xref: /dflybsd-src/sys/net/if.c (revision b5b0912b1891e95ccc48cad83f09239ccb7ffc16)
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 1.39 2005/06/05 12:35:24 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 #include <sys/domain.h>
57 #include <sys/thread.h>
58 
59 #include <net/if.h>
60 #include <net/if_arp.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/if_var.h>
64 #include <net/ifq_var.h>
65 #include <net/radix.h>
66 #include <net/route.h>
67 #include <machine/stdarg.h>
68 
69 #include <sys/thread2.h>
70 
71 #if defined(INET) || defined(INET6)
72 /*XXX*/
73 #include <netinet/in.h>
74 #include <netinet/in_var.h>
75 #include <netinet/if_ether.h>
76 #ifdef INET6
77 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
78 #include <netinet6/in6_var.h>
79 #include <netinet6/in6_ifattach.h>
80 #endif
81 #endif
82 
83 #if defined(COMPAT_43)
84 #include <emulation/43bsd/43bsd_socket.h>
85 #endif /* COMPAT_43 */
86 
87 /*
88  * Support for non-ALTQ interfaces.
89  */
90 static int	ifq_classic_enqueue(struct ifaltq *, struct mbuf *,
91 				    struct altq_pktattr *);
92 static struct mbuf *
93 		ifq_classic_dequeue(struct ifaltq *, int);
94 static int	ifq_classic_request(struct ifaltq *, int, void *);
95 
96 /*
97  * System initialization
98  */
99 
100 static void	if_attachdomain(void *);
101 static void	if_attachdomain1(struct ifnet *);
102 static int ifconf (u_long, caddr_t, struct thread *);
103 static void ifinit (void *);
104 static void if_slowtimo (void *);
105 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
106 static int  if_rtdel (struct radix_node *, void *);
107 
108 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
109 
110 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
111 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
112 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
113 
114 int	ifqmaxlen = IFQ_MAXLEN;
115 struct	ifnethead ifnet;	/* depend on static init XXX */
116 
117 #ifdef INET6
118 /*
119  * XXX: declare here to avoid to include many inet6 related files..
120  * should be more generalized?
121  */
122 extern void	nd6_setmtu (struct ifnet *);
123 #endif
124 
125 struct if_clone *if_clone_lookup (const char *, int *);
126 int if_clone_list (struct if_clonereq *);
127 
128 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
129 int if_cloners_count;
130 
131 struct callout if_slowtimo_timer;
132 
133 /*
134  * Network interface utility routines.
135  *
136  * Routines with ifa_ifwith* names take sockaddr *'s as
137  * parameters.
138  */
139 /* ARGSUSED*/
140 void
141 ifinit(void *dummy)
142 {
143 	struct ifnet *ifp;
144 	int s;
145 
146 	callout_init(&if_slowtimo_timer);
147 
148 	s = splimp();
149 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
150 		if (ifp->if_snd.ifq_maxlen == 0) {
151 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
152 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
153 		}
154 	}
155 	splx(s);
156 
157 	if_slowtimo(0);
158 }
159 
160 int if_index = 0;
161 struct ifnet **ifindex2ifnet = NULL;
162 
163 /*
164  * Attach an interface to the
165  * list of "active" interfaces.
166  */
167 void
168 if_attach(struct ifnet *ifp)
169 {
170 	unsigned socksize, ifasize;
171 	int namelen, masklen;
172 	struct sockaddr_dl *sdl;
173 	struct ifaddr *ifa;
174 	struct ifaltq *ifq;
175 
176 	static int if_indexlim = 8;
177 	static boolean_t inited;
178 
179 	if (!inited) {
180 		TAILQ_INIT(&ifnet);
181 		inited = TRUE;
182 	}
183 
184 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
185 	ifp->if_index = ++if_index;
186 	/*
187 	 * XXX -
188 	 * The old code would work if the interface passed a pre-existing
189 	 * chain of ifaddrs to this code.  We don't trust our callers to
190 	 * properly initialize the tailq, however, so we no longer allow
191 	 * this unlikely case.
192 	 */
193 	TAILQ_INIT(&ifp->if_addrhead);
194 	TAILQ_INIT(&ifp->if_prefixhead);
195 	LIST_INIT(&ifp->if_multiaddrs);
196 	getmicrotime(&ifp->if_lastchange);
197 	if (ifindex2ifnet == NULL || if_index >= if_indexlim) {
198 		unsigned int n;
199 		struct ifnet **q;
200 
201 		if_indexlim <<= 1;
202 
203 		/* grow ifindex2ifnet */
204 		n = if_indexlim * sizeof(*q);
205 		q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
206 		if (ifindex2ifnet) {
207 			bcopy(ifindex2ifnet, q, n/2);
208 			free(ifindex2ifnet, M_IFADDR);
209 		}
210 		ifindex2ifnet = q;
211 	}
212 
213 	ifindex2ifnet[if_index] = ifp;
214 
215 	/*
216 	 * create a Link Level name for this device
217 	 */
218 	namelen = strlen(ifp->if_xname);
219 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
220 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
221 	socksize = masklen + ifp->if_addrlen;
222 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
223 	if (socksize < sizeof(*sdl))
224 		socksize = sizeof(*sdl);
225 	socksize = ROUNDUP(socksize);
226 	ifasize = sizeof(struct ifaddr) + 2 * socksize;
227 	ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
228 	sdl = (struct sockaddr_dl *)(ifa + 1);
229 	sdl->sdl_len = socksize;
230 	sdl->sdl_family = AF_LINK;
231 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
232 	sdl->sdl_nlen = namelen;
233 	sdl->sdl_index = ifp->if_index;
234 	sdl->sdl_type = ifp->if_type;
235 	ifp->if_lladdr = ifa;
236 	ifa->ifa_ifp = ifp;
237 	ifa->ifa_rtrequest = link_rtrequest;
238 	ifa->ifa_addr = (struct sockaddr *)sdl;
239 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
240 	ifa->ifa_netmask = (struct sockaddr *)sdl;
241 	sdl->sdl_len = masklen;
242 	while (namelen != 0)
243 		sdl->sdl_data[--namelen] = 0xff;
244 	TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
245 
246 	EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
247 
248 	ifq = &ifp->if_snd;
249 	ifq->altq_type = 0;
250 	ifq->altq_disc = NULL;
251 	ifq->altq_flags &= ALTQF_CANTCHANGE;
252 	ifq->altq_tbr = NULL;
253 	ifq->altq_ifp = ifp;
254 	ifq_set_classic(ifq);
255 
256 	if (!SLIST_EMPTY(&domains))
257 		if_attachdomain1(ifp);
258 
259 	/* Announce the interface. */
260 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
261 }
262 
263 static void
264 if_attachdomain(void *dummy)
265 {
266 	struct ifnet *ifp;
267 	int s;
268 
269 	s = splnet();
270 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
271 		if_attachdomain1(ifp);
272 	splx(s);
273 }
274 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
275 	if_attachdomain, NULL);
276 
277 static void
278 if_attachdomain1(struct ifnet *ifp)
279 {
280 	struct domain *dp;
281 	int s;
282 
283 	s = splnet();
284 
285 	/* address family dependent data region */
286 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
287 	SLIST_FOREACH(dp, &domains, dom_next)
288 		if (dp->dom_ifattach)
289 			ifp->if_afdata[dp->dom_family] =
290 				(*dp->dom_ifattach)(ifp);
291 	splx(s);
292 }
293 
294 /*
295  * Detach an interface, removing it from the
296  * list of "active" interfaces.
297  */
298 void
299 if_detach(struct ifnet *ifp)
300 {
301 	struct ifaddr *ifa;
302 	struct radix_node_head	*rnh;
303 	int s;
304 	int i;
305 	struct domain *dp;
306 
307 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
308 
309 	/*
310 	 * Remove routes and flush queues.
311 	 */
312 	s = splnet();
313 #ifdef DEVICE_POLLING
314 	if (ifp->if_flags & IFF_POLLING)
315 		ether_poll_deregister(ifp);
316 #endif
317 	if_down(ifp);
318 
319 	if (ifq_is_enabled(&ifp->if_snd))
320 		altq_disable(&ifp->if_snd);
321 	if (ifq_is_attached(&ifp->if_snd))
322 		altq_detach(&ifp->if_snd);
323 
324 	/*
325 	 * Clean up all addresses.
326 	 */
327 	ifp->if_lladdr = NULL;
328 
329 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
330 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
331 #ifdef INET
332 		/* XXX: Ugly!! ad hoc just for INET */
333 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
334 			struct ifaliasreq ifr;
335 
336 			bzero(&ifr, sizeof ifr);
337 			ifr.ifra_addr = *ifa->ifa_addr;
338 			if (ifa->ifa_dstaddr)
339 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
340 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
341 				       NULL) == 0)
342 				continue;
343 		}
344 #endif /* INET */
345 #ifdef INET6
346 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
347 			in6_purgeaddr(ifa);
348 			/* ifp_addrhead is already updated */
349 			continue;
350 		}
351 #endif /* INET6 */
352 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
353 		IFAFREE(ifa);
354 	}
355 
356 #ifdef INET
357 	/*
358 	 * Remove all IPv4 kernel structures related to ifp.
359 	 */
360 	in_ifdetach(ifp);
361 #endif
362 
363 #ifdef INET6
364 	/*
365 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
366 	 * before removing routing entries below, since IPv6 interface direct
367 	 * routes are expected to be removed by the IPv6-specific kernel API.
368 	 * Otherwise, the kernel will detect some inconsistency and bark it.
369 	 */
370 	in6_ifdetach(ifp);
371 #endif
372 
373 	/*
374 	 * Delete all remaining routes using this interface
375 	 * Unfortuneatly the only way to do this is to slog through
376 	 * the entire routing table looking for routes which point
377 	 * to this interface...oh well...
378 	 */
379 	for (i = 1; i <= AF_MAX; i++) {
380 		if ((rnh = rt_tables[i]) == NULL)
381 			continue;
382 		rnh->rnh_walktree(rnh, if_rtdel, ifp);
383 	}
384 
385 	/* Announce that the interface is gone. */
386 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
387 
388 	SLIST_FOREACH(dp, &domains, dom_next)
389 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
390 			(*dp->dom_ifdetach)(ifp,
391 				ifp->if_afdata[dp->dom_family]);
392 
393 	/*
394 	 * Remove interface from ifindex2ifp[] and maybe decrement if_index.
395 	 */
396 	ifindex2ifnet[ifp->if_index] = NULL;
397 	while (if_index > 0 && ifindex2ifnet[if_index] == NULL)
398 		if_index--;
399 
400 	TAILQ_REMOVE(&ifnet, ifp, if_link);
401 	splx(s);
402 }
403 
404 /*
405  * Delete Routes for a Network Interface
406  *
407  * Called for each routing entry via the rnh->rnh_walktree() call above
408  * to delete all route entries referencing a detaching network interface.
409  *
410  * Arguments:
411  *	rn	pointer to node in the routing table
412  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
413  *
414  * Returns:
415  *	0	successful
416  *	errno	failed - reason indicated
417  *
418  */
419 static int
420 if_rtdel(struct radix_node *rn, void *arg)
421 {
422 	struct rtentry	*rt = (struct rtentry *)rn;
423 	struct ifnet	*ifp = arg;
424 	int		err;
425 
426 	if (rt->rt_ifp == ifp) {
427 
428 		/*
429 		 * Protect (sorta) against walktree recursion problems
430 		 * with cloned routes
431 		 */
432 		if (!(rt->rt_flags & RTF_UP))
433 			return (0);
434 
435 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
436 				rt_mask(rt), rt->rt_flags,
437 				(struct rtentry **) NULL);
438 		if (err) {
439 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
440 		}
441 	}
442 
443 	return (0);
444 }
445 
446 /*
447  * Create a clone network interface.
448  */
449 int
450 if_clone_create(char *name, int len)
451 {
452 	struct if_clone *ifc;
453 	char *dp;
454 	int wildcard, bytoff, bitoff;
455 	int unit;
456 	int err;
457 
458 	ifc = if_clone_lookup(name, &unit);
459 	if (ifc == NULL)
460 		return (EINVAL);
461 
462 	if (ifunit(name) != NULL)
463 		return (EEXIST);
464 
465 	bytoff = bitoff = 0;
466 	wildcard = (unit < 0);
467 	/*
468 	 * Find a free unit if none was given.
469 	 */
470 	if (wildcard) {
471 		while (bytoff < ifc->ifc_bmlen &&
472 		    ifc->ifc_units[bytoff] == 0xff)
473 			bytoff++;
474 		if (bytoff >= ifc->ifc_bmlen)
475 			return (ENOSPC);
476 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
477 			bitoff++;
478 		unit = (bytoff << 3) + bitoff;
479 	}
480 
481 	if (unit > ifc->ifc_maxunit)
482 		return (ENXIO);
483 
484 	err = (*ifc->ifc_create)(ifc, unit);
485 	if (err != 0)
486 		return (err);
487 
488 	if (!wildcard) {
489 		bytoff = unit >> 3;
490 		bitoff = unit - (bytoff << 3);
491 	}
492 
493 	/*
494 	 * Allocate the unit in the bitmap.
495 	 */
496 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
497 	    ("%s: bit is already set", __func__));
498 	ifc->ifc_units[bytoff] |= (1 << bitoff);
499 
500 	/* In the wildcard case, we need to update the name. */
501 	if (wildcard) {
502 		for (dp = name; *dp != '\0'; dp++);
503 		if (snprintf(dp, len - (dp-name), "%d", unit) >
504 		    len - (dp-name) - 1) {
505 			/*
506 			 * This can only be a programmer error and
507 			 * there's no straightforward way to recover if
508 			 * it happens.
509 			 */
510 			panic("if_clone_create(): interface name too long");
511 		}
512 
513 	}
514 
515 	EVENTHANDLER_INVOKE(if_clone_event, ifc);
516 
517 	return (0);
518 }
519 
520 /*
521  * Destroy a clone network interface.
522  */
523 int
524 if_clone_destroy(const char *name)
525 {
526 	struct if_clone *ifc;
527 	struct ifnet *ifp;
528 	int bytoff, bitoff;
529 	int unit;
530 
531 	ifc = if_clone_lookup(name, &unit);
532 	if (ifc == NULL)
533 		return (EINVAL);
534 
535 	if (unit < ifc->ifc_minifs)
536 		return (EINVAL);
537 
538 	ifp = ifunit(name);
539 	if (ifp == NULL)
540 		return (ENXIO);
541 
542 	if (ifc->ifc_destroy == NULL)
543 		return (EOPNOTSUPP);
544 
545 	(*ifc->ifc_destroy)(ifp);
546 
547 	/*
548 	 * Compute offset in the bitmap and deallocate the unit.
549 	 */
550 	bytoff = unit >> 3;
551 	bitoff = unit - (bytoff << 3);
552 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
553 	    ("%s: bit is already cleared", __func__));
554 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
555 	return (0);
556 }
557 
558 /*
559  * Look up a network interface cloner.
560  */
561 struct if_clone *
562 if_clone_lookup(const char *name, int *unitp)
563 {
564 	struct if_clone *ifc;
565 	const char *cp;
566 	int i;
567 
568 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
569 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
570 			if (ifc->ifc_name[i] != *cp)
571 				goto next_ifc;
572 		}
573 		goto found_name;
574  next_ifc:
575 		ifc = LIST_NEXT(ifc, ifc_list);
576 	}
577 
578 	/* No match. */
579 	return ((struct if_clone *)NULL);
580 
581  found_name:
582 	if (*cp == '\0') {
583 		i = -1;
584 	} else {
585 		for (i = 0; *cp != '\0'; cp++) {
586 			if (*cp < '0' || *cp > '9') {
587 				/* Bogus unit number. */
588 				return (NULL);
589 			}
590 			i = (i * 10) + (*cp - '0');
591 		}
592 	}
593 
594 	if (unitp != NULL)
595 		*unitp = i;
596 	return (ifc);
597 }
598 
599 /*
600  * Register a network interface cloner.
601  */
602 void
603 if_clone_attach(struct if_clone *ifc)
604 {
605 	int bytoff, bitoff;
606 	int err;
607 	int len, maxclone;
608 	int unit;
609 
610 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
611 	    ("%s: %s requested more units then allowed (%d > %d)",
612 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
613 	    ifc->ifc_maxunit + 1));
614 	/*
615 	 * Compute bitmap size and allocate it.
616 	 */
617 	maxclone = ifc->ifc_maxunit + 1;
618 	len = maxclone >> 3;
619 	if ((len << 3) < maxclone)
620 		len++;
621 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
622 	ifc->ifc_bmlen = len;
623 
624 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
625 	if_cloners_count++;
626 
627 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
628 		err = (*ifc->ifc_create)(ifc, unit);
629 		KASSERT(err == 0,
630 		    ("%s: failed to create required interface %s%d",
631 		    __func__, ifc->ifc_name, unit));
632 
633 		/* Allocate the unit in the bitmap. */
634 		bytoff = unit >> 3;
635 		bitoff = unit - (bytoff << 3);
636 		ifc->ifc_units[bytoff] |= (1 << bitoff);
637 	}
638 }
639 
640 /*
641  * Unregister a network interface cloner.
642  */
643 void
644 if_clone_detach(struct if_clone *ifc)
645 {
646 
647 	LIST_REMOVE(ifc, ifc_list);
648 	free(ifc->ifc_units, M_CLONE);
649 	if_cloners_count--;
650 }
651 
652 /*
653  * Provide list of interface cloners to userspace.
654  */
655 int
656 if_clone_list(struct if_clonereq *ifcr)
657 {
658 	char outbuf[IFNAMSIZ], *dst;
659 	struct if_clone *ifc;
660 	int count, error = 0;
661 
662 	ifcr->ifcr_total = if_cloners_count;
663 	if ((dst = ifcr->ifcr_buffer) == NULL) {
664 		/* Just asking how many there are. */
665 		return (0);
666 	}
667 
668 	if (ifcr->ifcr_count < 0)
669 		return (EINVAL);
670 
671 	count = (if_cloners_count < ifcr->ifcr_count) ?
672 	    if_cloners_count : ifcr->ifcr_count;
673 
674 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
675 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
676 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
677 		error = copyout(outbuf, dst, IFNAMSIZ);
678 		if (error)
679 			break;
680 	}
681 
682 	return (error);
683 }
684 
685 /*
686  * Locate an interface based on a complete address.
687  */
688 struct ifaddr *
689 ifa_ifwithaddr(struct sockaddr *addr)
690 {
691 	struct ifnet *ifp;
692 	struct ifaddr *ifa;
693 
694 	TAILQ_FOREACH(ifp, &ifnet, if_link)
695 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
696 		if (ifa->ifa_addr->sa_family != addr->sa_family)
697 			continue;
698 		if (sa_equal(addr, ifa->ifa_addr))
699 			return (ifa);
700 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
701 		    /* IPv6 doesn't have broadcast */
702 		    ifa->ifa_broadaddr->sa_len != 0 &&
703 		    sa_equal(ifa->ifa_broadaddr, addr))
704 			return (ifa);
705 	}
706 	return ((struct ifaddr *)NULL);
707 }
708 /*
709  * Locate the point to point interface with a given destination address.
710  */
711 struct ifaddr *
712 ifa_ifwithdstaddr(struct sockaddr *addr)
713 {
714 	struct ifnet *ifp;
715 	struct ifaddr *ifa;
716 
717 	TAILQ_FOREACH(ifp, &ifnet, if_link)
718 	    if (ifp->if_flags & IFF_POINTOPOINT)
719 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
720 			if (ifa->ifa_addr->sa_family != addr->sa_family)
721 				continue;
722 			if (ifa->ifa_dstaddr &&
723 			    sa_equal(addr, ifa->ifa_dstaddr))
724 				return (ifa);
725 	}
726 	return ((struct ifaddr *)NULL);
727 }
728 
729 /*
730  * Find an interface on a specific network.  If many, choice
731  * is most specific found.
732  */
733 struct ifaddr *
734 ifa_ifwithnet(struct sockaddr *addr)
735 {
736 	struct ifnet *ifp;
737 	struct ifaddr *ifa;
738 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
739 	u_int af = addr->sa_family;
740 	char *addr_data = addr->sa_data, *cplim;
741 
742 	/*
743 	 * AF_LINK addresses can be looked up directly by their index number,
744 	 * so do that if we can.
745 	 */
746 	if (af == AF_LINK) {
747 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
748 
749 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
750 		return (ifindex2ifnet[sdl->sdl_index]->if_lladdr);
751 	}
752 
753 	/*
754 	 * Scan though each interface, looking for ones that have
755 	 * addresses in this address family.
756 	 */
757 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
758 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
759 			char *cp, *cp2, *cp3;
760 
761 			if (ifa->ifa_addr->sa_family != af)
762 next:				continue;
763 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
764 				/*
765 				 * This is a bit broken as it doesn't
766 				 * take into account that the remote end may
767 				 * be a single node in the network we are
768 				 * looking for.
769 				 * The trouble is that we don't know the
770 				 * netmask for the remote end.
771 				 */
772 				if (ifa->ifa_dstaddr != NULL &&
773 				    sa_equal(addr, ifa->ifa_dstaddr))
774 					return (ifa);
775 			} else {
776 				/*
777 				 * if we have a special address handler,
778 				 * then use it instead of the generic one.
779 				 */
780 				if (ifa->ifa_claim_addr) {
781 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
782 						return (ifa);
783 					} else {
784 						continue;
785 					}
786 				}
787 
788 				/*
789 				 * Scan all the bits in the ifa's address.
790 				 * If a bit dissagrees with what we are
791 				 * looking for, mask it with the netmask
792 				 * to see if it really matters.
793 				 * (A byte at a time)
794 				 */
795 				if (ifa->ifa_netmask == 0)
796 					continue;
797 				cp = addr_data;
798 				cp2 = ifa->ifa_addr->sa_data;
799 				cp3 = ifa->ifa_netmask->sa_data;
800 				cplim = ifa->ifa_netmask->sa_len +
801 					(char *)ifa->ifa_netmask;
802 				while (cp3 < cplim)
803 					if ((*cp++ ^ *cp2++) & *cp3++)
804 						goto next; /* next address! */
805 				/*
806 				 * If the netmask of what we just found
807 				 * is more specific than what we had before
808 				 * (if we had one) then remember the new one
809 				 * before continuing to search
810 				 * for an even better one.
811 				 */
812 				if (ifa_maybe == 0 ||
813 				    rn_refines((char *)ifa->ifa_netmask,
814 					       (char *)ifa_maybe->ifa_netmask))
815 					ifa_maybe = ifa;
816 			}
817 		}
818 	}
819 	return (ifa_maybe);
820 }
821 
822 /*
823  * Find an interface address specific to an interface best matching
824  * a given address.
825  */
826 struct ifaddr *
827 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
828 {
829 	struct ifaddr *ifa;
830 	char *cp, *cp2, *cp3;
831 	char *cplim;
832 	struct ifaddr *ifa_maybe = 0;
833 	u_int af = addr->sa_family;
834 
835 	if (af >= AF_MAX)
836 		return (0);
837 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
838 		if (ifa->ifa_addr->sa_family != af)
839 			continue;
840 		if (ifa_maybe == 0)
841 			ifa_maybe = ifa;
842 		if (ifa->ifa_netmask == NULL) {
843 			if (sa_equal(addr, ifa->ifa_addr) ||
844 			    (ifa->ifa_dstaddr != NULL &&
845 			     sa_equal(addr, ifa->ifa_dstaddr)))
846 				return (ifa);
847 			continue;
848 		}
849 		if (ifp->if_flags & IFF_POINTOPOINT) {
850 			if (sa_equal(addr, ifa->ifa_dstaddr))
851 				return (ifa);
852 		} else {
853 			cp = addr->sa_data;
854 			cp2 = ifa->ifa_addr->sa_data;
855 			cp3 = ifa->ifa_netmask->sa_data;
856 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
857 			for (; cp3 < cplim; cp3++)
858 				if ((*cp++ ^ *cp2++) & *cp3)
859 					break;
860 			if (cp3 == cplim)
861 				return (ifa);
862 		}
863 	}
864 	return (ifa_maybe);
865 }
866 
867 #include <net/route.h>
868 
869 /*
870  * Default action when installing a route with a Link Level gateway.
871  * Lookup an appropriate real ifa to point to.
872  * This should be moved to /sys/net/link.c eventually.
873  */
874 static void
875 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
876 {
877 	struct ifaddr *ifa;
878 	struct sockaddr *dst;
879 	struct ifnet *ifp;
880 
881 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
882 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
883 		return;
884 	ifa = ifaof_ifpforaddr(dst, ifp);
885 	if (ifa != NULL) {
886 		IFAFREE(rt->rt_ifa);
887 		IFAREF(ifa);
888 		rt->rt_ifa = ifa;
889 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
890 			ifa->ifa_rtrequest(cmd, rt, info);
891 	}
892 }
893 
894 /*
895  * Mark an interface down and notify protocols of
896  * the transition.
897  * NOTE: must be called at splnet or eqivalent.
898  */
899 void
900 if_unroute(struct ifnet *ifp, int flag, int fam)
901 {
902 	struct ifaddr *ifa;
903 
904 	ifp->if_flags &= ~flag;
905 	getmicrotime(&ifp->if_lastchange);
906 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
907 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
908 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
909 	ifq_purge(&ifp->if_snd);
910 	rt_ifmsg(ifp);
911 }
912 
913 /*
914  * Mark an interface up and notify protocols of
915  * the transition.
916  * NOTE: must be called at splnet or eqivalent.
917  */
918 void
919 if_route(struct ifnet *ifp, int flag, int fam)
920 {
921 	struct ifaddr *ifa;
922 
923 	ifp->if_flags |= flag;
924 	getmicrotime(&ifp->if_lastchange);
925 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
926 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
927 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
928 	rt_ifmsg(ifp);
929 #ifdef INET6
930 	in6_if_up(ifp);
931 #endif
932 }
933 
934 /*
935  * Mark an interface down and notify protocols of the transition.  An
936  * interface going down is also considered to be a synchronizing event.
937  * We must ensure that all packet processing related to the interface
938  * has completed before we return so e.g. the caller can free the ifnet
939  * structure that the mbufs may be referencing.
940  *
941  * NOTE: must be called at splnet or eqivalent.
942  */
943 void
944 if_down(struct ifnet *ifp)
945 {
946 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
947 	netmsg_service_sync();
948 }
949 
950 /*
951  * Mark an interface up and notify protocols of
952  * the transition.
953  * NOTE: must be called at splnet or eqivalent.
954  */
955 void
956 if_up(struct ifnet *ifp)
957 {
958 
959 	if_route(ifp, IFF_UP, AF_UNSPEC);
960 }
961 
962 /*
963  * Handle interface watchdog timer routines.  Called
964  * from softclock, we decrement timers (if set) and
965  * call the appropriate interface routine on expiration.
966  */
967 static void
968 if_slowtimo(void *arg)
969 {
970 	struct ifnet *ifp;
971 	int s = splimp();
972 
973 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
974 		if (ifp->if_timer == 0 || --ifp->if_timer)
975 			continue;
976 		if (ifp->if_watchdog)
977 			(*ifp->if_watchdog)(ifp);
978 	}
979 	splx(s);
980 	callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
981 }
982 
983 /*
984  * Map interface name to
985  * interface structure pointer.
986  */
987 struct ifnet *
988 ifunit(const char *name)
989 {
990 	struct ifnet *ifp;
991 
992 	/*
993 	 * Search all the interfaces for this name/number
994 	 */
995 
996 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
997 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
998 			break;
999 	}
1000 	return (ifp);
1001 }
1002 
1003 
1004 /*
1005  * Map interface name in a sockaddr_dl to
1006  * interface structure pointer.
1007  */
1008 struct ifnet *
1009 if_withname(struct sockaddr *sa)
1010 {
1011 	char ifname[IFNAMSIZ+1];
1012 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1013 
1014 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1015 	     (sdl->sdl_nlen > IFNAMSIZ) )
1016 		return NULL;
1017 
1018 	/*
1019 	 * ifunit wants a null-terminated name.  It may not be null-terminated
1020 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
1021 	 * and there might not be room to put the trailing null anyway, so we
1022 	 * make a local copy that we know we can null terminate safely.
1023 	 */
1024 
1025 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1026 	ifname[sdl->sdl_nlen] = '\0';
1027 	return ifunit(ifname);
1028 }
1029 
1030 
1031 /*
1032  * Interface ioctls.
1033  */
1034 int
1035 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1036 {
1037 	struct ifnet *ifp;
1038 	struct ifreq *ifr;
1039 	struct ifstat *ifs;
1040 	int error;
1041 	short oif_flags;
1042 	int new_flags;
1043 	size_t namelen, onamelen;
1044 	char new_name[IFNAMSIZ];
1045 	struct ifaddr *ifa;
1046 	struct sockaddr_dl *sdl;
1047 
1048 	switch (cmd) {
1049 
1050 	case SIOCGIFCONF:
1051 	case OSIOCGIFCONF:
1052 		return (ifconf(cmd, data, td));
1053 	}
1054 	ifr = (struct ifreq *)data;
1055 
1056 	switch (cmd) {
1057 	case SIOCIFCREATE:
1058 	case SIOCIFDESTROY:
1059 		if ((error = suser(td)) != 0)
1060 			return (error);
1061 		return ((cmd == SIOCIFCREATE) ?
1062 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1063 			if_clone_destroy(ifr->ifr_name));
1064 
1065 	case SIOCIFGCLONERS:
1066 		return (if_clone_list((struct if_clonereq *)data));
1067 	}
1068 
1069 	ifp = ifunit(ifr->ifr_name);
1070 	if (ifp == 0)
1071 		return (ENXIO);
1072 	switch (cmd) {
1073 
1074 	case SIOCGIFFLAGS:
1075 		ifr->ifr_flags = ifp->if_flags;
1076 		ifr->ifr_flagshigh = ifp->if_flags >> 16;
1077 		break;
1078 
1079 	case SIOCGIFCAP:
1080 		ifr->ifr_reqcap = ifp->if_capabilities;
1081 		ifr->ifr_curcap = ifp->if_capenable;
1082 		break;
1083 
1084 	case SIOCGIFMETRIC:
1085 		ifr->ifr_metric = ifp->if_metric;
1086 		break;
1087 
1088 	case SIOCGIFMTU:
1089 		ifr->ifr_mtu = ifp->if_mtu;
1090 		break;
1091 
1092 	case SIOCGIFPHYS:
1093 		ifr->ifr_phys = ifp->if_physical;
1094 		break;
1095 
1096 	case SIOCSIFFLAGS:
1097 		error = suser(td);
1098 		if (error)
1099 			return (error);
1100 		new_flags = (ifr->ifr_flags & 0xffff) |
1101 		    (ifr->ifr_flagshigh << 16);
1102 		if (ifp->if_flags & IFF_SMART) {
1103 			/* Smart drivers twiddle their own routes */
1104 		} else if (ifp->if_flags & IFF_UP &&
1105 		    (new_flags & IFF_UP) == 0) {
1106 			int s = splimp();
1107 			if_down(ifp);
1108 			splx(s);
1109 		} else if (new_flags & IFF_UP &&
1110 		    (ifp->if_flags & IFF_UP) == 0) {
1111 			int s = splimp();
1112 			if_up(ifp);
1113 			splx(s);
1114 		}
1115 
1116 #ifdef DEVICE_POLLING
1117 		if ((new_flags ^ ifp->if_flags) & IFF_POLLING) {
1118 			if (new_flags & IFF_POLLING) {
1119 				ether_poll_register(ifp);
1120 			} else {
1121 				ether_poll_deregister(ifp);
1122 			}
1123 		}
1124 #endif
1125 
1126 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1127 			(new_flags &~ IFF_CANTCHANGE);
1128 		if (new_flags & IFF_PPROMISC) {
1129 			/* Permanently promiscuous mode requested */
1130 			ifp->if_flags |= IFF_PROMISC;
1131 		} else if (ifp->if_pcount == 0) {
1132 			ifp->if_flags &= ~IFF_PROMISC;
1133 		}
1134 		if (ifp->if_ioctl)
1135 			(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1136 		getmicrotime(&ifp->if_lastchange);
1137 		break;
1138 
1139 	case SIOCSIFCAP:
1140 		error = suser(td);
1141 		if (error)
1142 			return (error);
1143 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1144 			return (EINVAL);
1145 		(*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1146 		break;
1147 
1148 	case SIOCSIFNAME:
1149 		error = suser(td);
1150 		if (error != 0)
1151 			return (error);
1152 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1153 		if (error != 0)
1154 			return (error);
1155 		if (new_name[0] == '\0')
1156 			return (EINVAL);
1157 		if (ifunit(new_name) != NULL)
1158 			return (EEXIST);
1159 
1160 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1161 
1162 		/* Announce the departure of the interface. */
1163 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1164 
1165 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1166 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1167 		/* XXX IFA_LOCK(ifa); */
1168 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1169 		namelen = strlen(new_name);
1170 		onamelen = sdl->sdl_nlen;
1171 		/*
1172 		 * Move the address if needed.  This is safe because we
1173 		 * allocate space for a name of length IFNAMSIZ when we
1174 		 * create this in if_attach().
1175 		 */
1176 		if (namelen != onamelen) {
1177 			bcopy(sdl->sdl_data + onamelen,
1178 			    sdl->sdl_data + namelen, sdl->sdl_alen);
1179 		}
1180 		bcopy(new_name, sdl->sdl_data, namelen);
1181 		sdl->sdl_nlen = namelen;
1182 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1183 		bzero(sdl->sdl_data, onamelen);
1184 		while (namelen != 0)
1185 			sdl->sdl_data[--namelen] = 0xff;
1186 		/* XXX IFA_UNLOCK(ifa) */
1187 
1188 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1189 
1190 		/* Announce the return of the interface. */
1191 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1192 		break;
1193 
1194 	case SIOCSIFMETRIC:
1195 		error = suser(td);
1196 		if (error)
1197 			return (error);
1198 		ifp->if_metric = ifr->ifr_metric;
1199 		getmicrotime(&ifp->if_lastchange);
1200 		break;
1201 
1202 	case SIOCSIFPHYS:
1203 		error = suser(td);
1204 		if (error)
1205 			return error;
1206 		if (!ifp->if_ioctl)
1207 		        return EOPNOTSUPP;
1208 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1209 		if (error == 0)
1210 			getmicrotime(&ifp->if_lastchange);
1211 		return (error);
1212 
1213 	case SIOCSIFMTU:
1214 	{
1215 		u_long oldmtu = ifp->if_mtu;
1216 
1217 		error = suser(td);
1218 		if (error)
1219 			return (error);
1220 		if (ifp->if_ioctl == NULL)
1221 			return (EOPNOTSUPP);
1222 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1223 			return (EINVAL);
1224 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1225 		if (error == 0) {
1226 			getmicrotime(&ifp->if_lastchange);
1227 			rt_ifmsg(ifp);
1228 		}
1229 		/*
1230 		 * If the link MTU changed, do network layer specific procedure.
1231 		 */
1232 		if (ifp->if_mtu != oldmtu) {
1233 #ifdef INET6
1234 			nd6_setmtu(ifp);
1235 #endif
1236 		}
1237 		return (error);
1238 	}
1239 
1240 	case SIOCADDMULTI:
1241 	case SIOCDELMULTI:
1242 		error = suser(td);
1243 		if (error)
1244 			return (error);
1245 
1246 		/* Don't allow group membership on non-multicast interfaces. */
1247 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1248 			return EOPNOTSUPP;
1249 
1250 		/* Don't let users screw up protocols' entries. */
1251 		if (ifr->ifr_addr.sa_family != AF_LINK)
1252 			return EINVAL;
1253 
1254 		if (cmd == SIOCADDMULTI) {
1255 			struct ifmultiaddr *ifma;
1256 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1257 		} else {
1258 			error = if_delmulti(ifp, &ifr->ifr_addr);
1259 		}
1260 		if (error == 0)
1261 			getmicrotime(&ifp->if_lastchange);
1262 		return error;
1263 
1264 	case SIOCSIFPHYADDR:
1265 	case SIOCDIFPHYADDR:
1266 #ifdef INET6
1267 	case SIOCSIFPHYADDR_IN6:
1268 #endif
1269 	case SIOCSLIFPHYADDR:
1270         case SIOCSIFMEDIA:
1271 	case SIOCSIFGENERIC:
1272 		error = suser(td);
1273 		if (error)
1274 			return (error);
1275 		if (ifp->if_ioctl == 0)
1276 			return (EOPNOTSUPP);
1277 		error = (*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred);
1278 		if (error == 0)
1279 			getmicrotime(&ifp->if_lastchange);
1280 		return error;
1281 
1282 	case SIOCGIFSTATUS:
1283 		ifs = (struct ifstat *)data;
1284 		ifs->ascii[0] = '\0';
1285 
1286 	case SIOCGIFPSRCADDR:
1287 	case SIOCGIFPDSTADDR:
1288 	case SIOCGLIFPHYADDR:
1289 	case SIOCGIFMEDIA:
1290 	case SIOCGIFGENERIC:
1291 		if (ifp->if_ioctl == 0)
1292 			return (EOPNOTSUPP);
1293 		return ((*ifp->if_ioctl)(ifp, cmd, data, td->td_proc->p_ucred));
1294 
1295 	case SIOCSIFLLADDR:
1296 		error = suser(td);
1297 		if (error)
1298 			return (error);
1299 		return if_setlladdr(ifp,
1300 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1301 
1302 	default:
1303 		oif_flags = ifp->if_flags;
1304 		if (so->so_proto == 0)
1305 			return (EOPNOTSUPP);
1306 #ifndef COMPAT_43
1307 		error = so_pru_control(so, cmd, data, ifp, td);
1308 #else
1309 	    {
1310 		int ocmd = cmd;
1311 
1312 		switch (cmd) {
1313 
1314 		case SIOCSIFDSTADDR:
1315 		case SIOCSIFADDR:
1316 		case SIOCSIFBRDADDR:
1317 		case SIOCSIFNETMASK:
1318 #if BYTE_ORDER != BIG_ENDIAN
1319 			if (ifr->ifr_addr.sa_family == 0 &&
1320 			    ifr->ifr_addr.sa_len < 16) {
1321 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1322 				ifr->ifr_addr.sa_len = 16;
1323 			}
1324 #else
1325 			if (ifr->ifr_addr.sa_len == 0)
1326 				ifr->ifr_addr.sa_len = 16;
1327 #endif
1328 			break;
1329 
1330 		case OSIOCGIFADDR:
1331 			cmd = SIOCGIFADDR;
1332 			break;
1333 
1334 		case OSIOCGIFDSTADDR:
1335 			cmd = SIOCGIFDSTADDR;
1336 			break;
1337 
1338 		case OSIOCGIFBRDADDR:
1339 			cmd = SIOCGIFBRDADDR;
1340 			break;
1341 
1342 		case OSIOCGIFNETMASK:
1343 			cmd = SIOCGIFNETMASK;
1344 		}
1345 		error =  so_pru_control(so, cmd, data, ifp, td);
1346 		switch (ocmd) {
1347 
1348 		case OSIOCGIFADDR:
1349 		case OSIOCGIFDSTADDR:
1350 		case OSIOCGIFBRDADDR:
1351 		case OSIOCGIFNETMASK:
1352 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1353 
1354 		}
1355 	    }
1356 #endif /* COMPAT_43 */
1357 
1358 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1359 #ifdef INET6
1360 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1361 			if (ifp->if_flags & IFF_UP) {
1362 				int s = splimp();
1363 				in6_if_up(ifp);
1364 				splx(s);
1365 			}
1366 #endif
1367 		}
1368 		return (error);
1369 
1370 	}
1371 	return (0);
1372 }
1373 
1374 /*
1375  * Set/clear promiscuous mode on interface ifp based on the truth value
1376  * of pswitch.  The calls are reference counted so that only the first
1377  * "on" request actually has an effect, as does the final "off" request.
1378  * Results are undefined if the "off" and "on" requests are not matched.
1379  */
1380 int
1381 ifpromisc(struct ifnet *ifp, int pswitch)
1382 {
1383 	struct ifreq ifr;
1384 	int error;
1385 	int oldflags;
1386 
1387 	oldflags = ifp->if_flags;
1388 	if (ifp->if_flags & IFF_PPROMISC) {
1389 		/* Do nothing if device is in permanently promiscuous mode */
1390 		ifp->if_pcount += pswitch ? 1 : -1;
1391 		return (0);
1392 	}
1393 	if (pswitch) {
1394 		/*
1395 		 * If the device is not configured up, we cannot put it in
1396 		 * promiscuous mode.
1397 		 */
1398 		if ((ifp->if_flags & IFF_UP) == 0)
1399 			return (ENETDOWN);
1400 		if (ifp->if_pcount++ != 0)
1401 			return (0);
1402 		ifp->if_flags |= IFF_PROMISC;
1403 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1404 		    ifp->if_xname);
1405 	} else {
1406 		if (--ifp->if_pcount > 0)
1407 			return (0);
1408 		ifp->if_flags &= ~IFF_PROMISC;
1409 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1410 		    ifp->if_xname);
1411 	}
1412 	ifr.ifr_flags = ifp->if_flags;
1413 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
1414 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1415 				 (struct ucred *)NULL);
1416 	if (error == 0)
1417 		rt_ifmsg(ifp);
1418 	else
1419 		ifp->if_flags = oldflags;
1420 	return error;
1421 }
1422 
1423 /*
1424  * Return interface configuration
1425  * of system.  List may be used
1426  * in later ioctl's (above) to get
1427  * other information.
1428  */
1429 static int
1430 ifconf(u_long cmd, caddr_t data, struct thread *td)
1431 {
1432 	struct ifconf *ifc = (struct ifconf *)data;
1433 	struct ifnet *ifp;
1434 	struct ifaddr *ifa;
1435 	struct sockaddr *sa;
1436 	struct ifreq ifr, *ifrp;
1437 	int space = ifc->ifc_len, error = 0;
1438 
1439 	ifrp = ifc->ifc_req;
1440 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1441 		int addrs;
1442 
1443 		if (space <= sizeof ifr)
1444 			break;
1445 
1446 		/*
1447 		 * Zero the stack declared structure first to prevent
1448 		 * memory disclosure.
1449 		 */
1450 		bzero(&ifr, sizeof(ifr));
1451 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1452 		    >= sizeof(ifr.ifr_name)) {
1453 			error = ENAMETOOLONG;
1454 			break;
1455 		}
1456 
1457 		addrs = 0;
1458 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1459 			if (space <= sizeof ifr)
1460 				break;
1461 			sa = ifa->ifa_addr;
1462 			if (td->td_proc->p_ucred->cr_prison &&
1463 			    prison_if(td, sa))
1464 				continue;
1465 			addrs++;
1466 #ifdef COMPAT_43
1467 			if (cmd == OSIOCGIFCONF) {
1468 				struct osockaddr *osa =
1469 					 (struct osockaddr *)&ifr.ifr_addr;
1470 				ifr.ifr_addr = *sa;
1471 				osa->sa_family = sa->sa_family;
1472 				error = copyout(&ifr, ifrp, sizeof ifr);
1473 				ifrp++;
1474 			} else
1475 #endif
1476 			if (sa->sa_len <= sizeof(*sa)) {
1477 				ifr.ifr_addr = *sa;
1478 				error = copyout(&ifr, ifrp, sizeof ifr);
1479 				ifrp++;
1480 			} else {
1481 				if (space < (sizeof ifr) + sa->sa_len -
1482 					    sizeof(*sa))
1483 					break;
1484 				space -= sa->sa_len - sizeof(*sa);
1485 				error = copyout(&ifr, ifrp,
1486 						sizeof ifr.ifr_name);
1487 				if (error == 0)
1488 					error = copyout(sa, &ifrp->ifr_addr,
1489 							sa->sa_len);
1490 				ifrp = (struct ifreq *)
1491 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1492 			}
1493 			if (error)
1494 				break;
1495 			space -= sizeof ifr;
1496 		}
1497 		if (error)
1498 			break;
1499 		if (!addrs) {
1500 			bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1501 			error = copyout(&ifr, ifrp, sizeof ifr);
1502 			if (error)
1503 				break;
1504 			space -= sizeof ifr;
1505 			ifrp++;
1506 		}
1507 	}
1508 	ifc->ifc_len -= space;
1509 	return (error);
1510 }
1511 
1512 /*
1513  * Just like if_promisc(), but for all-multicast-reception mode.
1514  */
1515 int
1516 if_allmulti(struct ifnet *ifp, int onswitch)
1517 {
1518 	int error = 0;
1519 	int s = splimp();
1520 	struct ifreq ifr;
1521 
1522 	if (onswitch) {
1523 		if (ifp->if_amcount++ == 0) {
1524 			ifp->if_flags |= IFF_ALLMULTI;
1525 			ifr.ifr_flags = ifp->if_flags;
1526 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1527 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1528 					      (struct ucred *)NULL);
1529 		}
1530 	} else {
1531 		if (ifp->if_amcount > 1) {
1532 			ifp->if_amcount--;
1533 		} else {
1534 			ifp->if_amcount = 0;
1535 			ifp->if_flags &= ~IFF_ALLMULTI;
1536 			ifr.ifr_flags = ifp->if_flags;
1537 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1538 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1539 					      (struct ucred *)NULL);
1540 		}
1541 	}
1542 	splx(s);
1543 
1544 	if (error == 0)
1545 		rt_ifmsg(ifp);
1546 	return error;
1547 }
1548 
1549 /*
1550  * Add a multicast listenership to the interface in question.
1551  * The link layer provides a routine which converts
1552  */
1553 int
1554 if_addmulti(
1555 	struct ifnet *ifp,	/* interface to manipulate */
1556 	struct sockaddr *sa,	/* address to add */
1557 	struct ifmultiaddr **retifma)
1558 {
1559 	struct sockaddr *llsa, *dupsa;
1560 	int error, s;
1561 	struct ifmultiaddr *ifma;
1562 
1563 	/*
1564 	 * If the matching multicast address already exists
1565 	 * then don't add a new one, just add a reference
1566 	 */
1567 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1568 		if (sa_equal(sa, ifma->ifma_addr)) {
1569 			ifma->ifma_refcount++;
1570 			if (retifma)
1571 				*retifma = ifma;
1572 			return 0;
1573 		}
1574 	}
1575 
1576 	/*
1577 	 * Give the link layer a chance to accept/reject it, and also
1578 	 * find out which AF_LINK address this maps to, if it isn't one
1579 	 * already.
1580 	 */
1581 	if (ifp->if_resolvemulti) {
1582 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1583 		if (error) return error;
1584 	} else {
1585 		llsa = 0;
1586 	}
1587 
1588 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1589 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1590 	bcopy(sa, dupsa, sa->sa_len);
1591 
1592 	ifma->ifma_addr = dupsa;
1593 	ifma->ifma_lladdr = llsa;
1594 	ifma->ifma_ifp = ifp;
1595 	ifma->ifma_refcount = 1;
1596 	ifma->ifma_protospec = 0;
1597 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1598 
1599 	/*
1600 	 * Some network interfaces can scan the address list at
1601 	 * interrupt time; lock them out.
1602 	 */
1603 	s = splimp();
1604 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1605 	splx(s);
1606 	*retifma = ifma;
1607 
1608 	if (llsa != 0) {
1609 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1610 			if (sa_equal(ifma->ifma_addr, llsa))
1611 				break;
1612 		}
1613 		if (ifma) {
1614 			ifma->ifma_refcount++;
1615 		} else {
1616 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1617 			       M_IFMADDR, M_WAITOK);
1618 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1619 			       M_IFMADDR, M_WAITOK);
1620 			bcopy(llsa, dupsa, llsa->sa_len);
1621 			ifma->ifma_addr = dupsa;
1622 			ifma->ifma_ifp = ifp;
1623 			ifma->ifma_refcount = 1;
1624 			s = splimp();
1625 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1626 			splx(s);
1627 		}
1628 	}
1629 	/*
1630 	 * We are certain we have added something, so call down to the
1631 	 * interface to let them know about it.
1632 	 */
1633 	s = splimp();
1634 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1635 	splx(s);
1636 
1637 	return 0;
1638 }
1639 
1640 /*
1641  * Remove a reference to a multicast address on this interface.  Yell
1642  * if the request does not match an existing membership.
1643  */
1644 int
1645 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1646 {
1647 	struct ifmultiaddr *ifma;
1648 	int s;
1649 
1650 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1651 		if (sa_equal(sa, ifma->ifma_addr))
1652 			break;
1653 	if (ifma == 0)
1654 		return ENOENT;
1655 
1656 	if (ifma->ifma_refcount > 1) {
1657 		ifma->ifma_refcount--;
1658 		return 0;
1659 	}
1660 
1661 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1662 	sa = ifma->ifma_lladdr;
1663 	s = splimp();
1664 	LIST_REMOVE(ifma, ifma_link);
1665 	/*
1666 	 * Make sure the interface driver is notified
1667 	 * in the case of a link layer mcast group being left.
1668 	 */
1669 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0)
1670 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1671 	splx(s);
1672 	free(ifma->ifma_addr, M_IFMADDR);
1673 	free(ifma, M_IFMADDR);
1674 	if (sa == 0)
1675 		return 0;
1676 
1677 	/*
1678 	 * Now look for the link-layer address which corresponds to
1679 	 * this network address.  It had been squirreled away in
1680 	 * ifma->ifma_lladdr for this purpose (so we don't have
1681 	 * to call ifp->if_resolvemulti() again), and we saved that
1682 	 * value in sa above.  If some nasty deleted the
1683 	 * link-layer address out from underneath us, we can deal because
1684 	 * the address we stored was is not the same as the one which was
1685 	 * in the record for the link-layer address.  (So we don't complain
1686 	 * in that case.)
1687 	 */
1688 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1689 		if (sa_equal(sa, ifma->ifma_addr))
1690 			break;
1691 	if (ifma == 0)
1692 		return 0;
1693 
1694 	if (ifma->ifma_refcount > 1) {
1695 		ifma->ifma_refcount--;
1696 		return 0;
1697 	}
1698 
1699 	s = splimp();
1700 	LIST_REMOVE(ifma, ifma_link);
1701 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1702 	splx(s);
1703 	free(ifma->ifma_addr, M_IFMADDR);
1704 	free(sa, M_IFMADDR);
1705 	free(ifma, M_IFMADDR);
1706 
1707 	return 0;
1708 }
1709 
1710 /*
1711  * Set the link layer address on an interface.
1712  *
1713  * At this time we only support certain types of interfaces,
1714  * and we don't allow the length of the address to change.
1715  */
1716 int
1717 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1718 {
1719 	struct sockaddr_dl *sdl;
1720 	struct ifaddr *ifa;
1721 	struct ifreq ifr;
1722 
1723 	sdl = IF_LLSOCKADDR(ifp);
1724 	if (sdl == NULL)
1725 		return (EINVAL);
1726 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1727 		return (EINVAL);
1728 	switch (ifp->if_type) {
1729 	case IFT_ETHER:			/* these types use struct arpcom */
1730 	case IFT_FDDI:
1731 	case IFT_XETHER:
1732 	case IFT_ISO88025:
1733 	case IFT_L2VLAN:
1734 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1735 		/* FALLTHROUGH */
1736 	case IFT_ARCNET:
1737 		bcopy(lladdr, LLADDR(sdl), len);
1738 		break;
1739 	default:
1740 		return (ENODEV);
1741 	}
1742 	/*
1743 	 * If the interface is already up, we need
1744 	 * to re-init it in order to reprogram its
1745 	 * address filter.
1746 	 */
1747 	if ((ifp->if_flags & IFF_UP) != 0) {
1748 		ifp->if_flags &= ~IFF_UP;
1749 		ifr.ifr_flags = ifp->if_flags;
1750 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1751 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1752 				 (struct ucred *)NULL);
1753 		ifp->if_flags |= IFF_UP;
1754 		ifr.ifr_flags = ifp->if_flags;
1755 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1756 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1757 				 (struct ucred *)NULL);
1758 #ifdef INET
1759 		/*
1760 		 * Also send gratuitous ARPs to notify other nodes about
1761 		 * the address change.
1762 		 */
1763 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1764 			if (ifa->ifa_addr != NULL &&
1765 			    ifa->ifa_addr->sa_family == AF_INET)
1766 				arp_ifinit(ifp, ifa);
1767 		}
1768 #endif
1769 	}
1770 	return (0);
1771 }
1772 
1773 struct ifmultiaddr *
1774 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1775 {
1776 	struct ifmultiaddr *ifma;
1777 
1778 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1779 		if (sa_equal(ifma->ifma_addr, sa))
1780 			break;
1781 
1782 	return ifma;
1783 }
1784 
1785 /*
1786  * The name argument must be a pointer to storage which will last as
1787  * long as the interface does.  For physical devices, the result of
1788  * device_get_name(dev) is a good choice and for pseudo-devices a
1789  * static string works well.
1790  */
1791 void
1792 if_initname(struct ifnet *ifp, const char *name, int unit)
1793 {
1794 	ifp->if_dname = name;
1795 	ifp->if_dunit = unit;
1796 	if (unit != IF_DUNIT_NONE)
1797 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1798 	else
1799 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1800 }
1801 
1802 int
1803 if_printf(struct ifnet *ifp, const char *fmt, ...)
1804 {
1805 	__va_list ap;
1806 	int retval;
1807 
1808 	retval = printf("%s: ", ifp->if_xname);
1809 	__va_start(ap, fmt);
1810 	retval += vprintf(fmt, ap);
1811 	__va_end(ap);
1812 	return (retval);
1813 }
1814 
1815 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1816 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1817 
1818 void
1819 ifq_set_classic(struct ifaltq *ifq)
1820 {
1821 	ifq->altq_enqueue = ifq_classic_enqueue;
1822 	ifq->altq_dequeue = ifq_classic_dequeue;
1823 	ifq->altq_request = ifq_classic_request;
1824 }
1825 
1826 static int
1827 ifq_classic_enqueue(struct ifaltq *ifq, struct mbuf *m,
1828 		    struct altq_pktattr *pa __unused)
1829 {
1830 	crit_enter();
1831 	if (IF_QFULL(ifq)) {
1832 		m_freem(m);
1833 		crit_exit();
1834 		return(ENOBUFS);
1835 	} else {
1836 		IF_ENQUEUE(ifq, m);
1837 		crit_exit();
1838 		return(0);
1839 	}
1840 }
1841 
1842 static struct mbuf *
1843 ifq_classic_dequeue(struct ifaltq *ifq, int op)
1844 {
1845 	struct mbuf *m;
1846 
1847 	crit_enter();
1848 	switch (op) {
1849 	case ALTDQ_POLL:
1850 		IF_POLL(ifq, m);
1851 		break;
1852 	case ALTDQ_REMOVE:
1853 		IF_DEQUEUE(ifq, m);
1854 		break;
1855 	default:
1856 		panic("unsupported ALTQ dequeue op: %d", op);
1857 	}
1858 	crit_exit();
1859 	return(m);
1860 }
1861 
1862 static int
1863 ifq_classic_request(struct ifaltq *ifq, int req, void *arg)
1864 {
1865 	crit_enter();
1866 	switch (req) {
1867 	case ALTRQ_PURGE:
1868 		IF_DRAIN(ifq);
1869 		break;
1870 	default:
1871 		panic("unspported ALTQ request: %d", req);
1872 	}
1873 	crit_exit();
1874 	return(0);
1875 }
1876 
1877