 	 * Scan though each interface, looking for ones that have
682 	 * addresses in this address family.
683 	 */
684 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
685 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
686 			char *cp, *cp2, *cp3;
687 
688 			if (ifa->ifa_addr->sa_family != af)
689 next:				continue;
690 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
691 				/*
692 				 * This is a bit broken as it doesn't
693 				 * take into account that the remote end may
694 				 * be a single node in the network we are
695 				 * looking for.
696 				 * The trouble is that we don't know the
697 				 * netmask for the remote end.
698 				 */
699 				if (ifa->ifa_dstaddr != 0 &&
700 				    equal(addr, ifa->ifa_dstaddr))
701 					return (ifa);
702 			} else {
703 				/*
704 				 * if we have a special address handler,
705 				 * then use it instead of the generic one.
706 				 */
707 				if (ifa->ifa_claim_addr) {
708 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
709 						return (ifa);
710 					} else {
711 						continue;
712 					}
713 				}
714 
715 				/*
716 				 * Scan all the bits in the ifa's address.
717 				 * If a bit dissagrees with what we are
718 				 * looking for, mask it with the netmask
719 				 * to see if it really matters.
720 				 * (A byte at a time)
721 				 */
722 				if (ifa->ifa_netmask == 0)
723 					continue;
724 				cp = addr_data;
725 				cp2 = ifa->ifa_addr->sa_data;
726 				cp3 = ifa->ifa_netmask->sa_data;
727 				cplim = ifa->ifa_netmask->sa_len
728 					+ (char *)ifa->ifa_netmask;
729 				while (cp3 < cplim)
730 					if ((*cp++ ^ *cp2++) & *cp3++)
731 						goto next; /* next address! */
732 				/*
733 				 * If the netmask of what we just found
734 				 * is more specific than what we had before
735 				 * (if we had one) then remember the new one
736 				 * before continuing to search
737 				 * for an even better one.
738 				 */
739 				if (ifa_maybe == 0 ||
740 				    rn_refines((char *)ifa->ifa_netmask,
741 					       (char *)ifa_maybe->ifa_netmask))
742 					ifa_maybe = ifa;
743 			}
744 		}
745 	}
746 	return (ifa_maybe);
747 }
748 
749 /*
750  * Find an interface address specific to an interface best matching
751  * a given address.
752  */
753 struct ifaddr *
754 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
755 {
756 	struct ifaddr *ifa;
757 	char *cp, *cp2, *cp3;
758 	char *cplim;
759 	struct ifaddr *ifa_maybe = 0;
760 	u_int af = addr->sa_family;
761 
762 	if (af >= AF_MAX)
763 		return (0);
764 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
765 		if (ifa->ifa_addr->sa_family != af)
766 			continue;
767 		if (ifa_maybe == 0)
768 			ifa_maybe = ifa;
769 		if (ifa->ifa_netmask == 0) {
770 			if (equal(addr, ifa->ifa_addr) ||
771 			    (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
772 				return (ifa);
773 			continue;
774 		}
775 		if (ifp->if_flags & IFF_POINTOPOINT) {
776 			if (equal(addr, ifa->ifa_dstaddr))
777 				return (ifa);
778 		} else {
779 			cp = addr->sa_data;
780 			cp2 = ifa->ifa_addr->sa_data;
781 			cp3 = ifa->ifa_netmask->sa_data;
782 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
783 			for (; cp3 < cplim; cp3++)
784 				if ((*cp++ ^ *cp2++) & *cp3)
785 					break;
786 			if (cp3 == cplim)
787 				return (ifa);
788 		}
789 	}
790 	return (ifa_maybe);
791 }
792 
793 #include <net/route.h>
794 
795 /*
796  * Default action when installing a route with a Link Level gateway.
797  * Lookup an appropriate real ifa to point to.
798  * This should be moved to /sys/net/link.c eventually.
799  */
800 static void
801 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
802 {
803 	struct ifaddr *ifa;
804 	struct sockaddr *dst;
805 	struct ifnet *ifp;
806 
807 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
808 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
809 		return;
810 	ifa = ifaof_ifpforaddr(dst, ifp);
811 	if (ifa != NULL) {
812 		IFAFREE(rt->rt_ifa);
813 		IFAREF(ifa);
814 		rt->rt_ifa = ifa;
815 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
816 			ifa->ifa_rtrequest(cmd, rt, info);
817 	}
818 }
819 
820 /*
821  * Mark an interface down and notify protocols of
822  * the transition.
823  * NOTE: must be called at splnet or eqivalent.
824  */
825 void
826 if_unroute(struct ifnet *ifp, int flag, int fam)
827 {
828 	struct ifaddr *ifa;
829 
830 	ifp->if_flags &= ~flag;
831 	getmicrotime(&ifp->if_lastchange);
832 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
833 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
834 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
835 	if_qflush(&ifp->if_snd);
836 	rt_ifmsg(ifp);
837 }
838 
839 /*
840  * Mark an interface up and notify protocols of
841  * the transition.
842  * NOTE: must be called at splnet or eqivalent.
843  */
844 void
845 if_route(struct ifnet *ifp, int flag, int fam)
846 {
847 	struct ifaddr *ifa;
848 
849 	ifp->if_flags |= flag;
850 	getmicrotime(&ifp->if_lastchange);
851 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
852 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
853 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
854 	rt_ifmsg(ifp);
855 #ifdef INET6
856 	in6_if_up(ifp);
857 #endif
858 }
859 
860 /*
861  * Mark an interface down and notify protocols of
862  * the transition.
863  * NOTE: must be called at splnet or eqivalent.
864  */
865 void
866 if_down(struct ifnet *ifp)
867 {
868 
869 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
870 }
871 
872 /*
873  * Mark an interface up and notify protocols of
874  * the transition.
875  * NOTE: must be called at splnet or eqivalent.
876  */
877 void
878 if_up(struct ifnet *ifp)
879 {
880 
881 	if_route(ifp, IFF_UP, AF_UNSPEC);
882 }
883 
884 /*
885  * Flush an interface queue.
886  */
887 static void
888 if_qflush(struct ifqueue *ifq)
889 {
890 	struct mbuf *m, *n;
891 
892 	n = ifq->ifq_head;
893 	while ((m = n) != 0) {
894 		n = m->m_nextpkt;
895 		m_freem(m);
896 	}
897 	ifq->ifq_head = 0;
898 	ifq->ifq_tail = 0;
899 	ifq->ifq_len = 0;
900 }
901 
902 /*
903  * Handle interface watchdog timer routines.  Called
904  * from softclock, we decrement timers (if set) and
905  * call the appropriate interface routine on expiration.
906  */
907 static void
908 if_slowtimo(void *arg)
909 {
910 	struct ifnet *ifp;
911 	int s = splimp();
912 
913 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
914 		if (ifp->if_timer == 0 || --ifp->if_timer)
915 			continue;
916 		if (ifp->if_watchdog)
917 			(*ifp->if_watchdog)(ifp);
918 	}
919 	splx(s);
920 	callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
921 }
922 
923 /*
924  * Map interface name to
925  * interface structure pointer.
926  */
927 struct ifnet *
928 ifunit(const char *name)
929 {
930 	struct ifnet *ifp;
931 
932 	/*
933 	 * Search all the interfaces for this name/number
934 	 */
935 
936 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
937 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
938 			break;
939 	}
940 	return (ifp);
941 }
942 
943 
944 /*
945  * Map interface name in a sockaddr_dl to
946  * interface structure pointer.
947  */
948 struct ifnet *
949 if_withname(struct sockaddr *sa)
950 {
951 	char ifname[IFNAMSIZ+1];
952 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
953 
954 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
955 	     (sdl->sdl_nlen > IFNAMSIZ) )
956 		return NULL;
957 
958 	/*
959 	 * ifunit wants a null-terminated name.  It may not be null-terminated
960 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
961 	 * and there might not be room to put the trailing null anyway, so we
962 	 * make a local copy that we know we can null terminate safely.
963 	 */
964 
965 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
966 	ifname[sdl->sdl_nlen] = '\0';
967 	return ifunit(ifname);
968 }
969 
970 
971 /*
972  * Interface ioctls.
973  */
974 int
975 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
976 {
977 	struct ifnet *ifp;
978 	struct ifreq *ifr;
979 	struct ifstat *ifs;
980 	int error;
981 	short oif_flags;
982 	int new_flags;
983 	size_t namelen, onamelen;
984 	char new_name[IFNAMSIZ];
985 	struct ifaddr *ifa;
986 	struct sockaddr_dl *sdl;
987 
988 	switch (cmd) {
989 
990 	case SIOCGIFCONF:
991 	case OSIOCGIFCONF:
992 		return (ifconf(cmd, data, td));
993 	}
994 	ifr = (struct ifreq *)data;
995 
996 	switch (cmd) {
997 	case SIOCIFCREATE:
998 	case SIOCIFDESTROY:
999 		if ((error = suser(td)) != 0)
1000 			return (error);
1001 		return ((cmd == SIOCIFCREATE) ?
1002 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1003 			if_clone_destroy(ifr->ifr_name));
1004 
1005 	case SIOCIFGCLONERS:
1006 		return (if_clone_list((struct if_clonereq *)data));
1007 	}
1008 
1009 	ifp = ifunit(ifr->ifr_name);
1010 	if (ifp == 0)
1011 		return (ENXIO);
1012 	switch (cmd) {
1013 
1014 	case SIOCGIFFLAGS:
1015 		ifr->ifr_flags = ifp->if_flags;
1016 		ifr->ifr_flagshigh = ifp->if_flags >> 16;
1017 		break;
1018 
1019 	case SIOCGIFCAP:
1020 		ifr->ifr_reqcap = ifp->if_capabilities;
1021 		ifr->ifr_curcap = ifp->if_capenable;
1022 		break;
1023 
1024 	case SIOCGIFMETRIC:
1025 		ifr->ifr_metric = ifp->if_metric;
1026 		break;
1027 
1028 	case SIOCGIFMTU:
1029 		ifr->ifr_mtu = ifp->if_mtu;
1030 		break;
1031 
1032 	case SIOCGIFPHYS:
1033 		ifr->ifr_phys = ifp->if_physical;
1034 		break;
1035 
1036 	case SIOCSIFFLAGS:
1037 		error = suser(td);
1038 		if (error)
1039 			return (error);
1040 		new_flags = (ifr->ifr_flags & 0xffff) |
1041 		    (ifr->ifr_flagshigh << 16);
1042 		if (ifp->if_flags & IFF_SMART) {
1043 			/* Smart drivers twiddle their own routes */
1044 		} else if (ifp->if_flags & IFF_UP &&
1045 		    (new_flags & IFF_UP) == 0) {
1046 			int s = splimp();
1047 			if_down(ifp);
1048 			splx(s);
1049 		} else if (new_flags & IFF_UP &&
1050 		    (ifp->if_flags & IFF_UP) == 0) {
1051 			int s = splimp();
1052 			if_up(ifp);
1053 			splx(s);
1054 		}
1055 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1056 			(new_flags &~ IFF_CANTCHANGE);
1057 		if (new_flags & IFF_PPROMISC) {
1058 			/* Permanently promiscuous mode requested */
1059 			ifp->if_flags |= IFF_PROMISC;
1060 		} else if (ifp->if_pcount == 0) {
1061 			ifp->if_flags &= ~IFF_PROMISC;
1062 		}
1063 		if (ifp->if_ioctl)
1064 			(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1065 		getmicrotime(&ifp->if_lastchange);
1066 		break;
1067 
1068 	case SIOCSIFCAP:
1069 		error = suser(td);
1070 		if (error)
1071 			return (error);
1072 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1073 			return (EINVAL);
1074 		(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1075 		break;
1076 
1077 	case SIOCSIFNAME:
1078 		error = suser(td);
1079 		if (error != 0)
1080 			return (error);
1081 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1082 		if (error != 0)
1083 			return (error);
1084 		if (new_name[0] == '\0')
1085 			return (EINVAL);
1086 		if (ifunit(new_name) != NULL)
1087 			return (EEXIST);
1088 
1089 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1090 
1091 		/* Announce the departure of the interface. */
1092 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1093 
1094 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1095 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1096 		/* XXX IFA_LOCK(ifa); */
1097 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1098 		namelen = strlen(new_name);
1099 		onamelen = sdl->sdl_nlen;
1100 		/*
1101 		 * Move the address if needed.  This is safe because we
1102 		 * allocate space for a name of length IFNAMSIZ when we
1103 		 * create this in if_attach().
1104 		 */
1105 		if (namelen != onamelen) {
1106 			bcopy(sdl->sdl_data + onamelen,
1107 			    sdl->sdl_data + namelen, sdl->sdl_alen);
1108 		}
1109 		bcopy(new_name, sdl->sdl_data, namelen);
1110 		sdl->sdl_nlen = namelen;
1111 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1112 		bzero(sdl->sdl_data, onamelen);
1113 		while (namelen != 0)
1114 			sdl->sdl_data[--namelen] = 0xff;
1115 		/* XXX IFA_UNLOCK(ifa) */
1116 
1117 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1118 
1119 		/* Announce the return of the interface. */
1120 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1121 		break;
1122 
1123 	case SIOCSIFMETRIC:
1124 		error = suser(td);
1125 		if (error)
1126 			return (error);
1127 		ifp->if_metric = ifr->ifr_metric;
1128 		getmicrotime(&ifp->if_lastchange);
1129 		break;
1130 
1131 	case SIOCSIFPHYS:
1132 		error = suser(td);
1133 		if (error)
1134 			return error;
1135 		if (!ifp->if_ioctl)
1136 		        return EOPNOTSUPP;
1137 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1138 		if (error == 0)
1139 			getmicrotime(&ifp->if_lastchange);
1140 		return (error);
1141 
1142 	case SIOCSIFMTU:
1143 	{
1144 		u_long oldmtu = ifp->if_mtu;
1145 
1146 		error = suser(td);
1147 		if (error)
1148 			return (error);
1149 		if (ifp->if_ioctl == NULL)
1150 			return (EOPNOTSUPP);
1151 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1152 			return (EINVAL);
1153 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1154 		if (error == 0) {
1155 			getmicrotime(&ifp->if_lastchange);
1156 			rt_ifmsg(ifp);
1157 		}
1158 		/*
1159 		 * If the link MTU changed, do network layer specific procedure.
1160 		 */
1161 		if (ifp->if_mtu != oldmtu) {
1162 #ifdef INET6
1163 			nd6_setmtu(ifp);
1164 #endif
1165 		}
1166 		return (error);
1167 	}
1168 
1169 	case SIOCADDMULTI:
1170 	case SIOCDELMULTI:
1171 		error = suser(td);
1172 		if (error)
1173 			return (error);
1174 
1175 		/* Don't allow group membership on non-multicast interfaces. */
1176 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1177 			return EOPNOTSUPP;
1178 
1179 		/* Don't let users screw up protocols' entries. */
1180 		if (ifr->ifr_addr.sa_family != AF_LINK)
1181 			return EINVAL;
1182 
1183 		if (cmd == SIOCADDMULTI) {
1184 			struct ifmultiaddr *ifma;
1185 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1186 		} else {
1187 			error = if_delmulti(ifp, &ifr->ifr_addr);
1188 		}
1189 		if (error == 0)
1190 			getmicrotime(&ifp->if_lastchange);
1191 		return error;
1192 
1193 	case SIOCSIFPHYADDR:
1194 	case SIOCDIFPHYADDR:
1195 #ifdef INET6
1196 	case SIOCSIFPHYADDR_IN6:
1197 #endif
1198 	case SIOCSLIFPHYADDR:
1199         case SIOCSIFMEDIA:
1200 	case SIOCSIFGENERIC:
1201 		error = suser(td);
1202 		if (error)
1203 			return (error);
1204 		if (ifp->if_ioctl == 0)
1205 			return (EOPNOTSUPP);
1206 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1207 		if (error == 0)
1208 			getmicrotime(&ifp->if_lastchange);
1209 		return error;
1210 
1211 	case SIOCGIFSTATUS:
1212 		ifs = (struct ifstat *)data;
1213 		ifs->ascii[0] = '\0';
1214 
1215 	case SIOCGIFPSRCADDR:
1216 	case SIOCGIFPDSTADDR:
1217 	case SIOCGLIFPHYADDR:
1218 	case SIOCGIFMEDIA:
1219 	case SIOCGIFGENERIC:
1220 		if (ifp->if_ioctl == 0)
1221 			return (EOPNOTSUPP);
1222 		return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1223 
1224 	case SIOCSIFLLADDR:
1225 		error = suser(td);
1226 		if (error)
1227 			return (error);
1228 		return if_setlladdr(ifp,
1229 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1230 
1231 	default:
1232 		oif_flags = ifp->if_flags;
1233 		if (so->so_proto == 0)
1234 			return (EOPNOTSUPP);
1235 #ifndef COMPAT_43
1236 		error = so_pru_control(so, cmd, data, ifp, td);
1237 #else
1238 	    {
1239 		int ocmd = cmd;
1240 
1241 		switch (cmd) {
1242 
1243 		case SIOCSIFDSTADDR:
1244 		case SIOCSIFADDR:
1245 		case SIOCSIFBRDADDR:
1246 		case SIOCSIFNETMASK:
1247 #if BYTE_ORDER != BIG_ENDIAN
1248 			if (ifr->ifr_addr.sa_family == 0 &&
1249 			    ifr->ifr_addr.sa_len < 16) {
1250 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1251 				ifr->ifr_addr.sa_len = 16;
1252 			}
1253 #else
1254 			if (ifr->ifr_addr.sa_len == 0)
1255 				ifr->ifr_addr.sa_len = 16;
1256 #endif
1257 			break;
1258 
1259 		case OSIOCGIFADDR:
1260 			cmd = SIOCGIFADDR;
1261 			break;
1262 
1263 		case OSIOCGIFDSTADDR:
1264 			cmd = SIOCGIFDSTADDR;
1265 			break;
1266 
1267 		case OSIOCGIFBRDADDR:
1268 			cmd = SIOCGIFBRDADDR;
1269 			break;
1270 
1271 		case OSIOCGIFNETMASK:
1272 			cmd = SIOCGIFNETMASK;
1273 		}
1274 		error =  so_pru_control(so, cmd, data, ifp, td);
1275 		switch (ocmd) {
1276 
1277 		case OSIOCGIFADDR:
1278 		case OSIOCGIFDSTADDR:
1279 		case OSIOCGIFBRDADDR:
1280 		case OSIOCGIFNETMASK:
1281 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1282 
1283 		}
1284 	    }
1285 #endif /* COMPAT_43 */
1286 
1287 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1288 #ifdef INET6
1289 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1290 			if (ifp->if_flags & IFF_UP) {
1291 				int s = splimp();
1292 				in6_if_up(ifp);
1293 				splx(s);
1294 			}
1295 #endif
1296 		}
1297 		return (error);
1298 
1299 	}
1300 	return (0);
1301 }
1302 
1303 /*
1304  * Set/clear promiscuous mode on interface ifp based on the truth value
1305  * of pswitch.  The calls are reference counted so that only the first
1306  * "on" request actually has an effect, as does the final "off" request.
1307  * Results are undefined if the "off" and "on" requests are not matched.
1308  */
1309 int
1310 ifpromisc(struct ifnet *ifp, int pswitch)
1311 {
1312 	struct ifreq ifr;
1313 	int error;
1314 	int oldflags;
1315 
1316 	oldflags = ifp->if_flags;
1317 	if (ifp->if_flags & IFF_PPROMISC) {
1318 		/* Do nothing if device is in permanently promiscuous mode */
1319 		ifp->if_pcount += pswitch ? 1 : -1;
1320 		return (0);
1321 	}
1322 	if (pswitch) {
1323 		/*
1324 		 * If the device is not configured up, we cannot put it in
1325 		 * promiscuous mode.
1326 		 */
1327 		if ((ifp->if_flags & IFF_UP) == 0)
1328 			return (ENETDOWN);
1329 		if (ifp->if_pcount++ != 0)
1330 			return (0);
1331 		ifp->if_flags |= IFF_PROMISC;
1332 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1333 		    ifp->if_xname);
1334 	} else {
1335 		if (--ifp->if_pcount > 0)
1336 			return (0);
1337 		ifp->if_flags &= ~IFF_PROMISC;
1338 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1339 		    ifp->if_xname);
1340 	}
1341 	ifr.ifr_flags = ifp->if_flags;
1342 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
1343 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1344 				 (struct ucred *)NULL);
1345 	if (error == 0)
1346 		rt_ifmsg(ifp);
1347 	else
1348 		ifp->if_flags = oldflags;
1349 	return error;
1350 }
1351 
1352 /*
1353  * Return interface configuration
1354  * of system.  List may be used
1355  * in later ioctl's (above) to get
1356  * other information.
1357  */
1358 static int
1359 ifconf(u_long cmd, caddr_t data, struct thread *td)
1360 {
1361 	struct ifconf *ifc = (struct ifconf *)data;
1362 	struct ifnet *ifp;
1363 	struct ifaddr *ifa;
1364 	struct sockaddr *sa;
1365 	struct ifreq ifr, *ifrp;
1366 	int space = ifc->ifc_len, error = 0;
1367 
1368 	ifrp = ifc->ifc_req;
1369 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1370 		int addrs;
1371 
1372 		if (space <= sizeof ifr)
1373 			break;
1374 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1375 		    >= sizeof(ifr.ifr_name)) {
1376 			error = ENAMETOOLONG;
1377 			break;
1378 		}
1379 
1380 		addrs = 0;
1381 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1382 			if (space <= sizeof ifr)
1383 				break;
1384 			sa = ifa->ifa_addr;
1385 			if (td->td_proc->p_ucred->cr_prison &&
1386 			    prison_if(td, sa))
1387 				continue;
1388 			addrs++;
1389 #ifdef COMPAT_43
1390 			if (cmd == OSIOCGIFCONF) {
1391 				struct osockaddr *osa =
1392 					 (struct osockaddr *)&ifr.ifr_addr;
1393 				ifr.ifr_addr = *sa;
1394 				osa->sa_family = sa->sa_family;
1395 				error = copyout(&ifr, ifrp, sizeof ifr);
1396 				ifrp++;
1397 			} else
1398 #endif
1399 			if (sa->sa_len <= sizeof(*sa)) {
1400 				ifr.ifr_addr = *sa;
1401 				error = copyout(&ifr, ifrp, sizeof ifr);
1402 				ifrp++;
1403 			} else {
1404 				if (space < (sizeof ifr) + sa->sa_len -
1405 					    sizeof(*sa))
1406 					break;
1407 				space -= sa->sa_len - sizeof(*sa);
1408 				error = copyout(&ifr, ifrp,
1409 						sizeof ifr.ifr_name);
1410 				if (error == 0)
1411 					error = copyout(sa, &ifrp->ifr_addr,
1412 							sa->sa_len);
1413 				ifrp = (struct ifreq *)
1414 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1415 			}
1416 			if (error)
1417 				break;
1418 			space -= sizeof ifr;
1419 		}
1420 		if (error)
1421 			break;
1422 		if (!addrs) {
1423 			bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1424 			error = copyout(&ifr, ifrp, sizeof ifr);
1425 			if (error)
1426 				break;
1427 			space -= sizeof ifr;
1428 			ifrp++;
1429 		}
1430 	}
1431 	ifc->ifc_len -= space;
1432 	return (error);
1433 }
1434 
1435 /*
1436  * Just like if_promisc(), but for all-multicast-reception mode.
1437  */
1438 int
1439 if_allmulti(struct ifnet *ifp, int onswitch)
1440 {
1441 	int error = 0;
1442 	int s = splimp();
1443 	struct ifreq ifr;
1444 
1445 	if (onswitch) {
1446 		if (ifp->if_amcount++ == 0) {
1447 			ifp->if_flags |= IFF_ALLMULTI;
1448 			ifr.ifr_flags = ifp->if_flags;
1449 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1450 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1451 					      (struct ucred *)NULL);
1452 		}
1453 	} else {
1454 		if (ifp->if_amcount > 1) {
1455 			ifp->if_amcount--;
1456 		} else {
1457 			ifp->if_amcount = 0;
1458 			ifp->if_flags &= ~IFF_ALLMULTI;
1459 			ifr.ifr_flags = ifp->if_flags;
1460 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1461 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1462 					      (struct ucred *)NULL);
1463 		}
1464 	}
1465 	splx(s);
1466 
1467 	if (error == 0)
1468 		rt_ifmsg(ifp);
1469 	return error;
1470 }
1471 
1472 /*
1473  * Add a multicast listenership to the interface in question.
1474  * The link layer provides a routine which converts
1475  */
1476 int
1477 if_addmulti(
1478 	struct ifnet *ifp,	/* interface to manipulate */
1479 	struct sockaddr *sa,	/* address to add */
1480 	struct ifmultiaddr **retifma)
1481 {
1482 	struct sockaddr *llsa, *dupsa;
1483 	int error, s;
1484 	struct ifmultiaddr *ifma;
1485 
1486 	/*
1487 	 * If the matching multicast address already exists
1488 	 * then don't add a new one, just add a reference
1489 	 */
1490 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1491 		if (equal(sa, ifma->ifma_addr)) {
1492 			ifma->ifma_refcount++;
1493 			if (retifma)
1494 				*retifma = ifma;
1495 			return 0;
1496 		}
1497 	}
1498 
1499 	/*
1500 	 * Give the link layer a chance to accept/reject it, and also
1501 	 * find out which AF_LINK address this maps to, if it isn't one
1502 	 * already.
1503 	 */
1504 	if (ifp->if_resolvemulti) {
1505 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1506 		if (error) return error;
1507 	} else {
1508 		llsa = 0;
1509 	}
1510 
1511 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1512 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1513 	bcopy(sa, dupsa, sa->sa_len);
1514 
1515 	ifma->ifma_addr = dupsa;
1516 	ifma->ifma_lladdr = llsa;
1517 	ifma->ifma_ifp = ifp;
1518 	ifma->ifma_refcount = 1;
1519 	ifma->ifma_protospec = 0;
1520 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1521 
1522 	/*
1523 	 * Some network interfaces can scan the address list at
1524 	 * interrupt time; lock them out.
1525 	 */
1526 	s = splimp();
1527 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1528 	splx(s);
1529 	*retifma = ifma;
1530 
1531 	if (llsa != 0) {
1532 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1533 			if (equal(ifma->ifma_addr, llsa))
1534 				break;
1535 		}
1536 		if (ifma) {
1537 			ifma->ifma_refcount++;
1538 		} else {
1539 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1540 			       M_IFMADDR, M_WAITOK);
1541 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1542 			       M_IFMADDR, M_WAITOK);
1543 			bcopy(llsa, dupsa, llsa->sa_len);
1544 			ifma->ifma_addr = dupsa;
1545 			ifma->ifma_ifp = ifp;
1546 			ifma->ifma_refcount = 1;
1547 			s = splimp();
1548 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1549 			splx(s);
1550 		}
1551 	}
1552 	/*
1553 	 * We are certain we have added something, so call down to the
1554 	 * interface to let them know about it.
1555 	 */
1556 	s = splimp();
1557 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1558 	splx(s);
1559 
1560 	return 0;
1561 }
1562 
1563 /*
1564  * Remove a reference to a multicast address on this interface.  Yell
1565  * if the request does not match an existing membership.
1566  */
1567 int
1568 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1569 {
1570 	struct ifmultiaddr *ifma;
1571 	int s;
1572 
1573 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1574 		if (equal(sa, ifma->ifma_addr))
1575 			break;
1576 	if (ifma == 0)
1577 		return ENOENT;
1578 
1579 	if (ifma->ifma_refcount > 1) {
1580 		ifma->ifma_refcount--;
1581 		return 0;
1582 	}
1583 
1584 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1585 	sa = ifma->ifma_lladdr;
1586 	s = splimp();
1587 	LIST_REMOVE(ifma, ifma_link);
1588 	/*
1589 	 * Make sure the interface driver is notified
1590 	 * in the case of a link layer mcast group being left.
1591 	 */
1592 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1593 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1594 	splx(s);
1595 	free(ifma->ifma_addr, M_IFMADDR);
1596 	free(ifma, M_IFMADDR);
1597 	if (sa == 0)
1598 		return 0;
1599 
1600 	/*
1601 	 * Now look for the link-layer address which corresponds to
1602 	 * this network address.  It had been squirreled away in
1603 	 * ifma->ifma_lladdr for this purpose (so we don't have
1604 	 * to call ifp->if_resolvemulti() again), and we saved that
1605 	 * value in sa above.  If some nasty deleted the
1606 	 * link-layer address out from underneath us, we can deal because
1607 	 * the address we stored was is not the same as the one which was
1608 	 * in the record for the link-layer address.  (So we don't complain
1609 	 * in that case.)
1610 	 */
1611 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1612 		if (equal(sa, ifma->ifma_addr))
1613 			break;
1614 	if (ifma == 0)
1615 		return 0;
1616 
1617 	if (ifma->ifma_refcount > 1) {
1618 		ifma->ifma_refcount--;
1619 		return 0;
1620 	}
1621 
1622 	s = splimp();
1623 	LIST_REMOVE(ifma, ifma_link);
1624 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1625 	splx(s);
1626 	free(ifma->ifma_addr, M_IFMADDR);
1627 	free(sa, M_IFMADDR);
1628 	free(ifma, M_IFMADDR);
1629 
1630 	return 0;
1631 }
1632 
1633 /*
1634  * Set the link layer address on an interface.
1635  *
1636  * At this time we only support certain types of interfaces,
1637  * and we don't allow the length of the address to change.
1638  */
1639 int
1640 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1641 {
1642 	struct sockaddr_dl *sdl;
1643 	struct ifaddr *ifa;
1644 	struct ifreq ifr;
1645 
1646 	ifa = ifnet_addrs[ifp->if_index - 1];
1647 	if (ifa == NULL)
1648 		return (EINVAL);
1649 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1650 	if (sdl == NULL)
1651 		return (EINVAL);
1652 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1653 		return (EINVAL);
1654 	switch (ifp->if_type) {
1655 	case IFT_ETHER:			/* these types use struct arpcom */
1656 	case IFT_FDDI:
1657 	case IFT_XETHER:
1658 	case IFT_ISO88025:
1659 	case IFT_L2VLAN:
1660 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1661 		/* FALLTHROUGH */
1662 	case IFT_ARCNET:
1663 		bcopy(lladdr, LLADDR(sdl), len);
1664 		break;
1665 	default:
1666 		return (ENODEV);
1667 	}
1668 	/*
1669 	 * If the interface is already up, we need
1670 	 * to re-init it in order to reprogram its
1671 	 * address filter.
1672 	 */
1673 	if ((ifp->if_flags & IFF_UP) != 0) {
1674 		ifp->if_flags &= ~IFF_UP;
1675 		ifr.ifr_flags = ifp->if_flags;
1676 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1677 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1678 				 (struct ucred *)NULL);
1679 		ifp->if_flags |= IFF_UP;
1680 		ifr.ifr_flags = ifp->if_flags;
1681 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1682 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1683 				 (struct ucred *)NULL);
1684 #ifdef INET
1685 		/*
1686 		 * Also send gratuitous ARPs to notify other nodes about
1687 		 * the address change.
1688 		 */
1689 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1690 			if (ifa->ifa_addr != NULL &&
1691 			    ifa->ifa_addr->sa_family == AF_INET)
1692 				arp_ifinit(ifp, ifa);
1693 		}
1694 #endif
1695 	}
1696 	return (0);
1697 }
1698 
1699 struct ifmultiaddr *
1700 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1701 {
1702 	struct ifmultiaddr *ifma;
1703 
1704 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1705 		if (equal(ifma->ifma_addr, sa))
1706 			break;
1707 
1708 	return ifma;
1709 }
1710 
1711 /*
1712  * The name argument must be a pointer to storage which will last as
1713  * long as the interface does.  For physical devices, the result of
1714  * device_get_name(dev) is a good choice and for pseudo-devices a
1715  * static string works well.
1716  */
1717 void
1718 if_initname(struct ifnet *ifp, const char *name, int unit)
1719 {
1720 	ifp->if_dname = name;
1721 	ifp->if_dunit = unit;
1722 	if (unit != IF_DUNIT_NONE)
1723 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1724 	else
1725 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1726 }
1727 
1728 int
1729 if_printf(struct ifnet *ifp, const char *fmt, ...)
1730 {
1731 	__va_list ap;
1732 	int retval;
1733 
1734 	retval = printf("%s: ", ifp->if_xname);
1735 	__va_start(ap, fmt);
1736 	retval += vprintf(fmt, ap);
1737 	__va_end(ap);
1738 	return (retval);
1739 }
1740 
1741 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1742 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1743