xref: /dflybsd-src/sys/net/if.c (revision 78195a764d5e70464a6d4f49bc08332a2a8bb4d0)
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 1.43 2005/11/28 17:13:45 dillon Exp $
36  */
37 
38 #include "opt_compat.h"
39 #include "opt_inet6.h"
40 #include "opt_inet.h"
41 #include "opt_polling.h"
42 
43 #include <sys/param.h>
44 #include <sys/malloc.h>
45 #include <sys/mbuf.h>
46 #include <sys/systm.h>
47 #include <sys/proc.h>
48 #include <sys/protosw.h>
49 #include <sys/socket.h>
50 #include <sys/socketvar.h>
51 #include <sys/socketops.h>
52 #include <sys/protosw.h>
53 #include <sys/kernel.h>
54 #include <sys/sockio.h>
55 #include <sys/syslog.h>
56 #include <sys/sysctl.h>
57 #include <sys/domain.h>
58 #include <sys/thread.h>
59 #include <sys/serialize.h>
60 
61 #include <net/if.h>
62 #include <net/if_arp.h>
63 #include <net/if_dl.h>
64 #include <net/if_types.h>
65 #include <net/if_var.h>
66 #include <net/ifq_var.h>
67 #include <net/radix.h>
68 #include <net/route.h>
69 #include <machine/stdarg.h>
70 
71 #include <sys/thread2.h>
72 
73 #if defined(INET) || defined(INET6)
74 /*XXX*/
75 #include <netinet/in.h>
76 #include <netinet/in_var.h>
77 #include <netinet/if_ether.h>
78 #ifdef INET6
79 #include <machine/clock.h> /* XXX: temporal workaround for fxp issue */
80 #include <netinet6/in6_var.h>
81 #include <netinet6/in6_ifattach.h>
82 #endif
83 #endif
84 
85 #if defined(COMPAT_43)
86 #include <emulation/43bsd/43bsd_socket.h>
87 #endif /* COMPAT_43 */
88 
89 /*
90  * Support for non-ALTQ interfaces.
91  */
92 static int	ifq_classic_enqueue(struct ifaltq *, struct mbuf *,
93 				    struct altq_pktattr *);
94 static struct mbuf *
95 		ifq_classic_dequeue(struct ifaltq *, struct mbuf *, int);
96 static int	ifq_classic_request(struct ifaltq *, int, void *);
97 
98 /*
99  * System initialization
100  */
101 
102 static void	if_attachdomain(void *);
103 static void	if_attachdomain1(struct ifnet *);
104 static int ifconf (u_long, caddr_t, struct thread *);
105 static void ifinit (void *);
106 static void if_slowtimo (void *);
107 static void link_rtrequest (int, struct rtentry *, struct rt_addrinfo *);
108 static int  if_rtdel (struct radix_node *, void *);
109 
110 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
111 
112 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
113 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
114 MALLOC_DEFINE(M_CLONE, "clone", "interface cloning framework");
115 
116 int	ifqmaxlen = IFQ_MAXLEN;
117 struct	ifnethead ifnet;	/* depend on static init XXX */
118 
119 #ifdef INET6
120 /*
121  * XXX: declare here to avoid to include many inet6 related files..
122  * should be more generalized?
123  */
124 extern void	nd6_setmtu (struct ifnet *);
125 #endif
126 
127 struct if_clone *if_clone_lookup (const char *, int *);
128 int if_clone_list (struct if_clonereq *);
129 
130 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
131 int if_cloners_count;
132 
133 struct callout if_slowtimo_timer;
134 
135 /*
136  * Network interface utility routines.
137  *
138  * Routines with ifa_ifwith* names take sockaddr *'s as
139  * parameters.
140  */
141 /* ARGSUSED*/
142 void
143 ifinit(void *dummy)
144 {
145 	struct ifnet *ifp;
146 
147 	callout_init(&if_slowtimo_timer);
148 
149 	crit_enter();
150 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
151 		if (ifp->if_snd.ifq_maxlen == 0) {
152 			if_printf(ifp, "XXX: driver didn't set ifq_maxlen\n");
153 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
154 		}
155 	}
156 	crit_exit();
157 
158 	if_slowtimo(0);
159 }
160 
161 int if_index = 0;
162 struct ifnet **ifindex2ifnet = NULL;
163 
164 /*
165  * Attach an interface to the list of "active" interfaces.
166  *
167  * The serializer is optional.  If non-NULL access to the interface
168  * may be MPSAFE.
169  */
170 void
171 if_attach(struct ifnet *ifp, lwkt_serialize_t serializer)
172 {
173 	unsigned socksize, ifasize;
174 	int namelen, masklen;
175 	struct sockaddr_dl *sdl;
176 	struct ifaddr *ifa;
177 	struct ifaltq *ifq;
178 
179 	static int if_indexlim = 8;
180 	static boolean_t inited;
181 
182 	if (!inited) {
183 		TAILQ_INIT(&ifnet);
184 		inited = TRUE;
185 	}
186 
187 	/*
188 	 * The serializer can be passed in from the device, allowing the
189 	 * same serializer to be used for both the interrupt interlock and
190 	 * the device queue.  If not specified, the netif structure will
191 	 * use an embedded serializer.
192 	 */
193 	if (serializer == NULL) {
194 		serializer = &ifp->if_default_serializer;
195 		lwkt_serialize_init(serializer);
196 	}
197 	ifp->if_serializer = serializer;
198 
199 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
200 	ifp->if_index = ++if_index;
201 	/*
202 	 * XXX -
203 	 * The old code would work if the interface passed a pre-existing
204 	 * chain of ifaddrs to this code.  We don't trust our callers to
205 	 * properly initialize the tailq, however, so we no longer allow
206 	 * this unlikely case.
207 	 */
208 	TAILQ_INIT(&ifp->if_addrhead);
209 	TAILQ_INIT(&ifp->if_prefixhead);
210 	LIST_INIT(&ifp->if_multiaddrs);
211 	getmicrotime(&ifp->if_lastchange);
212 	if (ifindex2ifnet == NULL || if_index >= if_indexlim) {
213 		unsigned int n;
214 		struct ifnet **q;
215 
216 		if_indexlim <<= 1;
217 
218 		/* grow ifindex2ifnet */
219 		n = if_indexlim * sizeof(*q);
220 		q = malloc(n, M_IFADDR, M_WAITOK | M_ZERO);
221 		if (ifindex2ifnet) {
222 			bcopy(ifindex2ifnet, q, n/2);
223 			free(ifindex2ifnet, M_IFADDR);
224 		}
225 		ifindex2ifnet = q;
226 	}
227 
228 	ifindex2ifnet[if_index] = ifp;
229 
230 	/*
231 	 * create a Link Level name for this device
232 	 */
233 	namelen = strlen(ifp->if_xname);
234 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
235 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
236 	socksize = masklen + ifp->if_addrlen;
237 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
238 	if (socksize < sizeof(*sdl))
239 		socksize = sizeof(*sdl);
240 	socksize = ROUNDUP(socksize);
241 	ifasize = sizeof(struct ifaddr) + 2 * socksize;
242 	ifa = malloc(ifasize, M_IFADDR, M_WAITOK | M_ZERO);
243 	sdl = (struct sockaddr_dl *)(ifa + 1);
244 	sdl->sdl_len = socksize;
245 	sdl->sdl_family = AF_LINK;
246 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
247 	sdl->sdl_nlen = namelen;
248 	sdl->sdl_index = ifp->if_index;
249 	sdl->sdl_type = ifp->if_type;
250 	ifp->if_lladdr = ifa;
251 	ifa->ifa_ifp = ifp;
252 	ifa->ifa_rtrequest = link_rtrequest;
253 	ifa->ifa_addr = (struct sockaddr *)sdl;
254 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
255 	ifa->ifa_netmask = (struct sockaddr *)sdl;
256 	sdl->sdl_len = masklen;
257 	while (namelen != 0)
258 		sdl->sdl_data[--namelen] = 0xff;
259 	TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
260 
261 	EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
262 
263 	ifq = &ifp->if_snd;
264 	ifq->altq_type = 0;
265 	ifq->altq_disc = NULL;
266 	ifq->altq_flags &= ALTQF_CANTCHANGE;
267 	ifq->altq_tbr = NULL;
268 	ifq->altq_ifp = ifp;
269 	ifq_set_classic(ifq);
270 
271 	if (!SLIST_EMPTY(&domains))
272 		if_attachdomain1(ifp);
273 
274 	/* Announce the interface. */
275 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
276 }
277 
278 static void
279 if_attachdomain(void *dummy)
280 {
281 	struct ifnet *ifp;
282 
283 	crit_enter();
284 	TAILQ_FOREACH(ifp, &ifnet, if_list)
285 		if_attachdomain1(ifp);
286 	crit_exit();
287 }
288 SYSINIT(domainifattach, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST,
289 	if_attachdomain, NULL);
290 
291 static void
292 if_attachdomain1(struct ifnet *ifp)
293 {
294 	struct domain *dp;
295 
296 	crit_enter();
297 
298 	/* address family dependent data region */
299 	bzero(ifp->if_afdata, sizeof(ifp->if_afdata));
300 	SLIST_FOREACH(dp, &domains, dom_next)
301 		if (dp->dom_ifattach)
302 			ifp->if_afdata[dp->dom_family] =
303 				(*dp->dom_ifattach)(ifp);
304 	crit_exit();
305 }
306 
307 /*
308  * Detach an interface, removing it from the
309  * list of "active" interfaces.
310  */
311 void
312 if_detach(struct ifnet *ifp)
313 {
314 	struct ifaddr *ifa;
315 	struct radix_node_head	*rnh;
316 	int i;
317 	struct domain *dp;
318 
319 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
320 
321 	/*
322 	 * Remove routes and flush queues.
323 	 */
324 	crit_enter();
325 #ifdef DEVICE_POLLING
326 	if (ifp->if_flags & IFF_POLLING)
327 		ether_poll_deregister(ifp);
328 #endif
329 	if_down(ifp);
330 
331 	if (ifq_is_enabled(&ifp->if_snd))
332 		altq_disable(&ifp->if_snd);
333 	if (ifq_is_attached(&ifp->if_snd))
334 		altq_detach(&ifp->if_snd);
335 
336 	/*
337 	 * Clean up all addresses.
338 	 */
339 	ifp->if_lladdr = NULL;
340 
341 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
342 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
343 #ifdef INET
344 		/* XXX: Ugly!! ad hoc just for INET */
345 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
346 			struct ifaliasreq ifr;
347 
348 			bzero(&ifr, sizeof ifr);
349 			ifr.ifra_addr = *ifa->ifa_addr;
350 			if (ifa->ifa_dstaddr)
351 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
352 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
353 				       NULL) == 0)
354 				continue;
355 		}
356 #endif /* INET */
357 #ifdef INET6
358 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
359 			in6_purgeaddr(ifa);
360 			/* ifp_addrhead is already updated */
361 			continue;
362 		}
363 #endif /* INET6 */
364 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
365 		IFAFREE(ifa);
366 	}
367 
368 #ifdef INET
369 	/*
370 	 * Remove all IPv4 kernel structures related to ifp.
371 	 */
372 	in_ifdetach(ifp);
373 #endif
374 
375 #ifdef INET6
376 	/*
377 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
378 	 * before removing routing entries below, since IPv6 interface direct
379 	 * routes are expected to be removed by the IPv6-specific kernel API.
380 	 * Otherwise, the kernel will detect some inconsistency and bark it.
381 	 */
382 	in6_ifdetach(ifp);
383 #endif
384 
385 	/*
386 	 * Delete all remaining routes using this interface
387 	 * Unfortuneatly the only way to do this is to slog through
388 	 * the entire routing table looking for routes which point
389 	 * to this interface...oh well...
390 	 */
391 	for (i = 1; i <= AF_MAX; i++) {
392 		if ((rnh = rt_tables[i]) == NULL)
393 			continue;
394 		rnh->rnh_walktree(rnh, if_rtdel, ifp);
395 	}
396 
397 	/* Announce that the interface is gone. */
398 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
399 
400 	SLIST_FOREACH(dp, &domains, dom_next)
401 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
402 			(*dp->dom_ifdetach)(ifp,
403 				ifp->if_afdata[dp->dom_family]);
404 
405 	/*
406 	 * Remove interface from ifindex2ifp[] and maybe decrement if_index.
407 	 */
408 	ifindex2ifnet[ifp->if_index] = NULL;
409 	while (if_index > 0 && ifindex2ifnet[if_index] == NULL)
410 		if_index--;
411 
412 	TAILQ_REMOVE(&ifnet, ifp, if_link);
413 	crit_exit();
414 }
415 
416 /*
417  * Delete Routes for a Network Interface
418  *
419  * Called for each routing entry via the rnh->rnh_walktree() call above
420  * to delete all route entries referencing a detaching network interface.
421  *
422  * Arguments:
423  *	rn	pointer to node in the routing table
424  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
425  *
426  * Returns:
427  *	0	successful
428  *	errno	failed - reason indicated
429  *
430  */
431 static int
432 if_rtdel(struct radix_node *rn, void *arg)
433 {
434 	struct rtentry	*rt = (struct rtentry *)rn;
435 	struct ifnet	*ifp = arg;
436 	int		err;
437 
438 	if (rt->rt_ifp == ifp) {
439 
440 		/*
441 		 * Protect (sorta) against walktree recursion problems
442 		 * with cloned routes
443 		 */
444 		if (!(rt->rt_flags & RTF_UP))
445 			return (0);
446 
447 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
448 				rt_mask(rt), rt->rt_flags,
449 				(struct rtentry **) NULL);
450 		if (err) {
451 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
452 		}
453 	}
454 
455 	return (0);
456 }
457 
458 /*
459  * Create a clone network interface.
460  */
461 int
462 if_clone_create(char *name, int len)
463 {
464 	struct if_clone *ifc;
465 	char *dp;
466 	int wildcard, bytoff, bitoff;
467 	int unit;
468 	int err;
469 
470 	ifc = if_clone_lookup(name, &unit);
471 	if (ifc == NULL)
472 		return (EINVAL);
473 
474 	if (ifunit(name) != NULL)
475 		return (EEXIST);
476 
477 	bytoff = bitoff = 0;
478 	wildcard = (unit < 0);
479 	/*
480 	 * Find a free unit if none was given.
481 	 */
482 	if (wildcard) {
483 		while (bytoff < ifc->ifc_bmlen &&
484 		    ifc->ifc_units[bytoff] == 0xff)
485 			bytoff++;
486 		if (bytoff >= ifc->ifc_bmlen)
487 			return (ENOSPC);
488 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
489 			bitoff++;
490 		unit = (bytoff << 3) + bitoff;
491 	}
492 
493 	if (unit > ifc->ifc_maxunit)
494 		return (ENXIO);
495 
496 	err = (*ifc->ifc_create)(ifc, unit);
497 	if (err != 0)
498 		return (err);
499 
500 	if (!wildcard) {
501 		bytoff = unit >> 3;
502 		bitoff = unit - (bytoff << 3);
503 	}
504 
505 	/*
506 	 * Allocate the unit in the bitmap.
507 	 */
508 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
509 	    ("%s: bit is already set", __func__));
510 	ifc->ifc_units[bytoff] |= (1 << bitoff);
511 
512 	/* In the wildcard case, we need to update the name. */
513 	if (wildcard) {
514 		for (dp = name; *dp != '\0'; dp++);
515 		if (snprintf(dp, len - (dp-name), "%d", unit) >
516 		    len - (dp-name) - 1) {
517 			/*
518 			 * This can only be a programmer error and
519 			 * there's no straightforward way to recover if
520 			 * it happens.
521 			 */
522 			panic("if_clone_create(): interface name too long");
523 		}
524 
525 	}
526 
527 	EVENTHANDLER_INVOKE(if_clone_event, ifc);
528 
529 	return (0);
530 }
531 
532 /*
533  * Destroy a clone network interface.
534  */
535 int
536 if_clone_destroy(const char *name)
537 {
538 	struct if_clone *ifc;
539 	struct ifnet *ifp;
540 	int bytoff, bitoff;
541 	int unit;
542 
543 	ifc = if_clone_lookup(name, &unit);
544 	if (ifc == NULL)
545 		return (EINVAL);
546 
547 	if (unit < ifc->ifc_minifs)
548 		return (EINVAL);
549 
550 	ifp = ifunit(name);
551 	if (ifp == NULL)
552 		return (ENXIO);
553 
554 	if (ifc->ifc_destroy == NULL)
555 		return (EOPNOTSUPP);
556 
557 	(*ifc->ifc_destroy)(ifp);
558 
559 	/*
560 	 * Compute offset in the bitmap and deallocate the unit.
561 	 */
562 	bytoff = unit >> 3;
563 	bitoff = unit - (bytoff << 3);
564 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
565 	    ("%s: bit is already cleared", __func__));
566 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
567 	return (0);
568 }
569 
570 /*
571  * Look up a network interface cloner.
572  */
573 struct if_clone *
574 if_clone_lookup(const char *name, int *unitp)
575 {
576 	struct if_clone *ifc;
577 	const char *cp;
578 	int i;
579 
580 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
581 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
582 			if (ifc->ifc_name[i] != *cp)
583 				goto next_ifc;
584 		}
585 		goto found_name;
586  next_ifc:
587 		ifc = LIST_NEXT(ifc, ifc_list);
588 	}
589 
590 	/* No match. */
591 	return ((struct if_clone *)NULL);
592 
593  found_name:
594 	if (*cp == '\0') {
595 		i = -1;
596 	} else {
597 		for (i = 0; *cp != '\0'; cp++) {
598 			if (*cp < '0' || *cp > '9') {
599 				/* Bogus unit number. */
600 				return (NULL);
601 			}
602 			i = (i * 10) + (*cp - '0');
603 		}
604 	}
605 
606 	if (unitp != NULL)
607 		*unitp = i;
608 	return (ifc);
609 }
610 
611 /*
612  * Register a network interface cloner.
613  */
614 void
615 if_clone_attach(struct if_clone *ifc)
616 {
617 	int bytoff, bitoff;
618 	int err;
619 	int len, maxclone;
620 	int unit;
621 
622 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
623 	    ("%s: %s requested more units then allowed (%d > %d)",
624 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
625 	    ifc->ifc_maxunit + 1));
626 	/*
627 	 * Compute bitmap size and allocate it.
628 	 */
629 	maxclone = ifc->ifc_maxunit + 1;
630 	len = maxclone >> 3;
631 	if ((len << 3) < maxclone)
632 		len++;
633 	ifc->ifc_units = malloc(len, M_CLONE, M_WAITOK | M_ZERO);
634 	ifc->ifc_bmlen = len;
635 
636 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
637 	if_cloners_count++;
638 
639 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
640 		err = (*ifc->ifc_create)(ifc, unit);
641 		KASSERT(err == 0,
642 		    ("%s: failed to create required interface %s%d",
643 		    __func__, ifc->ifc_name, unit));
644 
645 		/* Allocate the unit in the bitmap. */
646 		bytoff = unit >> 3;
647 		bitoff = unit - (bytoff << 3);
648 		ifc->ifc_units[bytoff] |= (1 << bitoff);
649 	}
650 }
651 
652 /*
653  * Unregister a network interface cloner.
654  */
655 void
656 if_clone_detach(struct if_clone *ifc)
657 {
658 
659 	LIST_REMOVE(ifc, ifc_list);
660 	free(ifc->ifc_units, M_CLONE);
661 	if_cloners_count--;
662 }
663 
664 /*
665  * Provide list of interface cloners to userspace.
666  */
667 int
668 if_clone_list(struct if_clonereq *ifcr)
669 {
670 	char outbuf[IFNAMSIZ], *dst;
671 	struct if_clone *ifc;
672 	int count, error = 0;
673 
674 	ifcr->ifcr_total = if_cloners_count;
675 	if ((dst = ifcr->ifcr_buffer) == NULL) {
676 		/* Just asking how many there are. */
677 		return (0);
678 	}
679 
680 	if (ifcr->ifcr_count < 0)
681 		return (EINVAL);
682 
683 	count = (if_cloners_count < ifcr->ifcr_count) ?
684 	    if_cloners_count : ifcr->ifcr_count;
685 
686 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
687 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
688 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
689 		error = copyout(outbuf, dst, IFNAMSIZ);
690 		if (error)
691 			break;
692 	}
693 
694 	return (error);
695 }
696 
697 /*
698  * Locate an interface based on a complete address.
699  */
700 struct ifaddr *
701 ifa_ifwithaddr(struct sockaddr *addr)
702 {
703 	struct ifnet *ifp;
704 	struct ifaddr *ifa;
705 
706 	TAILQ_FOREACH(ifp, &ifnet, if_link)
707 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
708 		if (ifa->ifa_addr->sa_family != addr->sa_family)
709 			continue;
710 		if (sa_equal(addr, ifa->ifa_addr))
711 			return (ifa);
712 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
713 		    /* IPv6 doesn't have broadcast */
714 		    ifa->ifa_broadaddr->sa_len != 0 &&
715 		    sa_equal(ifa->ifa_broadaddr, addr))
716 			return (ifa);
717 	}
718 	return ((struct ifaddr *)NULL);
719 }
720 /*
721  * Locate the point to point interface with a given destination address.
722  */
723 struct ifaddr *
724 ifa_ifwithdstaddr(struct sockaddr *addr)
725 {
726 	struct ifnet *ifp;
727 	struct ifaddr *ifa;
728 
729 	TAILQ_FOREACH(ifp, &ifnet, if_link)
730 	    if (ifp->if_flags & IFF_POINTOPOINT)
731 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
732 			if (ifa->ifa_addr->sa_family != addr->sa_family)
733 				continue;
734 			if (ifa->ifa_dstaddr &&
735 			    sa_equal(addr, ifa->ifa_dstaddr))
736 				return (ifa);
737 	}
738 	return ((struct ifaddr *)NULL);
739 }
740 
741 /*
742  * Find an interface on a specific network.  If many, choice
743  * is most specific found.
744  */
745 struct ifaddr *
746 ifa_ifwithnet(struct sockaddr *addr)
747 {
748 	struct ifnet *ifp;
749 	struct ifaddr *ifa;
750 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
751 	u_int af = addr->sa_family;
752 	char *addr_data = addr->sa_data, *cplim;
753 
754 	/*
755 	 * AF_LINK addresses can be looked up directly by their index number,
756 	 * so do that if we can.
757 	 */
758 	if (af == AF_LINK) {
759 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
760 
761 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
762 		return (ifindex2ifnet[sdl->sdl_index]->if_lladdr);
763 	}
764 
765 	/*
766 	 * Scan though each interface, looking for ones that have
767 	 * addresses in this address family.
768 	 */
769 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
770 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
771 			char *cp, *cp2, *cp3;
772 
773 			if (ifa->ifa_addr->sa_family != af)
774 next:				continue;
775 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
776 				/*
777 				 * This is a bit broken as it doesn't
778 				 * take into account that the remote end may
779 				 * be a single node in the network we are
780 				 * looking for.
781 				 * The trouble is that we don't know the
782 				 * netmask for the remote end.
783 				 */
784 				if (ifa->ifa_dstaddr != NULL &&
785 				    sa_equal(addr, ifa->ifa_dstaddr))
786 					return (ifa);
787 			} else {
788 				/*
789 				 * if we have a special address handler,
790 				 * then use it instead of the generic one.
791 				 */
792 				if (ifa->ifa_claim_addr) {
793 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
794 						return (ifa);
795 					} else {
796 						continue;
797 					}
798 				}
799 
800 				/*
801 				 * Scan all the bits in the ifa's address.
802 				 * If a bit dissagrees with what we are
803 				 * looking for, mask it with the netmask
804 				 * to see if it really matters.
805 				 * (A byte at a time)
806 				 */
807 				if (ifa->ifa_netmask == 0)
808 					continue;
809 				cp = addr_data;
810 				cp2 = ifa->ifa_addr->sa_data;
811 				cp3 = ifa->ifa_netmask->sa_data;
812 				cplim = ifa->ifa_netmask->sa_len +
813 					(char *)ifa->ifa_netmask;
814 				while (cp3 < cplim)
815 					if ((*cp++ ^ *cp2++) & *cp3++)
816 						goto next; /* next address! */
817 				/*
818 				 * If the netmask of what we just found
819 				 * is more specific than what we had before
820 				 * (if we had one) then remember the new one
821 				 * before continuing to search
822 				 * for an even better one.
823 				 */
824 				if (ifa_maybe == 0 ||
825 				    rn_refines((char *)ifa->ifa_netmask,
826 					       (char *)ifa_maybe->ifa_netmask))
827 					ifa_maybe = ifa;
828 			}
829 		}
830 	}
831 	return (ifa_maybe);
832 }
833 
834 /*
835  * Find an interface address specific to an interface best matching
836  * a given address.
837  */
838 struct ifaddr *
839 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
840 {
841 	struct ifaddr *ifa;
842 	char *cp, *cp2, *cp3;
843 	char *cplim;
844 	struct ifaddr *ifa_maybe = 0;
845 	u_int af = addr->sa_family;
846 
847 	if (af >= AF_MAX)
848 		return (0);
849 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
850 		if (ifa->ifa_addr->sa_family != af)
851 			continue;
852 		if (ifa_maybe == 0)
853 			ifa_maybe = ifa;
854 		if (ifa->ifa_netmask == NULL) {
855 			if (sa_equal(addr, ifa->ifa_addr) ||
856 			    (ifa->ifa_dstaddr != NULL &&
857 			     sa_equal(addr, ifa->ifa_dstaddr)))
858 				return (ifa);
859 			continue;
860 		}
861 		if (ifp->if_flags & IFF_POINTOPOINT) {
862 			if (sa_equal(addr, ifa->ifa_dstaddr))
863 				return (ifa);
864 		} else {
865 			cp = addr->sa_data;
866 			cp2 = ifa->ifa_addr->sa_data;
867 			cp3 = ifa->ifa_netmask->sa_data;
868 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
869 			for (; cp3 < cplim; cp3++)
870 				if ((*cp++ ^ *cp2++) & *cp3)
871 					break;
872 			if (cp3 == cplim)
873 				return (ifa);
874 		}
875 	}
876 	return (ifa_maybe);
877 }
878 
879 #include <net/route.h>
880 
881 /*
882  * Default action when installing a route with a Link Level gateway.
883  * Lookup an appropriate real ifa to point to.
884  * This should be moved to /sys/net/link.c eventually.
885  */
886 static void
887 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
888 {
889 	struct ifaddr *ifa;
890 	struct sockaddr *dst;
891 	struct ifnet *ifp;
892 
893 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
894 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
895 		return;
896 	ifa = ifaof_ifpforaddr(dst, ifp);
897 	if (ifa != NULL) {
898 		IFAFREE(rt->rt_ifa);
899 		IFAREF(ifa);
900 		rt->rt_ifa = ifa;
901 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
902 			ifa->ifa_rtrequest(cmd, rt, info);
903 	}
904 }
905 
906 /*
907  * Mark an interface down and notify protocols of
908  * the transition.
909  * NOTE: must be called at splnet or eqivalent.
910  */
911 void
912 if_unroute(struct ifnet *ifp, int flag, int fam)
913 {
914 	struct ifaddr *ifa;
915 
916 	ifp->if_flags &= ~flag;
917 	getmicrotime(&ifp->if_lastchange);
918 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
919 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
920 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
921 	ifq_purge(&ifp->if_snd);
922 	rt_ifmsg(ifp);
923 }
924 
925 /*
926  * Mark an interface up and notify protocols of
927  * the transition.
928  * NOTE: must be called at splnet or eqivalent.
929  */
930 void
931 if_route(struct ifnet *ifp, int flag, int fam)
932 {
933 	struct ifaddr *ifa;
934 
935 	ifp->if_flags |= flag;
936 	getmicrotime(&ifp->if_lastchange);
937 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
938 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
939 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
940 	rt_ifmsg(ifp);
941 #ifdef INET6
942 	in6_if_up(ifp);
943 #endif
944 }
945 
946 /*
947  * Mark an interface down and notify protocols of the transition.  An
948  * interface going down is also considered to be a synchronizing event.
949  * We must ensure that all packet processing related to the interface
950  * has completed before we return so e.g. the caller can free the ifnet
951  * structure that the mbufs may be referencing.
952  *
953  * NOTE: must be called at splnet or eqivalent.
954  */
955 void
956 if_down(struct ifnet *ifp)
957 {
958 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
959 	netmsg_service_sync();
960 }
961 
962 /*
963  * Mark an interface up and notify protocols of
964  * the transition.
965  * NOTE: must be called at splnet or eqivalent.
966  */
967 void
968 if_up(struct ifnet *ifp)
969 {
970 
971 	if_route(ifp, IFF_UP, AF_UNSPEC);
972 }
973 
974 /*
975  * Handle interface watchdog timer routines.  Called
976  * from softclock, we decrement timers (if set) and
977  * call the appropriate interface routine on expiration.
978  */
979 static void
980 if_slowtimo(void *arg)
981 {
982 	struct ifnet *ifp;
983 
984 	crit_enter();
985 
986 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
987 		if (ifp->if_timer == 0 || --ifp->if_timer)
988 			continue;
989 		if (ifp->if_watchdog) {
990 			if (lwkt_serialize_try(ifp->if_serializer)) {
991 				(*ifp->if_watchdog)(ifp);
992 				lwkt_serialize_exit(ifp->if_serializer);
993 			} else {
994 				/* try again next timeout */
995 				++ifp->if_timer;
996 			}
997 		}
998 	}
999 
1000 	crit_exit();
1001 
1002 	callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1003 }
1004 
1005 /*
1006  * Map interface name to
1007  * interface structure pointer.
1008  */
1009 struct ifnet *
1010 ifunit(const char *name)
1011 {
1012 	struct ifnet *ifp;
1013 
1014 	/*
1015 	 * Search all the interfaces for this name/number
1016 	 */
1017 
1018 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1019 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
1020 			break;
1021 	}
1022 	return (ifp);
1023 }
1024 
1025 
1026 /*
1027  * Map interface name in a sockaddr_dl to
1028  * interface structure pointer.
1029  */
1030 struct ifnet *
1031 if_withname(struct sockaddr *sa)
1032 {
1033 	char ifname[IFNAMSIZ+1];
1034 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1035 
1036 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1037 	     (sdl->sdl_nlen > IFNAMSIZ) )
1038 		return NULL;
1039 
1040 	/*
1041 	 * ifunit wants a null-terminated name.  It may not be null-terminated
1042 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
1043 	 * and there might not be room to put the trailing null anyway, so we
1044 	 * make a local copy that we know we can null terminate safely.
1045 	 */
1046 
1047 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1048 	ifname[sdl->sdl_nlen] = '\0';
1049 	return ifunit(ifname);
1050 }
1051 
1052 
1053 /*
1054  * Interface ioctls.
1055  */
1056 int
1057 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct thread *td)
1058 {
1059 	struct ifnet *ifp;
1060 	struct ifreq *ifr;
1061 	struct ifstat *ifs;
1062 	int error;
1063 	short oif_flags;
1064 	int new_flags;
1065 	size_t namelen, onamelen;
1066 	char new_name[IFNAMSIZ];
1067 	struct ifaddr *ifa;
1068 	struct sockaddr_dl *sdl;
1069 
1070 	switch (cmd) {
1071 
1072 	case SIOCGIFCONF:
1073 	case OSIOCGIFCONF:
1074 		return (ifconf(cmd, data, td));
1075 	}
1076 	ifr = (struct ifreq *)data;
1077 
1078 	switch (cmd) {
1079 	case SIOCIFCREATE:
1080 	case SIOCIFDESTROY:
1081 		if ((error = suser(td)) != 0)
1082 			return (error);
1083 		return ((cmd == SIOCIFCREATE) ?
1084 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1085 			if_clone_destroy(ifr->ifr_name));
1086 
1087 	case SIOCIFGCLONERS:
1088 		return (if_clone_list((struct if_clonereq *)data));
1089 	}
1090 
1091 	ifp = ifunit(ifr->ifr_name);
1092 	if (ifp == 0)
1093 		return (ENXIO);
1094 	switch (cmd) {
1095 
1096 	case SIOCGIFFLAGS:
1097 		ifr->ifr_flags = ifp->if_flags;
1098 		ifr->ifr_flagshigh = ifp->if_flags >> 16;
1099 		break;
1100 
1101 	case SIOCGIFCAP:
1102 		ifr->ifr_reqcap = ifp->if_capabilities;
1103 		ifr->ifr_curcap = ifp->if_capenable;
1104 		break;
1105 
1106 	case SIOCGIFMETRIC:
1107 		ifr->ifr_metric = ifp->if_metric;
1108 		break;
1109 
1110 	case SIOCGIFMTU:
1111 		ifr->ifr_mtu = ifp->if_mtu;
1112 		break;
1113 
1114 	case SIOCGIFPHYS:
1115 		ifr->ifr_phys = ifp->if_physical;
1116 		break;
1117 
1118 	case SIOCSIFFLAGS:
1119 		error = suser(td);
1120 		if (error)
1121 			return (error);
1122 		new_flags = (ifr->ifr_flags & 0xffff) |
1123 		    (ifr->ifr_flagshigh << 16);
1124 		if (ifp->if_flags & IFF_SMART) {
1125 			/* Smart drivers twiddle their own routes */
1126 		} else if (ifp->if_flags & IFF_UP &&
1127 		    (new_flags & IFF_UP) == 0) {
1128 			crit_enter();
1129 			if_down(ifp);
1130 			crit_exit();
1131 		} else if (new_flags & IFF_UP &&
1132 		    (ifp->if_flags & IFF_UP) == 0) {
1133 			crit_enter();
1134 			if_up(ifp);
1135 			crit_exit();
1136 		}
1137 
1138 #ifdef DEVICE_POLLING
1139 		if ((new_flags ^ ifp->if_flags) & IFF_POLLING) {
1140 			if (new_flags & IFF_POLLING) {
1141 				ether_poll_register(ifp);
1142 			} else {
1143 				ether_poll_deregister(ifp);
1144 			}
1145 		}
1146 #endif
1147 
1148 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1149 			(new_flags &~ IFF_CANTCHANGE);
1150 		if (new_flags & IFF_PPROMISC) {
1151 			/* Permanently promiscuous mode requested */
1152 			ifp->if_flags |= IFF_PROMISC;
1153 		} else if (ifp->if_pcount == 0) {
1154 			ifp->if_flags &= ~IFF_PROMISC;
1155 		}
1156 		if (ifp->if_ioctl) {
1157 			lwkt_serialize_enter(ifp->if_serializer);
1158 			ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
1159 			lwkt_serialize_exit(ifp->if_serializer);
1160 		}
1161 		getmicrotime(&ifp->if_lastchange);
1162 		break;
1163 
1164 	case SIOCSIFCAP:
1165 		error = suser(td);
1166 		if (error)
1167 			return (error);
1168 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1169 			return (EINVAL);
1170 		lwkt_serialize_enter(ifp->if_serializer);
1171 		ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
1172 		lwkt_serialize_exit(ifp->if_serializer);
1173 		break;
1174 
1175 	case SIOCSIFNAME:
1176 		error = suser(td);
1177 		if (error != 0)
1178 			return (error);
1179 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1180 		if (error != 0)
1181 			return (error);
1182 		if (new_name[0] == '\0')
1183 			return (EINVAL);
1184 		if (ifunit(new_name) != NULL)
1185 			return (EEXIST);
1186 
1187 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1188 
1189 		/* Announce the departure of the interface. */
1190 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1191 
1192 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1193 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1194 		/* XXX IFA_LOCK(ifa); */
1195 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1196 		namelen = strlen(new_name);
1197 		onamelen = sdl->sdl_nlen;
1198 		/*
1199 		 * Move the address if needed.  This is safe because we
1200 		 * allocate space for a name of length IFNAMSIZ when we
1201 		 * create this in if_attach().
1202 		 */
1203 		if (namelen != onamelen) {
1204 			bcopy(sdl->sdl_data + onamelen,
1205 			    sdl->sdl_data + namelen, sdl->sdl_alen);
1206 		}
1207 		bcopy(new_name, sdl->sdl_data, namelen);
1208 		sdl->sdl_nlen = namelen;
1209 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1210 		bzero(sdl->sdl_data, onamelen);
1211 		while (namelen != 0)
1212 			sdl->sdl_data[--namelen] = 0xff;
1213 		/* XXX IFA_UNLOCK(ifa) */
1214 
1215 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1216 
1217 		/* Announce the return of the interface. */
1218 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1219 		break;
1220 
1221 	case SIOCSIFMETRIC:
1222 		error = suser(td);
1223 		if (error)
1224 			return (error);
1225 		ifp->if_metric = ifr->ifr_metric;
1226 		getmicrotime(&ifp->if_lastchange);
1227 		break;
1228 
1229 	case SIOCSIFPHYS:
1230 		error = suser(td);
1231 		if (error)
1232 			return error;
1233 		if (!ifp->if_ioctl)
1234 		        return EOPNOTSUPP;
1235 		lwkt_serialize_enter(ifp->if_serializer);
1236 		error = ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
1237 		lwkt_serialize_exit(ifp->if_serializer);
1238 		if (error == 0)
1239 			getmicrotime(&ifp->if_lastchange);
1240 		return (error);
1241 
1242 	case SIOCSIFMTU:
1243 	{
1244 		u_long oldmtu = ifp->if_mtu;
1245 
1246 		error = suser(td);
1247 		if (error)
1248 			return (error);
1249 		if (ifp->if_ioctl == NULL)
1250 			return (EOPNOTSUPP);
1251 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1252 			return (EINVAL);
1253 		lwkt_serialize_enter(ifp->if_serializer);
1254 		error = ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
1255 		lwkt_serialize_exit(ifp->if_serializer);
1256 		if (error == 0) {
1257 			getmicrotime(&ifp->if_lastchange);
1258 			rt_ifmsg(ifp);
1259 		}
1260 		/*
1261 		 * If the link MTU changed, do network layer specific procedure.
1262 		 */
1263 		if (ifp->if_mtu != oldmtu) {
1264 #ifdef INET6
1265 			nd6_setmtu(ifp);
1266 #endif
1267 		}
1268 		return (error);
1269 	}
1270 
1271 	case SIOCADDMULTI:
1272 	case SIOCDELMULTI:
1273 		error = suser(td);
1274 		if (error)
1275 			return (error);
1276 
1277 		/* Don't allow group membership on non-multicast interfaces. */
1278 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1279 			return EOPNOTSUPP;
1280 
1281 		/* Don't let users screw up protocols' entries. */
1282 		if (ifr->ifr_addr.sa_family != AF_LINK)
1283 			return EINVAL;
1284 
1285 		if (cmd == SIOCADDMULTI) {
1286 			struct ifmultiaddr *ifma;
1287 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1288 		} else {
1289 			error = if_delmulti(ifp, &ifr->ifr_addr);
1290 		}
1291 		if (error == 0)
1292 			getmicrotime(&ifp->if_lastchange);
1293 		return error;
1294 
1295 	case SIOCSIFPHYADDR:
1296 	case SIOCDIFPHYADDR:
1297 #ifdef INET6
1298 	case SIOCSIFPHYADDR_IN6:
1299 #endif
1300 	case SIOCSLIFPHYADDR:
1301         case SIOCSIFMEDIA:
1302 	case SIOCSIFGENERIC:
1303 		error = suser(td);
1304 		if (error)
1305 			return (error);
1306 		if (ifp->if_ioctl == 0)
1307 			return (EOPNOTSUPP);
1308 		lwkt_serialize_enter(ifp->if_serializer);
1309 		error = ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
1310 		lwkt_serialize_exit(ifp->if_serializer);
1311 		if (error == 0)
1312 			getmicrotime(&ifp->if_lastchange);
1313 		return error;
1314 
1315 	case SIOCGIFSTATUS:
1316 		ifs = (struct ifstat *)data;
1317 		ifs->ascii[0] = '\0';
1318 
1319 	case SIOCGIFPSRCADDR:
1320 	case SIOCGIFPDSTADDR:
1321 	case SIOCGLIFPHYADDR:
1322 	case SIOCGIFMEDIA:
1323 	case SIOCGIFGENERIC:
1324 		if (ifp->if_ioctl == NULL)
1325 			return (EOPNOTSUPP);
1326 		lwkt_serialize_enter(ifp->if_serializer);
1327 		error = ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
1328 		lwkt_serialize_exit(ifp->if_serializer);
1329 		return (error);
1330 
1331 	case SIOCSIFLLADDR:
1332 		error = suser(td);
1333 		if (error)
1334 			return (error);
1335 		return if_setlladdr(ifp,
1336 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1337 
1338 	default:
1339 		oif_flags = ifp->if_flags;
1340 		if (so->so_proto == 0)
1341 			return (EOPNOTSUPP);
1342 #ifndef COMPAT_43
1343 		error = so_pru_control(so, cmd, data, ifp, td);
1344 #else
1345 	    {
1346 		int ocmd = cmd;
1347 
1348 		switch (cmd) {
1349 
1350 		case SIOCSIFDSTADDR:
1351 		case SIOCSIFADDR:
1352 		case SIOCSIFBRDADDR:
1353 		case SIOCSIFNETMASK:
1354 #if BYTE_ORDER != BIG_ENDIAN
1355 			if (ifr->ifr_addr.sa_family == 0 &&
1356 			    ifr->ifr_addr.sa_len < 16) {
1357 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1358 				ifr->ifr_addr.sa_len = 16;
1359 			}
1360 #else
1361 			if (ifr->ifr_addr.sa_len == 0)
1362 				ifr->ifr_addr.sa_len = 16;
1363 #endif
1364 			break;
1365 
1366 		case OSIOCGIFADDR:
1367 			cmd = SIOCGIFADDR;
1368 			break;
1369 
1370 		case OSIOCGIFDSTADDR:
1371 			cmd = SIOCGIFDSTADDR;
1372 			break;
1373 
1374 		case OSIOCGIFBRDADDR:
1375 			cmd = SIOCGIFBRDADDR;
1376 			break;
1377 
1378 		case OSIOCGIFNETMASK:
1379 			cmd = SIOCGIFNETMASK;
1380 		}
1381 		error =  so_pru_control(so, cmd, data, ifp, td);
1382 		switch (ocmd) {
1383 
1384 		case OSIOCGIFADDR:
1385 		case OSIOCGIFDSTADDR:
1386 		case OSIOCGIFBRDADDR:
1387 		case OSIOCGIFNETMASK:
1388 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1389 
1390 		}
1391 	    }
1392 #endif /* COMPAT_43 */
1393 
1394 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1395 #ifdef INET6
1396 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1397 			if (ifp->if_flags & IFF_UP) {
1398 				crit_enter();
1399 				in6_if_up(ifp);
1400 				crit_exit();
1401 			}
1402 #endif
1403 		}
1404 		return (error);
1405 
1406 	}
1407 	return (0);
1408 }
1409 
1410 /*
1411  * Set/clear promiscuous mode on interface ifp based on the truth value
1412  * of pswitch.  The calls are reference counted so that only the first
1413  * "on" request actually has an effect, as does the final "off" request.
1414  * Results are undefined if the "off" and "on" requests are not matched.
1415  */
1416 int
1417 ifpromisc(struct ifnet *ifp, int pswitch)
1418 {
1419 	struct ifreq ifr;
1420 	int error;
1421 	int oldflags;
1422 
1423 	oldflags = ifp->if_flags;
1424 	if (ifp->if_flags & IFF_PPROMISC) {
1425 		/* Do nothing if device is in permanently promiscuous mode */
1426 		ifp->if_pcount += pswitch ? 1 : -1;
1427 		return (0);
1428 	}
1429 	if (pswitch) {
1430 		/*
1431 		 * If the device is not configured up, we cannot put it in
1432 		 * promiscuous mode.
1433 		 */
1434 		if ((ifp->if_flags & IFF_UP) == 0)
1435 			return (ENETDOWN);
1436 		if (ifp->if_pcount++ != 0)
1437 			return (0);
1438 		ifp->if_flags |= IFF_PROMISC;
1439 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1440 		    ifp->if_xname);
1441 	} else {
1442 		if (--ifp->if_pcount > 0)
1443 			return (0);
1444 		ifp->if_flags &= ~IFF_PROMISC;
1445 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1446 		    ifp->if_xname);
1447 	}
1448 	ifr.ifr_flags = ifp->if_flags;
1449 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
1450 	lwkt_serialize_enter(ifp->if_serializer);
1451 	error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1452 				 (struct ucred *)NULL);
1453 	lwkt_serialize_exit(ifp->if_serializer);
1454 	if (error == 0)
1455 		rt_ifmsg(ifp);
1456 	else
1457 		ifp->if_flags = oldflags;
1458 	return error;
1459 }
1460 
1461 /*
1462  * Return interface configuration
1463  * of system.  List may be used
1464  * in later ioctl's (above) to get
1465  * other information.
1466  */
1467 static int
1468 ifconf(u_long cmd, caddr_t data, struct thread *td)
1469 {
1470 	struct ifconf *ifc = (struct ifconf *)data;
1471 	struct ifnet *ifp;
1472 	struct ifaddr *ifa;
1473 	struct sockaddr *sa;
1474 	struct ifreq ifr, *ifrp;
1475 	int space = ifc->ifc_len, error = 0;
1476 
1477 	ifrp = ifc->ifc_req;
1478 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1479 		int addrs;
1480 
1481 		if (space <= sizeof ifr)
1482 			break;
1483 
1484 		/*
1485 		 * Zero the stack declared structure first to prevent
1486 		 * memory disclosure.
1487 		 */
1488 		bzero(&ifr, sizeof(ifr));
1489 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1490 		    >= sizeof(ifr.ifr_name)) {
1491 			error = ENAMETOOLONG;
1492 			break;
1493 		}
1494 
1495 		addrs = 0;
1496 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1497 			if (space <= sizeof ifr)
1498 				break;
1499 			sa = ifa->ifa_addr;
1500 			if (td->td_proc->p_ucred->cr_prison &&
1501 			    prison_if(td, sa))
1502 				continue;
1503 			addrs++;
1504 #ifdef COMPAT_43
1505 			if (cmd == OSIOCGIFCONF) {
1506 				struct osockaddr *osa =
1507 					 (struct osockaddr *)&ifr.ifr_addr;
1508 				ifr.ifr_addr = *sa;
1509 				osa->sa_family = sa->sa_family;
1510 				error = copyout(&ifr, ifrp, sizeof ifr);
1511 				ifrp++;
1512 			} else
1513 #endif
1514 			if (sa->sa_len <= sizeof(*sa)) {
1515 				ifr.ifr_addr = *sa;
1516 				error = copyout(&ifr, ifrp, sizeof ifr);
1517 				ifrp++;
1518 			} else {
1519 				if (space < (sizeof ifr) + sa->sa_len -
1520 					    sizeof(*sa))
1521 					break;
1522 				space -= sa->sa_len - sizeof(*sa);
1523 				error = copyout(&ifr, ifrp,
1524 						sizeof ifr.ifr_name);
1525 				if (error == 0)
1526 					error = copyout(sa, &ifrp->ifr_addr,
1527 							sa->sa_len);
1528 				ifrp = (struct ifreq *)
1529 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1530 			}
1531 			if (error)
1532 				break;
1533 			space -= sizeof ifr;
1534 		}
1535 		if (error)
1536 			break;
1537 		if (!addrs) {
1538 			bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1539 			error = copyout(&ifr, ifrp, sizeof ifr);
1540 			if (error)
1541 				break;
1542 			space -= sizeof ifr;
1543 			ifrp++;
1544 		}
1545 	}
1546 	ifc->ifc_len -= space;
1547 	return (error);
1548 }
1549 
1550 /*
1551  * Just like if_promisc(), but for all-multicast-reception mode.
1552  */
1553 int
1554 if_allmulti(struct ifnet *ifp, int onswitch)
1555 {
1556 	int error = 0;
1557 	struct ifreq ifr;
1558 
1559 	crit_enter();
1560 
1561 	if (onswitch) {
1562 		if (ifp->if_amcount++ == 0) {
1563 			ifp->if_flags |= IFF_ALLMULTI;
1564 			ifr.ifr_flags = ifp->if_flags;
1565 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1566 			lwkt_serialize_enter(ifp->if_serializer);
1567 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1568 					      (struct ucred *)NULL);
1569 			lwkt_serialize_exit(ifp->if_serializer);
1570 		}
1571 	} else {
1572 		if (ifp->if_amcount > 1) {
1573 			ifp->if_amcount--;
1574 		} else {
1575 			ifp->if_amcount = 0;
1576 			ifp->if_flags &= ~IFF_ALLMULTI;
1577 			ifr.ifr_flags = ifp->if_flags;
1578 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1579 			lwkt_serialize_enter(ifp->if_serializer);
1580 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1581 					      (struct ucred *)NULL);
1582 			lwkt_serialize_exit(ifp->if_serializer);
1583 		}
1584 	}
1585 
1586 	crit_exit();
1587 
1588 	if (error == 0)
1589 		rt_ifmsg(ifp);
1590 	return error;
1591 }
1592 
1593 /*
1594  * Add a multicast listenership to the interface in question.
1595  * The link layer provides a routine which converts
1596  */
1597 int
1598 if_addmulti(
1599 	struct ifnet *ifp,	/* interface to manipulate */
1600 	struct sockaddr *sa,	/* address to add */
1601 	struct ifmultiaddr **retifma)
1602 {
1603 	struct sockaddr *llsa, *dupsa;
1604 	int error;
1605 	struct ifmultiaddr *ifma;
1606 
1607 	/*
1608 	 * If the matching multicast address already exists
1609 	 * then don't add a new one, just add a reference
1610 	 */
1611 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1612 		if (sa_equal(sa, ifma->ifma_addr)) {
1613 			ifma->ifma_refcount++;
1614 			if (retifma)
1615 				*retifma = ifma;
1616 			return 0;
1617 		}
1618 	}
1619 
1620 	/*
1621 	 * Give the link layer a chance to accept/reject it, and also
1622 	 * find out which AF_LINK address this maps to, if it isn't one
1623 	 * already.
1624 	 */
1625 	if (ifp->if_resolvemulti) {
1626 		lwkt_serialize_enter(ifp->if_serializer);
1627 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1628 		lwkt_serialize_exit(ifp->if_serializer);
1629 		if (error)
1630 			return error;
1631 	} else {
1632 		llsa = 0;
1633 	}
1634 
1635 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1636 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1637 	bcopy(sa, dupsa, sa->sa_len);
1638 
1639 	ifma->ifma_addr = dupsa;
1640 	ifma->ifma_lladdr = llsa;
1641 	ifma->ifma_ifp = ifp;
1642 	ifma->ifma_refcount = 1;
1643 	ifma->ifma_protospec = 0;
1644 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1645 
1646 	/*
1647 	 * Some network interfaces can scan the address list at
1648 	 * interrupt time; lock them out.
1649 	 */
1650 	crit_enter();
1651 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1652 	crit_exit();
1653 	*retifma = ifma;
1654 
1655 	if (llsa != 0) {
1656 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1657 			if (sa_equal(ifma->ifma_addr, llsa))
1658 				break;
1659 		}
1660 		if (ifma) {
1661 			ifma->ifma_refcount++;
1662 		} else {
1663 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1664 			       M_IFMADDR, M_WAITOK);
1665 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1666 			       M_IFMADDR, M_WAITOK);
1667 			bcopy(llsa, dupsa, llsa->sa_len);
1668 			ifma->ifma_addr = dupsa;
1669 			ifma->ifma_ifp = ifp;
1670 			ifma->ifma_refcount = 1;
1671 			crit_enter();
1672 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1673 			crit_exit();
1674 		}
1675 	}
1676 	/*
1677 	 * We are certain we have added something, so call down to the
1678 	 * interface to let them know about it.
1679 	 */
1680 	crit_enter();
1681 	lwkt_serialize_enter(ifp->if_serializer);
1682 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1683 	lwkt_serialize_exit(ifp->if_serializer);
1684 	crit_exit();
1685 
1686 	return 0;
1687 }
1688 
1689 /*
1690  * Remove a reference to a multicast address on this interface.  Yell
1691  * if the request does not match an existing membership.
1692  */
1693 int
1694 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1695 {
1696 	struct ifmultiaddr *ifma;
1697 
1698 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1699 		if (sa_equal(sa, ifma->ifma_addr))
1700 			break;
1701 	if (ifma == 0)
1702 		return ENOENT;
1703 
1704 	if (ifma->ifma_refcount > 1) {
1705 		ifma->ifma_refcount--;
1706 		return 0;
1707 	}
1708 
1709 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1710 	sa = ifma->ifma_lladdr;
1711 	crit_enter();
1712 	LIST_REMOVE(ifma, ifma_link);
1713 	/*
1714 	 * Make sure the interface driver is notified
1715 	 * in the case of a link layer mcast group being left.
1716 	 */
1717 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0) {
1718 		lwkt_serialize_enter(ifp->if_serializer);
1719 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1720 		lwkt_serialize_exit(ifp->if_serializer);
1721 	}
1722 	crit_exit();
1723 	free(ifma->ifma_addr, M_IFMADDR);
1724 	free(ifma, M_IFMADDR);
1725 	if (sa == 0)
1726 		return 0;
1727 
1728 	/*
1729 	 * Now look for the link-layer address which corresponds to
1730 	 * this network address.  It had been squirreled away in
1731 	 * ifma->ifma_lladdr for this purpose (so we don't have
1732 	 * to call ifp->if_resolvemulti() again), and we saved that
1733 	 * value in sa above.  If some nasty deleted the
1734 	 * link-layer address out from underneath us, we can deal because
1735 	 * the address we stored was is not the same as the one which was
1736 	 * in the record for the link-layer address.  (So we don't complain
1737 	 * in that case.)
1738 	 */
1739 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1740 		if (sa_equal(sa, ifma->ifma_addr))
1741 			break;
1742 	if (ifma == 0)
1743 		return 0;
1744 
1745 	if (ifma->ifma_refcount > 1) {
1746 		ifma->ifma_refcount--;
1747 		return 0;
1748 	}
1749 
1750 	crit_enter();
1751 	lwkt_serialize_enter(ifp->if_serializer);
1752 	LIST_REMOVE(ifma, ifma_link);
1753 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1754 	lwkt_serialize_exit(ifp->if_serializer);
1755 	crit_exit();
1756 	free(ifma->ifma_addr, M_IFMADDR);
1757 	free(sa, M_IFMADDR);
1758 	free(ifma, M_IFMADDR);
1759 
1760 	return 0;
1761 }
1762 
1763 /*
1764  * Set the link layer address on an interface.
1765  *
1766  * At this time we only support certain types of interfaces,
1767  * and we don't allow the length of the address to change.
1768  */
1769 int
1770 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1771 {
1772 	struct sockaddr_dl *sdl;
1773 	struct ifaddr *ifa;
1774 	struct ifreq ifr;
1775 
1776 	sdl = IF_LLSOCKADDR(ifp);
1777 	if (sdl == NULL)
1778 		return (EINVAL);
1779 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1780 		return (EINVAL);
1781 	switch (ifp->if_type) {
1782 	case IFT_ETHER:			/* these types use struct arpcom */
1783 	case IFT_FDDI:
1784 	case IFT_XETHER:
1785 	case IFT_ISO88025:
1786 	case IFT_L2VLAN:
1787 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1788 		/* FALLTHROUGH */
1789 	case IFT_ARCNET:
1790 		bcopy(lladdr, LLADDR(sdl), len);
1791 		break;
1792 	default:
1793 		return (ENODEV);
1794 	}
1795 	/*
1796 	 * If the interface is already up, we need
1797 	 * to re-init it in order to reprogram its
1798 	 * address filter.
1799 	 */
1800 	lwkt_serialize_enter(ifp->if_serializer);
1801 	if ((ifp->if_flags & IFF_UP) != 0) {
1802 		ifp->if_flags &= ~IFF_UP;
1803 		ifr.ifr_flags = ifp->if_flags;
1804 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1805 		ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1806 			      (struct ucred *)NULL);
1807 		ifp->if_flags |= IFF_UP;
1808 		ifr.ifr_flags = ifp->if_flags;
1809 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1810 		ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1811 				 (struct ucred *)NULL);
1812 #ifdef INET
1813 		/*
1814 		 * Also send gratuitous ARPs to notify other nodes about
1815 		 * the address change.
1816 		 */
1817 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1818 			if (ifa->ifa_addr != NULL &&
1819 			    ifa->ifa_addr->sa_family == AF_INET)
1820 				arp_ifinit(ifp, ifa);
1821 		}
1822 #endif
1823 	}
1824 	lwkt_serialize_exit(ifp->if_serializer);
1825 	return (0);
1826 }
1827 
1828 struct ifmultiaddr *
1829 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1830 {
1831 	struct ifmultiaddr *ifma;
1832 
1833 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1834 		if (sa_equal(ifma->ifma_addr, sa))
1835 			break;
1836 
1837 	return ifma;
1838 }
1839 
1840 /*
1841  * The name argument must be a pointer to storage which will last as
1842  * long as the interface does.  For physical devices, the result of
1843  * device_get_name(dev) is a good choice and for pseudo-devices a
1844  * static string works well.
1845  */
1846 void
1847 if_initname(struct ifnet *ifp, const char *name, int unit)
1848 {
1849 	ifp->if_dname = name;
1850 	ifp->if_dunit = unit;
1851 	if (unit != IF_DUNIT_NONE)
1852 		snprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1853 	else
1854 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1855 }
1856 
1857 int
1858 if_printf(struct ifnet *ifp, const char *fmt, ...)
1859 {
1860 	__va_list ap;
1861 	int retval;
1862 
1863 	retval = printf("%s: ", ifp->if_xname);
1864 	__va_start(ap, fmt);
1865 	retval += vprintf(fmt, ap);
1866 	__va_end(ap);
1867 	return (retval);
1868 }
1869 
1870 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1871 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1872 
1873 void
1874 ifq_set_classic(struct ifaltq *ifq)
1875 {
1876 	ifq->altq_enqueue = ifq_classic_enqueue;
1877 	ifq->altq_dequeue = ifq_classic_dequeue;
1878 	ifq->altq_request = ifq_classic_request;
1879 }
1880 
1881 static int
1882 ifq_classic_enqueue(struct ifaltq *ifq, struct mbuf *m,
1883 		    struct altq_pktattr *pa __unused)
1884 {
1885 	crit_enter();
1886 	if (IF_QFULL(ifq)) {
1887 		m_freem(m);
1888 		crit_exit();
1889 		return(ENOBUFS);
1890 	} else {
1891 		IF_ENQUEUE(ifq, m);
1892 		crit_exit();
1893 		return(0);
1894 	}
1895 }
1896 
1897 static struct mbuf *
1898 ifq_classic_dequeue(struct ifaltq *ifq, struct mbuf *mpolled, int op)
1899 {
1900 	struct mbuf *m;
1901 
1902 	crit_enter();
1903 	switch (op) {
1904 	case ALTDQ_POLL:
1905 		IF_POLL(ifq, m);
1906 		break;
1907 	case ALTDQ_REMOVE:
1908 		IF_DEQUEUE(ifq, m);
1909 		break;
1910 	default:
1911 		panic("unsupported ALTQ dequeue op: %d", op);
1912 	}
1913 	crit_exit();
1914 	KKASSERT(mpolled == NULL || mpolled == m);
1915 	return(m);
1916 }
1917 
1918 static int
1919 ifq_classic_request(struct ifaltq *ifq, int req, void *arg)
1920 {
1921 	crit_enter();
1922 	switch (req) {
1923 	case ALTRQ_PURGE:
1924 		IF_DRAIN(ifq);
1925 		break;
1926 	default:
1927 		panic("unspported ALTQ request: %d", req);
1928 	}
1929 	crit_exit();
1930 	return(0);
1931 }
1932 
1933