xref: /dragonfly/sys/net/if.c (revision aa8d5dcb)
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 1.14 2004/03/15 22:37:40 hmp 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);
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 /*
117  * Network interface utility routines.
118  *
119  * Routines with ifa_ifwith* names take sockaddr *'s as
120  * parameters.
121  */
122 /* ARGSUSED*/
123 void
124 ifinit(dummy)
125 	void *dummy;
126 {
127 	struct ifnet *ifp;
128 	int s;
129 
130 	s = splimp();
131 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
132 		if (ifp->if_snd.ifq_maxlen == 0) {
133 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
134 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
135 		}
136 	}
137 	splx(s);
138 	if_slowtimo(0);
139 }
140 
141 int if_index = 0;
142 struct ifaddr **ifnet_addrs;
143 struct ifnet **ifindex2ifnet = NULL;
144 
145 
146 /*
147  * Attach an interface to the
148  * list of "active" interfaces.
149  */
150 void
151 if_attach(ifp)
152 	struct ifnet *ifp;
153 {
154 	unsigned socksize, ifasize;
155 	int namelen, masklen;
156 	struct sockaddr_dl *sdl;
157 	struct ifaddr *ifa;
158 	static int if_indexlim = 8;
159 	static int inited;
160 
161 	if (!inited) {
162 		TAILQ_INIT(&ifnet);
163 		inited = 1;
164 	}
165 
166 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
167 	ifp->if_index = ++if_index;
168 	/*
169 	 * XXX -
170 	 * The old code would work if the interface passed a pre-existing
171 	 * chain of ifaddrs to this code.  We don't trust our callers to
172 	 * properly initialize the tailq, however, so we no longer allow
173 	 * this unlikely case.
174 	 */
175 	TAILQ_INIT(&ifp->if_addrhead);
176 	TAILQ_INIT(&ifp->if_prefixhead);
177 	LIST_INIT(&ifp->if_multiaddrs);
178 	getmicrotime(&ifp->if_lastchange);
179 	if (ifnet_addrs == 0 || if_index >= if_indexlim) {
180 		unsigned n = (if_indexlim <<= 1) * sizeof(ifa);
181 		caddr_t q = malloc(n, M_IFADDR, M_WAITOK);
182 		bzero(q, n);
183 		if (ifnet_addrs) {
184 			bcopy((caddr_t)ifnet_addrs, (caddr_t)q, n/2);
185 			free((caddr_t)ifnet_addrs, M_IFADDR);
186 		}
187 		ifnet_addrs = (struct ifaddr **)q;
188 
189 		/* grow ifindex2ifnet */
190 		n = if_indexlim * sizeof(struct ifnet *);
191 		q = malloc(n, M_IFADDR, M_WAITOK);
192 		bzero(q, n);
193 		if (ifindex2ifnet) {
194 			bcopy((caddr_t)ifindex2ifnet, q, n/2);
195 			free((caddr_t)ifindex2ifnet, M_IFADDR);
196 		}
197 		ifindex2ifnet = (struct ifnet **)q;
198 	}
199 
200 	ifindex2ifnet[if_index] = ifp;
201 
202 	/*
203 	 * create a Link Level name for this device
204 	 */
205 	namelen = strlen(ifp->if_xname);
206 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
207 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
208 	socksize = masklen + ifp->if_addrlen;
209 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
210 	if (socksize < sizeof(*sdl))
211 		socksize = sizeof(*sdl);
212 	socksize = ROUNDUP(socksize);
213 	ifasize = sizeof(*ifa) + 2 * socksize;
214 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
215 	if (ifa) {
216 		bzero((caddr_t)ifa, ifasize);
217 		sdl = (struct sockaddr_dl *)(ifa + 1);
218 		sdl->sdl_len = socksize;
219 		sdl->sdl_family = AF_LINK;
220 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
221 		sdl->sdl_nlen = namelen;
222 		sdl->sdl_index = ifp->if_index;
223 		sdl->sdl_type = ifp->if_type;
224 		ifnet_addrs[if_index - 1] = ifa;
225 		ifa->ifa_ifp = ifp;
226 		ifa->ifa_rtrequest = link_rtrequest;
227 		ifa->ifa_addr = (struct sockaddr *)sdl;
228 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
229 		ifa->ifa_netmask = (struct sockaddr *)sdl;
230 		sdl->sdl_len = masklen;
231 		while (namelen != 0)
232 			sdl->sdl_data[--namelen] = 0xff;
233 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
234 	}
235 
236 	/* Announce the interface. */
237 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
238 }
239 
240 /*
241  * Detach an interface, removing it from the
242  * list of "active" interfaces.
243  */
244 void
245 if_detach(ifp)
246 	struct ifnet *ifp;
247 {
248 	struct ifaddr *ifa;
249 	struct radix_node_head	*rnh;
250 	int s;
251 	int i;
252 
253 	/*
254 	 * Remove routes and flush queues.
255 	 */
256 	s = splnet();
257 	if_down(ifp);
258 
259 	/*
260 	 * Remove address from ifnet_addrs[] and maybe decrement if_index.
261 	 * Clean up all addresses.
262 	 */
263 	ifnet_addrs[ifp->if_index - 1] = 0;
264 	while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
265 		if_index--;
266 
267 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
268 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
269 #ifdef INET
270 		/* XXX: Ugly!! ad hoc just for INET */
271 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
272 			struct ifaliasreq ifr;
273 
274 			bzero(&ifr, sizeof(ifr));
275 			ifr.ifra_addr = *ifa->ifa_addr;
276 			if (ifa->ifa_dstaddr)
277 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
278 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
279 			    NULL) == 0)
280 				continue;
281 		}
282 #endif /* INET */
283 #ifdef INET6
284 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
285 			in6_purgeaddr(ifa);
286 			/* ifp_addrhead is already updated */
287 			continue;
288 		}
289 #endif /* INET6 */
290 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
291 		IFAFREE(ifa);
292 	}
293 
294 #ifdef INET6
295 	/*
296 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
297 	 * before removing routing entries below, since IPv6 interface direct
298 	 * routes are expected to be removed by the IPv6-specific kernel API.
299 	 * Otherwise, the kernel will detect some inconsistency and bark it.
300 	 */
301 	in6_ifdetach(ifp);
302 #endif
303 
304 	/*
305 	 * Delete all remaining routes using this interface
306 	 * Unfortuneatly the only way to do this is to slog through
307 	 * the entire routing table looking for routes which point
308 	 * to this interface...oh well...
309 	 */
310 	for (i = 1; i <= AF_MAX; i++) {
311 		if ((rnh = rt_tables[i]) == NULL)
312 			continue;
313 		(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
314 	}
315 
316 	/* Announce that the interface is gone. */
317 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
318 
319 	TAILQ_REMOVE(&ifnet, ifp, if_link);
320 	splx(s);
321 }
322 
323 /*
324  * Delete Routes for a Network Interface
325  *
326  * Called for each routing entry via the rnh->rnh_walktree() call above
327  * to delete all route entries referencing a detaching network interface.
328  *
329  * Arguments:
330  *	rn	pointer to node in the routing table
331  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
332  *
333  * Returns:
334  *	0	successful
335  *	errno	failed - reason indicated
336  *
337  */
338 static int
339 if_rtdel(rn, arg)
340 	struct radix_node	*rn;
341 	void			*arg;
342 {
343 	struct rtentry	*rt = (struct rtentry *)rn;
344 	struct ifnet	*ifp = arg;
345 	int		err;
346 
347 	if (rt->rt_ifp == ifp) {
348 
349 		/*
350 		 * Protect (sorta) against walktree recursion problems
351 		 * with cloned routes
352 		 */
353 		if ((rt->rt_flags & RTF_UP) == 0)
354 			return (0);
355 
356 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
357 				rt_mask(rt), rt->rt_flags,
358 				(struct rtentry **) NULL);
359 		if (err) {
360 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
361 		}
362 	}
363 
364 	return (0);
365 }
366 
367 /*
368  * Create a clone network interface.
369  */
370 int
371 if_clone_create(name, len)
372 	char *name;
373 	int len;
374 {
375 	struct if_clone *ifc;
376 	char *dp;
377 	int wildcard, bytoff, bitoff;
378 	int unit;
379 	int err;
380 
381 	ifc = if_clone_lookup(name, &unit);
382 	if (ifc == NULL)
383 		return (EINVAL);
384 
385 	if (ifunit(name) != NULL)
386 		return (EEXIST);
387 
388 	bytoff = bitoff = 0;
389 	wildcard = (unit < 0);
390 	/*
391 	 * Find a free unit if none was given.
392 	 */
393 	if (wildcard) {
394 		while ((bytoff < ifc->ifc_bmlen)
395 		    && (ifc->ifc_units[bytoff] == 0xff))
396 			bytoff++;
397 		if (bytoff >= ifc->ifc_bmlen)
398 			return (ENOSPC);
399 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
400 			bitoff++;
401 		unit = (bytoff << 3) + bitoff;
402 	}
403 
404 	if (unit > ifc->ifc_maxunit)
405 		return (ENXIO);
406 
407 	err = (*ifc->ifc_create)(ifc, unit);
408 	if (err != 0)
409 		return (err);
410 
411 	if (!wildcard) {
412 		bytoff = unit >> 3;
413 		bitoff = unit - (bytoff << 3);
414 	}
415 
416 	/*
417 	 * Allocate the unit in the bitmap.
418 	 */
419 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
420 	    ("%s: bit is already set", __func__));
421 	ifc->ifc_units[bytoff] |= (1 << bitoff);
422 
423 	/* In the wildcard case, we need to update the name. */
424 	if (wildcard) {
425 		for (dp = name; *dp != '\0'; dp++);
426 		if (snprintf(dp, len - (dp-name), "%d", unit) >
427 		    len - (dp-name) - 1) {
428 			/*
429 			 * This can only be a programmer error and
430 			 * there's no straightforward way to recover if
431 			 * it happens.
432 			 */
433 			panic("if_clone_create(): interface name too long");
434 		}
435 
436 	}
437 
438 	return (0);
439 }
440 
441 /*
442  * Destroy a clone network interface.
443  */
444 int
445 if_clone_destroy(name)
446 	const char *name;
447 {
448 	struct if_clone *ifc;
449 	struct ifnet *ifp;
450 	int bytoff, bitoff;
451 	int unit;
452 
453 	ifc = if_clone_lookup(name, &unit);
454 	if (ifc == NULL)
455 		return (EINVAL);
456 
457 	if (unit < ifc->ifc_minifs)
458 		return (EINVAL);
459 
460 	ifp = ifunit(name);
461 	if (ifp == NULL)
462 		return (ENXIO);
463 
464 	if (ifc->ifc_destroy == NULL)
465 		return (EOPNOTSUPP);
466 
467 	(*ifc->ifc_destroy)(ifp);
468 
469 	/*
470 	 * Compute offset in the bitmap and deallocate the unit.
471 	 */
472 	bytoff = unit >> 3;
473 	bitoff = unit - (bytoff << 3);
474 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
475 	    ("%s: bit is already cleared", __func__));
476 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
477 	return (0);
478 }
479 
480 /*
481  * Look up a network interface cloner.
482  */
483 struct if_clone *
484 if_clone_lookup(name, unitp)
485 	const char *name;
486 	int *unitp;
487 {
488 	struct if_clone *ifc;
489 	const char *cp;
490 	int i;
491 
492 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
493 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
494 			if (ifc->ifc_name[i] != *cp)
495 				goto next_ifc;
496 		}
497 		goto found_name;
498  next_ifc:
499 		ifc = LIST_NEXT(ifc, ifc_list);
500 	}
501 
502 	/* No match. */
503 	return ((struct if_clone *)NULL);
504 
505  found_name:
506 	if (*cp == '\0') {
507 		i = -1;
508 	} else {
509 		for (i = 0; *cp != '\0'; cp++) {
510 			if (*cp < '0' || *cp > '9') {
511 				/* Bogus unit number. */
512 				return (NULL);
513 			}
514 			i = (i * 10) + (*cp - '0');
515 		}
516 	}
517 
518 	if (unitp != NULL)
519 		*unitp = i;
520 	return (ifc);
521 }
522 
523 /*
524  * Register a network interface cloner.
525  */
526 void
527 if_clone_attach(ifc)
528 	struct if_clone *ifc;
529 {
530 	int bytoff, bitoff;
531 	int err;
532 	int len, maxclone;
533 	int unit;
534 
535 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
536 	    ("%s: %s requested more units then allowed (%d > %d)",
537 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
538 	    ifc->ifc_maxunit + 1));
539 	/*
540 	 * Compute bitmap size and allocate it.
541 	 */
542 	maxclone = ifc->ifc_maxunit + 1;
543 	len = maxclone >> 3;
544 	if ((len << 3) < maxclone)
545 		len++;
546 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
547 	ifc->ifc_bmlen = len;
548 
549 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
550 	if_cloners_count++;
551 
552 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
553 		err = (*ifc->ifc_create)(ifc, unit);
554 		KASSERT(err == 0,
555 		    ("%s: failed to create required interface %s%d",
556 		    __func__, ifc->ifc_name, unit));
557 
558 		/* Allocate the unit in the bitmap. */
559 		bytoff = unit >> 3;
560 		bitoff = unit - (bytoff << 3);
561 		ifc->ifc_units[bytoff] |= (1 << bitoff);
562 	}
563 }
564 
565 /*
566  * Unregister a network interface cloner.
567  */
568 void
569 if_clone_detach(ifc)
570 	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(ifcr)
583 	struct if_clonereq *ifcr;
584 {
585 	char outbuf[IFNAMSIZ], *dst;
586 	struct if_clone *ifc;
587 	int count, error = 0;
588 
589 	ifcr->ifcr_total = if_cloners_count;
590 	if ((dst = ifcr->ifcr_buffer) == NULL) {
591 		/* Just asking how many there are. */
592 		return (0);
593 	}
594 
595 	if (ifcr->ifcr_count < 0)
596 		return (EINVAL);
597 
598 	count = (if_cloners_count < ifcr->ifcr_count) ?
599 	    if_cloners_count : ifcr->ifcr_count;
600 
601 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
602 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
603 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
604 		error = copyout(outbuf, dst, IFNAMSIZ);
605 		if (error)
606 			break;
607 	}
608 
609 	return (error);
610 }
611 
612 /*
613  * Locate an interface based on a complete address.
614  */
615 /*ARGSUSED*/
616 struct ifaddr *
617 ifa_ifwithaddr(addr)
618 	struct sockaddr *addr;
619 {
620 	struct ifnet *ifp;
621 	struct ifaddr *ifa;
622 
623 #define	equal(a1, a2) \
624   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
625 	TAILQ_FOREACH(ifp, &ifnet, if_link)
626 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
627 		if (ifa->ifa_addr->sa_family != addr->sa_family)
628 			continue;
629 		if (equal(addr, ifa->ifa_addr))
630 			return (ifa);
631 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
632 		    /* IP6 doesn't have broadcast */
633 		    ifa->ifa_broadaddr->sa_len != 0 &&
634 		    equal(ifa->ifa_broadaddr, addr))
635 			return (ifa);
636 	}
637 	return ((struct ifaddr *)0);
638 }
639 /*
640  * Locate the point to point interface with a given destination address.
641  */
642 /*ARGSUSED*/
643 struct ifaddr *
644 ifa_ifwithdstaddr(addr)
645 	struct sockaddr *addr;
646 {
647 	struct ifnet *ifp;
648 	struct ifaddr *ifa;
649 
650 	TAILQ_FOREACH(ifp, &ifnet, if_link)
651 	    if (ifp->if_flags & IFF_POINTOPOINT)
652 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
653 			if (ifa->ifa_addr->sa_family != addr->sa_family)
654 				continue;
655 			if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
656 				return (ifa);
657 	}
658 	return ((struct ifaddr *)0);
659 }
660 
661 /*
662  * Find an interface on a specific network.  If many, choice
663  * is most specific found.
664  */
665 struct ifaddr *
666 ifa_ifwithnet(addr)
667 	struct sockaddr *addr;
668 {
669 	struct ifnet *ifp;
670 	struct ifaddr *ifa;
671 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
672 	u_int af = addr->sa_family;
673 	char *addr_data = addr->sa_data, *cplim;
674 
675 	/*
676 	 * AF_LINK addresses can be looked up directly by their index number,
677 	 * so do that if we can.
678 	 */
679 	if (af == AF_LINK) {
680 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
681 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
682 		return (ifnet_addrs[sdl->sdl_index - 1]);
683 	}
684 
685 	/*
686 	 * Scan though each interface, looking for ones that have
687 	 * addresses in this address family.
688 	 */
689 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
690 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
691 			char *cp, *cp2, *cp3;
692 
693 			if (ifa->ifa_addr->sa_family != af)
694 next:				continue;
695 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
696 				/*
697 				 * This is a bit broken as it doesn't
698 				 * take into account that the remote end may
699 				 * be a single node in the network we are
700 				 * looking for.
701 				 * The trouble is that we don't know the
702 				 * netmask for the remote end.
703 				 */
704 				if (ifa->ifa_dstaddr != 0
705 				    && equal(addr, ifa->ifa_dstaddr))
706  					return (ifa);
707 			} else {
708 				/*
709 				 * if we have a special address handler,
710 				 * then use it instead of the generic one.
711 				 */
712 	          		if (ifa->ifa_claim_addr) {
713 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
714 						return (ifa);
715 					} else {
716 						continue;
717 					}
718 				}
719 
720 				/*
721 				 * Scan all the bits in the ifa's address.
722 				 * If a bit dissagrees with what we are
723 				 * looking for, mask it with the netmask
724 				 * to see if it really matters.
725 				 * (A byte at a time)
726 				 */
727 				if (ifa->ifa_netmask == 0)
728 					continue;
729 				cp = addr_data;
730 				cp2 = ifa->ifa_addr->sa_data;
731 				cp3 = ifa->ifa_netmask->sa_data;
732 				cplim = ifa->ifa_netmask->sa_len
733 					+ (char *)ifa->ifa_netmask;
734 				while (cp3 < cplim)
735 					if ((*cp++ ^ *cp2++) & *cp3++)
736 						goto next; /* next address! */
737 				/*
738 				 * If the netmask of what we just found
739 				 * is more specific than what we had before
740 				 * (if we had one) then remember the new one
741 				 * before continuing to search
742 				 * for an even better one.
743 				 */
744 				if (ifa_maybe == 0 ||
745 				    rn_refines((caddr_t)ifa->ifa_netmask,
746 				    (caddr_t)ifa_maybe->ifa_netmask))
747 					ifa_maybe = ifa;
748 			}
749 		}
750 	}
751 	return (ifa_maybe);
752 }
753 
754 /*
755  * Find an interface address specific to an interface best matching
756  * a given address.
757  */
758 struct ifaddr *
759 ifaof_ifpforaddr(addr, ifp)
760 	struct sockaddr *addr;
761 	struct ifnet *ifp;
762 {
763 	struct ifaddr *ifa;
764 	char *cp, *cp2, *cp3;
765 	char *cplim;
766 	struct ifaddr *ifa_maybe = 0;
767 	u_int af = addr->sa_family;
768 
769 	if (af >= AF_MAX)
770 		return (0);
771 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
772 		if (ifa->ifa_addr->sa_family != af)
773 			continue;
774 		if (ifa_maybe == 0)
775 			ifa_maybe = ifa;
776 		if (ifa->ifa_netmask == 0) {
777 			if (equal(addr, ifa->ifa_addr) ||
778 			    (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
779 				return (ifa);
780 			continue;
781 		}
782 		if (ifp->if_flags & IFF_POINTOPOINT) {
783 			if (equal(addr, ifa->ifa_dstaddr))
784 				return (ifa);
785 		} else {
786 			cp = addr->sa_data;
787 			cp2 = ifa->ifa_addr->sa_data;
788 			cp3 = ifa->ifa_netmask->sa_data;
789 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
790 			for (; cp3 < cplim; cp3++)
791 				if ((*cp++ ^ *cp2++) & *cp3)
792 					break;
793 			if (cp3 == cplim)
794 				return (ifa);
795 		}
796 	}
797 	return (ifa_maybe);
798 }
799 
800 #include <net/route.h>
801 
802 /*
803  * Default action when installing a route with a Link Level gateway.
804  * Lookup an appropriate real ifa to point to.
805  * This should be moved to /sys/net/link.c eventually.
806  */
807 static void
808 link_rtrequest(cmd, rt, info)
809 	int cmd;
810 	struct rtentry *rt;
811 	struct rt_addrinfo *info;
812 {
813 	struct ifaddr *ifa;
814 	struct sockaddr *dst;
815 	struct ifnet *ifp;
816 
817 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
818 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
819 		return;
820 	ifa = ifaof_ifpforaddr(dst, ifp);
821 	if (ifa) {
822 		IFAFREE(rt->rt_ifa);
823 		rt->rt_ifa = ifa;
824 		ifa->ifa_refcnt++;
825 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
826 			ifa->ifa_rtrequest(cmd, rt, info);
827 	}
828 }
829 
830 /*
831  * Mark an interface down and notify protocols of
832  * the transition.
833  * NOTE: must be called at splnet or eqivalent.
834  */
835 void
836 if_unroute(ifp, flag, fam)
837 	struct ifnet *ifp;
838 	int flag, fam;
839 {
840 	struct ifaddr *ifa;
841 
842 	ifp->if_flags &= ~flag;
843 	getmicrotime(&ifp->if_lastchange);
844 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
845 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
846 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
847 	if_qflush(&ifp->if_snd);
848 	rt_ifmsg(ifp);
849 }
850 
851 /*
852  * Mark an interface up and notify protocols of
853  * the transition.
854  * NOTE: must be called at splnet or eqivalent.
855  */
856 void
857 if_route(ifp, flag, fam)
858 	struct ifnet *ifp;
859 	int flag, fam;
860 {
861 	struct ifaddr *ifa;
862 
863 	ifp->if_flags |= flag;
864 	getmicrotime(&ifp->if_lastchange);
865 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
866 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
867 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
868 	rt_ifmsg(ifp);
869 #ifdef INET6
870 	in6_if_up(ifp);
871 #endif
872 }
873 
874 /*
875  * Mark an interface down and notify protocols of
876  * the transition.
877  * NOTE: must be called at splnet or eqivalent.
878  */
879 void
880 if_down(ifp)
881 	struct ifnet *ifp;
882 {
883 
884 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
885 }
886 
887 /*
888  * Mark an interface up and notify protocols of
889  * the transition.
890  * NOTE: must be called at splnet or eqivalent.
891  */
892 void
893 if_up(ifp)
894 	struct ifnet *ifp;
895 {
896 
897 	if_route(ifp, IFF_UP, AF_UNSPEC);
898 }
899 
900 /*
901  * Flush an interface queue.
902  */
903 static void
904 if_qflush(ifq)
905 	struct ifqueue *ifq;
906 {
907 	struct mbuf *m, *n;
908 
909 	n = ifq->ifq_head;
910 	while ((m = n) != 0) {
911 		n = m->m_act;
912 		m_freem(m);
913 	}
914 	ifq->ifq_head = 0;
915 	ifq->ifq_tail = 0;
916 	ifq->ifq_len = 0;
917 }
918 
919 /*
920  * Handle interface watchdog timer routines.  Called
921  * from softclock, we decrement timers (if set) and
922  * call the appropriate interface routine on expiration.
923  */
924 static void
925 if_slowtimo(arg)
926 	void *arg;
927 {
928 	struct ifnet *ifp;
929 	int s = splimp();
930 
931 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
932 		if (ifp->if_timer == 0 || --ifp->if_timer)
933 			continue;
934 		if (ifp->if_watchdog)
935 			(*ifp->if_watchdog)(ifp);
936 	}
937 	splx(s);
938 	timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
939 }
940 
941 /*
942  * Map interface name to
943  * interface structure pointer.
944  */
945 struct ifnet *
946 ifunit(const char *name)
947 {
948 	struct ifnet *ifp;
949 
950 	/*
951 	 * Search all the interfaces for this name/number
952 	 */
953 
954 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
955 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
956 			break;
957 	}
958 	return (ifp);
959 }
960 
961 
962 /*
963  * Map interface name in a sockaddr_dl to
964  * interface structure pointer.
965  */
966 struct ifnet *
967 if_withname(sa)
968 	struct sockaddr *sa;
969 {
970 	char ifname[IFNAMSIZ+1];
971 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
972 
973 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
974 	     (sdl->sdl_nlen > IFNAMSIZ) )
975 		return NULL;
976 
977 	/*
978 	 * ifunit wants a null-terminated name.  It may not be null-terminated
979 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
980 	 * and there might not be room to put the trailing null anyway, so we
981 	 * make a local copy that we know we can null terminate safely.
982 	 */
983 
984 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
985 	ifname[sdl->sdl_nlen] = '\0';
986 	return ifunit(ifname);
987 }
988 
989 
990 /*
991  * Interface ioctls.
992  */
993 int
994 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
995 {
996 	struct ifnet *ifp;
997 	struct ifreq *ifr;
998 	struct ifstat *ifs;
999 	int error;
1000 	short oif_flags;
1001 	int new_flags;
1002 	size_t namelen, onamelen;
1003 	char new_name[IFNAMSIZ];
1004 	struct ifaddr *ifa;
1005 	struct sockaddr_dl *sdl;
1006 
1007 	switch (cmd) {
1008 
1009 	case SIOCGIFCONF:
1010 	case OSIOCGIFCONF:
1011 		return (ifconf(cmd, data));
1012 	}
1013 	ifr = (struct ifreq *)data;
1014 
1015 	switch (cmd) {
1016 	case SIOCIFCREATE:
1017 	case SIOCIFDESTROY:
1018 		if ((error = suser(td)) != 0)
1019 			return (error);
1020 		return ((cmd == SIOCIFCREATE) ?
1021 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1022 			if_clone_destroy(ifr->ifr_name));
1023 
1024 	case SIOCIFGCLONERS:
1025 		return (if_clone_list((struct if_clonereq *)data));
1026 	}
1027 
1028 	ifp = ifunit(ifr->ifr_name);
1029 	if (ifp == 0)
1030 		return (ENXIO);
1031 	switch (cmd) {
1032 
1033 	case SIOCGIFFLAGS:
1034 		ifr->ifr_flags = ifp->if_flags;
1035 		ifr->ifr_flagshigh = ifp->if_ipending >> 16;
1036 		break;
1037 
1038 	case SIOCGIFCAP:
1039 		ifr->ifr_reqcap = ifp->if_capabilities;
1040 		ifr->ifr_curcap = ifp->if_capenable;
1041 		break;
1042 
1043 	case SIOCGIFMETRIC:
1044 		ifr->ifr_metric = ifp->if_metric;
1045 		break;
1046 
1047 	case SIOCGIFMTU:
1048 		ifr->ifr_mtu = ifp->if_mtu;
1049 		break;
1050 
1051 	case SIOCGIFPHYS:
1052 		ifr->ifr_phys = ifp->if_physical;
1053 		break;
1054 
1055 	case SIOCSIFFLAGS:
1056 		error = suser(td);
1057 		if (error)
1058 			return (error);
1059 		new_flags = (ifr->ifr_flags & 0xffff) |
1060 		    (ifr->ifr_flagshigh << 16);
1061 		if (ifp->if_flags & IFF_SMART) {
1062 			/* Smart drivers twiddle their own routes */
1063 		} else if (ifp->if_flags & IFF_UP &&
1064 		    (new_flags & IFF_UP) == 0) {
1065 			int s = splimp();
1066 			if_down(ifp);
1067 			splx(s);
1068 		} else if (new_flags & IFF_UP &&
1069 		    (ifp->if_flags & IFF_UP) == 0) {
1070 			int s = splimp();
1071 			if_up(ifp);
1072 			splx(s);
1073 		}
1074 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1075 			(new_flags &~ IFF_CANTCHANGE);
1076 		ifp->if_ipending = (ifp->if_ipending & IFF_CANTCHANGE) |
1077 			(new_flags &~ IFF_CANTCHANGE);
1078 		if (new_flags & IFF_PPROMISC) {
1079 			/* Permanently promiscuous mode requested */
1080 			ifp->if_flags |= IFF_PROMISC;
1081 		} else if (ifp->if_pcount == 0) {
1082 			ifp->if_flags &= ~IFF_PROMISC;
1083 		}
1084 		if (ifp->if_ioctl)
1085 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
1086 		getmicrotime(&ifp->if_lastchange);
1087 		break;
1088 
1089 	case SIOCSIFCAP:
1090 		error = suser(td);
1091 		if (error)
1092 			return (error);
1093 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1094 			return (EINVAL);
1095 		(void) (*ifp->if_ioctl)(ifp, cmd, data);
1096 		break;
1097 
1098  	case SIOCSIFNAME:
1099  		error = suser(td);
1100  		if (error != 0)
1101  			return (error);
1102  		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1103  		if (error != 0)
1104  			return (error);
1105 		if (new_name[0] == '\0')
1106 			return (EINVAL);
1107  		if (ifunit(new_name) != NULL)
1108  			return (EEXIST);
1109 
1110  		/* Announce the departure of the interface. */
1111  		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1112 
1113  		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1114  		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1115  		/* XXX IFA_LOCK(ifa); */
1116  		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1117  		namelen = strlen(new_name);
1118  		onamelen = sdl->sdl_nlen;
1119  		/*
1120  		 * Move the address if needed.  This is safe because we
1121  		 * allocate space for a name of length IFNAMSIZ when we
1122  		 * create this in if_attach().
1123  		 */
1124  		if (namelen != onamelen) {
1125  			bcopy(sdl->sdl_data + onamelen,
1126  			    sdl->sdl_data + namelen, sdl->sdl_alen);
1127  		}
1128  		bcopy(new_name, sdl->sdl_data, namelen);
1129  		sdl->sdl_nlen = namelen;
1130  		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1131  		bzero(sdl->sdl_data, onamelen);
1132  		while (namelen != 0)
1133  			sdl->sdl_data[--namelen] = 0xff;
1134  		/* XXX IFA_UNLOCK(ifa) */
1135 
1136  		/* Announce the return of the interface. */
1137  		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1138  		break;
1139 
1140 	case SIOCSIFMETRIC:
1141 		error = suser(td);
1142 		if (error)
1143 			return (error);
1144 		ifp->if_metric = ifr->ifr_metric;
1145 		getmicrotime(&ifp->if_lastchange);
1146 		break;
1147 
1148 	case SIOCSIFPHYS:
1149 		error = suser(td);
1150 		if (error)
1151 			return error;
1152 		if (!ifp->if_ioctl)
1153 		        return EOPNOTSUPP;
1154 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1155 		if (error == 0)
1156 			getmicrotime(&ifp->if_lastchange);
1157 		return(error);
1158 
1159 	case SIOCSIFMTU:
1160 	{
1161 		u_long oldmtu = ifp->if_mtu;
1162 
1163 		error = suser(td);
1164 		if (error)
1165 			return (error);
1166 		if (ifp->if_ioctl == NULL)
1167 			return (EOPNOTSUPP);
1168 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1169 			return (EINVAL);
1170 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1171 		if (error == 0) {
1172 			getmicrotime(&ifp->if_lastchange);
1173 			rt_ifmsg(ifp);
1174 		}
1175 		/*
1176 		 * If the link MTU changed, do network layer specific procedure.
1177 		 */
1178 		if (ifp->if_mtu != oldmtu) {
1179 #ifdef INET6
1180 			nd6_setmtu(ifp);
1181 #endif
1182 		}
1183 		return (error);
1184 	}
1185 
1186 	case SIOCADDMULTI:
1187 	case SIOCDELMULTI:
1188 		error = suser(td);
1189 		if (error)
1190 			return (error);
1191 
1192 		/* Don't allow group membership on non-multicast interfaces. */
1193 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1194 			return EOPNOTSUPP;
1195 
1196 		/* Don't let users screw up protocols' entries. */
1197 		if (ifr->ifr_addr.sa_family != AF_LINK)
1198 			return EINVAL;
1199 
1200 		if (cmd == SIOCADDMULTI) {
1201 			struct ifmultiaddr *ifma;
1202 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1203 		} else {
1204 			error = if_delmulti(ifp, &ifr->ifr_addr);
1205 		}
1206 		if (error == 0)
1207 			getmicrotime(&ifp->if_lastchange);
1208 		return error;
1209 
1210 	case SIOCSIFPHYADDR:
1211 	case SIOCDIFPHYADDR:
1212 #ifdef INET6
1213 	case SIOCSIFPHYADDR_IN6:
1214 #endif
1215 	case SIOCSLIFPHYADDR:
1216         case SIOCSIFMEDIA:
1217 	case SIOCSIFGENERIC:
1218 		error = suser(td);
1219 		if (error)
1220 			return (error);
1221 		if (ifp->if_ioctl == 0)
1222 			return (EOPNOTSUPP);
1223 		error = (*ifp->if_ioctl)(ifp, cmd, data);
1224 		if (error == 0)
1225 			getmicrotime(&ifp->if_lastchange);
1226 		return error;
1227 
1228 	case SIOCGIFSTATUS:
1229 		ifs = (struct ifstat *)data;
1230 		ifs->ascii[0] = '\0';
1231 
1232 	case SIOCGIFPSRCADDR:
1233 	case SIOCGIFPDSTADDR:
1234 	case SIOCGLIFPHYADDR:
1235 	case SIOCGIFMEDIA:
1236 	case SIOCGIFGENERIC:
1237 		if (ifp->if_ioctl == 0)
1238 			return (EOPNOTSUPP);
1239 		return ((*ifp->if_ioctl)(ifp, cmd, data));
1240 
1241 	case SIOCSIFLLADDR:
1242 		error = suser(td);
1243 		if (error)
1244 			return (error);
1245 		return if_setlladdr(ifp,
1246 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1247 
1248 	default:
1249 		oif_flags = ifp->if_flags;
1250 		if (so->so_proto == 0)
1251 			return (EOPNOTSUPP);
1252 #ifndef COMPAT_43
1253 		error = so_pru_control(so, cmd, data, ifp, td);
1254 #else
1255 	    {
1256 		int ocmd = cmd;
1257 
1258 		switch (cmd) {
1259 
1260 		case SIOCSIFDSTADDR:
1261 		case SIOCSIFADDR:
1262 		case SIOCSIFBRDADDR:
1263 		case SIOCSIFNETMASK:
1264 #if BYTE_ORDER != BIG_ENDIAN
1265 			if (ifr->ifr_addr.sa_family == 0 &&
1266 			    ifr->ifr_addr.sa_len < 16) {
1267 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1268 				ifr->ifr_addr.sa_len = 16;
1269 			}
1270 #else
1271 			if (ifr->ifr_addr.sa_len == 0)
1272 				ifr->ifr_addr.sa_len = 16;
1273 #endif
1274 			break;
1275 
1276 		case OSIOCGIFADDR:
1277 			cmd = SIOCGIFADDR;
1278 			break;
1279 
1280 		case OSIOCGIFDSTADDR:
1281 			cmd = SIOCGIFDSTADDR;
1282 			break;
1283 
1284 		case OSIOCGIFBRDADDR:
1285 			cmd = SIOCGIFBRDADDR;
1286 			break;
1287 
1288 		case OSIOCGIFNETMASK:
1289 			cmd = SIOCGIFNETMASK;
1290 		}
1291 		error =  so_pru_control(so, cmd, data, ifp, td);
1292 		switch (ocmd) {
1293 
1294 		case OSIOCGIFADDR:
1295 		case OSIOCGIFDSTADDR:
1296 		case OSIOCGIFBRDADDR:
1297 		case OSIOCGIFNETMASK:
1298 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1299 
1300 		}
1301 	    }
1302 #endif /* COMPAT_43 */
1303 
1304 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1305 #ifdef INET6
1306 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1307 			if (ifp->if_flags & IFF_UP) {
1308 				int s = splimp();
1309 				in6_if_up(ifp);
1310 				splx(s);
1311 			}
1312 #endif
1313 		}
1314 		return (error);
1315 
1316 	}
1317 	return (0);
1318 }
1319 
1320 /*
1321  * Set/clear promiscuous mode on interface ifp based on the truth value
1322  * of pswitch.  The calls are reference counted so that only the first
1323  * "on" request actually has an effect, as does the final "off" request.
1324  * Results are undefined if the "off" and "on" requests are not matched.
1325  */
1326 int
1327 ifpromisc(ifp, pswitch)
1328 	struct ifnet *ifp;
1329 	int pswitch;
1330 {
1331 	struct ifreq ifr;
1332 	int error;
1333 	int oldflags;
1334 
1335 	oldflags = ifp->if_flags;
1336 	if (ifp->if_ipending & IFF_PPROMISC) {
1337 		/* Do nothing if device is in permanently promiscuous mode */
1338 		ifp->if_pcount += pswitch ? 1 : -1;
1339 		return (0);
1340 	}
1341 	if (pswitch) {
1342 		/*
1343 		 * If the device is not configured up, we cannot put it in
1344 		 * promiscuous mode.
1345 		 */
1346 		if ((ifp->if_flags & IFF_UP) == 0)
1347 			return (ENETDOWN);
1348 		if (ifp->if_pcount++ != 0)
1349 			return (0);
1350 		ifp->if_flags |= IFF_PROMISC;
1351 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1352 		    ifp->if_xname);
1353 	} else {
1354 		if (--ifp->if_pcount > 0)
1355 			return (0);
1356 		ifp->if_flags &= ~IFF_PROMISC;
1357 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1358 		    ifp->if_xname);
1359 	}
1360 	ifr.ifr_flags = ifp->if_flags;
1361 	ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1362 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1363 	if (error == 0)
1364 		rt_ifmsg(ifp);
1365 	else
1366 		ifp->if_flags = oldflags;
1367 	return error;
1368 }
1369 
1370 /*
1371  * Return interface configuration
1372  * of system.  List may be used
1373  * in later ioctl's (above) to get
1374  * other information.
1375  */
1376 /*ARGSUSED*/
1377 static int
1378 ifconf(u_long cmd, caddr_t data)
1379 {
1380 	struct ifconf *ifc = (struct ifconf *)data;
1381 	struct ifnet *ifp;
1382 	struct ifaddr *ifa;
1383 	struct sockaddr *sa;
1384 	struct ifreq ifr, *ifrp;
1385 	int space = ifc->ifc_len, error = 0;
1386 
1387 	ifrp = ifc->ifc_req;
1388 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1389 		int addrs;
1390 
1391 		if (space <= sizeof (ifr))
1392 			break;
1393 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1394 		    >= sizeof(ifr.ifr_name)) {
1395 			error = ENAMETOOLONG;
1396 			break;
1397 		}
1398 
1399 		addrs = 0;
1400 		ifa = ifp->if_addrhead.tqh_first;
1401 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1402 			if (space <= sizeof(ifr))
1403 				break;
1404 			sa = ifa->ifa_addr;
1405 			if (curproc->p_ucred->cr_prison && prison_if(curthread, sa))
1406 				continue;
1407 			addrs++;
1408 #ifdef COMPAT_43
1409 			if (cmd == OSIOCGIFCONF) {
1410 				struct osockaddr *osa =
1411 					 (struct osockaddr *)&ifr.ifr_addr;
1412 				ifr.ifr_addr = *sa;
1413 				osa->sa_family = sa->sa_family;
1414 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1415 						sizeof (ifr));
1416 				ifrp++;
1417 			} else
1418 #endif
1419 			if (sa->sa_len <= sizeof(*sa)) {
1420 				ifr.ifr_addr = *sa;
1421 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1422 						sizeof (ifr));
1423 				ifrp++;
1424 			} else {
1425 				if (space < sizeof (ifr) + sa->sa_len -
1426 					    sizeof(*sa))
1427 					break;
1428 				space -= sa->sa_len - sizeof(*sa);
1429 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1430 						sizeof (ifr.ifr_name));
1431 				if (error == 0)
1432 				    error = copyout((caddr_t)sa,
1433 				      (caddr_t)&ifrp->ifr_addr, sa->sa_len);
1434 				ifrp = (struct ifreq *)
1435 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1436 			}
1437 			if (error)
1438 				break;
1439 			space -= sizeof (ifr);
1440 		}
1441 		if (error)
1442 			break;
1443 		if (!addrs) {
1444 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1445 			error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1446 			    sizeof (ifr));
1447 			if (error)
1448 				break;
1449 			space -= sizeof (ifr);
1450 			ifrp++;
1451 		}
1452 	}
1453 	ifc->ifc_len -= space;
1454 	return (error);
1455 }
1456 
1457 /*
1458  * Just like if_promisc(), but for all-multicast-reception mode.
1459  */
1460 int
1461 if_allmulti(ifp, onswitch)
1462 	struct ifnet *ifp;
1463 	int onswitch;
1464 {
1465 	int error = 0;
1466 	int s = splimp();
1467 	struct ifreq ifr;
1468 
1469 	if (onswitch) {
1470 		if (ifp->if_amcount++ == 0) {
1471 			ifp->if_flags |= IFF_ALLMULTI;
1472 			ifr.ifr_flags = ifp->if_flags;
1473 			ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1474 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1475 		}
1476 	} else {
1477 		if (ifp->if_amcount > 1) {
1478 			ifp->if_amcount--;
1479 		} else {
1480 			ifp->if_amcount = 0;
1481 			ifp->if_flags &= ~IFF_ALLMULTI;
1482 			ifr.ifr_flags = ifp->if_flags;
1483 			ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1484 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1485 		}
1486 	}
1487 	splx(s);
1488 
1489 	if (error == 0)
1490 		rt_ifmsg(ifp);
1491 	return error;
1492 }
1493 
1494 /*
1495  * Add a multicast listenership to the interface in question.
1496  * The link layer provides a routine which converts
1497  */
1498 int
1499 if_addmulti(ifp, sa, retifma)
1500 	struct ifnet *ifp;	/* interface to manipulate */
1501 	struct sockaddr *sa;	/* address to add */
1502 	struct ifmultiaddr **retifma;
1503 {
1504 	struct sockaddr *llsa, *dupsa;
1505 	int error, s;
1506 	struct ifmultiaddr *ifma;
1507 
1508 	/*
1509 	 * If the matching multicast address already exists
1510 	 * then don't add a new one, just add a reference
1511 	 */
1512 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1513 		if (equal(sa, ifma->ifma_addr)) {
1514 			ifma->ifma_refcount++;
1515 			if (retifma)
1516 				*retifma = ifma;
1517 			return 0;
1518 		}
1519 	}
1520 
1521 	/*
1522 	 * Give the link layer a chance to accept/reject it, and also
1523 	 * find out which AF_LINK address this maps to, if it isn't one
1524 	 * already.
1525 	 */
1526 	if (ifp->if_resolvemulti) {
1527 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1528 		if (error) return error;
1529 	} else {
1530 		llsa = 0;
1531 	}
1532 
1533 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1534 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1535 	bcopy(sa, dupsa, sa->sa_len);
1536 
1537 	ifma->ifma_addr = dupsa;
1538 	ifma->ifma_lladdr = llsa;
1539 	ifma->ifma_ifp = ifp;
1540 	ifma->ifma_refcount = 1;
1541 	ifma->ifma_protospec = 0;
1542 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1543 
1544 	/*
1545 	 * Some network interfaces can scan the address list at
1546 	 * interrupt time; lock them out.
1547 	 */
1548 	s = splimp();
1549 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1550 	splx(s);
1551 	*retifma = ifma;
1552 
1553 	if (llsa != 0) {
1554 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1555 			if (equal(ifma->ifma_addr, llsa))
1556 				break;
1557 		}
1558 		if (ifma) {
1559 			ifma->ifma_refcount++;
1560 		} else {
1561 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1562 			       M_IFMADDR, M_WAITOK);
1563 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1564 			       M_IFMADDR, M_WAITOK);
1565 			bcopy(llsa, dupsa, llsa->sa_len);
1566 			ifma->ifma_addr = dupsa;
1567 			ifma->ifma_ifp = ifp;
1568 			ifma->ifma_refcount = 1;
1569 			s = splimp();
1570 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1571 			splx(s);
1572 		}
1573 	}
1574 	/*
1575 	 * We are certain we have added something, so call down to the
1576 	 * interface to let them know about it.
1577 	 */
1578 	s = splimp();
1579 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0);
1580 	splx(s);
1581 
1582 	return 0;
1583 }
1584 
1585 /*
1586  * Remove a reference to a multicast address on this interface.  Yell
1587  * if the request does not match an existing membership.
1588  */
1589 int
1590 if_delmulti(ifp, sa)
1591 	struct ifnet *ifp;
1592 	struct sockaddr *sa;
1593 {
1594 	struct ifmultiaddr *ifma;
1595 	int s;
1596 
1597 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1598 		if (equal(sa, ifma->ifma_addr))
1599 			break;
1600 	if (ifma == 0)
1601 		return ENOENT;
1602 
1603 	if (ifma->ifma_refcount > 1) {
1604 		ifma->ifma_refcount--;
1605 		return 0;
1606 	}
1607 
1608 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1609 	sa = ifma->ifma_lladdr;
1610 	s = splimp();
1611 	LIST_REMOVE(ifma, ifma_link);
1612 	/*
1613 	 * Make sure the interface driver is notified
1614 	 * in the case of a link layer mcast group being left.
1615 	 */
1616 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1617 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1618 	splx(s);
1619 	free(ifma->ifma_addr, M_IFMADDR);
1620 	free(ifma, M_IFMADDR);
1621 	if (sa == 0)
1622 		return 0;
1623 
1624 	/*
1625 	 * Now look for the link-layer address which corresponds to
1626 	 * this network address.  It had been squirreled away in
1627 	 * ifma->ifma_lladdr for this purpose (so we don't have
1628 	 * to call ifp->if_resolvemulti() again), and we saved that
1629 	 * value in sa above.  If some nasty deleted the
1630 	 * link-layer address out from underneath us, we can deal because
1631 	 * the address we stored was is not the same as the one which was
1632 	 * in the record for the link-layer address.  (So we don't complain
1633 	 * in that case.)
1634 	 */
1635 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1636 		if (equal(sa, ifma->ifma_addr))
1637 			break;
1638 	if (ifma == 0)
1639 		return 0;
1640 
1641 	if (ifma->ifma_refcount > 1) {
1642 		ifma->ifma_refcount--;
1643 		return 0;
1644 	}
1645 
1646 	s = splimp();
1647 	LIST_REMOVE(ifma, ifma_link);
1648 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0);
1649 	splx(s);
1650 	free(ifma->ifma_addr, M_IFMADDR);
1651 	free(sa, M_IFMADDR);
1652 	free(ifma, M_IFMADDR);
1653 
1654 	return 0;
1655 }
1656 
1657 /*
1658  * Set the link layer address on an interface.
1659  *
1660  * At this time we only support certain types of interfaces,
1661  * and we don't allow the length of the address to change.
1662  */
1663 int
1664 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1665 {
1666 	struct sockaddr_dl *sdl;
1667 	struct ifaddr *ifa;
1668 	struct ifreq ifr;
1669 
1670 	ifa = ifnet_addrs[ifp->if_index - 1];
1671 	if (ifa == NULL)
1672 		return (EINVAL);
1673 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1674 	if (sdl == NULL)
1675 		return (EINVAL);
1676 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1677 		return (EINVAL);
1678 	switch (ifp->if_type) {
1679 	case IFT_ETHER:			/* these types use struct arpcom */
1680 	case IFT_FDDI:
1681 	case IFT_XETHER:
1682 	case IFT_ISO88025:
1683 	case IFT_L2VLAN:
1684 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1685 		/* FALLTHROUGH */
1686 	case IFT_ARCNET:
1687 		bcopy(lladdr, LLADDR(sdl), len);
1688 		break;
1689 	default:
1690 		return (ENODEV);
1691 	}
1692 	/*
1693 	 * If the interface is already up, we need
1694 	 * to re-init it in order to reprogram its
1695 	 * address filter.
1696 	 */
1697 	if ((ifp->if_flags & IFF_UP) != 0) {
1698 		ifp->if_flags &= ~IFF_UP;
1699 		ifr.ifr_flags = ifp->if_flags;
1700 		ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1701 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1702 		ifp->if_flags |= IFF_UP;
1703 		ifr.ifr_flags = ifp->if_flags;
1704 		ifr.ifr_flagshigh = ifp->if_ipending >> 16;
1705 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
1706 #ifdef INET
1707 		/*
1708 		 * Also send gratuitous ARPs to notify other nodes about
1709 		 * the address change.
1710 		 */
1711 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1712 			if (ifa->ifa_addr != NULL &&
1713 			    ifa->ifa_addr->sa_family == AF_INET)
1714 				arp_ifinit(ifp, ifa);
1715 		}
1716 #endif
1717 	}
1718 	return (0);
1719 }
1720 
1721 struct ifmultiaddr *
1722 ifmaof_ifpforaddr(sa, ifp)
1723 	struct sockaddr *sa;
1724 	struct ifnet *ifp;
1725 {
1726 	struct ifmultiaddr *ifma;
1727 
1728 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1729 		if (equal(ifma->ifma_addr, sa))
1730 			break;
1731 
1732 	return ifma;
1733 }
1734 
1735 /*
1736  * The name argument must be a pointer to storage which will last as
1737  * long as the interface does.  For physical devices, the result of
1738  * device_get_name(dev) is a good choice and for pseudo-devices a
1739  * static string works well.
1740  */
1741 void
1742 if_initname(struct ifnet *ifp, const char *name, int unit)
1743 {
1744 	ifp->if_dname = name;
1745 	ifp->if_dunit = unit;
1746 	if (unit != IF_DUNIT_NONE)
1747 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1748 	else
1749 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1750 }
1751 
1752 int
1753 if_printf(struct ifnet *ifp, const char *fmt, ...)
1754 {
1755 	__va_list ap;
1756 	int retval;
1757 
1758 	retval = printf("%s: ", ifp->if_xname);
1759 	__va_start(ap, fmt);
1760 	retval += vprintf(fmt, ap);
1761 	__va_end(ap);
1762 	return (retval);
1763 }
1764 
1765 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1766 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1767