xref: /freebsd/sys/net/if.c (revision 02f2e93b60c2b91feac8f45c4c889a5a8e40d8a2)
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  *	$Id: if.c,v 1.54 1997/10/07 07:40:35 joerg Exp $
35  */
36 
37 #include <sys/param.h>
38 #include <sys/queue.h>
39 #include <sys/malloc.h>
40 #include <sys/mbuf.h>
41 #include <sys/systm.h>
42 #include <sys/proc.h>
43 #include <sys/socket.h>
44 #include <sys/socketvar.h>
45 #include <sys/protosw.h>
46 #include <sys/kernel.h>
47 #include <sys/sockio.h>
48 #include <sys/errno.h>
49 #include <sys/syslog.h>
50 #include <sys/sysctl.h>
51 
52 #include <net/if.h>
53 #include <net/if_dl.h>
54 #include <net/if_types.h>
55 #include <net/radix.h>
56 
57 /*
58  * System initialization
59  */
60 
61 static int ifconf __P((int, caddr_t));
62 static void ifinit __P((void *));
63 static void if_qflush __P((struct ifqueue *));
64 static void if_slowtimo __P((void *));
65 static void link_rtrequest __P((int, struct rtentry *, struct sockaddr *));
66 
67 SYSINIT(interfaces, SI_SUB_PROTO_IF, SI_ORDER_FIRST, ifinit, NULL)
68 
69 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
70 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
71 
72 int	ifqmaxlen = IFQ_MAXLEN;
73 struct	ifnethead ifnet;	/* depend on static init XXX */
74 
75 /*
76  * Network interface utility routines.
77  *
78  * Routines with ifa_ifwith* names take sockaddr *'s as
79  * parameters.
80  *
81  * This routine assumes that it will be called at splimp() or higher.
82  */
83 /* ARGSUSED*/
84 void
85 ifinit(dummy)
86 	void *dummy;
87 {
88 	register struct ifnet *ifp;
89 
90 	for (ifp = ifnet.tqh_first; ifp; ifp = ifp->if_link.tqe_next)
91 		if (ifp->if_snd.ifq_maxlen == 0)
92 			ifp->if_snd.ifq_maxlen = ifqmaxlen;
93 	if_slowtimo(0);
94 }
95 
96 int if_index = 0;
97 struct ifaddr **ifnet_addrs;
98 
99 
100 /*
101  * Attach an interface to the
102  * list of "active" interfaces.
103  */
104 void
105 if_attach(ifp)
106 	struct ifnet *ifp;
107 {
108 	unsigned socksize, ifasize;
109 	int namelen, masklen;
110 	char workbuf[64];
111 	register struct sockaddr_dl *sdl;
112 	register struct ifaddr *ifa;
113 	static int if_indexlim = 8;
114 	static int inited;
115 
116 	if (!inited) {
117 		TAILQ_INIT(&ifnet);
118 		inited = 1;
119 	}
120 
121 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_link);
122 	ifp->if_index = ++if_index;
123 	/*
124 	 * XXX -
125 	 * The old code would work if the interface passed a pre-existing
126 	 * chain of ifaddrs to this code.  We don't trust our callers to
127 	 * properly initialize the tailq, however, so we no longer allow
128 	 * this unlikely case.
129 	 */
130 	TAILQ_INIT(&ifp->if_addrhead);
131 	LIST_INIT(&ifp->if_multiaddrs);
132 	microtime(&ifp->if_lastchange);
133 	if (ifnet_addrs == 0 || if_index >= if_indexlim) {
134 		unsigned n = (if_indexlim <<= 1) * sizeof(ifa);
135 		struct ifaddr **q = (struct ifaddr **)
136 					malloc(n, M_IFADDR, M_WAITOK);
137 		bzero((caddr_t)q, n);
138 		if (ifnet_addrs) {
139 			bcopy((caddr_t)ifnet_addrs, (caddr_t)q, n/2);
140 			free((caddr_t)ifnet_addrs, M_IFADDR);
141 		}
142 		ifnet_addrs = q;
143 	}
144 	/*
145 	 * create a Link Level name for this device
146 	 */
147 	namelen = sprintf(workbuf, "%s%d", ifp->if_name, ifp->if_unit);
148 #define _offsetof(t, m) ((int)((caddr_t)&((t *)0)->m))
149 	masklen = _offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
150 	socksize = masklen + ifp->if_addrlen;
151 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
152 	socksize = ROUNDUP(socksize);
153 	if (socksize < sizeof(*sdl))
154 		socksize = sizeof(*sdl);
155 	ifasize = sizeof(*ifa) + 2 * socksize;
156 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
157 	if (ifa) {
158 		bzero((caddr_t)ifa, ifasize);
159 		sdl = (struct sockaddr_dl *)(ifa + 1);
160 		sdl->sdl_len = socksize;
161 		sdl->sdl_family = AF_LINK;
162 		bcopy(workbuf, sdl->sdl_data, namelen);
163 		sdl->sdl_nlen = namelen;
164 		sdl->sdl_index = ifp->if_index;
165 		sdl->sdl_type = ifp->if_type;
166 		ifnet_addrs[if_index - 1] = ifa;
167 		ifa->ifa_ifp = ifp;
168 		ifa->ifa_rtrequest = link_rtrequest;
169 		ifa->ifa_addr = (struct sockaddr *)sdl;
170 		sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
171 		ifa->ifa_netmask = (struct sockaddr *)sdl;
172 		sdl->sdl_len = masklen;
173 		while (namelen != 0)
174 			sdl->sdl_data[--namelen] = 0xff;
175 		TAILQ_INSERT_HEAD(&ifp->if_addrhead, ifa, ifa_link);
176 	}
177 }
178 /*
179  * Locate an interface based on a complete address.
180  */
181 /*ARGSUSED*/
182 struct ifaddr *
183 ifa_ifwithaddr(addr)
184 	register struct sockaddr *addr;
185 {
186 	register struct ifnet *ifp;
187 	register struct ifaddr *ifa;
188 
189 #define	equal(a1, a2) \
190   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
191 	for (ifp = ifnet.tqh_first; ifp; ifp = ifp->if_link.tqe_next)
192 	    for (ifa = ifp->if_addrhead.tqh_first; ifa;
193 		 ifa = ifa->ifa_link.tqe_next) {
194 		if (ifa->ifa_addr->sa_family != addr->sa_family)
195 			continue;
196 		if (equal(addr, ifa->ifa_addr))
197 			return (ifa);
198 		if ((ifp->if_flags & IFF_BROADCAST) && ifa->ifa_broadaddr &&
199 		    equal(ifa->ifa_broadaddr, addr))
200 			return (ifa);
201 	}
202 	return ((struct ifaddr *)0);
203 }
204 /*
205  * Locate the point to point interface with a given destination address.
206  */
207 /*ARGSUSED*/
208 struct ifaddr *
209 ifa_ifwithdstaddr(addr)
210 	register struct sockaddr *addr;
211 {
212 	register struct ifnet *ifp;
213 	register struct ifaddr *ifa;
214 
215 	for (ifp = ifnet.tqh_first; ifp; ifp = ifp->if_link.tqe_next)
216 	    if (ifp->if_flags & IFF_POINTOPOINT)
217 		for (ifa = ifp->if_addrhead.tqh_first; ifa;
218 		     ifa = ifa->ifa_link.tqe_next) {
219 			if (ifa->ifa_addr->sa_family != addr->sa_family)
220 				continue;
221 			if (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr))
222 				return (ifa);
223 	}
224 	return ((struct ifaddr *)0);
225 }
226 
227 /*
228  * Find an interface on a specific network.  If many, choice
229  * is most specific found.
230  */
231 struct ifaddr *
232 ifa_ifwithnet(addr)
233 	struct sockaddr *addr;
234 {
235 	register struct ifnet *ifp;
236 	register struct ifaddr *ifa;
237 	struct ifaddr *ifa_maybe = (struct ifaddr *) 0;
238 	u_int af = addr->sa_family;
239 	char *addr_data = addr->sa_data, *cplim;
240 
241 	/*
242 	 * AF_LINK addresses can be looked up directly by their index number,
243 	 * so do that if we can.
244 	 */
245 	if (af == AF_LINK) {
246 	    register struct sockaddr_dl *sdl = (struct sockaddr_dl *)addr;
247 	    if (sdl->sdl_index && sdl->sdl_index <= if_index)
248 		return (ifnet_addrs[sdl->sdl_index - 1]);
249 	}
250 
251 	/*
252 	 * Scan though each interface, looking for ones that have
253 	 * addresses in this address family.
254 	 */
255 	for (ifp = ifnet.tqh_first; ifp; ifp = ifp->if_link.tqe_next) {
256 		for (ifa = ifp->if_addrhead.tqh_first; ifa;
257 		     ifa = ifa->ifa_link.tqe_next) {
258 			register char *cp, *cp2, *cp3;
259 
260 			if (ifa->ifa_addr->sa_family != af)
261 next:				continue;
262 			if (ifp->if_flags & IFF_POINTOPOINT) {
263 				/*
264 				 * This is a bit broken as it doesn't
265 				 * take into account that the remote end may
266 				 * be a single node in the network we are
267 				 * looking for.
268 				 * The trouble is that we don't know the
269 				 * netmask for the remote end.
270 				 */
271 				if (ifa->ifa_dstaddr != 0
272 				    && equal(addr, ifa->ifa_dstaddr))
273  					return (ifa);
274 			} else {
275 				/*
276 				 * if we have a special address handler,
277 				 * then use it instead of the generic one.
278 				 */
279 	          		if (ifa->ifa_claim_addr) {
280 					if ((*ifa->ifa_claim_addr)(ifa, addr)) {
281 						return (ifa);
282 					} else {
283 						continue;
284 					}
285 				}
286 
287 				/*
288 				 * Scan all the bits in the ifa's address.
289 				 * If a bit dissagrees with what we are
290 				 * looking for, mask it with the netmask
291 				 * to see if it really matters.
292 				 * (A byte at a time)
293 				 */
294 				if (ifa->ifa_netmask == 0)
295 					continue;
296 				cp = addr_data;
297 				cp2 = ifa->ifa_addr->sa_data;
298 				cp3 = ifa->ifa_netmask->sa_data;
299 				cplim = ifa->ifa_netmask->sa_len
300 					+ (char *)ifa->ifa_netmask;
301 				while (cp3 < cplim)
302 					if ((*cp++ ^ *cp2++) & *cp3++)
303 						goto next; /* next address! */
304 				/*
305 				 * If the netmask of what we just found
306 				 * is more specific than what we had before
307 				 * (if we had one) then remember the new one
308 				 * before continuing to search
309 				 * for an even better one.
310 				 */
311 				if (ifa_maybe == 0 ||
312 				    rn_refines((caddr_t)ifa->ifa_netmask,
313 				    (caddr_t)ifa_maybe->ifa_netmask))
314 					ifa_maybe = ifa;
315 			}
316 		}
317 	}
318 	return (ifa_maybe);
319 }
320 
321 /*
322  * Find an interface address specific to an interface best matching
323  * a given address.
324  */
325 struct ifaddr *
326 ifaof_ifpforaddr(addr, ifp)
327 	struct sockaddr *addr;
328 	register struct ifnet *ifp;
329 {
330 	register struct ifaddr *ifa;
331 	register char *cp, *cp2, *cp3;
332 	register char *cplim;
333 	struct ifaddr *ifa_maybe = 0;
334 	u_int af = addr->sa_family;
335 
336 	if (af >= AF_MAX)
337 		return (0);
338 	for (ifa = ifp->if_addrhead.tqh_first; ifa;
339 	     ifa = ifa->ifa_link.tqe_next) {
340 		if (ifa->ifa_addr->sa_family != af)
341 			continue;
342 		if (ifa_maybe == 0)
343 			ifa_maybe = ifa;
344 		if (ifa->ifa_netmask == 0) {
345 			if (equal(addr, ifa->ifa_addr) ||
346 			    (ifa->ifa_dstaddr && equal(addr, ifa->ifa_dstaddr)))
347 				return (ifa);
348 			continue;
349 		}
350 		if (ifp->if_flags & IFF_POINTOPOINT) {
351 			if (equal(addr, ifa->ifa_dstaddr))
352 				return (ifa);
353 		} else {
354 			cp = addr->sa_data;
355 			cp2 = ifa->ifa_addr->sa_data;
356 			cp3 = ifa->ifa_netmask->sa_data;
357 			cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
358 			for (; cp3 < cplim; cp3++)
359 				if ((*cp++ ^ *cp2++) & *cp3)
360 					break;
361 			if (cp3 == cplim)
362 				return (ifa);
363 		}
364 	}
365 	return (ifa_maybe);
366 }
367 
368 #include <net/route.h>
369 
370 /*
371  * Default action when installing a route with a Link Level gateway.
372  * Lookup an appropriate real ifa to point to.
373  * This should be moved to /sys/net/link.c eventually.
374  */
375 static void
376 link_rtrequest(cmd, rt, sa)
377 	int cmd;
378 	register struct rtentry *rt;
379 	struct sockaddr *sa;
380 {
381 	register struct ifaddr *ifa;
382 	struct sockaddr *dst;
383 	struct ifnet *ifp;
384 
385 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
386 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
387 		return;
388 	ifa = ifaof_ifpforaddr(dst, ifp);
389 	if (ifa) {
390 		IFAFREE(rt->rt_ifa);
391 		rt->rt_ifa = ifa;
392 		ifa->ifa_refcnt++;
393 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
394 			ifa->ifa_rtrequest(cmd, rt, sa);
395 	}
396 }
397 
398 /*
399  * Mark an interface down and notify protocols of
400  * the transition.
401  * NOTE: must be called at splnet or eqivalent.
402  */
403 void
404 if_down(ifp)
405 	register struct ifnet *ifp;
406 {
407 	register struct ifaddr *ifa;
408 
409 	ifp->if_flags &= ~IFF_UP;
410 	microtime(&ifp->if_lastchange);
411 	for (ifa = ifp->if_addrhead.tqh_first; ifa;
412 	     ifa = ifa->ifa_link.tqe_next)
413 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
414 	if_qflush(&ifp->if_snd);
415 	rt_ifmsg(ifp);
416 }
417 
418 /*
419  * Mark an interface up and notify protocols of
420  * the transition.
421  * NOTE: must be called at splnet or eqivalent.
422  */
423 void
424 if_up(ifp)
425 	register struct ifnet *ifp;
426 {
427 	register struct ifaddr *ifa;
428 
429 	ifp->if_flags |= IFF_UP;
430 	microtime(&ifp->if_lastchange);
431 	for (ifa = ifp->if_addrhead.tqh_first; ifa;
432 	     ifa = ifa->ifa_link.tqe_next)
433 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
434 	rt_ifmsg(ifp);
435 }
436 
437 /*
438  * Flush an interface queue.
439  */
440 static void
441 if_qflush(ifq)
442 	register struct ifqueue *ifq;
443 {
444 	register struct mbuf *m, *n;
445 
446 	n = ifq->ifq_head;
447 	while ((m = n) != 0) {
448 		n = m->m_act;
449 		m_freem(m);
450 	}
451 	ifq->ifq_head = 0;
452 	ifq->ifq_tail = 0;
453 	ifq->ifq_len = 0;
454 }
455 
456 /*
457  * Handle interface watchdog timer routines.  Called
458  * from softclock, we decrement timers (if set) and
459  * call the appropriate interface routine on expiration.
460  */
461 static void
462 if_slowtimo(arg)
463 	void *arg;
464 {
465 	register struct ifnet *ifp;
466 	int s = splimp();
467 
468 	for (ifp = ifnet.tqh_first; ifp; ifp = ifp->if_link.tqe_next) {
469 		if (ifp->if_timer == 0 || --ifp->if_timer)
470 			continue;
471 		if (ifp->if_watchdog)
472 			(*ifp->if_watchdog)(ifp);
473 	}
474 	splx(s);
475 	timeout(if_slowtimo, (void *)0, hz / IFNET_SLOWHZ);
476 }
477 
478 /*
479  * Map interface name to
480  * interface structure pointer.
481  */
482 struct ifnet *
483 ifunit(name)
484 	register char *name;
485 {
486 	register char *cp;
487 	register struct ifnet *ifp;
488 	int unit;
489 	unsigned len;
490 	char *ep, c;
491 
492 	for (cp = name; cp < name + IFNAMSIZ && *cp; cp++)
493 		if (*cp >= '0' && *cp <= '9')
494 			break;
495 	if (*cp == '\0' || cp == name + IFNAMSIZ)
496 		return ((struct ifnet *)0);
497 	/*
498 	 * Save first char of unit, and pointer to it,
499 	 * so we can put a null there to avoid matching
500 	 * initial substrings of interface names.
501 	 */
502 	len = cp - name + 1;
503 	c = *cp;
504 	ep = cp;
505 	for (unit = 0; *cp >= '0' && *cp <= '9'; )
506 		unit = unit * 10 + *cp++ - '0';
507 	if (*cp != '\0')
508 		return 0;	/* no trailing garbage allowed */
509 	*ep = 0;
510 	for (ifp = ifnet.tqh_first; ifp; ifp = ifp->if_link.tqe_next) {
511 		if (bcmp(ifp->if_name, name, len))
512 			continue;
513 		if (unit == ifp->if_unit)
514 			break;
515 	}
516 	*ep = c;
517 	return (ifp);
518 }
519 
520 /*
521  * Interface ioctls.
522  */
523 int
524 ifioctl(so, cmd, data, p)
525 	struct socket *so;
526 	int cmd;
527 	caddr_t data;
528 	struct proc *p;
529 {
530 	register struct ifnet *ifp;
531 	register struct ifreq *ifr;
532 	int error;
533 
534 	switch (cmd) {
535 
536 	case SIOCGIFCONF:
537 	case OSIOCGIFCONF:
538 		return (ifconf(cmd, data));
539 	}
540 	ifr = (struct ifreq *)data;
541 	ifp = ifunit(ifr->ifr_name);
542 	if (ifp == 0)
543 		return (ENXIO);
544 	switch (cmd) {
545 
546 	case SIOCGIFFLAGS:
547 		ifr->ifr_flags = ifp->if_flags;
548 		break;
549 
550 	case SIOCGIFMETRIC:
551 		ifr->ifr_metric = ifp->if_metric;
552 		break;
553 
554 	case SIOCGIFMTU:
555 		ifr->ifr_mtu = ifp->if_mtu;
556 		break;
557 
558 	case SIOCGIFPHYS:
559 		ifr->ifr_phys = ifp->if_physical;
560 		break;
561 
562 	case SIOCSIFFLAGS:
563 		error = suser(p->p_ucred, &p->p_acflag);
564 		if (error)
565 			return (error);
566 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
567 			int s = splimp();
568 			if_down(ifp);
569 			splx(s);
570 		}
571 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
572 			int s = splimp();
573 			if_up(ifp);
574 			splx(s);
575 		}
576 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
577 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
578 		if (ifp->if_ioctl)
579 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
580 		microtime(&ifp->if_lastchange);
581 		break;
582 
583 	case SIOCSIFMETRIC:
584 		error = suser(p->p_ucred, &p->p_acflag);
585 		if (error)
586 			return (error);
587 		ifp->if_metric = ifr->ifr_metric;
588 		microtime(&ifp->if_lastchange);
589 		break;
590 
591 	case SIOCSIFPHYS:
592 		error = suser(p->p_ucred, &p->p_acflag);
593 		if (error)
594 			return error;
595 		if (!ifp->if_ioctl)
596 		        return EOPNOTSUPP;
597 		error = (*ifp->if_ioctl)(ifp, cmd, data);
598 		if (error == 0)
599 			microtime(&ifp->if_lastchange);
600 		return(error);
601 
602 	case SIOCSIFMTU:
603 		error = suser(p->p_ucred, &p->p_acflag);
604 		if (error)
605 			return (error);
606 		if (ifp->if_ioctl == NULL)
607 			return (EOPNOTSUPP);
608 		/*
609 		 * 72 was chosen below because it is the size of a TCP/IP
610 		 * header (40) + the minimum mss (32).
611 		 */
612 		if (ifr->ifr_mtu < 72 || ifr->ifr_mtu > 65535)
613 			return (EINVAL);
614 		error = (*ifp->if_ioctl)(ifp, cmd, data);
615 		if (error == 0)
616 			microtime(&ifp->if_lastchange);
617 		return(error);
618 
619 	case SIOCADDMULTI:
620 	case SIOCDELMULTI:
621 		error = suser(p->p_ucred, &p->p_acflag);
622 		if (error)
623 			return (error);
624 
625 		/* Don't allow group membership on non-multicast interfaces. */
626 		if ((ifp->if_flags & IFF_MULTICAST) == 0)
627 			return EOPNOTSUPP;
628 
629 		/* Don't let users screw up protocols' entries. */
630 		if (ifr->ifr_addr.sa_family != AF_LINK)
631 			return EINVAL;
632 
633 		if (cmd == SIOCADDMULTI) {
634 			struct ifmultiaddr *ifma;
635 			error = if_addmulti(ifp, &ifr->ifr_addr, &ifma);
636 		} else {
637 			error = if_delmulti(ifp, &ifr->ifr_addr);
638 		}
639 		if (error == 0)
640 			microtime(&ifp->if_lastchange);
641 		return error;
642 
643         case SIOCSIFMEDIA:
644 	case SIOCSIFGENERIC:
645 		error = suser(p->p_ucred, &p->p_acflag);
646 		if (error)
647 			return (error);
648 		if (ifp->if_ioctl == 0)
649 			return (EOPNOTSUPP);
650 		error = (*ifp->if_ioctl)(ifp, cmd, data);
651 		if (error == 0)
652 			microtime(&ifp->if_lastchange);
653 		return error;
654 
655 	case SIOCGIFMEDIA:
656 	case SIOCGIFGENERIC:
657 		if (ifp->if_ioctl == 0)
658 			return (EOPNOTSUPP);
659 		return ((*ifp->if_ioctl)(ifp, cmd, data));
660 
661 	default:
662 		if (so->so_proto == 0)
663 			return (EOPNOTSUPP);
664 #ifndef COMPAT_43
665 		return ((*so->so_proto->pr_usrreqs->pru_control)(so, cmd,
666 								 data,
667 								 ifp, p));
668 #else
669 	    {
670 		int ocmd = cmd;
671 
672 		switch (cmd) {
673 
674 		case SIOCSIFDSTADDR:
675 		case SIOCSIFADDR:
676 		case SIOCSIFBRDADDR:
677 		case SIOCSIFNETMASK:
678 #if BYTE_ORDER != BIG_ENDIAN
679 			if (ifr->ifr_addr.sa_family == 0 &&
680 			    ifr->ifr_addr.sa_len < 16) {
681 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
682 				ifr->ifr_addr.sa_len = 16;
683 			}
684 #else
685 			if (ifr->ifr_addr.sa_len == 0)
686 				ifr->ifr_addr.sa_len = 16;
687 #endif
688 			break;
689 
690 		case OSIOCGIFADDR:
691 			cmd = SIOCGIFADDR;
692 			break;
693 
694 		case OSIOCGIFDSTADDR:
695 			cmd = SIOCGIFDSTADDR;
696 			break;
697 
698 		case OSIOCGIFBRDADDR:
699 			cmd = SIOCGIFBRDADDR;
700 			break;
701 
702 		case OSIOCGIFNETMASK:
703 			cmd = SIOCGIFNETMASK;
704 		}
705 		error =  ((*so->so_proto->pr_usrreqs->pru_control)(so,
706 								   cmd,
707 								   data,
708 								   ifp, p));
709 		switch (ocmd) {
710 
711 		case OSIOCGIFADDR:
712 		case OSIOCGIFDSTADDR:
713 		case OSIOCGIFBRDADDR:
714 		case OSIOCGIFNETMASK:
715 			*(u_short *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
716 		}
717 		return (error);
718 
719 	    }
720 #endif
721 	}
722 	return (0);
723 }
724 
725 /*
726  * Set/clear promiscuous mode on interface ifp based on the truth value
727  * of pswitch.  The calls are reference counted so that only the first
728  * "on" request actually has an effect, as does the final "off" request.
729  * Results are undefined if the "off" and "on" requests are not matched.
730  */
731 int
732 ifpromisc(ifp, pswitch)
733 	struct ifnet *ifp;
734 	int pswitch;
735 {
736 	struct ifreq ifr;
737 	int error;
738 
739 	if (pswitch) {
740 		/*
741 		 * If the device is not configured up, we cannot put it in
742 		 * promiscuous mode.
743 		 */
744 		if ((ifp->if_flags & IFF_UP) == 0)
745 			return (ENETDOWN);
746 		if (ifp->if_pcount++ != 0)
747 			return (0);
748 		ifp->if_flags |= IFF_PROMISC;
749 		log(LOG_INFO, "%s%d: promiscuous mode enabled\n",
750 		    ifp->if_name, ifp->if_unit);
751 	} else {
752 		if (--ifp->if_pcount > 0)
753 			return (0);
754 		ifp->if_flags &= ~IFF_PROMISC;
755 	}
756 	ifr.ifr_flags = ifp->if_flags;
757 	error = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
758 	if (error == 0)
759 		rt_ifmsg(ifp);
760 	return error;
761 }
762 
763 /*
764  * Return interface configuration
765  * of system.  List may be used
766  * in later ioctl's (above) to get
767  * other information.
768  */
769 /*ARGSUSED*/
770 static int
771 ifconf(cmd, data)
772 	int cmd;
773 	caddr_t data;
774 {
775 	register struct ifconf *ifc = (struct ifconf *)data;
776 	register struct ifnet *ifp = ifnet.tqh_first;
777 	register struct ifaddr *ifa;
778 	struct ifreq ifr, *ifrp;
779 	int space = ifc->ifc_len, error = 0;
780 
781 	ifrp = ifc->ifc_req;
782 	for (; space > sizeof (ifr) && ifp; ifp = ifp->if_link.tqe_next) {
783 		char workbuf[64];
784 		int ifnlen;
785 
786 		ifnlen = sprintf(workbuf, "%s%d", ifp->if_name, ifp->if_unit);
787 		if(ifnlen + 1 > sizeof ifr.ifr_name) {
788 			error = ENAMETOOLONG;
789 		} else {
790 			strcpy(ifr.ifr_name, workbuf);
791 		}
792 
793 		if ((ifa = ifp->if_addrhead.tqh_first) == 0) {
794 			bzero((caddr_t)&ifr.ifr_addr, sizeof(ifr.ifr_addr));
795 			error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
796 			    sizeof (ifr));
797 			if (error)
798 				break;
799 			space -= sizeof (ifr), ifrp++;
800 		} else
801 		    for ( ; space > sizeof (ifr) && ifa;
802 			 ifa = ifa->ifa_link.tqe_next) {
803 			register struct sockaddr *sa = ifa->ifa_addr;
804 #ifdef COMPAT_43
805 			if (cmd == OSIOCGIFCONF) {
806 				struct osockaddr *osa =
807 					 (struct osockaddr *)&ifr.ifr_addr;
808 				ifr.ifr_addr = *sa;
809 				osa->sa_family = sa->sa_family;
810 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
811 						sizeof (ifr));
812 				ifrp++;
813 			} else
814 #endif
815 			if (sa->sa_len <= sizeof(*sa)) {
816 				ifr.ifr_addr = *sa;
817 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
818 						sizeof (ifr));
819 				ifrp++;
820 			} else {
821 				space -= sa->sa_len - sizeof(*sa);
822 				if (space < sizeof (ifr))
823 					break;
824 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
825 						sizeof (ifr.ifr_name));
826 				if (error == 0)
827 				    error = copyout((caddr_t)sa,
828 				      (caddr_t)&ifrp->ifr_addr, sa->sa_len);
829 				ifrp = (struct ifreq *)
830 					(sa->sa_len + (caddr_t)&ifrp->ifr_addr);
831 			}
832 			if (error)
833 				break;
834 			space -= sizeof (ifr);
835 		}
836 	}
837 	ifc->ifc_len -= space;
838 	return (error);
839 }
840 
841 /*
842  * Just like if_promisc(), but for all-multicast-reception mode.
843  */
844 int
845 if_allmulti(ifp, onswitch)
846 	struct ifnet *ifp;
847 	int onswitch;
848 {
849 	int error = 0;
850 	int s = splimp();
851 
852 	if (onswitch) {
853 		if (ifp->if_amcount++ == 0) {
854 			ifp->if_flags |= IFF_ALLMULTI;
855 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, 0);
856 		}
857 	} else {
858 		if (ifp->if_amcount > 1) {
859 			ifp->if_amcount--;
860 		} else {
861 			ifp->if_amcount = 0;
862 			ifp->if_flags &= ~IFF_ALLMULTI;
863 			error = ifp->if_ioctl(ifp, SIOCSIFFLAGS, 0);
864 		}
865 	}
866 	splx(s);
867 
868 	if (error == 0)
869 		rt_ifmsg(ifp);
870 	return error;
871 }
872 
873 /*
874  * Add a multicast listenership to the interface in question.
875  * The link layer provides a routine which converts
876  */
877 int
878 if_addmulti(ifp, sa, retifma)
879 	struct ifnet *ifp;	/* interface to manipulate */
880 	struct sockaddr *sa;	/* address to add */
881 	struct ifmultiaddr **retifma;
882 {
883 	struct sockaddr *llsa, *dupsa;
884 	int error, s;
885 	struct ifmultiaddr *ifma;
886 
887 	/*
888 	 * If the matching multicast address already exists
889 	 * then don't add a new one, just add a reference
890 	 */
891 	for (ifma = ifp->if_multiaddrs.lh_first; ifma;
892 	     ifma = ifma->ifma_link.le_next) {
893 		if (equal(sa, ifma->ifma_addr)) {
894 			ifma->ifma_refcount++;
895 			if (retifma)
896 				*retifma = ifma;
897 			return 0;
898 		}
899 	}
900 
901 	/*
902 	 * Give the link layer a chance to accept/reject it, and also
903 	 * find out which AF_LINK address this maps to, if it isn't one
904 	 * already.
905 	 */
906 	if (ifp->if_resolvemulti) {
907 		error = ifp->if_resolvemulti(ifp, &llsa, sa);
908 		if (error) return error;
909 	} else {
910 		llsa = 0;
911 	}
912 
913 	MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma, M_IFMADDR, M_WAITOK);
914 	MALLOC(dupsa, struct sockaddr *, sa->sa_len, M_IFMADDR, M_WAITOK);
915 	bcopy(sa, dupsa, sa->sa_len);
916 
917 	ifma->ifma_addr = dupsa;
918 	ifma->ifma_lladdr = llsa;
919 	ifma->ifma_ifp = ifp;
920 	ifma->ifma_refcount = 1;
921 	ifma->ifma_protospec = 0;
922 	rt_newmaddrmsg(RTM_NEWMADDR, ifma);
923 
924 	/*
925 	 * Some network interfaces can scan the address list at
926 	 * interrupt time; lock them out.
927 	 */
928 	s = splimp();
929 	LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
930 	splx(s);
931 	*retifma = ifma;
932 
933 	if (llsa != 0) {
934 		for (ifma = ifp->if_multiaddrs.lh_first; ifma;
935 		     ifma = ifma->ifma_link.le_next) {
936 			if (equal(ifma->ifma_addr, llsa))
937 				break;
938 		}
939 		if (ifma) {
940 			ifma->ifma_refcount++;
941 		} else {
942 			MALLOC(ifma, struct ifmultiaddr *, sizeof *ifma,
943 			       M_IFMADDR, M_WAITOK);
944 			MALLOC(dupsa, struct sockaddr *, llsa->sa_len,
945 			       M_IFMADDR, M_WAITOK);
946 			bcopy(llsa, dupsa, llsa->sa_len);
947 			ifma->ifma_addr = dupsa;
948 			ifma->ifma_ifp = ifp;
949 			ifma->ifma_refcount = 1;
950 			s = splimp();
951 			LIST_INSERT_HEAD(&ifp->if_multiaddrs, ifma, ifma_link);
952 			splx(s);
953 		}
954 	}
955 	/*
956 	 * We are certain we have added something, so call down to the
957 	 * interface to let them know about it.
958 	 */
959 	s = splimp();
960 	ifp->if_ioctl(ifp, SIOCADDMULTI, 0);
961 	splx(s);
962 
963 	return 0;
964 }
965 
966 /*
967  * Remove a reference to a multicast address on this interface.  Yell
968  * if the request does not match an existing membership.
969  */
970 int
971 if_delmulti(ifp, sa)
972 	struct ifnet *ifp;
973 	struct sockaddr *sa;
974 {
975 	struct ifmultiaddr *ifma;
976 	int s;
977 
978 	for (ifma = ifp->if_multiaddrs.lh_first; ifma;
979 	     ifma = ifma->ifma_link.le_next)
980 		if (equal(sa, ifma->ifma_addr))
981 			break;
982 	if (ifma == 0)
983 		return ENOENT;
984 
985 	if (ifma->ifma_refcount > 1) {
986 		ifma->ifma_refcount--;
987 		return 0;
988 	}
989 
990 	rt_newmaddrmsg(RTM_DELMADDR, ifma);
991 	sa = ifma->ifma_lladdr;
992 	s = splimp();
993 	LIST_REMOVE(ifma, ifma_link);
994 	splx(s);
995 	free(ifma->ifma_addr, M_IFMADDR);
996 	free(ifma, M_IFMADDR);
997 	if (sa == 0)
998 		return 0;
999 
1000 	/*
1001 	 * Now look for the link-layer address which corresponds to
1002 	 * this network address.  It had been squirreled away in
1003 	 * ifma->ifma_lladdr for this purpose (so we don't have
1004 	 * to call ifp->if_resolvemulti() again), and we saved that
1005 	 * value in sa above.  If some nasty deleted the
1006 	 * link-layer address out from underneath us, we can deal because
1007 	 * the address we stored was is not the same as the one which was
1008 	 * in the record for the link-layer address.  (So we don't complain
1009 	 * in that case.)
1010 	 */
1011 	for (ifma = ifp->if_multiaddrs.lh_first; ifma;
1012 	     ifma = ifma->ifma_link.le_next)
1013 		if (equal(sa, ifma->ifma_addr))
1014 			break;
1015 	if (ifma == 0)
1016 		return 0;
1017 
1018 	if (ifma->ifma_refcount > 1) {
1019 		ifma->ifma_refcount--;
1020 		return 0;
1021 	}
1022 
1023 	s = splimp();
1024 	LIST_REMOVE(ifma, ifma_link);
1025 	splx(s);
1026 	free(ifma->ifma_addr, M_IFMADDR);
1027 	free(sa, M_IFMADDR);
1028 	free(ifma, M_IFMADDR);
1029 
1030 	return 0;
1031 }
1032 
1033 struct ifmultiaddr *
1034 ifmaof_ifpforaddr(sa, ifp)
1035 	struct sockaddr *sa;
1036 	struct ifnet *ifp;
1037 {
1038 	struct ifmultiaddr *ifma;
1039 
1040 	for (ifma = ifp->if_multiaddrs.lh_first; ifma;
1041 	     ifma = ifma->ifma_link.le_next)
1042 		if (equal(ifma->ifma_addr, sa))
1043 			break;
1044 
1045 	return ifma;
1046 }
1047 
1048 SYSCTL_NODE(_net, PF_LINK, link, CTLFLAG_RW, 0, "Link layers");
1049 SYSCTL_NODE(_net_link, 0, generic, CTLFLAG_RW, 0, "Generic link-management");
1050