xref: /dragonfly/sys/net/if.c (revision 23265324)
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 1.49 2006/12/22 23:44:54 swildner 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 ucred *);
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 = kmalloc(n, M_IFADDR, M_WAITOK | M_ZERO);
221 		if (ifindex2ifnet) {
222 			bcopy(ifindex2ifnet, q, n/2);
223 			kfree(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 = kmalloc(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 	int cpu, origcpu;
318 	struct domain *dp;
319 
320 	EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
321 
322 	/*
323 	 * Remove routes and flush queues.
324 	 */
325 	crit_enter();
326 #ifdef DEVICE_POLLING
327 	if (ifp->if_flags & IFF_POLLING)
328 		ether_poll_deregister(ifp);
329 #endif
330 	if_down(ifp);
331 
332 	if (ifq_is_enabled(&ifp->if_snd))
333 		altq_disable(&ifp->if_snd);
334 	if (ifq_is_attached(&ifp->if_snd))
335 		altq_detach(&ifp->if_snd);
336 
337 	/*
338 	 * Clean up all addresses.
339 	 */
340 	ifp->if_lladdr = NULL;
341 
342 	for (ifa = TAILQ_FIRST(&ifp->if_addrhead); ifa;
343 	     ifa = TAILQ_FIRST(&ifp->if_addrhead)) {
344 #ifdef INET
345 		/* XXX: Ugly!! ad hoc just for INET */
346 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) {
347 			struct ifaliasreq ifr;
348 
349 			bzero(&ifr, sizeof ifr);
350 			ifr.ifra_addr = *ifa->ifa_addr;
351 			if (ifa->ifa_dstaddr)
352 				ifr.ifra_broadaddr = *ifa->ifa_dstaddr;
353 			if (in_control(NULL, SIOCDIFADDR, (caddr_t)&ifr, ifp,
354 				       NULL) == 0)
355 				continue;
356 		}
357 #endif /* INET */
358 #ifdef INET6
359 		if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET6) {
360 			in6_purgeaddr(ifa);
361 			/* ifp_addrhead is already updated */
362 			continue;
363 		}
364 #endif /* INET6 */
365 		TAILQ_REMOVE(&ifp->if_addrhead, ifa, ifa_link);
366 		IFAFREE(ifa);
367 	}
368 
369 #ifdef INET
370 	/*
371 	 * Remove all IPv4 kernel structures related to ifp.
372 	 */
373 	in_ifdetach(ifp);
374 #endif
375 
376 #ifdef INET6
377 	/*
378 	 * Remove all IPv6 kernel structs related to ifp.  This should be done
379 	 * before removing routing entries below, since IPv6 interface direct
380 	 * routes are expected to be removed by the IPv6-specific kernel API.
381 	 * Otherwise, the kernel will detect some inconsistency and bark it.
382 	 */
383 	in6_ifdetach(ifp);
384 #endif
385 
386 	/*
387 	 * Delete all remaining routes using this interface
388 	 * Unfortuneatly the only way to do this is to slog through
389 	 * the entire routing table looking for routes which point
390 	 * to this interface...oh well...
391 	 */
392 	origcpu = mycpuid;
393 	for (cpu = 0; cpu < ncpus2; cpu++) {
394 		lwkt_migratecpu(cpu);
395 		for (i = 1; i <= AF_MAX; i++) {
396 			if ((rnh = rt_tables[mycpuid][i]) == NULL)
397 				continue;
398 			rnh->rnh_walktree(rnh, if_rtdel, ifp);
399 		}
400 	}
401 	lwkt_migratecpu(origcpu);
402 
403 	/* Announce that the interface is gone. */
404 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
405 
406 	SLIST_FOREACH(dp, &domains, dom_next)
407 		if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family])
408 			(*dp->dom_ifdetach)(ifp,
409 				ifp->if_afdata[dp->dom_family]);
410 
411 	/*
412 	 * Remove interface from ifindex2ifp[] and maybe decrement if_index.
413 	 */
414 	ifindex2ifnet[ifp->if_index] = NULL;
415 	while (if_index > 0 && ifindex2ifnet[if_index] == NULL)
416 		if_index--;
417 
418 	TAILQ_REMOVE(&ifnet, ifp, if_link);
419 	crit_exit();
420 }
421 
422 /*
423  * Delete Routes for a Network Interface
424  *
425  * Called for each routing entry via the rnh->rnh_walktree() call above
426  * to delete all route entries referencing a detaching network interface.
427  *
428  * Arguments:
429  *	rn	pointer to node in the routing table
430  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
431  *
432  * Returns:
433  *	0	successful
434  *	errno	failed - reason indicated
435  *
436  */
437 static int
438 if_rtdel(struct radix_node *rn, void *arg)
439 {
440 	struct rtentry	*rt = (struct rtentry *)rn;
441 	struct ifnet	*ifp = arg;
442 	int		err;
443 
444 	if (rt->rt_ifp == ifp) {
445 
446 		/*
447 		 * Protect (sorta) against walktree recursion problems
448 		 * with cloned routes
449 		 */
450 		if (!(rt->rt_flags & RTF_UP))
451 			return (0);
452 
453 		err = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
454 				rt_mask(rt), rt->rt_flags,
455 				(struct rtentry **) NULL);
456 		if (err) {
457 			log(LOG_WARNING, "if_rtdel: error %d\n", err);
458 		}
459 	}
460 
461 	return (0);
462 }
463 
464 /*
465  * Create a clone network interface.
466  */
467 int
468 if_clone_create(char *name, int len)
469 {
470 	struct if_clone *ifc;
471 	char *dp;
472 	int wildcard, bytoff, bitoff;
473 	int unit;
474 	int err;
475 
476 	ifc = if_clone_lookup(name, &unit);
477 	if (ifc == NULL)
478 		return (EINVAL);
479 
480 	if (ifunit(name) != NULL)
481 		return (EEXIST);
482 
483 	bytoff = bitoff = 0;
484 	wildcard = (unit < 0);
485 	/*
486 	 * Find a free unit if none was given.
487 	 */
488 	if (wildcard) {
489 		while (bytoff < ifc->ifc_bmlen &&
490 		    ifc->ifc_units[bytoff] == 0xff)
491 			bytoff++;
492 		if (bytoff >= ifc->ifc_bmlen)
493 			return (ENOSPC);
494 		while ((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0)
495 			bitoff++;
496 		unit = (bytoff << 3) + bitoff;
497 	}
498 
499 	if (unit > ifc->ifc_maxunit)
500 		return (ENXIO);
501 
502 	err = (*ifc->ifc_create)(ifc, unit);
503 	if (err != 0)
504 		return (err);
505 
506 	if (!wildcard) {
507 		bytoff = unit >> 3;
508 		bitoff = unit - (bytoff << 3);
509 	}
510 
511 	/*
512 	 * Allocate the unit in the bitmap.
513 	 */
514 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) == 0,
515 	    ("%s: bit is already set", __func__));
516 	ifc->ifc_units[bytoff] |= (1 << bitoff);
517 
518 	/* In the wildcard case, we need to update the name. */
519 	if (wildcard) {
520 		for (dp = name; *dp != '\0'; dp++);
521 		if (ksnprintf(dp, len - (dp-name), "%d", unit) >
522 		    len - (dp-name) - 1) {
523 			/*
524 			 * This can only be a programmer error and
525 			 * there's no straightforward way to recover if
526 			 * it happens.
527 			 */
528 			panic("if_clone_create(): interface name too long");
529 		}
530 
531 	}
532 
533 	EVENTHANDLER_INVOKE(if_clone_event, ifc);
534 
535 	return (0);
536 }
537 
538 /*
539  * Destroy a clone network interface.
540  */
541 int
542 if_clone_destroy(const char *name)
543 {
544 	struct if_clone *ifc;
545 	struct ifnet *ifp;
546 	int bytoff, bitoff;
547 	int unit;
548 
549 	ifc = if_clone_lookup(name, &unit);
550 	if (ifc == NULL)
551 		return (EINVAL);
552 
553 	if (unit < ifc->ifc_minifs)
554 		return (EINVAL);
555 
556 	ifp = ifunit(name);
557 	if (ifp == NULL)
558 		return (ENXIO);
559 
560 	if (ifc->ifc_destroy == NULL)
561 		return (EOPNOTSUPP);
562 
563 	(*ifc->ifc_destroy)(ifp);
564 
565 	/*
566 	 * Compute offset in the bitmap and deallocate the unit.
567 	 */
568 	bytoff = unit >> 3;
569 	bitoff = unit - (bytoff << 3);
570 	KASSERT((ifc->ifc_units[bytoff] & (1 << bitoff)) != 0,
571 	    ("%s: bit is already cleared", __func__));
572 	ifc->ifc_units[bytoff] &= ~(1 << bitoff);
573 	return (0);
574 }
575 
576 /*
577  * Look up a network interface cloner.
578  */
579 struct if_clone *
580 if_clone_lookup(const char *name, int *unitp)
581 {
582 	struct if_clone *ifc;
583 	const char *cp;
584 	int i;
585 
586 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
587 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
588 			if (ifc->ifc_name[i] != *cp)
589 				goto next_ifc;
590 		}
591 		goto found_name;
592  next_ifc:
593 		ifc = LIST_NEXT(ifc, ifc_list);
594 	}
595 
596 	/* No match. */
597 	return ((struct if_clone *)NULL);
598 
599  found_name:
600 	if (*cp == '\0') {
601 		i = -1;
602 	} else {
603 		for (i = 0; *cp != '\0'; cp++) {
604 			if (*cp < '0' || *cp > '9') {
605 				/* Bogus unit number. */
606 				return (NULL);
607 			}
608 			i = (i * 10) + (*cp - '0');
609 		}
610 	}
611 
612 	if (unitp != NULL)
613 		*unitp = i;
614 	return (ifc);
615 }
616 
617 /*
618  * Register a network interface cloner.
619  */
620 void
621 if_clone_attach(struct if_clone *ifc)
622 {
623 	int bytoff, bitoff;
624 	int err;
625 	int len, maxclone;
626 	int unit;
627 
628 	KASSERT(ifc->ifc_minifs - 1 <= ifc->ifc_maxunit,
629 	    ("%s: %s requested more units then allowed (%d > %d)",
630 	    __func__, ifc->ifc_name, ifc->ifc_minifs,
631 	    ifc->ifc_maxunit + 1));
632 	/*
633 	 * Compute bitmap size and allocate it.
634 	 */
635 	maxclone = ifc->ifc_maxunit + 1;
636 	len = maxclone >> 3;
637 	if ((len << 3) < maxclone)
638 		len++;
639 	ifc->ifc_units = kmalloc(len, M_CLONE, M_WAITOK | M_ZERO);
640 	ifc->ifc_bmlen = len;
641 
642 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
643 	if_cloners_count++;
644 
645 	for (unit = 0; unit < ifc->ifc_minifs; unit++) {
646 		err = (*ifc->ifc_create)(ifc, unit);
647 		KASSERT(err == 0,
648 		    ("%s: failed to create required interface %s%d",
649 		    __func__, ifc->ifc_name, unit));
650 
651 		/* Allocate the unit in the bitmap. */
652 		bytoff = unit >> 3;
653 		bitoff = unit - (bytoff << 3);
654 		ifc->ifc_units[bytoff] |= (1 << bitoff);
655 	}
656 }
657 
658 /*
659  * Unregister a network interface cloner.
660  */
661 void
662 if_clone_detach(struct if_clone *ifc)
663 {
664 
665 	LIST_REMOVE(ifc, ifc_list);
666 	kfree(ifc->ifc_units, M_CLONE);
667 	if_cloners_count--;
668 }
669 
670 /*
671  * Provide list of interface cloners to userspace.
672  */
673 int
674 if_clone_list(struct if_clonereq *ifcr)
675 {
676 	char outbuf[IFNAMSIZ], *dst;
677 	struct if_clone *ifc;
678 	int count, error = 0;
679 
680 	ifcr->ifcr_total = if_cloners_count;
681 	if ((dst = ifcr->ifcr_buffer) == NULL) {
682 		/* Just asking how many there are. */
683 		return (0);
684 	}
685 
686 	if (ifcr->ifcr_count < 0)
687 		return (EINVAL);
688 
689 	count = (if_cloners_count < ifcr->ifcr_count) ?
690 	    if_cloners_count : ifcr->ifcr_count;
691 
692 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
693 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
694 		strlcpy(outbuf, ifc->ifc_name, IFNAMSIZ);
695 		error = copyout(outbuf, dst, IFNAMSIZ);
696 		if (error)
697 			break;
698 	}
699 
700 	return (error);
701 }
702 
703 /*
704  * Locate an interface based on a complete address.
705  */
706 struct ifaddr *
707 ifa_ifwithaddr(struct sockaddr *addr)
708 {
709 	struct ifnet *ifp;
710 	struct ifaddr *ifa;
711 
712 	TAILQ_FOREACH(ifp, &ifnet, if_link)
713 	    TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
714 		if (ifa->ifa_addr->sa_family != addr->sa_family)
715 			continue;
716 		if (sa_equal(addr, ifa->ifa_addr))
717 			return (ifa);
718 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
719 		    /* IPv6 doesn't have broadcast */
720 		    ifa->ifa_broadaddr->sa_len != 0 &&
721 		    sa_equal(ifa->ifa_broadaddr, addr))
722 			return (ifa);
723 	}
724 	return ((struct ifaddr *)NULL);
725 }
726 /*
727  * Locate the point to point interface with a given destination address.
728  */
729 struct ifaddr *
730 ifa_ifwithdstaddr(struct sockaddr *addr)
731 {
732 	struct ifnet *ifp;
733 	struct ifaddr *ifa;
734 
735 	TAILQ_FOREACH(ifp, &ifnet, if_link)
736 	    if (ifp->if_flags & IFF_POINTOPOINT)
737 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
738 			if (ifa->ifa_addr->sa_family != addr->sa_family)
739 				continue;
740 			if (ifa->ifa_dstaddr &&
741 			    sa_equal(addr, ifa->ifa_dstaddr))
742 				return (ifa);
743 	}
744 	return ((struct ifaddr *)NULL);
745 }
746 
747 /*
748  * Find an interface on a specific network.  If many, choice
749  * is most specific found.
750  */
751 struct ifaddr *
752 ifa_ifwithnet(struct sockaddr *addr)
753 {
754 	struct ifnet *ifp;
755 	struct ifaddr *ifa;
756 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
757 	u_int af = addr->sa_family;
758 	char *addr_data = addr->sa_data, *cplim;
759 
760 	/*
761 	 * AF_LINK addresses can be looked up directly by their index number,
762 	 * so do that if we can.
763 	 */
764 	if (af == AF_LINK) {
765 	    struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
766 
767 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
768 		return (ifindex2ifnet[sdl->sdl_index]->if_lladdr);
769 	}
770 
771 	/*
772 	 * Scan though each interface, looking for ones that have
773 	 * addresses in this address family.
774 	 */
775 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
776 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
777 			char *cp, *cp2, *cp3;
778 
779 			if (ifa->ifa_addr->sa_family != af)
780 next:				continue;
781 			if (af == AF_INET && ifp->if_flags & IFF_POINTOPOINT) {
782 				/*
783 				 * This is a bit broken as it doesn't
784 				 * take into account that the remote end may
785 				 * be a single node in the network we are
786 				 * looking for.
787 				 * The trouble is that we don't know the
788 				 * netmask for the remote end.
789 				 */
790 				if (ifa->ifa_dstaddr != NULL &&
791 				    sa_equal(addr, ifa->ifa_dstaddr))
792 					return (ifa);
793 			} else {
794 				/*
795 				 * if we have a special address handler,
796 				 * then use it instead of the generic one.
797 				 */
798 				if (ifa->ifa_claim_addr) {
799 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
800 						return (ifa);
801 					} else {
802 						continue;
803 					}
804 				}
805 
806 				/*
807 				 * Scan all the bits in the ifa's address.
808 				 * If a bit dissagrees with what we are
809 				 * looking for, mask it with the netmask
810 				 * to see if it really matters.
811 				 * (A byte at a time)
812 				 */
813 				if (ifa->ifa_netmask == 0)
814 					continue;
815 				cp = addr_data;
816 				cp2 = ifa->ifa_addr->sa_data;
817 				cp3 = ifa->ifa_netmask->sa_data;
818 				cplim = ifa->ifa_netmask->sa_len +
819 					(char *)ifa->ifa_netmask;
820 				while (cp3 < cplim)
821 					if ((*cp++ ^ *cp2++) & *cp3++)
822 						goto next; /* next address! */
823 				/*
824 				 * If the netmask of what we just found
825 				 * is more specific than what we had before
826 				 * (if we had one) then remember the new one
827 				 * before continuing to search
828 				 * for an even better one.
829 				 */
830 				if (ifa_maybe == 0 ||
831 				    rn_refines((char *)ifa->ifa_netmask,
832 					       (char *)ifa_maybe->ifa_netmask))
833 					ifa_maybe = ifa;
834 			}
835 		}
836 	}
837 	return (ifa_maybe);
838 }
839 
840 /*
841  * Find an interface address specific to an interface best matching
842  * a given address.
843  */
844 struct ifaddr *
845 ifaof_ifpforaddr(struct sockaddr *addr, struct ifnet *ifp)
846 {
847 	struct ifaddr *ifa;
848 	char *cp, *cp2, *cp3;
849 	char *cplim;
850 	struct ifaddr *ifa_maybe = 0;
851 	u_int af = addr->sa_family;
852 
853 	if (af >= AF_MAX)
854 		return (0);
855 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
856 		if (ifa->ifa_addr->sa_family != af)
857 			continue;
858 		if (ifa_maybe == 0)
859 			ifa_maybe = ifa;
860 		if (ifa->ifa_netmask == NULL) {
861 			if (sa_equal(addr, ifa->ifa_addr) ||
862 			    (ifa->ifa_dstaddr != NULL &&
863 			     sa_equal(addr, ifa->ifa_dstaddr)))
864 				return (ifa);
865 			continue;
866 		}
867 		if (ifp->if_flags & IFF_POINTOPOINT) {
868 			if (sa_equal(addr, ifa->ifa_dstaddr))
869 				return (ifa);
870 		} else {
871 			cp = addr->sa_data;
872 			cp2 = ifa->ifa_addr->sa_data;
873 			cp3 = ifa->ifa_netmask->sa_data;
874 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
875 			for (; cp3 < cplim; cp3++)
876 				if ((*cp++ ^ *cp2++) & *cp3)
877 					break;
878 			if (cp3 == cplim)
879 				return (ifa);
880 		}
881 	}
882 	return (ifa_maybe);
883 }
884 
885 #include <net/route.h>
886 
887 /*
888  * Default action when installing a route with a Link Level gateway.
889  * Lookup an appropriate real ifa to point to.
890  * This should be moved to /sys/net/link.c eventually.
891  */
892 static void
893 link_rtrequest(int cmd, struct rtentry *rt, struct rt_addrinfo *info)
894 {
895 	struct ifaddr *ifa;
896 	struct sockaddr *dst;
897 	struct ifnet *ifp;
898 
899 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
900 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_key(rt)) == NULL)
901 		return;
902 	ifa = ifaof_ifpforaddr(dst, ifp);
903 	if (ifa != NULL) {
904 		IFAFREE(rt->rt_ifa);
905 		IFAREF(ifa);
906 		rt->rt_ifa = ifa;
907 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
908 			ifa->ifa_rtrequest(cmd, rt, info);
909 	}
910 }
911 
912 /*
913  * Mark an interface down and notify protocols of
914  * the transition.
915  * NOTE: must be called at splnet or eqivalent.
916  */
917 void
918 if_unroute(struct ifnet *ifp, int flag, int fam)
919 {
920 	struct ifaddr *ifa;
921 
922 	ifp->if_flags &= ~flag;
923 	getmicrotime(&ifp->if_lastchange);
924 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
925 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
926 			pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
927 	ifq_purge(&ifp->if_snd);
928 	rt_ifmsg(ifp);
929 }
930 
931 /*
932  * Mark an interface up and notify protocols of
933  * the transition.
934  * NOTE: must be called at splnet or eqivalent.
935  */
936 void
937 if_route(struct ifnet *ifp, int flag, int fam)
938 {
939 	struct ifaddr *ifa;
940 
941 	ifp->if_flags |= flag;
942 	getmicrotime(&ifp->if_lastchange);
943 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
944 		if (fam == PF_UNSPEC || (fam == ifa->ifa_addr->sa_family))
945 			pfctlinput(PRC_IFUP, ifa->ifa_addr);
946 	rt_ifmsg(ifp);
947 #ifdef INET6
948 	in6_if_up(ifp);
949 #endif
950 }
951 
952 /*
953  * Mark an interface down and notify protocols of the transition.  An
954  * interface going down is also considered to be a synchronizing event.
955  * We must ensure that all packet processing related to the interface
956  * has completed before we return so e.g. the caller can free the ifnet
957  * structure that the mbufs may be referencing.
958  *
959  * NOTE: must be called at splnet or eqivalent.
960  */
961 void
962 if_down(struct ifnet *ifp)
963 {
964 	if_unroute(ifp, IFF_UP, AF_UNSPEC);
965 	netmsg_service_sync();
966 }
967 
968 /*
969  * Mark an interface up and notify protocols of
970  * the transition.
971  * NOTE: must be called at splnet or eqivalent.
972  */
973 void
974 if_up(struct ifnet *ifp)
975 {
976 
977 	if_route(ifp, IFF_UP, AF_UNSPEC);
978 }
979 
980 /*
981  * Handle interface watchdog timer routines.  Called
982  * from softclock, we decrement timers (if set) and
983  * call the appropriate interface routine on expiration.
984  */
985 static void
986 if_slowtimo(void *arg)
987 {
988 	struct ifnet *ifp;
989 
990 	crit_enter();
991 
992 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
993 		if (ifp->if_timer == 0 || --ifp->if_timer)
994 			continue;
995 		if (ifp->if_watchdog) {
996 			if (lwkt_serialize_try(ifp->if_serializer)) {
997 				(*ifp->if_watchdog)(ifp);
998 				lwkt_serialize_exit(ifp->if_serializer);
999 			} else {
1000 				/* try again next timeout */
1001 				++ifp->if_timer;
1002 			}
1003 		}
1004 	}
1005 
1006 	crit_exit();
1007 
1008 	callout_reset(&if_slowtimo_timer, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
1009 }
1010 
1011 /*
1012  * Map interface name to
1013  * interface structure pointer.
1014  */
1015 struct ifnet *
1016 ifunit(const char *name)
1017 {
1018 	struct ifnet *ifp;
1019 
1020 	/*
1021 	 * Search all the interfaces for this name/number
1022 	 */
1023 
1024 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1025 		if (strncmp(ifp->if_xname, name, IFNAMSIZ) == 0)
1026 			break;
1027 	}
1028 	return (ifp);
1029 }
1030 
1031 
1032 /*
1033  * Map interface name in a sockaddr_dl to
1034  * interface structure pointer.
1035  */
1036 struct ifnet *
1037 if_withname(struct sockaddr *sa)
1038 {
1039 	char ifname[IFNAMSIZ+1];
1040 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)sa;
1041 
1042 	if ( (sa->sa_family != AF_LINK) || (sdl->sdl_nlen == 0) ||
1043 	     (sdl->sdl_nlen > IFNAMSIZ) )
1044 		return NULL;
1045 
1046 	/*
1047 	 * ifunit wants a null-terminated name.  It may not be null-terminated
1048 	 * in the sockaddr.  We don't want to change the caller's sockaddr,
1049 	 * and there might not be room to put the trailing null anyway, so we
1050 	 * make a local copy that we know we can null terminate safely.
1051 	 */
1052 
1053 	bcopy(sdl->sdl_data, ifname, sdl->sdl_nlen);
1054 	ifname[sdl->sdl_nlen] = '\0';
1055 	return ifunit(ifname);
1056 }
1057 
1058 
1059 /*
1060  * Interface ioctls.
1061  */
1062 int
1063 ifioctl(struct socket *so, u_long cmd, caddr_t data, struct ucred *cred)
1064 {
1065 	struct ifnet *ifp;
1066 	struct ifreq *ifr;
1067 	struct ifstat *ifs;
1068 	int error;
1069 	short oif_flags;
1070 	int new_flags;
1071 	size_t namelen, onamelen;
1072 	char new_name[IFNAMSIZ];
1073 	struct ifaddr *ifa;
1074 	struct sockaddr_dl *sdl;
1075 
1076 	switch (cmd) {
1077 
1078 	case SIOCGIFCONF:
1079 	case OSIOCGIFCONF:
1080 		return (ifconf(cmd, data, cred));
1081 	}
1082 	ifr = (struct ifreq *)data;
1083 
1084 	switch (cmd) {
1085 	case SIOCIFCREATE:
1086 	case SIOCIFDESTROY:
1087 		if ((error = suser_cred(cred, 0)) != 0)
1088 			return (error);
1089 		return ((cmd == SIOCIFCREATE) ?
1090 			if_clone_create(ifr->ifr_name, sizeof(ifr->ifr_name)) :
1091 			if_clone_destroy(ifr->ifr_name));
1092 
1093 	case SIOCIFGCLONERS:
1094 		return (if_clone_list((struct if_clonereq *)data));
1095 	}
1096 
1097 	ifp = ifunit(ifr->ifr_name);
1098 	if (ifp == 0)
1099 		return (ENXIO);
1100 	switch (cmd) {
1101 
1102 	case SIOCGIFFLAGS:
1103 		ifr->ifr_flags = ifp->if_flags;
1104 		ifr->ifr_flagshigh = ifp->if_flags >> 16;
1105 		break;
1106 
1107 	case SIOCGIFCAP:
1108 		ifr->ifr_reqcap = ifp->if_capabilities;
1109 		ifr->ifr_curcap = ifp->if_capenable;
1110 		break;
1111 
1112 	case SIOCGIFMETRIC:
1113 		ifr->ifr_metric = ifp->if_metric;
1114 		break;
1115 
1116 	case SIOCGIFMTU:
1117 		ifr->ifr_mtu = ifp->if_mtu;
1118 		break;
1119 
1120 	case SIOCGIFPHYS:
1121 		ifr->ifr_phys = ifp->if_physical;
1122 		break;
1123 
1124 	case SIOCSIFFLAGS:
1125 		error = suser_cred(cred, 0);
1126 		if (error)
1127 			return (error);
1128 		new_flags = (ifr->ifr_flags & 0xffff) |
1129 		    (ifr->ifr_flagshigh << 16);
1130 		if (ifp->if_flags & IFF_SMART) {
1131 			/* Smart drivers twiddle their own routes */
1132 		} else if (ifp->if_flags & IFF_UP &&
1133 		    (new_flags & IFF_UP) == 0) {
1134 			crit_enter();
1135 			if_down(ifp);
1136 			crit_exit();
1137 		} else if (new_flags & IFF_UP &&
1138 		    (ifp->if_flags & IFF_UP) == 0) {
1139 			crit_enter();
1140 			if_up(ifp);
1141 			crit_exit();
1142 		}
1143 
1144 #ifdef DEVICE_POLLING
1145 		if ((new_flags ^ ifp->if_flags) & IFF_POLLING) {
1146 			if (new_flags & IFF_POLLING) {
1147 				ether_poll_register(ifp);
1148 			} else {
1149 				ether_poll_deregister(ifp);
1150 			}
1151 		}
1152 #endif
1153 
1154 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
1155 			(new_flags &~ IFF_CANTCHANGE);
1156 		if (new_flags & IFF_PPROMISC) {
1157 			/* Permanently promiscuous mode requested */
1158 			ifp->if_flags |= IFF_PROMISC;
1159 		} else if (ifp->if_pcount == 0) {
1160 			ifp->if_flags &= ~IFF_PROMISC;
1161 		}
1162 		if (ifp->if_ioctl) {
1163 			lwkt_serialize_enter(ifp->if_serializer);
1164 			ifp->if_ioctl(ifp, cmd, data, cred);
1165 			lwkt_serialize_exit(ifp->if_serializer);
1166 		}
1167 		getmicrotime(&ifp->if_lastchange);
1168 		break;
1169 
1170 	case SIOCSIFCAP:
1171 		error = suser_cred(cred, 0);
1172 		if (error)
1173 			return (error);
1174 		if (ifr->ifr_reqcap & ~ifp->if_capabilities)
1175 			return (EINVAL);
1176 		lwkt_serialize_enter(ifp->if_serializer);
1177 		ifp->if_ioctl(ifp, cmd, data, cred);
1178 		lwkt_serialize_exit(ifp->if_serializer);
1179 		break;
1180 
1181 	case SIOCSIFNAME:
1182 		error = suser_cred(cred, 0);
1183 		if (error != 0)
1184 			return (error);
1185 		error = copyinstr(ifr->ifr_data, new_name, IFNAMSIZ, NULL);
1186 		if (error != 0)
1187 			return (error);
1188 		if (new_name[0] == '\0')
1189 			return (EINVAL);
1190 		if (ifunit(new_name) != NULL)
1191 			return (EEXIST);
1192 
1193 		EVENTHANDLER_INVOKE(ifnet_detach_event, ifp);
1194 
1195 		/* Announce the departure of the interface. */
1196 		rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1197 
1198 		strlcpy(ifp->if_xname, new_name, sizeof(ifp->if_xname));
1199 		ifa = TAILQ_FIRST(&ifp->if_addrhead);
1200 		/* XXX IFA_LOCK(ifa); */
1201 		sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1202 		namelen = strlen(new_name);
1203 		onamelen = sdl->sdl_nlen;
1204 		/*
1205 		 * Move the address if needed.  This is safe because we
1206 		 * allocate space for a name of length IFNAMSIZ when we
1207 		 * create this in if_attach().
1208 		 */
1209 		if (namelen != onamelen) {
1210 			bcopy(sdl->sdl_data + onamelen,
1211 			    sdl->sdl_data + namelen, sdl->sdl_alen);
1212 		}
1213 		bcopy(new_name, sdl->sdl_data, namelen);
1214 		sdl->sdl_nlen = namelen;
1215 		sdl = (struct sockaddr_dl *)ifa->ifa_netmask;
1216 		bzero(sdl->sdl_data, onamelen);
1217 		while (namelen != 0)
1218 			sdl->sdl_data[--namelen] = 0xff;
1219 		/* XXX IFA_UNLOCK(ifa) */
1220 
1221 		EVENTHANDLER_INVOKE(ifnet_attach_event, ifp);
1222 
1223 		/* Announce the return of the interface. */
1224 		rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1225 		break;
1226 
1227 	case SIOCSIFMETRIC:
1228 		error = suser_cred(cred, 0);
1229 		if (error)
1230 			return (error);
1231 		ifp->if_metric = ifr->ifr_metric;
1232 		getmicrotime(&ifp->if_lastchange);
1233 		break;
1234 
1235 	case SIOCSIFPHYS:
1236 		error = suser_cred(cred, 0);
1237 		if (error)
1238 			return error;
1239 		if (!ifp->if_ioctl)
1240 		        return EOPNOTSUPP;
1241 		lwkt_serialize_enter(ifp->if_serializer);
1242 		error = ifp->if_ioctl(ifp, cmd, data, cred);
1243 		lwkt_serialize_exit(ifp->if_serializer);
1244 		if (error == 0)
1245 			getmicrotime(&ifp->if_lastchange);
1246 		return (error);
1247 
1248 	case SIOCSIFMTU:
1249 	{
1250 		u_long oldmtu = ifp->if_mtu;
1251 
1252 		error = suser_cred(cred, 0);
1253 		if (error)
1254 			return (error);
1255 		if (ifp->if_ioctl == NULL)
1256 			return (EOPNOTSUPP);
1257 		if (ifr->ifr_mtu < IF_MINMTU || ifr->ifr_mtu > IF_MAXMTU)
1258 			return (EINVAL);
1259 		lwkt_serialize_enter(ifp->if_serializer);
1260 		error = ifp->if_ioctl(ifp, cmd, data, cred);
1261 		lwkt_serialize_exit(ifp->if_serializer);
1262 		if (error == 0) {
1263 			getmicrotime(&ifp->if_lastchange);
1264 			rt_ifmsg(ifp);
1265 		}
1266 		/*
1267 		 * If the link MTU changed, do network layer specific procedure.
1268 		 */
1269 		if (ifp->if_mtu != oldmtu) {
1270 #ifdef INET6
1271 			nd6_setmtu(ifp);
1272 #endif
1273 		}
1274 		return (error);
1275 	}
1276 
1277 	case SIOCADDMULTI:
1278 	case SIOCDELMULTI:
1279 		error = suser_cred(cred, 0);
1280 		if (error)
1281 			return (error);
1282 
1283 		/* Don't allow group membership on non-multicast interfaces. */
1284 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
1285 			return EOPNOTSUPP;
1286 
1287 		/* Don't let users screw up protocols' entries. */
1288 		if (ifr->ifr_addr.sa_family != AF_LINK)
1289 			return EINVAL;
1290 
1291 		if (cmd == SIOCADDMULTI) {
1292 			struct ifmultiaddr *ifma;
1293 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
1294 		} else {
1295 			error = if_delmulti(ifp, &ifr->ifr_addr);
1296 		}
1297 		if (error == 0)
1298 			getmicrotime(&ifp->if_lastchange);
1299 		return error;
1300 
1301 	case SIOCSIFPHYADDR:
1302 	case SIOCDIFPHYADDR:
1303 #ifdef INET6
1304 	case SIOCSIFPHYADDR_IN6:
1305 #endif
1306 	case SIOCSLIFPHYADDR:
1307         case SIOCSIFMEDIA:
1308 	case SIOCSIFGENERIC:
1309 		error = suser_cred(cred, 0);
1310 		if (error)
1311 			return (error);
1312 		if (ifp->if_ioctl == 0)
1313 			return (EOPNOTSUPP);
1314 		lwkt_serialize_enter(ifp->if_serializer);
1315 		error = ifp->if_ioctl(ifp, cmd, data, cred);
1316 		lwkt_serialize_exit(ifp->if_serializer);
1317 		if (error == 0)
1318 			getmicrotime(&ifp->if_lastchange);
1319 		return error;
1320 
1321 	case SIOCGIFSTATUS:
1322 		ifs = (struct ifstat *)data;
1323 		ifs->ascii[0] = '\0';
1324 
1325 	case SIOCGIFPSRCADDR:
1326 	case SIOCGIFPDSTADDR:
1327 	case SIOCGLIFPHYADDR:
1328 	case SIOCGIFMEDIA:
1329 	case SIOCGIFGENERIC:
1330 		if (ifp->if_ioctl == NULL)
1331 			return (EOPNOTSUPP);
1332 		lwkt_serialize_enter(ifp->if_serializer);
1333 		error = ifp->if_ioctl(ifp, cmd, data, cred);
1334 		lwkt_serialize_exit(ifp->if_serializer);
1335 		return (error);
1336 
1337 	case SIOCSIFLLADDR:
1338 		error = suser_cred(cred, 0);
1339 		if (error)
1340 			return (error);
1341 		return if_setlladdr(ifp,
1342 		    ifr->ifr_addr.sa_data, ifr->ifr_addr.sa_len);
1343 
1344 	default:
1345 		oif_flags = ifp->if_flags;
1346 		if (so->so_proto == 0)
1347 			return (EOPNOTSUPP);
1348 #ifndef COMPAT_43
1349 		error = so_pru_control(so, cmd, data, ifp);
1350 #else
1351 	    {
1352 		int ocmd = cmd;
1353 
1354 		switch (cmd) {
1355 
1356 		case SIOCSIFDSTADDR:
1357 		case SIOCSIFADDR:
1358 		case SIOCSIFBRDADDR:
1359 		case SIOCSIFNETMASK:
1360 #if BYTE_ORDER != BIG_ENDIAN
1361 			if (ifr->ifr_addr.sa_family == 0 &&
1362 			    ifr->ifr_addr.sa_len < 16) {
1363 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
1364 				ifr->ifr_addr.sa_len = 16;
1365 			}
1366 #else
1367 			if (ifr->ifr_addr.sa_len == 0)
1368 				ifr->ifr_addr.sa_len = 16;
1369 #endif
1370 			break;
1371 
1372 		case OSIOCGIFADDR:
1373 			cmd = SIOCGIFADDR;
1374 			break;
1375 
1376 		case OSIOCGIFDSTADDR:
1377 			cmd = SIOCGIFDSTADDR;
1378 			break;
1379 
1380 		case OSIOCGIFBRDADDR:
1381 			cmd = SIOCGIFBRDADDR;
1382 			break;
1383 
1384 		case OSIOCGIFNETMASK:
1385 			cmd = SIOCGIFNETMASK;
1386 		}
1387 		error =  so_pru_control(so, cmd, data, ifp);
1388 		switch (ocmd) {
1389 
1390 		case OSIOCGIFADDR:
1391 		case OSIOCGIFDSTADDR:
1392 		case OSIOCGIFBRDADDR:
1393 		case OSIOCGIFNETMASK:
1394 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
1395 
1396 		}
1397 	    }
1398 #endif /* COMPAT_43 */
1399 
1400 		if ((oif_flags ^ ifp->if_flags) & IFF_UP) {
1401 #ifdef INET6
1402 			DELAY(100);/* XXX: temporary workaround for fxp issue*/
1403 			if (ifp->if_flags & IFF_UP) {
1404 				crit_enter();
1405 				in6_if_up(ifp);
1406 				crit_exit();
1407 			}
1408 #endif
1409 		}
1410 		return (error);
1411 
1412 	}
1413 	return (0);
1414 }
1415 
1416 /*
1417  * Set/clear promiscuous mode on interface ifp based on the truth value
1418  * of pswitch.  The calls are reference counted so that only the first
1419  * "on" request actually has an effect, as does the final "off" request.
1420  * Results are undefined if the "off" and "on" requests are not matched.
1421  */
1422 int
1423 ifpromisc(struct ifnet *ifp, int pswitch)
1424 {
1425 	struct ifreq ifr;
1426 	int error;
1427 	int oldflags;
1428 
1429 	oldflags = ifp->if_flags;
1430 	if (ifp->if_flags & IFF_PPROMISC) {
1431 		/* Do nothing if device is in permanently promiscuous mode */
1432 		ifp->if_pcount += pswitch ? 1 : -1;
1433 		return (0);
1434 	}
1435 	if (pswitch) {
1436 		/*
1437 		 * If the device is not configured up, we cannot put it in
1438 		 * promiscuous mode.
1439 		 */
1440 		if ((ifp->if_flags & IFF_UP) == 0)
1441 			return (ENETDOWN);
1442 		if (ifp->if_pcount++ != 0)
1443 			return (0);
1444 		ifp->if_flags |= IFF_PROMISC;
1445 		log(LOG_INFO, "%s: promiscuous mode enabled\n",
1446 		    ifp->if_xname);
1447 	} else {
1448 		if (--ifp->if_pcount > 0)
1449 			return (0);
1450 		ifp->if_flags &= ~IFF_PROMISC;
1451 		log(LOG_INFO, "%s: promiscuous mode disabled\n",
1452 		    ifp->if_xname);
1453 	}
1454 	ifr.ifr_flags = ifp->if_flags;
1455 	ifr.ifr_flagshigh = ifp->if_flags >> 16;
1456 	lwkt_serialize_enter(ifp->if_serializer);
1457 	error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1458 				 (struct ucred *)NULL);
1459 	lwkt_serialize_exit(ifp->if_serializer);
1460 	if (error == 0)
1461 		rt_ifmsg(ifp);
1462 	else
1463 		ifp->if_flags = oldflags;
1464 	return error;
1465 }
1466 
1467 /*
1468  * Return interface configuration
1469  * of system.  List may be used
1470  * in later ioctl's (above) to get
1471  * other information.
1472  */
1473 static int
1474 ifconf(u_long cmd, caddr_t data, struct ucred *cred)
1475 {
1476 	struct ifconf *ifc = (struct ifconf *)data;
1477 	struct ifnet *ifp;
1478 	struct ifaddr *ifa;
1479 	struct sockaddr *sa;
1480 	struct ifreq ifr, *ifrp;
1481 	int space = ifc->ifc_len, error = 0;
1482 
1483 	ifrp = ifc->ifc_req;
1484 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1485 		int addrs;
1486 
1487 		if (space <= sizeof ifr)
1488 			break;
1489 
1490 		/*
1491 		 * Zero the stack declared structure first to prevent
1492 		 * memory disclosure.
1493 		 */
1494 		bzero(&ifr, sizeof(ifr));
1495 		if (strlcpy(ifr.ifr_name, ifp->if_xname, sizeof(ifr.ifr_name))
1496 		    >= sizeof(ifr.ifr_name)) {
1497 			error = ENAMETOOLONG;
1498 			break;
1499 		}
1500 
1501 		addrs = 0;
1502 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1503 			if (space <= sizeof ifr)
1504 				break;
1505 			sa = ifa->ifa_addr;
1506 			if (cred->cr_prison &&
1507 			    prison_if(cred, sa))
1508 				continue;
1509 			addrs++;
1510 #ifdef COMPAT_43
1511 			if (cmd == OSIOCGIFCONF) {
1512 				struct osockaddr *osa =
1513 					 (struct osockaddr *)&ifr.ifr_addr;
1514 				ifr.ifr_addr = *sa;
1515 				osa->sa_family = sa->sa_family;
1516 				error = copyout(&ifr, ifrp, sizeof ifr);
1517 				ifrp++;
1518 			} else
1519 #endif
1520 			if (sa->sa_len <= sizeof(*sa)) {
1521 				ifr.ifr_addr = *sa;
1522 				error = copyout(&ifr, ifrp, sizeof ifr);
1523 				ifrp++;
1524 			} else {
1525 				if (space < (sizeof ifr) + sa->sa_len -
1526 					    sizeof(*sa))
1527 					break;
1528 				space -= sa->sa_len - sizeof(*sa);
1529 				error = copyout(&ifr, ifrp,
1530 						sizeof ifr.ifr_name);
1531 				if (error == 0)
1532 					error = copyout(sa, &ifrp->ifr_addr,
1533 							sa->sa_len);
1534 				ifrp = (struct ifreq *)
1535 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
1536 			}
1537 			if (error)
1538 				break;
1539 			space -= sizeof ifr;
1540 		}
1541 		if (error)
1542 			break;
1543 		if (!addrs) {
1544 			bzero(&ifr.ifr_addr, sizeof ifr.ifr_addr);
1545 			error = copyout(&ifr, ifrp, sizeof ifr);
1546 			if (error)
1547 				break;
1548 			space -= sizeof ifr;
1549 			ifrp++;
1550 		}
1551 	}
1552 	ifc->ifc_len -= space;
1553 	return (error);
1554 }
1555 
1556 /*
1557  * Just like if_promisc(), but for all-multicast-reception mode.
1558  */
1559 int
1560 if_allmulti(struct ifnet *ifp, int onswitch)
1561 {
1562 	int error = 0;
1563 	struct ifreq ifr;
1564 
1565 	crit_enter();
1566 
1567 	if (onswitch) {
1568 		if (ifp->if_amcount++ == 0) {
1569 			ifp->if_flags |= IFF_ALLMULTI;
1570 			ifr.ifr_flags = ifp->if_flags;
1571 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1572 			lwkt_serialize_enter(ifp->if_serializer);
1573 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1574 					      (struct ucred *)NULL);
1575 			lwkt_serialize_exit(ifp->if_serializer);
1576 		}
1577 	} else {
1578 		if (ifp->if_amcount > 1) {
1579 			ifp->if_amcount--;
1580 		} else {
1581 			ifp->if_amcount = 0;
1582 			ifp->if_flags &= ~IFF_ALLMULTI;
1583 			ifr.ifr_flags = ifp->if_flags;
1584 			ifr.ifr_flagshigh = ifp->if_flags >> 16;
1585 			lwkt_serialize_enter(ifp->if_serializer);
1586 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1587 					      (struct ucred *)NULL);
1588 			lwkt_serialize_exit(ifp->if_serializer);
1589 		}
1590 	}
1591 
1592 	crit_exit();
1593 
1594 	if (error == 0)
1595 		rt_ifmsg(ifp);
1596 	return error;
1597 }
1598 
1599 /*
1600  * Add a multicast listenership to the interface in question.
1601  * The link layer provides a routine which converts
1602  */
1603 int
1604 if_addmulti(
1605 	struct ifnet *ifp,	/* interface to manipulate */
1606 	struct sockaddr *sa,	/* address to add */
1607 	struct ifmultiaddr **retifma)
1608 {
1609 	struct sockaddr *llsa, *dupsa;
1610 	int error;
1611 	struct ifmultiaddr *ifma;
1612 
1613 	/*
1614 	 * If the matching multicast address already exists
1615 	 * then don't add a new one, just add a reference
1616 	 */
1617 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1618 		if (sa_equal(sa, ifma->ifma_addr)) {
1619 			ifma->ifma_refcount++;
1620 			if (retifma)
1621 				*retifma = ifma;
1622 			return 0;
1623 		}
1624 	}
1625 
1626 	/*
1627 	 * Give the link layer a chance to accept/reject it, and also
1628 	 * find out which AF_LINK address this maps to, if it isn't one
1629 	 * already.
1630 	 */
1631 	if (ifp->if_resolvemulti) {
1632 		lwkt_serialize_enter(ifp->if_serializer);
1633 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
1634 		lwkt_serialize_exit(ifp->if_serializer);
1635 		if (error)
1636 			return error;
1637 	} else {
1638 		llsa = 0;
1639 	}
1640 
1641 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
1642 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
1643 	bcopy(sa, dupsa, sa->sa_len);
1644 
1645 	ifma->ifma_addr = dupsa;
1646 	ifma->ifma_lladdr = llsa;
1647 	ifma->ifma_ifp = ifp;
1648 	ifma->ifma_refcount = 1;
1649 	ifma->ifma_protospec = 0;
1650 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
1651 
1652 	/*
1653 	 * Some network interfaces can scan the address list at
1654 	 * interrupt time; lock them out.
1655 	 */
1656 	crit_enter();
1657 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1658 	crit_exit();
1659 	*retifma = ifma;
1660 
1661 	if (llsa != 0) {
1662 		LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1663 			if (sa_equal(ifma->ifma_addr, llsa))
1664 				break;
1665 		}
1666 		if (ifma) {
1667 			ifma->ifma_refcount++;
1668 		} else {
1669 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
1670 			       M_IFMADDR, M_WAITOK);
1671 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
1672 			       M_IFMADDR, M_WAITOK);
1673 			bcopy(llsa, dupsa, llsa->sa_len);
1674 			ifma->ifma_addr = dupsa;
1675 			ifma->ifma_ifp = ifp;
1676 			ifma->ifma_refcount = 1;
1677 			crit_enter();
1678 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
1679 			crit_exit();
1680 		}
1681 	}
1682 	/*
1683 	 * We are certain we have added something, so call down to the
1684 	 * interface to let them know about it.
1685 	 */
1686 	crit_enter();
1687 	lwkt_serialize_enter(ifp->if_serializer);
1688 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0, (struct ucred *)NULL);
1689 	lwkt_serialize_exit(ifp->if_serializer);
1690 	crit_exit();
1691 
1692 	return 0;
1693 }
1694 
1695 /*
1696  * Remove a reference to a multicast address on this interface.  Yell
1697  * if the request does not match an existing membership.
1698  */
1699 int
1700 if_delmulti(struct ifnet *ifp, struct sockaddr *sa)
1701 {
1702 	struct ifmultiaddr *ifma;
1703 
1704 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1705 		if (sa_equal(sa, ifma->ifma_addr))
1706 			break;
1707 	if (ifma == 0)
1708 		return ENOENT;
1709 
1710 	if (ifma->ifma_refcount > 1) {
1711 		ifma->ifma_refcount--;
1712 		return 0;
1713 	}
1714 
1715 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
1716 	sa = ifma->ifma_lladdr;
1717 	crit_enter();
1718 	LIST_REMOVE(ifma, ifma_link);
1719 	/*
1720 	 * Make sure the interface driver is notified
1721 	 * in the case of a link layer mcast group being left.
1722 	 */
1723 	if (ifma->ifma_addr->sa_family == AF_LINK && sa == 0) {
1724 		lwkt_serialize_enter(ifp->if_serializer);
1725 		ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1726 		lwkt_serialize_exit(ifp->if_serializer);
1727 	}
1728 	crit_exit();
1729 	kfree(ifma->ifma_addr, M_IFMADDR);
1730 	kfree(ifma, M_IFMADDR);
1731 	if (sa == 0)
1732 		return 0;
1733 
1734 	/*
1735 	 * Now look for the link-layer address which corresponds to
1736 	 * this network address.  It had been squirreled away in
1737 	 * ifma->ifma_lladdr for this purpose (so we don't have
1738 	 * to call ifp->if_resolvemulti() again), and we saved that
1739 	 * value in sa above.  If some nasty deleted the
1740 	 * link-layer address out from underneath us, we can deal because
1741 	 * the address we stored was is not the same as the one which was
1742 	 * in the record for the link-layer address.  (So we don't complain
1743 	 * in that case.)
1744 	 */
1745 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1746 		if (sa_equal(sa, ifma->ifma_addr))
1747 			break;
1748 	if (ifma == 0)
1749 		return 0;
1750 
1751 	if (ifma->ifma_refcount > 1) {
1752 		ifma->ifma_refcount--;
1753 		return 0;
1754 	}
1755 
1756 	crit_enter();
1757 	lwkt_serialize_enter(ifp->if_serializer);
1758 	LIST_REMOVE(ifma, ifma_link);
1759 	ifp->if_ioctl(ifp, SIOCDELMULTI, 0, (struct ucred *)NULL);
1760 	lwkt_serialize_exit(ifp->if_serializer);
1761 	crit_exit();
1762 	kfree(ifma->ifma_addr, M_IFMADDR);
1763 	kfree(sa, M_IFMADDR);
1764 	kfree(ifma, M_IFMADDR);
1765 
1766 	return 0;
1767 }
1768 
1769 /*
1770  * Set the link layer address on an interface.
1771  *
1772  * At this time we only support certain types of interfaces,
1773  * and we don't allow the length of the address to change.
1774  */
1775 int
1776 if_setlladdr(struct ifnet *ifp, const u_char *lladdr, int len)
1777 {
1778 	struct sockaddr_dl *sdl;
1779 	struct ifaddr *ifa;
1780 	struct ifreq ifr;
1781 
1782 	sdl = IF_LLSOCKADDR(ifp);
1783 	if (sdl == NULL)
1784 		return (EINVAL);
1785 	if (len != sdl->sdl_alen)	/* don't allow length to change */
1786 		return (EINVAL);
1787 	switch (ifp->if_type) {
1788 	case IFT_ETHER:			/* these types use struct arpcom */
1789 	case IFT_FDDI:
1790 	case IFT_XETHER:
1791 	case IFT_ISO88025:
1792 	case IFT_L2VLAN:
1793 		bcopy(lladdr, ((struct arpcom *)ifp->if_softc)->ac_enaddr, len);
1794 		/* FALLTHROUGH */
1795 	case IFT_ARCNET:
1796 		bcopy(lladdr, LLADDR(sdl), len);
1797 		break;
1798 	default:
1799 		return (ENODEV);
1800 	}
1801 	/*
1802 	 * If the interface is already up, we need
1803 	 * to re-init it in order to reprogram its
1804 	 * address filter.
1805 	 */
1806 	lwkt_serialize_enter(ifp->if_serializer);
1807 	if ((ifp->if_flags & IFF_UP) != 0) {
1808 		ifp->if_flags &= ~IFF_UP;
1809 		ifr.ifr_flags = ifp->if_flags;
1810 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1811 		ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1812 			      (struct ucred *)NULL);
1813 		ifp->if_flags |= IFF_UP;
1814 		ifr.ifr_flags = ifp->if_flags;
1815 		ifr.ifr_flagshigh = ifp->if_flags >> 16;
1816 		ifp->if_ioctl(ifp, SIOCSIFFLAGS, (caddr_t)&ifr,
1817 				 (struct ucred *)NULL);
1818 #ifdef INET
1819 		/*
1820 		 * Also send gratuitous ARPs to notify other nodes about
1821 		 * the address change.
1822 		 */
1823 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1824 			if (ifa->ifa_addr != NULL &&
1825 			    ifa->ifa_addr->sa_family == AF_INET)
1826 				arp_ifinit(ifp, ifa);
1827 		}
1828 #endif
1829 	}
1830 	lwkt_serialize_exit(ifp->if_serializer);
1831 	return (0);
1832 }
1833 
1834 struct ifmultiaddr *
1835 ifmaof_ifpforaddr(struct sockaddr *sa, struct ifnet *ifp)
1836 {
1837 	struct ifmultiaddr *ifma;
1838 
1839 	LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link)
1840 		if (sa_equal(ifma->ifma_addr, sa))
1841 			break;
1842 
1843 	return ifma;
1844 }
1845 
1846 /*
1847  * The name argument must be a pointer to storage which will last as
1848  * long as the interface does.  For physical devices, the result of
1849  * device_get_name(dev) is a good choice and for pseudo-devices a
1850  * static string works well.
1851  */
1852 void
1853 if_initname(struct ifnet *ifp, const char *name, int unit)
1854 {
1855 	ifp->if_dname = name;
1856 	ifp->if_dunit = unit;
1857 	if (unit != IF_DUNIT_NONE)
1858 		ksnprintf(ifp->if_xname, IFNAMSIZ, "%s%d", name, unit);
1859 	else
1860 		strlcpy(ifp->if_xname, name, IFNAMSIZ);
1861 }
1862 
1863 int
1864 if_printf(struct ifnet *ifp, const char *fmt, ...)
1865 {
1866 	__va_list ap;
1867 	int retval;
1868 
1869 	retval = kprintf("%s: ", ifp->if_xname);
1870 	__va_start(ap, fmt);
1871 	retval += kvprintf(fmt, ap);
1872 	__va_end(ap);
1873 	return (retval);
1874 }
1875 
1876 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1877 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1878 
1879 void
1880 ifq_set_classic(struct ifaltq *ifq)
1881 {
1882 	ifq->altq_enqueue = ifq_classic_enqueue;
1883 	ifq->altq_dequeue = ifq_classic_dequeue;
1884 	ifq->altq_request = ifq_classic_request;
1885 }
1886 
1887 static int
1888 ifq_classic_enqueue(struct ifaltq *ifq, struct mbuf *m,
1889 		    struct altq_pktattr *pa __unused)
1890 {
1891 	crit_enter();
1892 	if (IF_QFULL(ifq)) {
1893 		m_freem(m);
1894 		crit_exit();
1895 		return(ENOBUFS);
1896 	} else {
1897 		IF_ENQUEUE(ifq, m);
1898 		crit_exit();
1899 		return(0);
1900 	}
1901 }
1902 
1903 static struct mbuf *
1904 ifq_classic_dequeue(struct ifaltq *ifq, struct mbuf *mpolled, int op)
1905 {
1906 	struct mbuf *m;
1907 
1908 	crit_enter();
1909 	switch (op) {
1910 	case ALTDQ_POLL:
1911 		IF_POLL(ifq, m);
1912 		break;
1913 	case ALTDQ_REMOVE:
1914 		IF_DEQUEUE(ifq, m);
1915 		break;
1916 	default:
1917 		panic("unsupported ALTQ dequeue op: %d", op);
1918 	}
1919 	crit_exit();
1920 	KKASSERT(mpolled == NULL || mpolled == m);
1921 	return(m);
1922 }
1923 
1924 static int
1925 ifq_classic_request(struct ifaltq *ifq, int req, void *arg)
1926 {
1927 	crit_enter();
1928 	switch (req) {
1929 	case ALTRQ_PURGE:
1930 		IF_DRAIN(ifq);
1931 		break;
1932 	default:
1933 		panic("unspported ALTQ request: %d", req);
1934 	}
1935 	crit_exit();
1936 	return(0);
1937 }
1938 
1939