xref: /dragonfly/sys/net/if.c (revision 71126e33)
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 1.22 2004/09/15 19:34:55 joerg 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(dummy)
127 	void *dummy;
128 {
129 	struct ifnet *ifp;
130 	int s;
131 
132 	callout_init(&if_slowtimo_timer);
133 
134 	s = splimp();
135 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
136 		if (ifp->if_snd.ifq_maxlen == 0) {
137 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
138 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
139 		}
140 	}
141 	splx(s);
142 
143 	if_slowtimo(0);
144 }
145 
146 int if_index = 0;
147 struct ifaddr **ifnet_addrs;
148 struct ifnet **ifindex2ifnet = NULL;
149 
150 
151 /*
152  * Attach an interface to the
153  * list of "active" interfaces.
154  */
155 void
156 if_attach(ifp)
157 	struct ifnet *ifp;
158 {
159 	unsigned socksize, ifasize;
160 	int namelen, masklen;
161 	struct sockaddr_dl *sdl;
162 	struct ifaddr *ifa;
163 	static int if_indexlim = 8;
164 	static int inited;
165 
166 	if (!inited) {
167 		TAILQ_INIT(&ifnet);
168 		inited = 1;
169 	}
170 
171 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
172 	ifp->if_index = ++if_index;
173 	/*
174 	 * XXX -
175 	 * The old code would work if the interface passed a pre-existing
176 	 * chain of ifaddrs to this code.  We don't trust our callers to
177 	 * properly initialize the tailq, however, so we no longer allow
178 	 * this unlikely case.
179 	 */
180 	TAILQ_INIT(&ifp->if_addrhead);
181 	TAILQ_INIT(&ifp->if_prefixhead);
182 	LIST_INIT(&ifp->if_multiaddrs);
183 	getmicrotime(&ifp->if_lastchange);
184 	if (ifnet_addrs == 0 || if_index >= if_indexlim) {
185 		unsigned n = (if_indexlim <<= 1) * sizeof(ifa);
186 		caddr_t q = malloc(n, M_IFADDR, M_WAITOK);
187 		bzero(q, n);
188 		if (ifnet_addrs) {
189 			bcopy((caddr_t)ifnet_addrs, (caddr_t)q, n/2);
190 			free((caddr_t)ifnet_addrs, M_IFADDR);
191 		}
192 		ifnet_addrs = (struct ifaddr **)q;
193 
194 		/* grow ifindex2ifnet */
195 		n = if_indexlim * sizeof(struct ifnet *);
196 		q = malloc(n, M_IFADDR, M_WAITOK);
197 		bzero(q, n);
198 		if (ifindex2ifnet) {
199 			bcopy((caddr_t)ifindex2ifnet, q, n/2);
200 			free((caddr_t)ifindex2ifnet, M_IFADDR);
201 		}
202 		ifindex2ifnet = (struct ifnet **)q;
203 	}
204 
205 	ifindex2ifnet[if_index] = ifp;
206 
207 	/*
208 	 * create a Link Level name for this device
209 	 */
210 	namelen = strlen(ifp->if_xname);
211 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
212 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
213 	socksize = masklen + ifp->if_addrlen;
214 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
215 	if (socksize < sizeof(*sdl))
216 		socksize = sizeof(*sdl);
217 	socksize = ROUNDUP(socksize);
218 	ifasize = sizeof(*ifa) + 2 * socksize;
219 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
220 	if (ifa) {
221 		bzero((caddr_t)ifa, ifasize);
222 		sdl = (struct sockaddr_dl *)(ifa + 1);
223 		sdl->sdl_len = socksize;
224 		sdl->sdl_family = AF_LINK;
225 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
226 		sdl->sdl_nlen = namelen;
227 		sdl->sdl_index = ifp->if_index;
228 		sdl->sdl_type = ifp->if_type;
229 		ifnet_addrs[if_index - 1] = ifa;
230 		ifa->ifa_ifp = ifp;
231 		ifa->ifa_rtrequest = link_rtrequest;
232 		ifa->ifa_addr = (struct sockaddr *)sdl;
233 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
234 		ifa->ifa_netmask = (struct sockaddr *)sdl;
235 		sdl->sdl_len = masklen;
236 		while (namelen != 0)
237 			sdl->sdl_data[--namelen] = 0xff;
238 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
239 	}
240 
241 	EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
242 
243 	/* Announce the interface. */
244 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
245 }
246 
247 /*
248  * Detach an interface, removing it from the
249  * list of "active" interfaces.
250  */
251 void
252 if_detach(ifp)
253 	struct ifnet *ifp;
254 {
255 	struct ifaddr *ifa;
256 	struct radix_node_head	*rnh;
257 	int s;
258 	int i;
259 
260 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
261 
262 	/*
263 	 * Remove routes and flush queues.
264 	 */
265 	s = splnet();
266 	if_down(ifp);
267 
268 	/*
269 	 * Remove address from ifnet_addrs[] and maybe decrement if_index.
270 	 * Clean up all addresses.
271 	 */
272 	ifnet_addrs[ifp->if_index - 1] = 0;
273 	while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
274 		if_index--;
275 
276 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
277 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
278 #ifdef INET
279 		/* XXX: Ugly!! ad hoc just for INET */
280 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
281 			struct ifaliasreq ifr;
282 
283 			bzero(&ifr, sizeof(ifr));
284 			ifr.ifra_addr = *ifa->ifa_addr;
285 			if (ifa->ifa_dstaddr)
286 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
287 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
288 			    NULL) == 0)
289 				continue;
290 		}
291 #endif /* INET */
292 #ifdef INET6
293 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
294 			in6_purgeaddr(ifa);
295 			/* ifp_addrhead is already updated */
296 			continue;
297 		}
298 #endif /* INET6 */
299 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
300 		IFAFREE(ifa);
301 	}
302 
303 #ifdef INET6
304 	/*
305 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
306 	 * before removing routing entries below, since IPv6 interface direct
307 	 * routes are expected to be removed by the IPv6-specific kernel API.
308 	 * Otherwise, the kernel will detect some inconsistency and bark it.
309 	 */
310 	in6_ifdetach(ifp);
311 #endif
312 
313 	/*
314 	 * Delete all remaining routes using this interface
315 	 * Unfortuneatly the only way to do this is to slog through
316 	 * the entire routing table looking for routes which point
317 	 * to this interface...oh well...
318 	 */
319 	for (i = 1; i <= AF_MAX; i++) {
320 		if ((rnh = rt_tables[i]) == NULL)
321 			continue;
322 		(void) rnh->rnh_walktree(rnh, if_rtdel, ifp);
323 	}
324 
325 	/* Announce that the interface is gone. */
326 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
327 
328 	ifindex2ifnet[ifp->if_index] = NULL;
329 
330 	TAILQ_REMOVE(&ifnet, ifp, if_link);
331 	splx(s);
332 }
333 
334 /*
335  * Delete Routes for a Network Interface
336  *
337  * Called for each routing entry via the rnh->rnh_walktree() call above
338  * to delete all route entries referencing a detaching network interface.
339  *
340  * Arguments:
341  *	rn	pointer to node in the routing table
342  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
343  *
344  * Returns:
345  *	0	successful
346  *	errno	failed - reason indicated
347  *
348  */
349 static int
350 if_rtdel(rn, arg)
351 	struct radix_node	*rn;
352 	void			*arg;
353 {
354 	struct rtentry	*rt = (struct rtentry *)rn;
355 	struct ifnet	*ifp = arg;
356 	int		err;
357 
358 	if (rt->rt_ifp == ifp) {
359 
360 		/*
361 		 * Protect (sorta) against walktree recursion problems
362 		 * with cloned routes
363 		 */
364 		if ((rt->rt_flags & RTF_UP) == 0)
365 			return (0);
366 
367 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
368 				rt_mask(rt), rt->rt_flags,
369 				(struct rtentry **) NULL);
370 		if (err) {
371 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
372 		}
373 	}
374 
375 	return (0);
376 }
377 
378 /*
379  * Create a clone network interface.
380  */
381 int
382 if_clone_create(name, len)
383 	char *name;
384 	int len;
385 {
386 	struct if_clone *ifc;
387 	char *dp;
388 	int wildcard, bytoff, bitoff;
389 	int unit;
390 	int err;
391 
392 	ifc = if_clone_lookup(name, &unit);
393 	if (ifc == NULL)
394 		return (EINVAL);
395 
396 	if (ifunit(name) != NULL)
397 		return (EEXIST);
398 
399 	bytoff = bitoff = 0;
400 	wildcard = (unit < 0);
401 	/*
402 	 * Find a free unit if none was given.
403 	 */
404 	if (wildcard) {
405 		while ((bytoff < ifc->ifc_bmlen)
406 		    && (ifc->ifc_units[bytoff] == 0xff))
407 			bytoff++;
408 		if (bytoff >= ifc->ifc_bmlen)
409 			return (ENOSPC);
410 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
411 			bitoff++;
412 		unit = (bytoff << 3) + bitoff;
413 	}
414 
415 	if (unit > ifc->ifc_maxunit)
416 		return (ENXIO);
417 
418 	err = (*ifc->ifc_create)(ifc, unit);
419 	if (err != 0)
420 		return (err);
421 
422 	if (!wildcard) {
423 		bytoff = unit >> 3;
424 		bitoff = unit - (bytoff << 3);
425 	}
426 
427 	/*
428 	 * Allocate the unit in the bitmap.
429 	 */
430 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
431 	    ("%s: bit is already set", __func__));
432 	ifc->ifc_units[bytoff] |= (1 << bitoff);
433 
434 	/* In the wildcard case, we need to update the name. */
435 	if (wildcard) {
436 		for (dp = name; *dp != '\0'; dp++);
437 		if (snprintf(dp, len - (dp-name), "%d", unit) >
438 		    len - (dp-name) - 1) {
439 			/*
440 			 * This can only be a programmer error and
441 			 * there's no straightforward way to recover if
442 			 * it happens.
443 			 */
444 			panic("if_clone_create(): interface name too long");
445 		}
446 
447 	}
448 
449 	EVENTHANDLER_INVOKE(if_clone_event, ifc);
450 
451 	return (0);
452 }
453 
454 /*
455  * Destroy a clone network interface.
456  */
457 int
458 if_clone_destroy(name)
459 	const char *name;
460 {
461 	struct if_clone *ifc;
462 	struct ifnet *ifp;
463 	int bytoff, bitoff;
464 	int unit;
465 
466 	ifc = if_clone_lookup(name, &unit);
467 	if (ifc == NULL)
468 		return (EINVAL);
469 
470 	if (unit < ifc->ifc_minifs)
471 		return (EINVAL);
472 
473 	ifp = ifunit(name);
474 	if (ifp == NULL)
475 		return (ENXIO);
476 
477 	if (ifc->ifc_destroy == NULL)
478 		return (EOPNOTSUPP);
479 
480 	(*ifc->ifc_destroy)(ifp);
481 
482 	/*
483 	 * Compute offset in the bitmap and deallocate the unit.
484 	 */
485 	bytoff = unit >> 3;
486 	bitoff = unit - (bytoff << 3);
487 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
488 	    ("%s: bit is already cleared", __func__));
489 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
490 	return (0);
491 }
492 
493 /*
494  * Look up a network interface cloner.
495  */
496 struct if_clone *
497 if_clone_lookup(name, unitp)
498 	const char *name;
499 	int *unitp;
500 {
501 	struct if_clone *ifc;
502 	const char *cp;
503 	int i;
504 
505 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
506 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
507 			if (ifc->ifc_name[i] != *cp)
508 				goto next_ifc;
509 		}
510 		goto found_name;
511  next_ifc:
512 		ifc = LIST_NEXT(ifc, ifc_list);
513 	}
514 
515 	/* No match. */
516 	return ((struct if_clone *)NULL);
517 
518  found_name:
519 	if (*cp == '\0') {
520 		i = -1;
521 	} else {
522 		for (i = 0; *cp != '\0'; cp++) {
523 			if (*cp < '0' || *cp > '9') {
524 				/* Bogus unit number. */
525 				return (NULL);
526 			}
527 			i = (i * 10) + (*cp - '0');
528 		}
529 	}
530 
531 	if (unitp != NULL)
532 		*unitp = i;
533 	return (ifc);
534 }
535 
536 /*
537  * Register a network interface cloner.
538  */
539 void
540 if_clone_attach(ifc)
541 	struct if_clone *ifc;
542 {
543 	int bytoff, bitoff;
544 	int err;
545 	int len, maxclone;
546 	int unit;
547 
548 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
549 	    ("%s: %s requested more units then allowed (%d > %d)",
550 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
551 	    ifc->ifc_maxunit + 1));
552 	/*
553 	 * Compute bitmap size and allocate it.
554 	 */
555 	maxclone = ifc->ifc_maxunit + 1;
556 	len = maxclone >> 3;
557 	if ((len << 3) < maxclone)
558 		len++;
559 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
560 	ifc->ifc_bmlen = len;
561 
562 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
563 	if_cloners_count++;
564 
565 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
566 		err = (*ifc->ifc_create)(ifc, unit);
567 		KASSERT(err == 0,
568 		    ("%s: failed to create required interface %s%d",
569 		    __func__, ifc->ifc_name, unit));
570 
571 		/* Allocate the unit in the bitmap. */
572 		bytoff = unit >> 3;
573 		bitoff = unit - (bytoff << 3);
574 		ifc->ifc_units[bytoff] |= (1 << bitoff);
575 	}
576 }
577 
578 /*
579  * Unregister a network interface cloner.
580  */
581 void
582 if_clone_detach(ifc)
583 	struct if_clone *ifc;
584 {
585 
586 	LIST_REMOVE(ifc, ifc_list);
587 	free(ifc->ifc_units, M_CLONE);
588 	if_cloners_count--;
589 }
590 
591 /*
592  * Provide list of interface cloners to userspace.
593  */
594 int
595 if_clone_list(ifcr)
596 	struct if_clonereq *ifcr;
597 {
598 	char outbuf[IFNAMSIZ], *dst;
599 	struct if_clone *ifc;
600 	int count, error = 0;
601 
602 	ifcr->ifcr_total = if_cloners_count;
603 	if ((dst = ifcr->ifcr_buffer) == NULL) {
604 		/* Just asking how many there are. */
605 		return (0);
606 	}
607 
608 	if (ifcr->ifcr_count < 0)
609 		return (EINVAL);
610 
611 	count = (if_cloners_count < ifcr->ifcr_count) ?
612 	    if_cloners_count : ifcr->ifcr_count;
613 
614 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
615 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
616 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
617 		error = copyout(outbuf, dst, IFNAMSIZ);
618 		if (error)
619 			break;
620 	}
621 
622 	return (error);
623 }
624 
625 /*
626  * Locate an interface based on a complete address.
627  */
628 /*ARGSUSED*/
629 struct ifaddr *
630 ifa_ifwithaddr(addr)
631 	struct sockaddr *addr;
632 {
633 	struct ifnet *ifp;
634 	struct ifaddr *ifa;
635 
636 #define	equal(a1, a2) \
637   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
638 	TAILQ_FOREACH(ifp, &ifnet, if_link)
639 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
640 		if (ifa->ifa_addr->sa_family != addr->sa_family)
641 			continue;
642 		if (equal(addr, ifa->ifa_addr))
643 			return (ifa);
644 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
645 		    /* IP6 doesn't have broadcast */
646 		    ifa->ifa_broadaddr->sa_len != 0 &&
647 		    equal(ifa->ifa_broadaddr, addr))
648 			return (ifa);
649 	}
650 	return ((struct ifaddr *)0);
651 }
652 /*
653  * Locate the point to point interface with a given destination address.
654  */
655 /*ARGSUSED*/
656 struct ifaddr *
657 ifa_ifwithdstaddr(addr)
658 	struct sockaddr *addr;
659 {
660 	struct ifnet *ifp;
661 	struct ifaddr *ifa;
662 
663 	TAILQ_FOREACH(ifp, &ifnet, if_link)
664 	    if (ifp->if_flags & IFF_POINTOPOINT)
665 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
666 			if (ifa->ifa_addr->sa_family != addr->sa_family)
667 				continue;
668 			if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
669 				return (ifa);
670 	}
671 	return ((struct ifaddr *)0);
672 }
673 
674 /*
675  * Find an interface on a specific network.  If many, choice
676  * is most specific found.
677  */
678 struct ifaddr *
679 ifa_ifwithnet(addr)
680 	struct sockaddr *addr;
681 {
682 	struct ifnet *ifp;
683 	struct ifaddr *ifa;
684 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
685 	u_int af = addr->sa_family;
686 	char *addr_data = addr->sa_data, *cplim;
687 
688 	/*
689 	 * AF_LINK addresses can be looked up directly by their index number,
690 	 * so do that if we can.
691 	 */
692 	if (af == AF_LINK) {
693 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
694 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
695 		return (ifnet_addrs[sdl->sdl_index - 1]);
696 	}
697 
698 	/*
699 	 * Scan though each interface, looking for ones that have
700 	 * addresses in this address family.
701 	 */
702 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
703 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
704 			char *cp, *cp2, *cp3;
705 
706 			if (ifa->ifa_addr->sa_family != af)
707 next:				continue;
708 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
709 				/*
710 				 * This is a bit broken as it doesn't
711 				 * take into account that the remote end may
712 				 * be a single node in the network we are
713 				 * looking for.
714 				 * The trouble is that we don't know the
715 				 * netmask for the remote end.
716 				 */
717 				if (ifa->ifa_dstaddr != 0
718 				    && equal(addr, ifa->ifa_dstaddr))
719  					return (ifa);
720 			} else {
721 				/*
722 				 * if we have a special address handler,
723 				 * then use it instead of the generic one.
724 				 */
725 	          		if (ifa->ifa_claim_addr) {
726 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
727 						return (ifa);
728 					} else {
729 						continue;
730 					}
731 				}
732 
733 				/*
734 				 * Scan all the bits in the ifa's address.
735 				 * If a bit dissagrees with what we are
736 				 * looking for, mask it with the netmask
737 				 * to see if it really matters.
738 				 * (A byte at a time)
739 				 */
740 				if (ifa->ifa_netmask == 0)
741 					continue;
742 				cp = addr_data;
743 				cp2 = ifa->ifa_addr->sa_data;
744 				cp3 = ifa->ifa_netmask->sa_data;
745 				cplim = ifa->ifa_netmask->sa_len
746 					+ (char *)ifa->ifa_netmask;
747 				while (cp3 < cplim)
748 					if ((*cp++ ^ *cp2++) & *cp3++)
749 						goto next; /* next address! */
750 				/*
751 				 * If the netmask of what we just found
752 				 * is more specific than what we had before
753 				 * (if we had one) then remember the new one
754 				 * before continuing to search
755 				 * for an even better one.
756 				 */
757 				if (ifa_maybe == 0 ||
758 				    rn_refines((caddr_t)ifa->ifa_netmask,
759 				    (caddr_t)ifa_maybe->ifa_netmask))
760 					ifa_maybe = ifa;
761 			}
762 		}
763 	}
764 	return (ifa_maybe);
765 }
766 
767 /*
768  * Find an interface address specific to an interface best matching
769  * a given address.
770  */
771 struct ifaddr *
772 ifaof_ifpforaddr(addr, ifp)
773 	struct sockaddr *addr;
774 	struct ifnet *ifp;
775 {
776 	struct ifaddr *ifa;
777 	char *cp, *cp2, *cp3;
778 	char *cplim;
779 	struct ifaddr *ifa_maybe = 0;
780 	u_int af = addr->sa_family;
781 
782 	if (af >= AF_MAX)
783 		return (0);
784 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
785 		if (ifa->ifa_addr->sa_family != af)
786 			continue;
787 		if (ifa_maybe == 0)
788 			ifa_maybe = ifa;
789 		if (ifa->ifa_netmask == 0) {
790 			if (equal(addr, ifa->ifa_addr) ||
791 			    (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
792 				return (ifa);
793 			continue;
794 		}
795 		if (ifp->if_flags & IFF_POINTOPOINT) {
796 			if (equal(addr, ifa->ifa_dstaddr))
797 				return (ifa);
798 		} else {
799 			cp = addr->sa_data;
800 			cp2 = ifa->ifa_addr->sa_data;
801 			cp3 = ifa->ifa_netmask->sa_data;
802 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
803 			for (; cp3 < cplim; cp3++)
804 				if ((*cp++ ^ *cp2++) & *cp3)
805 					break;
806 			if (cp3 == cplim)
807 				return (ifa);
808 		}
809 	}
810 	return (ifa_maybe);
811 }
812 
813 #include <net/route.h>
814 
815 /*
816  * Default action when installing a route with a Link Level gateway.
817  * Lookup an appropriate real ifa to point to.
818  * This should be moved to /sys/net/link.c eventually.
819  */
820 static void
821 link_rtrequest(cmd, rt, info)
822 	int cmd;
823 	struct rtentry *rt;
824 	struct rt_addrinfo *info;
825 {
826 	struct ifaddr *ifa;
827 	struct sockaddr *dst;
828 	struct ifnet *ifp;
829 
830 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
831 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
832 		return;
833 	ifa = ifaof_ifpforaddr(dst, ifp);
834 	if (ifa) {
835 		IFAFREE(rt->rt_ifa);
836 		rt->rt_ifa = ifa;
837 		ifa->ifa_refcnt++;
838 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
839 			ifa->ifa_rtrequest(cmd, rt, info);
840 	}
841 }
842 
843 /*
844  * Mark an interface down and notify protocols of
845  * the transition.
846  * NOTE: must be called at splnet or eqivalent.
847  */
848 void
849 if_unroute(ifp, flag, fam)
850 	struct ifnet *ifp;
851 	int flag, fam;
852 {
853 	struct ifaddr *ifa;
854 
855 	ifp->if_flags &= ~flag;
856 	getmicrotime(&ifp->if_lastchange);
857 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
858 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
859 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
860 	if_qflush(&ifp->if_snd);
861 	rt_ifmsg(ifp);
862 }
863 
864 /*
865  * Mark an interface up and notify protocols of
866  * the transition.
867  * NOTE: must be called at splnet or eqivalent.
868  */
869 void
870 if_route(ifp, flag, fam)
871 	struct ifnet *ifp;
872 	int flag, fam;
873 {
874 	struct ifaddr *ifa;
875 
876 	ifp->if_flags |= flag;
877 	getmicrotime(&ifp->if_lastchange);
878 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
879 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
880 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
881 	rt_ifmsg(ifp);
882 #ifdef INET6
883 	in6_if_up(ifp);
884 #endif
885 }
886 
887 /*
888  * Mark an interface down and notify protocols of
889  * the transition.
890  * NOTE: must be called at splnet or eqivalent.
891  */
892 void
893 if_down(ifp)
894 	struct ifnet *ifp;
895 {
896 
897 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
898 }
899 
900 /*
901  * Mark an interface up and notify protocols of
902  * the transition.
903  * NOTE: must be called at splnet or eqivalent.
904  */
905 void
906 if_up(ifp)
907 	struct ifnet *ifp;
908 {
909 
910 	if_route(ifp, IFF_UP, AF_UNSPEC);
911 }
912 
913 /*
914  * Flush an interface queue.
915  */
916 static void
917 if_qflush(ifq)
918 	struct ifqueue *ifq;
919 {
920 	struct mbuf *m, *n;
921 
922 	n = ifq->ifq_head;
923 	while ((m = n) != 0) {
924 		n = m->m_nextpkt;
925 		m_freem(m);
926 	}
927 	ifq->ifq_head = 0;
928 	ifq->ifq_tail = 0;
929 	ifq->ifq_len = 0;
930 }
931 
932 /*
933  * Handle interface watchdog timer routines.  Called
934  * from softclock, we decrement timers (if set) and
935  * call the appropriate interface routine on expiration.
936  */
937 static void
938 if_slowtimo(arg)
939 	void *arg;
940 {
941 	struct ifnet *ifp;
942 	int s = splimp();
943 
944 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
945 		if (ifp->if_timer == 0 || --ifp->if_timer)
946 			continue;
947 		if (ifp->if_watchdog)
948 			(*ifp->if_watchdog)(ifp);
949 	}
950 	splx(s);
951 	callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
952 }
953 
954 /*
955  * Map interface name to
956  * interface structure pointer.
957  */
958 struct ifnet *
959 ifunit(const char *name)
960 {
961 	struct ifnet *ifp;
962 
963 	/*
964 	 * Search all the interfaces for this name/number
965 	 */
966 
967 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
968 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
969 			break;
970 	}
971 	return (ifp);
972 }
973 
974 
975 /*
976  * Map interface name in a sockaddr_dl to
977  * interface structure pointer.
978  */
979 struct ifnet *
980 if_withname(sa)
981 	struct sockaddr *sa;
982 {
983 	char ifname[IFNAMSIZ+1];
984 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
985 
986 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
987 	     (sdl->sdl_nlen > IFNAMSIZ) )
988 		return NULL;
989 
990 	/*
991 	 * ifunit wants a null-terminated name.  It may not be null-terminated
992 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
993 	 * and there might not be room to put the trailing null anyway, so we
994 	 * make a local copy that we know we can null terminate safely.
995 	 */
996 
997 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
998 	ifname[sdl->sdl_nlen] = '\0';
999 	return ifunit(ifname);
1000 }
1001 
1002 
1003 /*
1004  * Interface ioctls.
1005  */
1006 int
1007 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1008 {
1009 	struct ifnet *ifp;
1010 	struct ifreq *ifr;
1011 	struct ifstat *ifs;
1012 	int error;
1013 	short oif_flags;
1014 	int new_flags;
1015 	size_t namelen, onamelen;
1016 	char new_name[IFNAMSIZ];
1017 	struct ifaddr *ifa;
1018 	struct sockaddr_dl *sdl;
1019 
1020 	switch (cmd) {
1021 
1022 	case SIOCGIFCONF:
1023 	case OSIOCGIFCONF:
1024 		return (ifconf(cmd, data, td));
1025 	}
1026 	ifr = (struct ifreq *)data;
1027 
1028 	switch (cmd) {
1029 	case SIOCIFCREATE:
1030 	case SIOCIFDESTROY:
1031 		if ((error = suser(td)) != 0)
1032 			return (error);
1033 		return ((cmd == SIOCIFCREATE) ?
1034 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1035 			if_clone_destroy(ifr->ifr_name));
1036 
1037 	case SIOCIFGCLONERS:
1038 		return (if_clone_list((struct if_clonereq *)data));
1039 	}
1040 
1041 	ifp = ifunit(ifr->ifr_name);
1042 	if (ifp == 0)
1043 		return (ENXIO);
1044 	switch (cmd) {
1045 
1046 	case SIOCGIFFLAGS:
1047 		ifr->ifr_flags = ifp->if_flags;
1048 		ifr->ifr_flagshigh = ifp->if_flags >> 16;
1049 		break;
1050 
1051 	case SIOCGIFCAP:
1052 		ifr->ifr_reqcap = ifp->if_capabilities;
1053 		ifr->ifr_curcap = ifp->if_capenable;
1054 		break;
1055 
1056 	case SIOCGIFMETRIC:
1057 		ifr->ifr_metric = ifp->if_metric;
1058 		break;
1059 
1060 	case SIOCGIFMTU:
1061 		ifr->ifr_mtu = ifp->if_mtu;
1062 		break;
1063 
1064 	case SIOCGIFPHYS:
1065 		ifr->ifr_phys = ifp->if_physical;
1066 		break;
1067 
1068 	case SIOCSIFFLAGS:
1069 		error = suser(td);
1070 		if (error)
1071 			return (error);
1072 		new_flags = (ifr->ifr_flags & 0xffff) |
1073 		    (ifr->ifr_flagshigh << 16);
1074 		if (ifp->if_flags & IFF_SMART) {
1075 			/* Smart drivers twiddle their own routes */
1076 		} else if (ifp->if_flags & IFF_UP &&
1077 		    (new_flags & IFF_UP) == 0) {
1078 			int s = splimp();
1079 			if_down(ifp);
1080 			splx(s);
1081 		} else if (new_flags & IFF_UP &&
1082 		    (ifp->if_flags & IFF_UP) == 0) {
1083 			int s = splimp();
1084 			if_up(ifp);
1085 			splx(s);
1086 		}
1087 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1088 			(new_flags &~ IFF_CANTCHANGE);
1089 		if (new_flags & IFF_PPROMISC) {
1090 			/* Permanently promiscuous mode requested */
1091 			ifp->if_flags |= IFF_PROMISC;
1092 		} else if (ifp->if_pcount == 0) {
1093 			ifp->if_flags &= ~IFF_PROMISC;
1094 		}
1095 		if (ifp->if_ioctl)
1096 			(void) (*ifp->if_ioctl)(ifp, cmd, data,
1097 						td->td_proc->p_ucred);
1098 		getmicrotime(&ifp->if_lastchange);
1099 		break;
1100 
1101 	case SIOCSIFCAP:
1102 		error = suser(td);
1103 		if (error)
1104 			return (error);
1105 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1106 			return (EINVAL);
1107 		(void) (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1108 		break;
1109 
1110  	case SIOCSIFNAME:
1111  		error = suser(td);
1112  		if (error != 0)
1113  			return (error);
1114  		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1115  		if (error != 0)
1116  			return (error);
1117 		if (new_name[0] == '\0')
1118 			return (EINVAL);
1119  		if (ifunit(new_name) != NULL)
1120  			return (EEXIST);
1121 
1122 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1123 
1124  		/* Announce the departure of the interface. */
1125  		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1126 
1127  		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1128  		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1129  		/* XXX IFA_LOCK(ifa); */
1130  		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1131  		namelen = strlen(new_name);
1132  		onamelen = sdl->sdl_nlen;
1133  		/*
1134  		 * Move the address if needed.  This is safe because we
1135  		 * allocate space for a name of length IFNAMSIZ when we
1136  		 * create this in if_attach().
1137  		 */
1138  		if (namelen != onamelen) {
1139  			bcopy(sdl->sdl_data + onamelen,
1140  			    sdl->sdl_data + namelen, sdl->sdl_alen);
1141  		}
1142  		bcopy(new_name, sdl->sdl_data, namelen);
1143  		sdl->sdl_nlen = namelen;
1144  		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1145  		bzero(sdl->sdl_data, onamelen);
1146  		while (namelen != 0)
1147  			sdl->sdl_data[--namelen] = 0xff;
1148  		/* XXX IFA_UNLOCK(ifa) */
1149 
1150 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1151 
1152  		/* Announce the return of the interface. */
1153  		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1154  		break;
1155 
1156 	case SIOCSIFMETRIC:
1157 		error = suser(td);
1158 		if (error)
1159 			return (error);
1160 		ifp->if_metric = ifr->ifr_metric;
1161 		getmicrotime(&ifp->if_lastchange);
1162 		break;
1163 
1164 	case SIOCSIFPHYS:
1165 		error = suser(td);
1166 		if (error)
1167 			return error;
1168 		if (!ifp->if_ioctl)
1169 		        return EOPNOTSUPP;
1170 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1171 		if (error == 0)
1172 			getmicrotime(&ifp->if_lastchange);
1173 		return(error);
1174 
1175 	case SIOCSIFMTU:
1176 	{
1177 		u_long oldmtu = ifp->if_mtu;
1178 
1179 		error = suser(td);
1180 		if (error)
1181 			return (error);
1182 		if (ifp->if_ioctl == NULL)
1183 			return (EOPNOTSUPP);
1184 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1185 			return (EINVAL);
1186 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1187 		if (error == 0) {
1188 			getmicrotime(&ifp->if_lastchange);
1189 			rt_ifmsg(ifp);
1190 		}
1191 		/*
1192 		 * If the link MTU changed, do network layer specific procedure.
1193 		 */
1194 		if (ifp->if_mtu != oldmtu) {
1195 #ifdef INET6
1196 			nd6_setmtu(ifp);
1197 #endif
1198 		}
1199 		return (error);
1200 	}
1201 
1202 	case SIOCADDMULTI:
1203 	case SIOCDELMULTI:
1204 		error = suser(td);
1205 		if (error)
1206 			return (error);
1207 
1208 		/* Don't allow group membership on non-multicast interfaces. */
1209 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1210 			return EOPNOTSUPP;
1211 
1212 		/* Don't let users screw up protocols' entries. */
1213 		if (ifr->ifr_addr.sa_family != AF_LINK)
1214 			return EINVAL;
1215 
1216 		if (cmd == SIOCADDMULTI) {
1217 			struct ifmultiaddr *ifma;
1218 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1219 		} else {
1220 			error = if_delmulti(ifp, &ifr->ifr_addr);
1221 		}
1222 		if (error == 0)
1223 			getmicrotime(&ifp->if_lastchange);
1224 		return error;
1225 
1226 	case SIOCSIFPHYADDR:
1227 	case SIOCDIFPHYADDR:
1228 #ifdef INET6
1229 	case SIOCSIFPHYADDR_IN6:
1230 #endif
1231 	case SIOCSLIFPHYADDR:
1232         case SIOCSIFMEDIA:
1233 	case SIOCSIFGENERIC:
1234 		error = suser(td);
1235 		if (error)
1236 			return (error);
1237 		if (ifp->if_ioctl == 0)
1238 			return (EOPNOTSUPP);
1239 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1240 		if (error == 0)
1241 			getmicrotime(&ifp->if_lastchange);
1242 		return error;
1243 
1244 	case SIOCGIFSTATUS:
1245 		ifs = (struct ifstat *)data;
1246 		ifs->ascii[0] = '\0';
1247 
1248 	case SIOCGIFPSRCADDR:
1249 	case SIOCGIFPDSTADDR:
1250 	case SIOCGLIFPHYADDR:
1251 	case SIOCGIFMEDIA:
1252 	case SIOCGIFGENERIC:
1253 		if (ifp->if_ioctl == 0)
1254 			return (EOPNOTSUPP);
1255 		return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1256 
1257 	case SIOCSIFLLADDR:
1258 		error = suser(td);
1259 		if (error)
1260 			return (error);
1261 		return if_setlladdr(ifp,
1262 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1263 
1264 	default:
1265 		oif_flags = ifp->if_flags;
1266 		if (so->so_proto == 0)
1267 			return (EOPNOTSUPP);
1268 #ifndef COMPAT_43
1269 		error = so_pru_control(so, cmd, data, ifp, td);
1270 #else
1271 	    {
1272 		int ocmd = cmd;
1273 
1274 		switch (cmd) {
1275 
1276 		case SIOCSIFDSTADDR:
1277 		case SIOCSIFADDR:
1278 		case SIOCSIFBRDADDR:
1279 		case SIOCSIFNETMASK:
1280 #if BYTE_ORDER != BIG_ENDIAN
1281 			if (ifr->ifr_addr.sa_family == 0 &&
1282 			    ifr->ifr_addr.sa_len < 16) {
1283 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1284 				ifr->ifr_addr.sa_len = 16;
1285 			}
1286 #else
1287 			if (ifr->ifr_addr.sa_len == 0)
1288 				ifr->ifr_addr.sa_len = 16;
1289 #endif
1290 			break;
1291 
1292 		case OSIOCGIFADDR:
1293 			cmd = SIOCGIFADDR;
1294 			break;
1295 
1296 		case OSIOCGIFDSTADDR:
1297 			cmd = SIOCGIFDSTADDR;
1298 			break;
1299 
1300 		case OSIOCGIFBRDADDR:
1301 			cmd = SIOCGIFBRDADDR;
1302 			break;
1303 
1304 		case OSIOCGIFNETMASK:
1305 			cmd = SIOCGIFNETMASK;
1306 		}
1307 		error =  so_pru_control(so, cmd, data, ifp, td);
1308 		switch (ocmd) {
1309 
1310 		case OSIOCGIFADDR:
1311 		case OSIOCGIFDSTADDR:
1312 		case OSIOCGIFBRDADDR:
1313 		case OSIOCGIFNETMASK:
1314 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1315 
1316 		}
1317 	    }
1318 #endif /* COMPAT_43 */
1319 
1320 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1321 #ifdef INET6
1322 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1323 			if (ifp->if_flags & IFF_UP) {
1324 				int s = splimp();
1325 				in6_if_up(ifp);
1326 				splx(s);
1327 			}
1328 #endif
1329 		}
1330 		return (error);
1331 
1332 	}
1333 	return (0);
1334 }
1335 
1336 /*
1337  * Set/clear promiscuous mode on interface ifp based on the truth value
1338  * of pswitch.  The calls are reference counted so that only the first
1339  * "on" request actually has an effect, as does the final "off" request.
1340  * Results are undefined if the "off" and "on" requests are not matched.
1341  */
1342 int
1343 ifpromisc(ifp, pswitch)
1344 	struct ifnet *ifp;
1345 	int pswitch;
1346 {
1347 	struct ifreq ifr;
1348 	int error;
1349 	int oldflags;
1350 
1351 	oldflags = ifp->if_flags;
1352 	if (ifp->if_flags & IFF_PPROMISC) {
1353 		/* Do nothing if device is in permanently promiscuous mode */
1354 		ifp->if_pcount += pswitch ? 1 : -1;
1355 		return (0);
1356 	}
1357 	if (pswitch) {
1358 		/*
1359 		 * If the device is not configured up, we cannot put it in
1360 		 * promiscuous mode.
1361 		 */
1362 		if ((ifp->if_flags & IFF_UP) == 0)
1363 			return (ENETDOWN);
1364 		if (ifp->if_pcount++ != 0)
1365 			return (0);
1366 		ifp->if_flags |= IFF_PROMISC;
1367 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1368 		    ifp->if_xname);
1369 	} else {
1370 		if (--ifp->if_pcount > 0)
1371 			return (0);
1372 		ifp->if_flags &= ~IFF_PROMISC;
1373 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1374 		    ifp->if_xname);
1375 	}
1376 	ifr.ifr_flags = ifp->if_flags;
1377 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
1378 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1379 				 (struct ucred *)NULL);
1380 	if (error == 0)
1381 		rt_ifmsg(ifp);
1382 	else
1383 		ifp->if_flags = oldflags;
1384 	return error;
1385 }
1386 
1387 /*
1388  * Return interface configuration
1389  * of system.  List may be used
1390  * in later ioctl's (above) to get
1391  * other information.
1392  */
1393 /*ARGSUSED*/
1394 static int
1395 ifconf(u_long cmd, caddr_t data, struct thread *td)
1396 {
1397 	struct ifconf *ifc = (struct ifconf *)data;
1398 	struct ifnet *ifp;
1399 	struct ifaddr *ifa;
1400 	struct sockaddr *sa;
1401 	struct ifreq ifr, *ifrp;
1402 	int space = ifc->ifc_len, error = 0;
1403 
1404 	ifrp = ifc->ifc_req;
1405 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1406 		int addrs;
1407 
1408 		if (space <= sizeof (ifr))
1409 			break;
1410 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1411 		    >= sizeof(ifr.ifr_name)) {
1412 			error = ENAMETOOLONG;
1413 			break;
1414 		}
1415 
1416 		addrs = 0;
1417 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1418 			if (space <= sizeof(ifr))
1419 				break;
1420 			sa = ifa->ifa_addr;
1421 			if (td->td_proc->p_ucred->cr_prison &&
1422 			    prison_if(td, sa))
1423 				continue;
1424 			addrs++;
1425 #ifdef COMPAT_43
1426 			if (cmd == OSIOCGIFCONF) {
1427 				struct osockaddr *osa =
1428 					 (struct osockaddr *)&ifr.ifr_addr;
1429 				ifr.ifr_addr = *sa;
1430 				osa->sa_family = sa->sa_family;
1431 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1432 						sizeof (ifr));
1433 				ifrp++;
1434 			} else
1435 #endif
1436 			if (sa->sa_len <= sizeof(*sa)) {
1437 				ifr.ifr_addr = *sa;
1438 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1439 						sizeof (ifr));
1440 				ifrp++;
1441 			} else {
1442 				if (space < sizeof (ifr) + sa->sa_len -
1443 					    sizeof(*sa))
1444 					break;
1445 				space -= sa->sa_len - sizeof(*sa);
1446 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1447 						sizeof (ifr.ifr_name));
1448 				if (error == 0)
1449 				    error = copyout((caddr_t)sa,
1450 				      (caddr_t)&ifrp->ifr_addr, sa->sa_len);
1451 				ifrp = (struct ifreq *)
1452 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1453 			}
1454 			if (error)
1455 				break;
1456 			space -= sizeof (ifr);
1457 		}
1458 		if (error)
1459 			break;
1460 		if (!addrs) {
1461 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
1462 			error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
1463 			    sizeof (ifr));
1464 			if (error)
1465 				break;
1466 			space -= sizeof (ifr);
1467 			ifrp++;
1468 		}
1469 	}
1470 	ifc->ifc_len -= space;
1471 	return (error);
1472 }
1473 
1474 /*
1475  * Just like if_promisc(), but for all-multicast-reception mode.
1476  */
1477 int
1478 if_allmulti(ifp, onswitch)
1479 	struct ifnet *ifp;
1480 	int onswitch;
1481 {
1482 	int error = 0;
1483 	int s = splimp();
1484 	struct ifreq ifr;
1485 
1486 	if (onswitch) {
1487 		if (ifp->if_amcount++ == 0) {
1488 			ifp->if_flags |= IFF_ALLMULTI;
1489 			ifr.ifr_flags = ifp->if_flags;
1490 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1491 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1492 					      (struct ucred *)NULL);
1493 		}
1494 	} else {
1495 		if (ifp->if_amcount > 1) {
1496 			ifp->if_amcount--;
1497 		} else {
1498 			ifp->if_amcount = 0;
1499 			ifp->if_flags &= ~IFF_ALLMULTI;
1500 			ifr.ifr_flags = ifp->if_flags;
1501 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1502 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1503 					      (struct ucred *)NULL);
1504 		}
1505 	}
1506 	splx(s);
1507 
1508 	if (error == 0)
1509 		rt_ifmsg(ifp);
1510 	return error;
1511 }
1512 
1513 /*
1514  * Add a multicast listenership to the interface in question.
1515  * The link layer provides a routine which converts
1516  */
1517 int
1518 if_addmulti(ifp, sa, retifma)
1519 	struct ifnet *ifp;	/* interface to manipulate */
1520 	struct sockaddr *sa;	/* address to add */
1521 	struct ifmultiaddr **retifma;
1522 {
1523 	struct sockaddr *llsa, *dupsa;
1524 	int error, s;
1525 	struct ifmultiaddr *ifma;
1526 
1527 	/*
1528 	 * If the matching multicast address already exists
1529 	 * then don't add a new one, just add a reference
1530 	 */
1531 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1532 		if (equal(sa, ifma->ifma_addr)) {
1533 			ifma->ifma_refcount++;
1534 			if (retifma)
1535 				*retifma = ifma;
1536 			return 0;
1537 		}
1538 	}
1539 
1540 	/*
1541 	 * Give the link layer a chance to accept/reject it, and also
1542 	 * find out which AF_LINK address this maps to, if it isn't one
1543 	 * already.
1544 	 */
1545 	if (ifp->if_resolvemulti) {
1546 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1547 		if (error) return error;
1548 	} else {
1549 		llsa = 0;
1550 	}
1551 
1552 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1553 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1554 	bcopy(sa, dupsa, sa->sa_len);
1555 
1556 	ifma->ifma_addr = dupsa;
1557 	ifma->ifma_lladdr = llsa;
1558 	ifma->ifma_ifp = ifp;
1559 	ifma->ifma_refcount = 1;
1560 	ifma->ifma_protospec = 0;
1561 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1562 
1563 	/*
1564 	 * Some network interfaces can scan the address list at
1565 	 * interrupt time; lock them out.
1566 	 */
1567 	s = splimp();
1568 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1569 	splx(s);
1570 	*retifma = ifma;
1571 
1572 	if (llsa != 0) {
1573 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1574 			if (equal(ifma->ifma_addr, llsa))
1575 				break;
1576 		}
1577 		if (ifma) {
1578 			ifma->ifma_refcount++;
1579 		} else {
1580 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1581 			       M_IFMADDR, M_WAITOK);
1582 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1583 			       M_IFMADDR, M_WAITOK);
1584 			bcopy(llsa, dupsa, llsa->sa_len);
1585 			ifma->ifma_addr = dupsa;
1586 			ifma->ifma_ifp = ifp;
1587 			ifma->ifma_refcount = 1;
1588 			s = splimp();
1589 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1590 			splx(s);
1591 		}
1592 	}
1593 	/*
1594 	 * We are certain we have added something, so call down to the
1595 	 * interface to let them know about it.
1596 	 */
1597 	s = splimp();
1598 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1599 	splx(s);
1600 
1601 	return 0;
1602 }
1603 
1604 /*
1605  * Remove a reference to a multicast address on this interface.  Yell
1606  * if the request does not match an existing membership.
1607  */
1608 int
1609 if_delmulti(ifp, sa)
1610 	struct ifnet *ifp;
1611 	struct sockaddr *sa;
1612 {
1613 	struct ifmultiaddr *ifma;
1614 	int s;
1615 
1616 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1617 		if (equal(sa, ifma->ifma_addr))
1618 			break;
1619 	if (ifma == 0)
1620 		return ENOENT;
1621 
1622 	if (ifma->ifma_refcount > 1) {
1623 		ifma->ifma_refcount--;
1624 		return 0;
1625 	}
1626 
1627 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1628 	sa = ifma->ifma_lladdr;
1629 	s = splimp();
1630 	LIST_REMOVE(ifma, ifma_link);
1631 	/*
1632 	 * Make sure the interface driver is notified
1633 	 * in the case of a link layer mcast group being left.
1634 	 */
1635 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1636 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1637 	splx(s);
1638 	free(ifma->ifma_addr, M_IFMADDR);
1639 	free(ifma, M_IFMADDR);
1640 	if (sa == 0)
1641 		return 0;
1642 
1643 	/*
1644 	 * Now look for the link-layer address which corresponds to
1645 	 * this network address.  It had been squirreled away in
1646 	 * ifma->ifma_lladdr for this purpose (so we don't have
1647 	 * to call ifp->if_resolvemulti() again), and we saved that
1648 	 * value in sa above.  If some nasty deleted the
1649 	 * link-layer address out from underneath us, we can deal because
1650 	 * the address we stored was is not the same as the one which was
1651 	 * in the record for the link-layer address.  (So we don't complain
1652 	 * in that case.)
1653 	 */
1654 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1655 		if (equal(sa, ifma->ifma_addr))
1656 			break;
1657 	if (ifma == 0)
1658 		return 0;
1659 
1660 	if (ifma->ifma_refcount > 1) {
1661 		ifma->ifma_refcount--;
1662 		return 0;
1663 	}
1664 
1665 	s = splimp();
1666 	LIST_REMOVE(ifma, ifma_link);
1667 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1668 	splx(s);
1669 	free(ifma->ifma_addr, M_IFMADDR);
1670 	free(sa, M_IFMADDR);
1671 	free(ifma, M_IFMADDR);
1672 
1673 	return 0;
1674 }
1675 
1676 /*
1677  * Set the link layer address on an interface.
1678  *
1679  * At this time we only support certain types of interfaces,
1680  * and we don't allow the length of the address to change.
1681  */
1682 int
1683 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1684 {
1685 	struct sockaddr_dl *sdl;
1686 	struct ifaddr *ifa;
1687 	struct ifreq ifr;
1688 
1689 	ifa = ifnet_addrs[ifp->if_index - 1];
1690 	if (ifa == NULL)
1691 		return (EINVAL);
1692 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1693 	if (sdl == NULL)
1694 		return (EINVAL);
1695 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1696 		return (EINVAL);
1697 	switch (ifp->if_type) {
1698 	case IFT_ETHER:			/* these types use struct arpcom */
1699 	case IFT_FDDI:
1700 	case IFT_XETHER:
1701 	case IFT_ISO88025:
1702 	case IFT_L2VLAN:
1703 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1704 		/* FALLTHROUGH */
1705 	case IFT_ARCNET:
1706 		bcopy(lladdr, LLADDR(sdl), len);
1707 		break;
1708 	default:
1709 		return (ENODEV);
1710 	}
1711 	/*
1712 	 * If the interface is already up, we need
1713 	 * to re-init it in order to reprogram its
1714 	 * address filter.
1715 	 */
1716 	if ((ifp->if_flags & IFF_UP) != 0) {
1717 		ifp->if_flags &= ~IFF_UP;
1718 		ifr.ifr_flags = ifp->if_flags;
1719 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1720 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1721 				 (struct ucred *)NULL);
1722 		ifp->if_flags |= IFF_UP;
1723 		ifr.ifr_flags = ifp->if_flags;
1724 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1725 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1726 				 (struct ucred *)NULL);
1727 #ifdef INET
1728 		/*
1729 		 * Also send gratuitous ARPs to notify other nodes about
1730 		 * the address change.
1731 		 */
1732 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1733 			if (ifa->ifa_addr != NULL &&
1734 			    ifa->ifa_addr->sa_family == AF_INET)
1735 				arp_ifinit(ifp, ifa);
1736 		}
1737 #endif
1738 	}
1739 	return (0);
1740 }
1741 
1742 struct ifmultiaddr *
1743 ifmaof_ifpforaddr(sa, ifp)
1744 	struct sockaddr *sa;
1745 	struct ifnet *ifp;
1746 {
1747 	struct ifmultiaddr *ifma;
1748 
1749 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1750 		if (equal(ifma->ifma_addr, sa))
1751 			break;
1752 
1753 	return ifma;
1754 }
1755 
1756 /*
1757  * The name argument must be a pointer to storage which will last as
1758  * long as the interface does.  For physical devices, the result of
1759  * device_get_name(dev) is a good choice and for pseudo-devices a
1760  * static string works well.
1761  */
1762 void
1763 if_initname(struct ifnet *ifp, const char *name, int unit)
1764 {
1765 	ifp->if_dname = name;
1766 	ifp->if_dunit = unit;
1767 	if (unit != IF_DUNIT_NONE)
1768 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1769 	else
1770 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1771 }
1772 
1773 int
1774 if_printf(struct ifnet *ifp, const char *fmt, ...)
1775 {
1776 	__va_list ap;
1777 	int retval;
1778 
1779 	retval = printf("%s: ", ifp->if_xname);
1780 	__va_start(ap, fmt);
1781 	retval += vprintf(fmt, ap);
1782 	__va_end(ap);
1783 	return (retval);
1784 }
1785 
1786 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1787 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1788