xref: /dragonfly/sys/net/if.c (revision 5153f92b)
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 1.24 2005/01/06 09:14:13 hsu Exp $
36  */
37 
38 #include "opt_compat.h"
39 #include "opt_inet6.h"
40 #include "opt_inet.h"
41 
42 #include <sys/param.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/systm.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/socketops.h>
51 #include <sys/protosw.h>
52 #include <sys/kernel.h>
53 #include <sys/sockio.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 
57 #include <net/if.h>
58 #include <net/if_arp.h>
59 #include <net/if_dl.h>
60 #include <net/if_types.h>
61 #include <net/if_var.h>
62 #include <net/radix.h>
63 #include <net/route.h>
64 #include <machine/stdarg.h>
65 
66 #if defined(INET) || defined(INET6)
67 /*XXX*/
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <netinet/if_ether.h>
71 #ifdef INET6
72 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
73 #include <netinet6/in6_var.h>
74 #include <netinet6/in6_ifattach.h>
75 #endif
76 #endif
77 
78 #if defined(COMPAT_43)
79 #include <emulation/43bsd/43bsd_socket.h>
80 #endif /* COMPAT_43 */
81 
82 /*
83  * System initialization
84  */
85 
86 static int ifconf (u_long, caddr_t, struct thread *);
87 static void ifinit (void *);
88 static void if_qflush (struct ifqueue *);
89 static void if_slowtimo (void *);
90 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
91 static int  if_rtdel (struct radix_node *, void *);
92 
93 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
94 
95 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
96 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
97 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
98 
99 int	ifqmaxlen = IFQ_MAXLEN;
100 struct	ifnethead ifnet;	/* depend on static init XXX */
101 
102 #ifdef INET6
103 /*
104  * XXX: declare here to avoid to include many inet6 related files..
105  * should be more generalized?
106  */
107 extern void	nd6_setmtu (struct ifnet *);
108 #endif
109 
110 struct if_clone *if_clone_lookup (const char *, int *);
111 int if_clone_list (struct if_clonereq *);
112 
113 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
114 int if_cloners_count;
115 
116 struct callout if_slowtimo_timer;
117 
118 /*
119  * Network interface utility routines.
120  *
121  * Routines with ifa_ifwith* names take sockaddr *'s as
122  * parameters.
123  */
124 /* ARGSUSED*/
125 void
126 ifinit(void *dummy)
127 {
128 	struct ifnet *ifp;
129 	int s;
130 
131 	callout_init(&if_slowtimo_timer);
132 
133 	s = splimp();
134 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
135 		if (ifp->if_snd.ifq_maxlen == 0) {
136 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
137 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
138 		}
139 	}
140 	splx(s);
141 
142 	if_slowtimo(0);
143 }
144 
145 int if_index = 0;
146 struct ifaddr **ifnet_addrs;
147 struct ifnet **ifindex2ifnet = NULL;
148 
149 /*
150  * Attach an interface to the
151  * list of "active" interfaces.
152  */
153 void
154 if_attach(struct ifnet *ifp)
155 {
156 	unsigned socksize, ifasize;
157 	int namelen, masklen;
158 	struct sockaddr_dl *sdl;
159 	struct ifaddr *ifa;
160 	static int if_indexlim = 8;
161 	static int inited;
162 
163 	if (!inited) {
164 		TAILQ_INIT(&ifnet);
165 		inited = 1;
166 	}
167 
168 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
169 	ifp->if_index = ++if_index;
170 	/*
171 	 * XXX -
172 	 * The old code would work if the interface passed a pre-existing
173 	 * chain of ifaddrs to this code.  We don't trust our callers to
174 	 * properly initialize the tailq, however, so we no longer allow
175 	 * this unlikely case.
176 	 */
177 	TAILQ_INIT(&ifp->if_addrhead);
178 	TAILQ_INIT(&ifp->if_prefixhead);
179 	LIST_INIT(&ifp->if_multiaddrs);
180 	getmicrotime(&ifp->if_lastchange);
181 	if (ifnet_addrs == NULL || if_index >= if_indexlim) {
182 		unsigned n = (if_indexlim <<= 1) * sizeof(ifa);
183 		caddr_t q = malloc(n, M_IFADDR, M_WAITOK);
184 		bzero(q, n);
185 		if (ifnet_addrs != NULL) {
186 			bcopy(ifnet_addrs, q, n/2);
187 			free(ifnet_addrs, M_IFADDR);
188 		}
189 		ifnet_addrs = (struct ifaddr **)q;
190 
191 		/* grow ifindex2ifnet */
192 		n = if_indexlim * sizeof(struct ifnet *);
193 		q = malloc(n, M_IFADDR, M_WAITOK);
194 		bzero(q, n);
195 		if (ifindex2ifnet) {
196 			bcopy(ifindex2ifnet, q, n/2);
197 			free(ifindex2ifnet, M_IFADDR);
198 		}
199 		ifindex2ifnet = (struct ifnet **)q;
200 	}
201 
202 	ifindex2ifnet[if_index] = ifp;
203 
204 	/*
205 	 * create a Link Level name for this device
206 	 */
207 	namelen = strlen(ifp->if_xname);
208 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
209 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
210 	socksize = masklen + ifp->if_addrlen;
211 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
212 	if (socksize < sizeof(*sdl))
213 		socksize = sizeof(*sdl);
214 	socksize = ROUNDUP(socksize);
215 	ifasize = sizeof(*ifa) + 2 * socksize;
216 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
217 	if (ifa) {
218 		bzero(ifa, ifasize);
219 		sdl = (struct sockaddr_dl *)(ifa + 1);
220 		sdl->sdl_len = socksize;
221 		sdl->sdl_family = AF_LINK;
222 		bcopy(ifp->if_xname, sdl->sdl_data, namelen);
223 		sdl->sdl_nlen = namelen;
224 		sdl->sdl_index = ifp->if_index;
225 		sdl->sdl_type = ifp->if_type;
226 		ifnet_addrs[if_index - 1] = ifa;
227 		ifa->ifa_ifp = ifp;
228 		ifa->ifa_rtrequest = link_rtrequest;
229 		ifa->ifa_addr = (struct sockaddr *)sdl;
230 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
231 		ifa->ifa_netmask = (struct sockaddr *)sdl;
232 		sdl->sdl_len = masklen;
233 		while (namelen != 0)
234 			sdl->sdl_data[--namelen] = 0xff;
235 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
236 	}
237 
238 	EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
239 
240 	/* Announce the interface. */
241 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
242 }
243 
244 /*
245  * Detach an interface, removing it from the
246  * list of "active" interfaces.
247  */
248 void
249 if_detach(struct ifnet *ifp)
250 {
251 	struct ifaddr *ifa;
252 	struct radix_node_head	*rnh;
253 	int s;
254 	int i;
255 
256 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
257 
258 	/*
259 	 * Remove routes and flush queues.
260 	 */
261 	s = splnet();
262 	if_down(ifp);
263 
264 	/*
265 	 * Remove address from ifnet_addrs[] and maybe decrement if_index.
266 	 * Clean up all addresses.
267 	 */
268 	ifnet_addrs[ifp->if_index - 1] = 0;
269 	while (if_index > 0 && ifnet_addrs[if_index - 1] == 0)
270 		if_index--;
271 
272 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
273 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
274 #ifdef INET
275 		/* XXX: Ugly!! ad hoc just for INET */
276 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
277 			struct ifaliasreq ifr;
278 
279 			bzero(&ifr, sizeof ifr);
280 			ifr.ifra_addr = *ifa->ifa_addr;
281 			if (ifa->ifa_dstaddr)
282 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
283 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
284 			    NULL) == 0)
285 				continue;
286 		}
287 #endif /* INET */
288 #ifdef INET6
289 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
290 			in6_purgeaddr(ifa);
291 			/* ifp_addrhead is already updated */
292 			continue;
293 		}
294 #endif /* INET6 */
295 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
296 		IFAFREE(ifa);
297 	}
298 
299 #ifdef INET6
300 	/*
301 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
302 	 * before removing routing entries below, since IPv6 interface direct
303 	 * routes are expected to be removed by the IPv6-specific kernel API.
304 	 * Otherwise, the kernel will detect some inconsistency and bark it.
305 	 */
306 	in6_ifdetach(ifp);
307 #endif
308 
309 	/*
310 	 * Delete all remaining routes using this interface
311 	 * Unfortuneatly the only way to do this is to slog through
312 	 * the entire routing table looking for routes which point
313 	 * to this interface...oh well...
314 	 */
315 	for (i = 1; i <= AF_MAX; i++) {
316 		if ((rnh = rt_tables[i]) == NULL)
317 			continue;
318 		rnh->rnh_walktree(rnh, if_rtdel, ifp);
319 	}
320 
321 	/* Announce that the interface is gone. */
322 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
323 
324 	ifindex2ifnet[ifp->if_index] = NULL;
325 
326 	TAILQ_REMOVE(&ifnet, ifp, if_link);
327 	splx(s);
328 }
329 
330 /*
331  * Delete Routes for a Network Interface
332  *
333  * Called for each routing entry via the rnh->rnh_walktree() call above
334  * to delete all route entries referencing a detaching network interface.
335  *
336  * Arguments:
337  *	rn	pointer to node in the routing table
338  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
339  *
340  * Returns:
341  *	0	successful
342  *	errno	failed - reason indicated
343  *
344  */
345 static int
346 if_rtdel(struct radix_node *rn, void *arg)
347 {
348 	struct rtentry	*rt = (struct rtentry *)rn;
349 	struct ifnet	*ifp = arg;
350 	int		err;
351 
352 	if (rt->rt_ifp == ifp) {
353 
354 		/*
355 		 * Protect (sorta) against walktree recursion problems
356 		 * with cloned routes
357 		 */
358 		if (!(rt->rt_flags & RTF_UP))
359 			return (0);
360 
361 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
362 				rt_mask(rt), rt->rt_flags,
363 				(struct rtentry **) NULL);
364 		if (err) {
365 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
366 		}
367 	}
368 
369 	return (0);
370 }
371 
372 /*
373  * Create a clone network interface.
374  */
375 int
376 if_clone_create(char *name, int len)
377 {
378 	struct if_clone *ifc;
379 	char *dp;
380 	int wildcard, bytoff, bitoff;
381 	int unit;
382 	int err;
383 
384 	ifc = if_clone_lookup(name, &unit);
385 	if (ifc == NULL)
386 		return (EINVAL);
387 
388 	if (ifunit(name) != NULL)
389 		return (EEXIST);
390 
391 	bytoff = bitoff = 0;
392 	wildcard = (unit < 0);
393 	/*
394 	 * Find a free unit if none was given.
395 	 */
396 	if (wildcard) {
397 		while (bytoff < ifc->ifc_bmlen &&
398 		    ifc->ifc_units[bytoff] == 0xff)
399 			bytoff++;
400 		if (bytoff >= ifc->ifc_bmlen)
401 			return (ENOSPC);
402 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
403 			bitoff++;
404 		unit = (bytoff << 3) + bitoff;
405 	}
406 
407 	if (unit > ifc->ifc_maxunit)
408 		return (ENXIO);
409 
410 	err = (*ifc->ifc_create)(ifc, unit);
411 	if (err != 0)
412 		return (err);
413 
414 	if (!wildcard) {
415 		bytoff = unit >> 3;
416 		bitoff = unit - (bytoff << 3);
417 	}
418 
419 	/*
420 	 * Allocate the unit in the bitmap.
421 	 */
422 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
423 	    ("%s: bit is already set", __func__));
424 	ifc->ifc_units[bytoff] |= (1 << bitoff);
425 
426 	/* In the wildcard case, we need to update the name. */
427 	if (wildcard) {
428 		for (dp = name; *dp != '\0'; dp++);
429 		if (snprintf(dp, len - (dp-name), "%d", unit) >
430 		    len - (dp-name) - 1) {
431 			/*
432 			 * This can only be a programmer error and
433 			 * there's no straightforward way to recover if
434 			 * it happens.
435 			 */
436 			panic("if_clone_create(): interface name too long");
437 		}
438 
439 	}
440 
441 	EVENTHANDLER_INVOKE(if_clone_event, ifc);
442 
443 	return (0);
444 }
445 
446 /*
447  * Destroy a clone network interface.
448  */
449 int
450 if_clone_destroy(const char *name)
451 {
452 	struct if_clone *ifc;
453 	struct ifnet *ifp;
454 	int bytoff, bitoff;
455 	int unit;
456 
457 	ifc = if_clone_lookup(name, &unit);
458 	if (ifc == NULL)
459 		return (EINVAL);
460 
461 	if (unit < ifc->ifc_minifs)
462 		return (EINVAL);
463 
464 	ifp = ifunit(name);
465 	if (ifp == NULL)
466 		return (ENXIO);
467 
468 	if (ifc->ifc_destroy == NULL)
469 		return (EOPNOTSUPP);
470 
471 	(*ifc->ifc_destroy)(ifp);
472 
473 	/*
474 	 * Compute offset in the bitmap and deallocate the unit.
475 	 */
476 	bytoff = unit >> 3;
477 	bitoff = unit - (bytoff << 3);
478 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
479 	    ("%s: bit is already cleared", __func__));
480 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
481 	return (0);
482 }
483 
484 /*
485  * Look up a network interface cloner.
486  */
487 struct if_clone *
488 if_clone_lookup(const char *name, int *unitp)
489 {
490 	struct if_clone *ifc;
491 	const char *cp;
492 	int i;
493 
494 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
495 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
496 			if (ifc->ifc_name[i] != *cp)
497 				goto next_ifc;
498 		}
499 		goto found_name;
500  next_ifc:
501 		ifc = LIST_NEXT(ifc, ifc_list);
502 	}
503 
504 	/* No match. */
505 	return ((struct if_clone *)NULL);
506 
507  found_name:
508 	if (*cp == '\0') {
509 		i = -1;
510 	} else {
511 		for (i = 0; *cp != '\0'; cp++) {
512 			if (*cp < '0' || *cp > '9') {
513 				/* Bogus unit number. */
514 				return (NULL);
515 			}
516 			i = (i * 10) + (*cp - '0');
517 		}
518 	}
519 
520 	if (unitp != NULL)
521 		*unitp = i;
522 	return (ifc);
523 }
524 
525 /*
526  * Register a network interface cloner.
527  */
528 void
529 if_clone_attach(struct if_clone *ifc)
530 {
531 	int bytoff, bitoff;
532 	int err;
533 	int len, maxclone;
534 	int unit;
535 
536 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
537 	    ("%s: %s requested more units then allowed (%d > %d)",
538 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
539 	    ifc->ifc_maxunit + 1));
540 	/*
541 	 * Compute bitmap size and allocate it.
542 	 */
543 	maxclone = ifc->ifc_maxunit + 1;
544 	len = maxclone >> 3;
545 	if ((len << 3) < maxclone)
546 		len++;
547 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
548 	ifc->ifc_bmlen = len;
549 
550 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
551 	if_cloners_count++;
552 
553 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
554 		err = (*ifc->ifc_create)(ifc, unit);
555 		KASSERT(err == 0,
556 		    ("%s: failed to create required interface %s%d",
557 		    __func__, ifc->ifc_name, unit));
558 
559 		/* Allocate the unit in the bitmap. */
560 		bytoff = unit >> 3;
561 		bitoff = unit - (bytoff << 3);
562 		ifc->ifc_units[bytoff] |= (1 << bitoff);
563 	}
564 }
565 
566 /*
567  * Unregister a network interface cloner.
568  */
569 void
570 if_clone_detach(struct if_clone *ifc)
571 {
572 
573 	LIST_REMOVE(ifc, ifc_list);
574 	free(ifc->ifc_units, M_CLONE);
575 	if_cloners_count--;
576 }
577 
578 /*
579  * Provide list of interface cloners to userspace.
580  */
581 int
582 if_clone_list(struct if_clonereq *ifcr)
583 {
584 	char outbuf[IFNAMSIZ], *dst;
585 	struct if_clone *ifc;
586 	int count, error = 0;
587 
588 	ifcr->ifcr_total = if_cloners_count;
589 	if ((dst = ifcr->ifcr_buffer) == NULL) {
590 		/* Just asking how many there are. */
591 		return (0);
592 	}
593 
594 	if (ifcr->ifcr_count < 0)
595 		return (EINVAL);
596 
597 	count = (if_cloners_count < ifcr->ifcr_count) ?
598 	    if_cloners_count : ifcr->ifcr_count;
599 
600 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
601 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
602 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
603 		error = copyout(outbuf, dst, IFNAMSIZ);
604 		if (error)
605 			break;
606 	}
607 
608 	return (error);
609 }
610 
611 /*
612  * Locate an interface based on a complete address.
613  */
614 struct ifaddr *
615 ifa_ifwithaddr(struct sockaddr *addr)
616 {
617 	struct ifnet *ifp;
618 	struct ifaddr *ifa;
619 
620 	TAILQ_FOREACH(ifp, &ifnet, if_link)
621 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
622 		if (ifa->ifa_addr->sa_family != addr->sa_family)
623 			continue;
624 		if (sa_equal(addr, ifa->ifa_addr))
625 			return (ifa);
626 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
627 		    /* IPv6 doesn't have broadcast */
628 		    ifa->ifa_broadaddr->sa_len != 0 &&
629 		    sa_equal(ifa->ifa_broadaddr, addr))
630 			return (ifa);
631 	}
632 	return ((struct ifaddr *)NULL);
633 }
634 /*
635  * Locate the point to point interface with a given destination address.
636  */
637 struct ifaddr *
638 ifa_ifwithdstaddr(struct sockaddr *addr)
639 {
640 	struct ifnet *ifp;
641 	struct ifaddr *ifa;
642 
643 	TAILQ_FOREACH(ifp, &ifnet, if_link)
644 	    if (ifp->if_flags & IFF_POINTOPOINT)
645 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
646 			if (ifa->ifa_addr->sa_family != addr->sa_family)
647 				continue;
648 			if (ifa->ifa_dstaddr &&
649 			    sa_equal(addr, ifa->ifa_dstaddr))
650 				return (ifa);
651 	}
652 	return ((struct ifaddr *)NULL);
653 }
654 
655 /*
656  * Find an interface on a specific network.  If many, choice
657  * is most specific found.
658  */
659 struct ifaddr *
660 ifa_ifwithnet(struct sockaddr *addr)
661 {
662 	struct ifnet *ifp;
663 	struct ifaddr *ifa;
664 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
665 	u_int af = addr->sa_family;
666 	char *addr_data = addr->sa_data, *cplim;
667 
668 	/*
669 	 * AF_LINK addresses can be looked up directly by their index number,
670 	 * so do that if we can.
671 	 */
672 	if (af == AF_LINK) {
673 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
674 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
675 		return (ifnet_addrs[sdl->sdl_index - 1]);
676 	}
677 
678 	/*
679 	 * Scan though each interface, looking for ones that have
680 	 * addresses in this address family.
681 	 */
682 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
683 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
684 			char *cp, *cp2, *cp3;
685 
686 			if (ifa->ifa_addr->sa_family != af)
687 next:				continue;
688 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
689 				/*
690 				 * This is a bit broken as it doesn't
691 				 * take into account that the remote end may
692 				 * be a single node in the network we are
693 				 * looking for.
694 				 * The trouble is that we don't know the
695 				 * netmask for the remote end.
696 				 */
697 				if (ifa->ifa_dstaddr != NULL &&
698 				    sa_equal(addr, ifa->ifa_dstaddr))
699 					return (ifa);
700 			} else {
701 				/*
702 				 * if we have a special address handler,
703 				 * then use it instead of the generic one.
704 				 */
705 				if (ifa->ifa_claim_addr) {
706 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
707 						return (ifa);
708 					} else {
709 						continue;
710 					}
711 				}
712 
713 				/*
714 				 * Scan all the bits in the ifa's address.
715 				 * If a bit dissagrees with what we are
716 				 * looking for, mask it with the netmask
717 				 * to see if it really matters.
718 				 * (A byte at a time)
719 				 */
720 				if (ifa->ifa_netmask == 0)
721 					continue;
722 				cp = addr_data;
723 				cp2 = ifa->ifa_addr->sa_data;
724 				cp3 = ifa->ifa_netmask->sa_data;
725 				cplim = ifa->ifa_netmask->sa_len
726 					+ (char *)ifa->ifa_netmask;
727 				while (cp3 < cplim)
728 					if ((*cp++ ^ *cp2++) & *cp3++)
729 						goto next; /* next address! */
730 				/*
731 				 * If the netmask of what we just found
732 				 * is more specific than what we had before
733 				 * (if we had one) then remember the new one
734 				 * before continuing to search
735 				 * for an even better one.
736 				 */
737 				if (ifa_maybe == 0 ||
738 				    rn_refines((char *)ifa->ifa_netmask,
739 					       (char *)ifa_maybe->ifa_netmask))
740 					ifa_maybe = ifa;
741 			}
742 		}
743 	}
744 	return (ifa_maybe);
745 }
746 
747 /*
748  * Find an interface address specific to an interface best matching
749  * a given address.
750  */
751 struct ifaddr *
752 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
753 {
754 	struct ifaddr *ifa;
755 	char *cp, *cp2, *cp3;
756 	char *cplim;
757 	struct ifaddr *ifa_maybe = 0;
758 	u_int af = addr->sa_family;
759 
760 	if (af >= AF_MAX)
761 		return (0);
762 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
763 		if (ifa->ifa_addr->sa_family != af)
764 			continue;
765 		if (ifa_maybe == 0)
766 			ifa_maybe = ifa;
767 		if (ifa->ifa_netmask == NULL) {
768 			if (sa_equal(addr, ifa->ifa_addr) ||
769 			    (ifa->ifa_dstaddr != NULL &&
770 			     sa_equal(addr, ifa->ifa_dstaddr)))
771 				return (ifa);
772 			continue;
773 		}
774 		if (ifp->if_flags & IFF_POINTOPOINT) {
775 			if (sa_equal(addr, ifa->ifa_dstaddr))
776 				return (ifa);
777 		} else {
778 			cp = addr->sa_data;
779 			cp2 = ifa->ifa_addr->sa_data;
780 			cp3 = ifa->ifa_netmask->sa_data;
781 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
782 			for (; cp3 < cplim; cp3++)
783 				if ((*cp++ ^ *cp2++) & *cp3)
784 					break;
785 			if (cp3 == cplim)
786 				return (ifa);
787 		}
788 	}
789 	return (ifa_maybe);
790 }
791 
792 #include <net/route.h>
793 
794 /*
795  * Default action when installing a route with a Link Level gateway.
796  * Lookup an appropriate real ifa to point to.
797  * This should be moved to /sys/net/link.c eventually.
798  */
799 static void
800 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
801 {
802 	struct ifaddr *ifa;
803 	struct sockaddr *dst;
804 	struct ifnet *ifp;
805 
806 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
807 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
808 		return;
809 	ifa = ifaof_ifpforaddr(dst, ifp);
810 	if (ifa != NULL) {
811 		IFAFREE(rt->rt_ifa);
812 		IFAREF(ifa);
813 		rt->rt_ifa = ifa;
814 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
815 			ifa->ifa_rtrequest(cmd, rt, info);
816 	}
817 }
818 
819 /*
820  * Mark an interface down and notify protocols of
821  * the transition.
822  * NOTE: must be called at splnet or eqivalent.
823  */
824 void
825 if_unroute(struct ifnet *ifp, int flag, int fam)
826 {
827 	struct ifaddr *ifa;
828 
829 	ifp->if_flags &= ~flag;
830 	getmicrotime(&ifp->if_lastchange);
831 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
832 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
833 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
834 	if_qflush(&ifp->if_snd);
835 	rt_ifmsg(ifp);
836 }
837 
838 /*
839  * Mark an interface up and notify protocols of
840  * the transition.
841  * NOTE: must be called at splnet or eqivalent.
842  */
843 void
844 if_route(struct ifnet *ifp, int flag, int fam)
845 {
846 	struct ifaddr *ifa;
847 
848 	ifp->if_flags |= flag;
849 	getmicrotime(&ifp->if_lastchange);
850 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
851 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
852 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
853 	rt_ifmsg(ifp);
854 #ifdef INET6
855 	in6_if_up(ifp);
856 #endif
857 }
858 
859 /*
860  * Mark an interface down and notify protocols of
861  * the transition.
862  * NOTE: must be called at splnet or eqivalent.
863  */
864 void
865 if_down(struct ifnet *ifp)
866 {
867 
868 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
869 }
870 
871 /*
872  * Mark an interface up and notify protocols of
873  * the transition.
874  * NOTE: must be called at splnet or eqivalent.
875  */
876 void
877 if_up(struct ifnet *ifp)
878 {
879 
880 	if_route(ifp, IFF_UP, AF_UNSPEC);
881 }
882 
883 /*
884  * Flush an interface queue.
885  */
886 static void
887 if_qflush(struct ifqueue *ifq)
888 {
889 	struct mbuf *m, *n;
890 
891 	n = ifq->ifq_head;
892 	while ((m = n) != 0) {
893 		n = m->m_nextpkt;
894 		m_freem(m);
895 	}
896 	ifq->ifq_head = 0;
897 	ifq->ifq_tail = 0;
898 	ifq->ifq_len = 0;
899 }
900 
901 /*
902  * Handle interface watchdog timer routines.  Called
903  * from softclock, we decrement timers (if set) and
904  * call the appropriate interface routine on expiration.
905  */
906 static void
907 if_slowtimo(void *arg)
908 {
909 	struct ifnet *ifp;
910 	int s = splimp();
911 
912 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
913 		if (ifp->if_timer == 0 || --ifp->if_timer)
914 			continue;
915 		if (ifp->if_watchdog)
916 			(*ifp->if_watchdog)(ifp);
917 	}
918 	splx(s);
919 	callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
920 }
921 
922 /*
923  * Map interface name to
924  * interface structure pointer.
925  */
926 struct ifnet *
927 ifunit(const char *name)
928 {
929 	struct ifnet *ifp;
930 
931 	/*
932 	 * Search all the interfaces for this name/number
933 	 */
934 
935 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
936 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
937 			break;
938 	}
939 	return (ifp);
940 }
941 
942 
943 /*
944  * Map interface name in a sockaddr_dl to
945  * interface structure pointer.
946  */
947 struct ifnet *
948 if_withname(struct sockaddr *sa)
949 {
950 	char ifname[IFNAMSIZ+1];
951 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
952 
953 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
954 	     (sdl->sdl_nlen > IFNAMSIZ) )
955 		return NULL;
956 
957 	/*
958 	 * ifunit wants a null-terminated name.  It may not be null-terminated
959 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
960 	 * and there might not be room to put the trailing null anyway, so we
961 	 * make a local copy that we know we can null terminate safely.
962 	 */
963 
964 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
965 	ifname[sdl->sdl_nlen] = '\0';
966 	return ifunit(ifname);
967 }
968 
969 
970 /*
971  * Interface ioctls.
972  */
973 int
974 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
975 {
976 	struct ifnet *ifp;
977 	struct ifreq *ifr;
978 	struct ifstat *ifs;
979 	int error;
980 	short oif_flags;
981 	int new_flags;
982 	size_t namelen, onamelen;
983 	char new_name[IFNAMSIZ];
984 	struct ifaddr *ifa;
985 	struct sockaddr_dl *sdl;
986 
987 	switch (cmd) {
988 
989 	case SIOCGIFCONF:
990 	case OSIOCGIFCONF:
991 		return (ifconf(cmd, data, td));
992 	}
993 	ifr = (struct ifreq *)data;
994 
995 	switch (cmd) {
996 	case SIOCIFCREATE:
997 	case SIOCIFDESTROY:
998 		if ((error = suser(td)) != 0)
999 			return (error);
1000 		return ((cmd == SIOCIFCREATE) ?
1001 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1002 			if_clone_destroy(ifr->ifr_name));
1003 
1004 	case SIOCIFGCLONERS:
1005 		return (if_clone_list((struct if_clonereq *)data));
1006 	}
1007 
1008 	ifp = ifunit(ifr->ifr_name);
1009 	if (ifp == 0)
1010 		return (ENXIO);
1011 	switch (cmd) {
1012 
1013 	case SIOCGIFFLAGS:
1014 		ifr->ifr_flags = ifp->if_flags;
1015 		ifr->ifr_flagshigh = ifp->if_flags >> 16;
1016 		break;
1017 
1018 	case SIOCGIFCAP:
1019 		ifr->ifr_reqcap = ifp->if_capabilities;
1020 		ifr->ifr_curcap = ifp->if_capenable;
1021 		break;
1022 
1023 	case SIOCGIFMETRIC:
1024 		ifr->ifr_metric = ifp->if_metric;
1025 		break;
1026 
1027 	case SIOCGIFMTU:
1028 		ifr->ifr_mtu = ifp->if_mtu;
1029 		break;
1030 
1031 	case SIOCGIFPHYS:
1032 		ifr->ifr_phys = ifp->if_physical;
1033 		break;
1034 
1035 	case SIOCSIFFLAGS:
1036 		error = suser(td);
1037 		if (error)
1038 			return (error);
1039 		new_flags = (ifr->ifr_flags & 0xffff) |
1040 		    (ifr->ifr_flagshigh << 16);
1041 		if (ifp->if_flags & IFF_SMART) {
1042 			/* Smart drivers twiddle their own routes */
1043 		} else if (ifp->if_flags & IFF_UP &&
1044 		    (new_flags & IFF_UP) == 0) {
1045 			int s = splimp();
1046 			if_down(ifp);
1047 			splx(s);
1048 		} else if (new_flags & IFF_UP &&
1049 		    (ifp->if_flags & IFF_UP) == 0) {
1050 			int s = splimp();
1051 			if_up(ifp);
1052 			splx(s);
1053 		}
1054 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1055 			(new_flags &~ IFF_CANTCHANGE);
1056 		if (new_flags & IFF_PPROMISC) {
1057 			/* Permanently promiscuous mode requested */
1058 			ifp->if_flags |= IFF_PROMISC;
1059 		} else if (ifp->if_pcount == 0) {
1060 			ifp->if_flags &= ~IFF_PROMISC;
1061 		}
1062 		if (ifp->if_ioctl)
1063 			(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1064 		getmicrotime(&ifp->if_lastchange);
1065 		break;
1066 
1067 	case SIOCSIFCAP:
1068 		error = suser(td);
1069 		if (error)
1070 			return (error);
1071 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1072 			return (EINVAL);
1073 		(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1074 		break;
1075 
1076 	case SIOCSIFNAME:
1077 		error = suser(td);
1078 		if (error != 0)
1079 			return (error);
1080 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1081 		if (error != 0)
1082 			return (error);
1083 		if (new_name[0] == '\0')
1084 			return (EINVAL);
1085 		if (ifunit(new_name) != NULL)
1086 			return (EEXIST);
1087 
1088 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1089 
1090 		/* Announce the departure of the interface. */
1091 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1092 
1093 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1094 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1095 		/* XXX IFA_LOCK(ifa); */
1096 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1097 		namelen = strlen(new_name);
1098 		onamelen = sdl->sdl_nlen;
1099 		/*
1100 		 * Move the address if needed.  This is safe because we
1101 		 * allocate space for a name of length IFNAMSIZ when we
1102 		 * create this in if_attach().
1103 		 */
1104 		if (namelen != onamelen) {
1105 			bcopy(sdl->sdl_data + onamelen,
1106 			    sdl->sdl_data + namelen, sdl->sdl_alen);
1107 		}
1108 		bcopy(new_name, sdl->sdl_data, namelen);
1109 		sdl->sdl_nlen = namelen;
1110 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1111 		bzero(sdl->sdl_data, onamelen);
1112 		while (namelen != 0)
1113 			sdl->sdl_data[--namelen] = 0xff;
1114 		/* XXX IFA_UNLOCK(ifa) */
1115 
1116 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1117 
1118 		/* Announce the return of the interface. */
1119 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1120 		break;
1121 
1122 	case SIOCSIFMETRIC:
1123 		error = suser(td);
1124 		if (error)
1125 			return (error);
1126 		ifp->if_metric = ifr->ifr_metric;
1127 		getmicrotime(&ifp->if_lastchange);
1128 		break;
1129 
1130 	case SIOCSIFPHYS:
1131 		error = suser(td);
1132 		if (error)
1133 			return error;
1134 		if (!ifp->if_ioctl)
1135 		        return EOPNOTSUPP;
1136 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1137 		if (error == 0)
1138 			getmicrotime(&ifp->if_lastchange);
1139 		return (error);
1140 
1141 	case SIOCSIFMTU:
1142 	{
1143 		u_long oldmtu = ifp->if_mtu;
1144 
1145 		error = suser(td);
1146 		if (error)
1147 			return (error);
1148 		if (ifp->if_ioctl == NULL)
1149 			return (EOPNOTSUPP);
1150 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1151 			return (EINVAL);
1152 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1153 		if (error == 0) {
1154 			getmicrotime(&ifp->if_lastchange);
1155 			rt_ifmsg(ifp);
1156 		}
1157 		/*
1158 		 * If the link MTU changed, do network layer specific procedure.
1159 		 */
1160 		if (ifp->if_mtu != oldmtu) {
1161 #ifdef INET6
1162 			nd6_setmtu(ifp);
1163 #endif
1164 		}
1165 		return (error);
1166 	}
1167 
1168 	case SIOCADDMULTI:
1169 	case SIOCDELMULTI:
1170 		error = suser(td);
1171 		if (error)
1172 			return (error);
1173 
1174 		/* Don't allow group membership on non-multicast interfaces. */
1175 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1176 			return EOPNOTSUPP;
1177 
1178 		/* Don't let users screw up protocols' entries. */
1179 		if (ifr->ifr_addr.sa_family != AF_LINK)
1180 			return EINVAL;
1181 
1182 		if (cmd == SIOCADDMULTI) {
1183 			struct ifmultiaddr *ifma;
1184 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1185 		} else {
1186 			error = if_delmulti(ifp, &ifr->ifr_addr);
1187 		}
1188 		if (error == 0)
1189 			getmicrotime(&ifp->if_lastchange);
1190 		return error;
1191 
1192 	case SIOCSIFPHYADDR:
1193 	case SIOCDIFPHYADDR:
1194 #ifdef INET6
1195 	case SIOCSIFPHYADDR_IN6:
1196 #endif
1197 	case SIOCSLIFPHYADDR:
1198         case SIOCSIFMEDIA:
1199 	case SIOCSIFGENERIC:
1200 		error = suser(td);
1201 		if (error)
1202 			return (error);
1203 		if (ifp->if_ioctl == 0)
1204 			return (EOPNOTSUPP);
1205 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1206 		if (error == 0)
1207 			getmicrotime(&ifp->if_lastchange);
1208 		return error;
1209 
1210 	case SIOCGIFSTATUS:
1211 		ifs = (struct ifstat *)data;
1212 		ifs->ascii[0] = '\0';
1213 
1214 	case SIOCGIFPSRCADDR:
1215 	case SIOCGIFPDSTADDR:
1216 	case SIOCGLIFPHYADDR:
1217 	case SIOCGIFMEDIA:
1218 	case SIOCGIFGENERIC:
1219 		if (ifp->if_ioctl == 0)
1220 			return (EOPNOTSUPP);
1221 		return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1222 
1223 	case SIOCSIFLLADDR:
1224 		error = suser(td);
1225 		if (error)
1226 			return (error);
1227 		return if_setlladdr(ifp,
1228 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1229 
1230 	default:
1231 		oif_flags = ifp->if_flags;
1232 		if (so->so_proto == 0)
1233 			return (EOPNOTSUPP);
1234 #ifndef COMPAT_43
1235 		error = so_pru_control(so, cmd, data, ifp, td);
1236 #else
1237 	    {
1238 		int ocmd = cmd;
1239 
1240 		switch (cmd) {
1241 
1242 		case SIOCSIFDSTADDR:
1243 		case SIOCSIFADDR:
1244 		case SIOCSIFBRDADDR:
1245 		case SIOCSIFNETMASK:
1246 #if BYTE_ORDER != BIG_ENDIAN
1247 			if (ifr->ifr_addr.sa_family == 0 &&
1248 			    ifr->ifr_addr.sa_len < 16) {
1249 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1250 				ifr->ifr_addr.sa_len = 16;
1251 			}
1252 #else
1253 			if (ifr->ifr_addr.sa_len == 0)
1254 				ifr->ifr_addr.sa_len = 16;
1255 #endif
1256 			break;
1257 
1258 		case OSIOCGIFADDR:
1259 			cmd = SIOCGIFADDR;
1260 			break;
1261 
1262 		case OSIOCGIFDSTADDR:
1263 			cmd = SIOCGIFDSTADDR;
1264 			break;
1265 
1266 		case OSIOCGIFBRDADDR:
1267 			cmd = SIOCGIFBRDADDR;
1268 			break;
1269 
1270 		case OSIOCGIFNETMASK:
1271 			cmd = SIOCGIFNETMASK;
1272 		}
1273 		error =  so_pru_control(so, cmd, data, ifp, td);
1274 		switch (ocmd) {
1275 
1276 		case OSIOCGIFADDR:
1277 		case OSIOCGIFDSTADDR:
1278 		case OSIOCGIFBRDADDR:
1279 		case OSIOCGIFNETMASK:
1280 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1281 
1282 		}
1283 	    }
1284 #endif /* COMPAT_43 */
1285 
1286 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1287 #ifdef INET6
1288 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1289 			if (ifp->if_flags & IFF_UP) {
1290 				int s = splimp();
1291 				in6_if_up(ifp);
1292 				splx(s);
1293 			}
1294 #endif
1295 		}
1296 		return (error);
1297 
1298 	}
1299 	return (0);
1300 }
1301 
1302 /*
1303  * Set/clear promiscuous mode on interface ifp based on the truth value
1304  * of pswitch.  The calls are reference counted so that only the first
1305  * "on" request actually has an effect, as does the final "off" request.
1306  * Results are undefined if the "off" and "on" requests are not matched.
1307  */
1308 int
1309 ifpromisc(struct ifnet *ifp, int pswitch)
1310 {
1311 	struct ifreq ifr;
1312 	int error;
1313 	int oldflags;
1314 
1315 	oldflags = ifp->if_flags;
1316 	if (ifp->if_flags & IFF_PPROMISC) {
1317 		/* Do nothing if device is in permanently promiscuous mode */
1318 		ifp->if_pcount += pswitch ? 1 : -1;
1319 		return (0);
1320 	}
1321 	if (pswitch) {
1322 		/*
1323 		 * If the device is not configured up, we cannot put it in
1324 		 * promiscuous mode.
1325 		 */
1326 		if ((ifp->if_flags & IFF_UP) == 0)
1327 			return (ENETDOWN);
1328 		if (ifp->if_pcount++ != 0)
1329 			return (0);
1330 		ifp->if_flags |= IFF_PROMISC;
1331 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1332 		    ifp->if_xname);
1333 	} else {
1334 		if (--ifp->if_pcount > 0)
1335 			return (0);
1336 		ifp->if_flags &= ~IFF_PROMISC;
1337 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1338 		    ifp->if_xname);
1339 	}
1340 	ifr.ifr_flags = ifp->if_flags;
1341 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
1342 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1343 				 (struct ucred *)NULL);
1344 	if (error == 0)
1345 		rt_ifmsg(ifp);
1346 	else
1347 		ifp->if_flags = oldflags;
1348 	return error;
1349 }
1350 
1351 /*
1352  * Return interface configuration
1353  * of system.  List may be used
1354  * in later ioctl's (above) to get
1355  * other information.
1356  */
1357 static int
1358 ifconf(u_long cmd, caddr_t data, struct thread *td)
1359 {
1360 	struct ifconf *ifc = (struct ifconf *)data;
1361 	struct ifnet *ifp;
1362 	struct ifaddr *ifa;
1363 	struct sockaddr *sa;
1364 	struct ifreq ifr, *ifrp;
1365 	int space = ifc->ifc_len, error = 0;
1366 
1367 	ifrp = ifc->ifc_req;
1368 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1369 		int addrs;
1370 
1371 		if (space <= sizeof ifr)
1372 			break;
1373 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1374 		    >= sizeof(ifr.ifr_name)) {
1375 			error = ENAMETOOLONG;
1376 			break;
1377 		}
1378 
1379 		addrs = 0;
1380 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1381 			if (space <= sizeof ifr)
1382 				break;
1383 			sa = ifa->ifa_addr;
1384 			if (td->td_proc->p_ucred->cr_prison &&
1385 			    prison_if(td, sa))
1386 				continue;
1387 			addrs++;
1388 #ifdef COMPAT_43
1389 			if (cmd == OSIOCGIFCONF) {
1390 				struct osockaddr *osa =
1391 					 (struct osockaddr *)&ifr.ifr_addr;
1392 				ifr.ifr_addr = *sa;
1393 				osa->sa_family = sa->sa_family;
1394 				error = copyout(&ifr, ifrp, sizeof ifr);
1395 				ifrp++;
1396 			} else
1397 #endif
1398 			if (sa->sa_len <= sizeof(*sa)) {
1399 				ifr.ifr_addr = *sa;
1400 				error = copyout(&ifr, ifrp, sizeof ifr);
1401 				ifrp++;
1402 			} else {
1403 				if (space < (sizeof ifr) + sa->sa_len -
1404 					    sizeof(*sa))
1405 					break;
1406 				space -= sa->sa_len - sizeof(*sa);
1407 				error = copyout(&ifr, ifrp,
1408 						sizeof ifr.ifr_name);
1409 				if (error == 0)
1410 					error = copyout(sa, &ifrp->ifr_addr,
1411 							sa->sa_len);
1412 				ifrp = (struct ifreq *)
1413 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1414 			}
1415 			if (error)
1416 				break;
1417 			space -= sizeof ifr;
1418 		}
1419 		if (error)
1420 			break;
1421 		if (!addrs) {
1422 			bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1423 			error = copyout(&ifr, ifrp, sizeof ifr);
1424 			if (error)
1425 				break;
1426 			space -= sizeof ifr;
1427 			ifrp++;
1428 		}
1429 	}
1430 	ifc->ifc_len -= space;
1431 	return (error);
1432 }
1433 
1434 /*
1435  * Just like if_promisc(), but for all-multicast-reception mode.
1436  */
1437 int
1438 if_allmulti(struct ifnet *ifp, int onswitch)
1439 {
1440 	int error = 0;
1441 	int s = splimp();
1442 	struct ifreq ifr;
1443 
1444 	if (onswitch) {
1445 		if (ifp->if_amcount++ == 0) {
1446 			ifp->if_flags |= IFF_ALLMULTI;
1447 			ifr.ifr_flags = ifp->if_flags;
1448 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1449 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1450 					      (struct ucred *)NULL);
1451 		}
1452 	} else {
1453 		if (ifp->if_amcount > 1) {
1454 			ifp->if_amcount--;
1455 		} else {
1456 			ifp->if_amcount = 0;
1457 			ifp->if_flags &= ~IFF_ALLMULTI;
1458 			ifr.ifr_flags = ifp->if_flags;
1459 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1460 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1461 					      (struct ucred *)NULL);
1462 		}
1463 	}
1464 	splx(s);
1465 
1466 	if (error == 0)
1467 		rt_ifmsg(ifp);
1468 	return error;
1469 }
1470 
1471 /*
1472  * Add a multicast listenership to the interface in question.
1473  * The link layer provides a routine which converts
1474  */
1475 int
1476 if_addmulti(
1477 	struct ifnet *ifp,	/* interface to manipulate */
1478 	struct sockaddr *sa,	/* address to add */
1479 	struct ifmultiaddr **retifma)
1480 {
1481 	struct sockaddr *llsa, *dupsa;
1482 	int error, s;
1483 	struct ifmultiaddr *ifma;
1484 
1485 	/*
1486 	 * If the matching multicast address already exists
1487 	 * then don't add a new one, just add a reference
1488 	 */
1489 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1490 		if (sa_equal(sa, ifma->ifma_addr)) {
1491 			ifma->ifma_refcount++;
1492 			if (retifma)
1493 				*retifma = ifma;
1494 			return 0;
1495 		}
1496 	}
1497 
1498 	/*
1499 	 * Give the link layer a chance to accept/reject it, and also
1500 	 * find out which AF_LINK address this maps to, if it isn't one
1501 	 * already.
1502 	 */
1503 	if (ifp->if_resolvemulti) {
1504 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1505 		if (error) return error;
1506 	} else {
1507 		llsa = 0;
1508 	}
1509 
1510 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1511 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1512 	bcopy(sa, dupsa, sa->sa_len);
1513 
1514 	ifma->ifma_addr = dupsa;
1515 	ifma->ifma_lladdr = llsa;
1516 	ifma->ifma_ifp = ifp;
1517 	ifma->ifma_refcount = 1;
1518 	ifma->ifma_protospec = 0;
1519 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1520 
1521 	/*
1522 	 * Some network interfaces can scan the address list at
1523 	 * interrupt time; lock them out.
1524 	 */
1525 	s = splimp();
1526 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1527 	splx(s);
1528 	*retifma = ifma;
1529 
1530 	if (llsa != 0) {
1531 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1532 			if (sa_equal(ifma->ifma_addr, llsa))
1533 				break;
1534 		}
1535 		if (ifma) {
1536 			ifma->ifma_refcount++;
1537 		} else {
1538 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1539 			       M_IFMADDR, M_WAITOK);
1540 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1541 			       M_IFMADDR, M_WAITOK);
1542 			bcopy(llsa, dupsa, llsa->sa_len);
1543 			ifma->ifma_addr = dupsa;
1544 			ifma->ifma_ifp = ifp;
1545 			ifma->ifma_refcount = 1;
1546 			s = splimp();
1547 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1548 			splx(s);
1549 		}
1550 	}
1551 	/*
1552 	 * We are certain we have added something, so call down to the
1553 	 * interface to let them know about it.
1554 	 */
1555 	s = splimp();
1556 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1557 	splx(s);
1558 
1559 	return 0;
1560 }
1561 
1562 /*
1563  * Remove a reference to a multicast address on this interface.  Yell
1564  * if the request does not match an existing membership.
1565  */
1566 int
1567 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1568 {
1569 	struct ifmultiaddr *ifma;
1570 	int s;
1571 
1572 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1573 		if (sa_equal(sa, ifma->ifma_addr))
1574 			break;
1575 	if (ifma == 0)
1576 		return ENOENT;
1577 
1578 	if (ifma->ifma_refcount > 1) {
1579 		ifma->ifma_refcount--;
1580 		return 0;
1581 	}
1582 
1583 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1584 	sa = ifma->ifma_lladdr;
1585 	s = splimp();
1586 	LIST_REMOVE(ifma, ifma_link);
1587 	/*
1588 	 * Make sure the interface driver is notified
1589 	 * in the case of a link layer mcast group being left.
1590 	 */
1591 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1592 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1593 	splx(s);
1594 	free(ifma->ifma_addr, M_IFMADDR);
1595 	free(ifma, M_IFMADDR);
1596 	if (sa == 0)
1597 		return 0;
1598 
1599 	/*
1600 	 * Now look for the link-layer address which corresponds to
1601 	 * this network address.  It had been squirreled away in
1602 	 * ifma->ifma_lladdr for this purpose (so we don't have
1603 	 * to call ifp->if_resolvemulti() again), and we saved that
1604 	 * value in sa above.  If some nasty deleted the
1605 	 * link-layer address out from underneath us, we can deal because
1606 	 * the address we stored was is not the same as the one which was
1607 	 * in the record for the link-layer address.  (So we don't complain
1608 	 * in that case.)
1609 	 */
1610 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1611 		if (sa_equal(sa, ifma->ifma_addr))
1612 			break;
1613 	if (ifma == 0)
1614 		return 0;
1615 
1616 	if (ifma->ifma_refcount > 1) {
1617 		ifma->ifma_refcount--;
1618 		return 0;
1619 	}
1620 
1621 	s = splimp();
1622 	LIST_REMOVE(ifma, ifma_link);
1623 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1624 	splx(s);
1625 	free(ifma->ifma_addr, M_IFMADDR);
1626 	free(sa, M_IFMADDR);
1627 	free(ifma, M_IFMADDR);
1628 
1629 	return 0;
1630 }
1631 
1632 /*
1633  * Set the link layer address on an interface.
1634  *
1635  * At this time we only support certain types of interfaces,
1636  * and we don't allow the length of the address to change.
1637  */
1638 int
1639 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1640 {
1641 	struct sockaddr_dl *sdl;
1642 	struct ifaddr *ifa;
1643 	struct ifreq ifr;
1644 
1645 	ifa = ifnet_addrs[ifp->if_index - 1];
1646 	if (ifa == NULL)
1647 		return (EINVAL);
1648 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1649 	if (sdl == NULL)
1650 		return (EINVAL);
1651 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1652 		return (EINVAL);
1653 	switch (ifp->if_type) {
1654 	case IFT_ETHER:			/* these types use struct arpcom */
1655 	case IFT_FDDI:
1656 	case IFT_XETHER:
1657 	case IFT_ISO88025:
1658 	case IFT_L2VLAN:
1659 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1660 		/* FALLTHROUGH */
1661 	case IFT_ARCNET:
1662 		bcopy(lladdr, LLADDR(sdl), len);
1663 		break;
1664 	default:
1665 		return (ENODEV);
1666 	}
1667 	/*
1668 	 * If the interface is already up, we need
1669 	 * to re-init it in order to reprogram its
1670 	 * address filter.
1671 	 */
1672 	if ((ifp->if_flags & IFF_UP) != 0) {
1673 		ifp->if_flags &= ~IFF_UP;
1674 		ifr.ifr_flags = ifp->if_flags;
1675 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1676 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1677 				 (struct ucred *)NULL);
1678 		ifp->if_flags |= IFF_UP;
1679 		ifr.ifr_flags = ifp->if_flags;
1680 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1681 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1682 				 (struct ucred *)NULL);
1683 #ifdef INET
1684 		/*
1685 		 * Also send gratuitous ARPs to notify other nodes about
1686 		 * the address change.
1687 		 */
1688 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1689 			if (ifa->ifa_addr != NULL &&
1690 			    ifa->ifa_addr->sa_family == AF_INET)
1691 				arp_ifinit(ifp, ifa);
1692 		}
1693 #endif
1694 	}
1695 	return (0);
1696 }
1697 
1698 struct ifmultiaddr *
1699 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1700 {
1701 	struct ifmultiaddr *ifma;
1702 
1703 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1704 		if (sa_equal(ifma->ifma_addr, sa))
1705 			break;
1706 
1707 	return ifma;
1708 }
1709 
1710 /*
1711  * The name argument must be a pointer to storage which will last as
1712  * long as the interface does.  For physical devices, the result of
1713  * device_get_name(dev) is a good choice and for pseudo-devices a
1714  * static string works well.
1715  */
1716 void
1717 if_initname(struct ifnet *ifp, const char *name, int unit)
1718 {
1719 	ifp->if_dname = name;
1720 	ifp->if_dunit = unit;
1721 	if (unit != IF_DUNIT_NONE)
1722 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1723 	else
1724 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1725 }
1726 
1727 int
1728 if_printf(struct ifnet *ifp, const char *fmt, ...)
1729 {
1730 	__va_list ap;
1731 	int retval;
1732 
1733 	retval = printf("%s: ", ifp->if_xname);
1734 	__va_start(ap, fmt);
1735 	retval += vprintf(fmt, ap);
1736 	__va_end(ap);
1737 	return (retval);
1738 }
1739 
1740 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1741 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1742