Home | History | Annotate | Line # | Download | only in net
if.c revision 1.104
      1 /*	$NetBSD: if.c,v 1.104 2002/05/12 20:40:11 matt Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1999, 2000, 2001 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by William Studnemund and Jason R. Thorpe.
      9  *
     10  * Redistribution and use in source and binary forms, with or without
     11  * modification, are permitted provided that the following conditions
     12  * are met:
     13  * 1. Redistributions of source code must retain the above copyright
     14  *    notice, this list of conditions and the following disclaimer.
     15  * 2. Redistributions in binary form must reproduce the above copyright
     16  *    notice, this list of conditions and the following disclaimer in the
     17  *    documentation and/or other materials provided with the distribution.
     18  * 3. All advertising materials mentioning features or use of this software
     19  *    must display the following acknowledgement:
     20  *	This product includes software developed by the NetBSD
     21  *	Foundation, Inc. and its contributors.
     22  * 4. Neither the name of The NetBSD Foundation nor the names of its
     23  *    contributors may be used to endorse or promote products derived
     24  *    from this software without specific prior written permission.
     25  *
     26  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     27  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     28  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     29  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     30  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     31  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     32  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     33  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     34  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     35  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     36  * POSSIBILITY OF SUCH DAMAGE.
     37  */
     38 
     39 /*
     40  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
     41  * All rights reserved.
     42  *
     43  * Redistribution and use in source and binary forms, with or without
     44  * modification, are permitted provided that the following conditions
     45  * are met:
     46  * 1. Redistributions of source code must retain the above copyright
     47  *    notice, this list of conditions and the following disclaimer.
     48  * 2. Redistributions in binary form must reproduce the above copyright
     49  *    notice, this list of conditions and the following disclaimer in the
     50  *    documentation and/or other materials provided with the distribution.
     51  * 3. Neither the name of the project nor the names of its contributors
     52  *    may be used to endorse or promote products derived from this software
     53  *    without specific prior written permission.
     54  *
     55  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     56  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     57  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     58  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     59  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     60  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     61  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     62  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     63  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     64  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     65  * SUCH DAMAGE.
     66  */
     67 
     68 /*
     69  * Copyright (c) 1980, 1986, 1993
     70  *	The Regents of the University of California.  All rights reserved.
     71  *
     72  * Redistribution and use in source and binary forms, with or without
     73  * modification, are permitted provided that the following conditions
     74  * are met:
     75  * 1. Redistributions of source code must retain the above copyright
     76  *    notice, this list of conditions and the following disclaimer.
     77  * 2. Redistributions in binary form must reproduce the above copyright
     78  *    notice, this list of conditions and the following disclaimer in the
     79  *    documentation and/or other materials provided with the distribution.
     80  * 3. All advertising materials mentioning features or use of this software
     81  *    must display the following acknowledgement:
     82  *	This product includes software developed by the University of
     83  *	California, Berkeley and its contributors.
     84  * 4. Neither the name of the University nor the names of its contributors
     85  *    may be used to endorse or promote products derived from this software
     86  *    without specific prior written permission.
     87  *
     88  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     89  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     90  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     91  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     92  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     93  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     94  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     95  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     96  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     97  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     98  * SUCH DAMAGE.
     99  *
    100  *	@(#)if.c	8.5 (Berkeley) 1/9/95
    101  */
    102 
    103 #include <sys/cdefs.h>
    104 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.104 2002/05/12 20:40:11 matt Exp $");
    105 
    106 #include "opt_inet.h"
    107 
    108 #include "opt_compat_linux.h"
    109 #include "opt_compat_svr4.h"
    110 #include "opt_compat_43.h"
    111 #include "opt_atalk.h"
    112 #include "opt_pfil_hooks.h"
    113 
    114 #include <sys/param.h>
    115 #include <sys/mbuf.h>
    116 #include <sys/systm.h>
    117 #include <sys/callout.h>
    118 #include <sys/proc.h>
    119 #include <sys/socket.h>
    120 #include <sys/socketvar.h>
    121 #include <sys/domain.h>
    122 #include <sys/protosw.h>
    123 #include <sys/kernel.h>
    124 #include <sys/ioctl.h>
    125 
    126 #include <net/if.h>
    127 #include <net/if_dl.h>
    128 #include <net/if_ether.h>
    129 #include <net/if_ieee80211.h>
    130 #include <net/if_types.h>
    131 #include <net/radix.h>
    132 #include <net/route.h>
    133 #include <net/netisr.h>
    134 #ifdef NETATALK
    135 #include <netatalk/at_extern.h>
    136 #include <netatalk/at.h>
    137 #endif
    138 
    139 #ifdef INET6
    140 /*XXX*/
    141 #include <netinet/in.h>
    142 #include <netinet6/in6_var.h>
    143 #endif
    144 
    145 int	ifqmaxlen = IFQ_MAXLEN;
    146 struct	callout if_slowtimo_ch;
    147 
    148 #ifdef INET6
    149 /*
    150  * XXX: declare here to avoid to include many inet6 related files..
    151  * should be more generalized?
    152  */
    153 void nd6_setmtu __P((struct ifnet *));
    154 #endif
    155 
    156 int netisr;			/* scheduling bits for network */
    157 
    158 int if_rt_walktree __P((struct radix_node *, void *));
    159 
    160 struct if_clone *if_clone_lookup __P((const char *, int *));
    161 int if_clone_list __P((struct if_clonereq *));
    162 
    163 LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
    164 int if_cloners_count;
    165 
    166 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
    167     defined(ISO) || defined(CCITT) || defined(NATM)
    168 static void if_detach_queues __P((struct ifnet *, struct ifqueue *));
    169 #endif
    170 
    171 /*
    172  * Network interface utility routines.
    173  *
    174  * Routines with ifa_ifwith* names take sockaddr *'s as
    175  * parameters.
    176  */
    177 void
    178 ifinit()
    179 {
    180 
    181 	callout_init(&if_slowtimo_ch);
    182 	if_slowtimo(NULL);
    183 }
    184 
    185 /*
    186  * Null routines used while an interface is going away.  These routines
    187  * just return an error.
    188  */
    189 
    190 int
    191 if_nulloutput(ifp, m, so, rt)
    192 	struct ifnet *ifp;
    193 	struct mbuf *m;
    194 	struct sockaddr *so;
    195 	struct rtentry *rt;
    196 {
    197 
    198 	return (ENXIO);
    199 }
    200 
    201 void
    202 if_nullinput(ifp, m)
    203 	struct ifnet *ifp;
    204 	struct mbuf *m;
    205 {
    206 
    207 	/* Nothing. */
    208 }
    209 
    210 void
    211 if_nullstart(ifp)
    212 	struct ifnet *ifp;
    213 {
    214 
    215 	/* Nothing. */
    216 }
    217 
    218 int
    219 if_nullioctl(ifp, cmd, data)
    220 	struct ifnet *ifp;
    221 	u_long cmd;
    222 	caddr_t data;
    223 {
    224 
    225 	return (ENXIO);
    226 }
    227 
    228 int
    229 if_nullinit(ifp)
    230 	struct ifnet *ifp;
    231 {
    232 
    233 	return (ENXIO);
    234 }
    235 
    236 void
    237 if_nullstop(ifp, disable)
    238 	struct ifnet *ifp;
    239 	int disable;
    240 {
    241 
    242 	/* Nothing. */
    243 }
    244 
    245 void
    246 if_nullwatchdog(ifp)
    247 	struct ifnet *ifp;
    248 {
    249 
    250 	/* Nothing. */
    251 }
    252 
    253 void
    254 if_nulldrain(ifp)
    255 	struct ifnet *ifp;
    256 {
    257 
    258 	/* Nothing. */
    259 }
    260 
    261 int if_index = 1;
    262 struct ifnet_head ifnet;
    263 struct ifaddr **ifnet_addrs = NULL;
    264 struct ifnet **ifindex2ifnet = NULL;
    265 
    266 /*
    267  * Allocate the link level name for the specified interface.  This
    268  * is an attachment helper.  It must be called after ifp->if_addrlen
    269  * is initialized, which may not be the case when if_attach() is
    270  * called.
    271  */
    272 void
    273 if_alloc_sadl(struct ifnet *ifp)
    274 {
    275 	unsigned socksize, ifasize;
    276 	int namelen, masklen;
    277 	struct sockaddr_dl *sdl;
    278 	struct ifaddr *ifa;
    279 
    280 	/*
    281 	 * If the interface already has a link name, release it
    282 	 * now.  This is useful for interfaces that can change
    283 	 * link types, and thus switch link names often.
    284 	 */
    285 	if (ifp->if_sadl != NULL)
    286 		if_free_sadl(ifp);
    287 
    288 	namelen = strlen(ifp->if_xname);
    289 	masklen = offsetof(struct sockaddr_dl, sdl_data[0]) + namelen;
    290 	socksize = masklen + ifp->if_addrlen;
    291 #define ROUNDUP(a) (1 + (((a) - 1) | (sizeof(long) - 1)))
    292 	if (socksize < sizeof(*sdl))
    293 		socksize = sizeof(*sdl);
    294 	socksize = ROUNDUP(socksize);
    295 	ifasize = sizeof(*ifa) + 2 * socksize;
    296 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK);
    297 	memset((caddr_t)ifa, 0, ifasize);
    298 	sdl = (struct sockaddr_dl *)(ifa + 1);
    299 	sdl->sdl_len = socksize;
    300 	sdl->sdl_family = AF_LINK;
    301 	bcopy(ifp->if_xname, sdl->sdl_data, namelen);
    302 	sdl->sdl_nlen = namelen;
    303 	sdl->sdl_alen = ifp->if_addrlen;
    304 	sdl->sdl_index = ifp->if_index;
    305 	sdl->sdl_type = ifp->if_type;
    306 	ifnet_addrs[ifp->if_index] = ifa;
    307 	IFAREF(ifa);
    308 	ifa->ifa_ifp = ifp;
    309 	ifa->ifa_rtrequest = link_rtrequest;
    310 	TAILQ_INSERT_HEAD(&ifp->if_addrlist, ifa, ifa_list);
    311 	IFAREF(ifa);
    312 	ifa->ifa_addr = (struct sockaddr *)sdl;
    313 	ifp->if_sadl = sdl;
    314 	sdl = (struct sockaddr_dl *)(socksize + (caddr_t)sdl);
    315 	ifa->ifa_netmask = (struct sockaddr *)sdl;
    316 	sdl->sdl_len = masklen;
    317 	while (namelen != 0)
    318 		sdl->sdl_data[--namelen] = 0xff;
    319 }
    320 
    321 /*
    322  * Free the link level name for the specified interface.  This is
    323  * a detach helper.  This is called from if_detach() or from
    324  * link layer type specific detach functions.
    325  */
    326 void
    327 if_free_sadl(struct ifnet *ifp)
    328 {
    329 	struct ifaddr *ifa;
    330 	int s;
    331 
    332 	ifa = ifnet_addrs[ifp->if_index];
    333 	if (ifa == NULL) {
    334 		KASSERT(ifp->if_sadl == NULL);
    335 		return;
    336 	}
    337 
    338 	KASSERT(ifp->if_sadl != NULL);
    339 
    340 	s = splnet();
    341 	rtinit(ifa, RTM_DELETE, 0);
    342 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
    343 	IFAFREE(ifa);
    344 
    345 	ifp->if_sadl = NULL;
    346 
    347 	ifnet_addrs[ifp->if_index] = NULL;
    348 	IFAFREE(ifa);
    349 	splx(s);
    350 }
    351 
    352 /*
    353  * Attach an interface to the
    354  * list of "active" interfaces.
    355  */
    356 void
    357 if_attach(ifp)
    358 	struct ifnet *ifp;
    359 {
    360 	static size_t if_indexlim = 0;
    361 	int indexlim = 0;
    362 
    363 	if (if_indexlim == 0) {
    364 		TAILQ_INIT(&ifnet);
    365 		if_indexlim = 8;
    366 	}
    367 	TAILQ_INIT(&ifp->if_addrlist);
    368 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
    369 	ifp->if_index = if_index;
    370 	if (ifindex2ifnet == 0)
    371 		if_index++;
    372 	else
    373 		while (ifindex2ifnet[ifp->if_index] != NULL) {
    374 			++if_index;
    375 			if (if_index == 0)
    376 				if_index = 1;
    377 			/*
    378 			 * If we hit USHRT_MAX, we skip back to 0 since
    379 			 * there are a number of places where the value
    380 			 * of if_index or if_index itself is compared
    381 			 * to to or stored in an unsigned short.  By
    382 			 * jumping back, we won't botch those assignments
    383 			 * or comparisons.
    384 			 */
    385 			else if (if_index == USHRT_MAX) {
    386 				/*
    387 				 * However, if we have to jump back to
    388 				 * zero *twice* without finding an empty
    389 				 * slot in ifindex2ifnet[], then there
    390 				 * there are too many (>65535) interfaces.
    391 				 */
    392 				if (indexlim++)
    393 					panic("too many interfaces");
    394 				else
    395 					if_index = 1;
    396 			}
    397 			ifp->if_index = if_index;
    398 		}
    399 
    400 	/*
    401 	 * We have some arrays that should be indexed by if_index.
    402 	 * since if_index will grow dynamically, they should grow too.
    403 	 *	struct ifadd **ifnet_addrs
    404 	 *	struct ifnet **ifindex2ifnet
    405 	 */
    406 	if (ifnet_addrs == 0 || ifindex2ifnet == 0 ||
    407 	    ifp->if_index >= if_indexlim) {
    408 		size_t n;
    409 		caddr_t q;
    410 
    411 		while (ifp->if_index >= if_indexlim)
    412 			if_indexlim <<= 1;
    413 
    414 		/* grow ifnet_addrs */
    415 		n = if_indexlim * sizeof(struct ifaddr *);
    416 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
    417 		memset(q, 0, n);
    418 		if (ifnet_addrs) {
    419 			bcopy((caddr_t)ifnet_addrs, q, n/2);
    420 			free((caddr_t)ifnet_addrs, M_IFADDR);
    421 		}
    422 		ifnet_addrs = (struct ifaddr **)q;
    423 
    424 		/* grow ifindex2ifnet */
    425 		n = if_indexlim * sizeof(struct ifnet *);
    426 		q = (caddr_t)malloc(n, M_IFADDR, M_WAITOK);
    427 		memset(q, 0, n);
    428 		if (ifindex2ifnet) {
    429 			bcopy((caddr_t)ifindex2ifnet, q, n/2);
    430 			free((caddr_t)ifindex2ifnet, M_IFADDR);
    431 		}
    432 		ifindex2ifnet = (struct ifnet **)q;
    433 	}
    434 
    435 	ifindex2ifnet[ifp->if_index] = ifp;
    436 
    437 	/*
    438 	 * Link level name is allocated later by a separate call to
    439 	 * if_alloc_sadl().
    440 	 */
    441 
    442 	if (ifp->if_snd.ifq_maxlen == 0)
    443 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
    444 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
    445 
    446 	ifp->if_link_state = LINK_STATE_UNKNOWN;
    447 
    448 	ifp->if_capenable = 0;
    449 	ifp->if_csum_flags_tx = 0;
    450 	ifp->if_csum_flags_rx = 0;
    451 
    452 #ifdef ALTQ
    453 	ifp->if_snd.altq_type = 0;
    454 	ifp->if_snd.altq_disc = NULL;
    455 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
    456 	ifp->if_snd.altq_tbr  = NULL;
    457 	ifp->if_snd.altq_ifp  = ifp;
    458 #endif
    459 
    460 #ifdef PFIL_HOOKS
    461 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
    462 	ifp->if_pfil.ph_ifnet = ifp;
    463 	if (pfil_head_register(&ifp->if_pfil) != 0)
    464 		printf("%s: WARNING: unable to register pfil hook\n",
    465 		    ifp->if_xname);
    466 #endif
    467 
    468 	/* Announce the interface. */
    469 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
    470 }
    471 
    472 /*
    473  * Deactivate an interface.  This points all of the procedure
    474  * handles at error stubs.  May be called from interrupt context.
    475  */
    476 void
    477 if_deactivate(ifp)
    478 	struct ifnet *ifp;
    479 {
    480 	int s;
    481 
    482 	s = splnet();
    483 
    484 	ifp->if_output	 = if_nulloutput;
    485 	ifp->if_input	 = if_nullinput;
    486 	ifp->if_start	 = if_nullstart;
    487 	ifp->if_ioctl	 = if_nullioctl;
    488 	ifp->if_init	 = if_nullinit;
    489 	ifp->if_stop	 = if_nullstop;
    490 	ifp->if_watchdog = if_nullwatchdog;
    491 	ifp->if_drain	 = if_nulldrain;
    492 
    493 	/* No more packets may be enqueued. */
    494 	ifp->if_snd.ifq_maxlen = 0;
    495 
    496 	splx(s);
    497 }
    498 
    499 /*
    500  * Detach an interface from the list of "active" interfaces,
    501  * freeing any resources as we go along.
    502  *
    503  * NOTE: This routine must be called with a valid thread context,
    504  * as it may block.
    505  */
    506 void
    507 if_detach(ifp)
    508 	struct ifnet *ifp;
    509 {
    510 	struct socket so;
    511 	struct ifaddr *ifa;
    512 #ifdef IFAREF_DEBUG
    513 	struct ifaddr *last_ifa = NULL;
    514 #endif
    515 	struct domain *dp;
    516 	struct protosw *pr;
    517 	struct radix_node_head *rnh;
    518 	int s, i, family, purged;
    519 
    520 	/*
    521 	 * XXX It's kind of lame that we have to have the
    522 	 * XXX socket structure...
    523 	 */
    524 	memset(&so, 0, sizeof(so));
    525 
    526 	s = splnet();
    527 
    528 	/*
    529 	 * Do an if_down() to give protocols a chance to do something.
    530 	 */
    531 	if_down(ifp);
    532 
    533 #ifdef ALTQ
    534 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
    535 		altq_disable(&ifp->if_snd);
    536 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
    537 		altq_detach(&ifp->if_snd);
    538 #endif
    539 
    540 #ifdef PFIL_HOOKS
    541 	(void) pfil_head_unregister(&ifp->if_pfil);
    542 #endif
    543 
    544 	if_free_sadl(ifp);
    545 
    546 	/*
    547 	 * Rip all the addresses off the interface.  This should make
    548 	 * all of the routes go away.
    549 	 */
    550 	while ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) != NULL) {
    551 		family = ifa->ifa_addr->sa_family;
    552 #ifdef IFAREF_DEBUG
    553 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
    554 		    ifa, family, ifa->ifa_refcnt);
    555 		if (last_ifa != NULL && ifa == last_ifa)
    556 			panic("if_detach: loop detected");
    557 		last_ifa = ifa;
    558 #endif
    559 		if (family == AF_LINK) {
    560 			/*
    561 			 * XXX This case may now be obsolete by
    562 			 * XXX the call to if_free_sadl().
    563 			 */
    564 			rtinit(ifa, RTM_DELETE, 0);
    565 			TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
    566 			IFAFREE(ifa);
    567 		} else {
    568 			dp = pffinddomain(family);
    569 #ifdef DIAGNOSTIC
    570 			if (dp == NULL)
    571 				panic("if_detach: no domain for AF %d\n",
    572 				    family);
    573 #endif
    574 			purged = 0;
    575 			for (pr = dp->dom_protosw;
    576 			     pr < dp->dom_protoswNPROTOSW; pr++) {
    577 				so.so_proto = pr;
    578 				if (pr->pr_usrreq != NULL) {
    579 					(void) (*pr->pr_usrreq)(&so,
    580 					    PRU_PURGEIF, NULL, NULL,
    581 					    (struct mbuf *) ifp, curproc);
    582 					purged = 1;
    583 				}
    584 			}
    585 			if (purged == 0) {
    586 				/*
    587 				 * XXX What's really the best thing to do
    588 				 * XXX here?  --thorpej (at) netbsd.org
    589 				 */
    590 				printf("if_detach: WARNING: AF %d not purged\n",
    591 				    family);
    592 			}
    593 		}
    594 	}
    595 
    596 	/* Walk the routing table looking for straglers. */
    597 	for (i = 0; i <= AF_MAX; i++) {
    598 		if ((rnh = rt_tables[i]) != NULL)
    599 			(void) (*rnh->rnh_walktree)(rnh, if_rt_walktree, ifp);
    600 	}
    601 
    602 	/* Announce that the interface is gone. */
    603 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
    604 
    605 	ifindex2ifnet[ifp->if_index] = NULL;
    606 
    607 	TAILQ_REMOVE(&ifnet, ifp, if_list);
    608 
    609 	/*
    610 	 * remove packets came from ifp, from software interrupt queues.
    611 	 * net/netisr_dispatch.h is not usable, as some of them use
    612 	 * strange queue names.
    613 	 */
    614 #define IF_DETACH_QUEUES(x) \
    615 do { \
    616 	extern struct ifqueue x; \
    617 	if_detach_queues(ifp, & x); \
    618 } while (0)
    619 #ifdef INET
    620 #if NARP > 0
    621 	IF_DETACH_QUEUES(arpintrq);
    622 #endif
    623 	IF_DETACH_QUEUES(ipintrq);
    624 #endif
    625 #ifdef INET6
    626 	IF_DETACH_QUEUES(ip6intrq);
    627 #endif
    628 #ifdef NETATALK
    629 	IF_DETACH_QUEUES(atintrq1);
    630 	IF_DETACH_QUEUES(atintrq2);
    631 #endif
    632 #ifdef NS
    633 	IF_DETACH_QUEUES(nsintrq);
    634 #endif
    635 #ifdef ISO
    636 	IF_DETACH_QUEUES(clnlintrq);
    637 #endif
    638 #ifdef CCITT
    639 	IF_DETACH_QUEUES(llcintrq);
    640 	IF_DETACH_QUEUES(hdintrq);
    641 #endif
    642 #ifdef NATM
    643 	IF_DETACH_QUEUES(natmintrq);
    644 #endif
    645 #undef IF_DETACH_QUEUES
    646 
    647 	splx(s);
    648 }
    649 
    650 #if defined(INET) || defined(INET6) || defined(NETATALK) || defined(NS) || \
    651     defined(ISO) || defined(CCITT) || defined(NATM)
    652 static void
    653 if_detach_queues(ifp, q)
    654 	struct ifnet *ifp;
    655 	struct ifqueue *q;
    656 {
    657 	struct mbuf *m, *prev, *next;
    658 
    659 	prev = NULL;
    660 	for (m = q->ifq_head; m; m = next) {
    661 		next = m->m_nextpkt;
    662 #ifdef DIAGNOSTIC
    663 		if ((m->m_flags & M_PKTHDR) == 0) {
    664 			prev = m;
    665 			continue;
    666 		}
    667 #endif
    668 		if (m->m_pkthdr.rcvif != ifp) {
    669 			prev = m;
    670 			continue;
    671 		}
    672 
    673 		if (prev)
    674 			prev->m_nextpkt = m->m_nextpkt;
    675 		else
    676 			q->ifq_head = m->m_nextpkt;
    677 		if (q->ifq_tail == m)
    678 			q->ifq_tail = prev;
    679 		q->ifq_len--;
    680 
    681 		m->m_nextpkt = NULL;
    682 		m_freem(m);
    683 		IF_DROP(q);
    684 	}
    685 }
    686 #endif /* defined(INET) || ... */
    687 
    688 /*
    689  * Callback for a radix tree walk to delete all references to an
    690  * ifnet.
    691  */
    692 int
    693 if_rt_walktree(rn, v)
    694 	struct radix_node *rn;
    695 	void *v;
    696 {
    697 	struct ifnet *ifp = (struct ifnet *)v;
    698 	struct rtentry *rt = (struct rtentry *)rn;
    699 	int error;
    700 
    701 	if (rt->rt_ifp == ifp) {
    702 		/* Delete the entry. */
    703 		error = rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
    704 		    rt_mask(rt), rt->rt_flags, NULL);
    705 		if (error)
    706 			printf("%s: warning: unable to delete rtentry @ %p, "
    707 			    "error = %d\n", ifp->if_xname, rt, error);
    708 	}
    709 	return (0);
    710 }
    711 
    712 /*
    713  * Create a clone network interface.
    714  */
    715 int
    716 if_clone_create(name)
    717 	const char *name;
    718 {
    719 	struct if_clone *ifc;
    720 	int unit;
    721 
    722 	ifc = if_clone_lookup(name, &unit);
    723 	if (ifc == NULL)
    724 		return (EINVAL);
    725 
    726 	if (ifunit(name) != NULL)
    727 		return (EEXIST);
    728 
    729 	return ((*ifc->ifc_create)(ifc, unit));
    730 }
    731 
    732 /*
    733  * Destroy a clone network interface.
    734  */
    735 int
    736 if_clone_destroy(name)
    737 	const char *name;
    738 {
    739 	struct if_clone *ifc;
    740 	struct ifnet *ifp;
    741 
    742 	ifc = if_clone_lookup(name, NULL);
    743 	if (ifc == NULL)
    744 		return (EINVAL);
    745 
    746 	ifp = ifunit(name);
    747 	if (ifp == NULL)
    748 		return (ENXIO);
    749 
    750 	if (ifc->ifc_destroy == NULL)
    751 		return (EOPNOTSUPP);
    752 
    753 	(*ifc->ifc_destroy)(ifp);
    754 	return (0);
    755 }
    756 
    757 /*
    758  * Look up a network interface cloner.
    759  */
    760 struct if_clone *
    761 if_clone_lookup(name, unitp)
    762 	const char *name;
    763 	int *unitp;
    764 {
    765 	struct if_clone *ifc;
    766 	const char *cp;
    767 	int i;
    768 
    769 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL;) {
    770 		for (cp = name, i = 0; i < ifc->ifc_namelen; i++, cp++) {
    771 			if (ifc->ifc_name[i] != *cp)
    772 				goto next_ifc;
    773 		}
    774 		goto found_name;
    775  next_ifc:
    776 		ifc = LIST_NEXT(ifc, ifc_list);
    777 	}
    778 
    779 	/* No match. */
    780 	return (NULL);
    781 
    782  found_name:
    783 	for (i = 0; *cp != '\0'; cp++) {
    784 		if (*cp < '0' || *cp > '9') {
    785 			/* Bogus unit number. */
    786 			return (NULL);
    787 		}
    788 		i = (i * 10) + (*cp - '0');
    789 	}
    790 
    791 	if (unitp != NULL)
    792 		*unitp = i;
    793 	return (ifc);
    794 }
    795 
    796 /*
    797  * Register a network interface cloner.
    798  */
    799 void
    800 if_clone_attach(ifc)
    801 	struct if_clone *ifc;
    802 {
    803 
    804 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
    805 	if_cloners_count++;
    806 }
    807 
    808 /*
    809  * Unregister a network interface cloner.
    810  */
    811 void
    812 if_clone_detach(ifc)
    813 	struct if_clone *ifc;
    814 {
    815 
    816 	LIST_REMOVE(ifc, ifc_list);
    817 	if_cloners_count--;
    818 }
    819 
    820 /*
    821  * Provide list of interface cloners to userspace.
    822  */
    823 int
    824 if_clone_list(ifcr)
    825 	struct if_clonereq *ifcr;
    826 {
    827 	char outbuf[IFNAMSIZ], *dst;
    828 	struct if_clone *ifc;
    829 	int count, error = 0;
    830 
    831 	ifcr->ifcr_total = if_cloners_count;
    832 	if ((dst = ifcr->ifcr_buffer) == NULL) {
    833 		/* Just asking how many there are. */
    834 		return (0);
    835 	}
    836 
    837 	if (ifcr->ifcr_count < 0)
    838 		return (EINVAL);
    839 
    840 	count = (if_cloners_count < ifcr->ifcr_count) ?
    841 	    if_cloners_count : ifcr->ifcr_count;
    842 
    843 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
    844 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
    845 		strncpy(outbuf, ifc->ifc_name, IFNAMSIZ);
    846 		outbuf[IFNAMSIZ - 1] = '\0';	/* sanity */
    847 		error = copyout(outbuf, dst, IFNAMSIZ);
    848 		if (error)
    849 			break;
    850 	}
    851 
    852 	return (error);
    853 }
    854 
    855 /*
    856  * Locate an interface based on a complete address.
    857  */
    858 /*ARGSUSED*/
    859 struct ifaddr *
    860 ifa_ifwithaddr(addr)
    861 	struct sockaddr *addr;
    862 {
    863 	struct ifnet *ifp;
    864 	struct ifaddr *ifa;
    865 
    866 #define	equal(a1, a2) \
    867   (bcmp((caddr_t)(a1), (caddr_t)(a2), ((struct sockaddr *)(a1))->sa_len) == 0)
    868 
    869 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
    870 	     ifp = TAILQ_NEXT(ifp, if_list)) {
    871 		if (ifp->if_output == if_nulloutput)
    872 			continue;
    873 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
    874 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
    875 			if (ifa->ifa_addr->sa_family != addr->sa_family)
    876 				continue;
    877 			if (equal(addr, ifa->ifa_addr))
    878 				return (ifa);
    879 			if ((ifp->if_flags & IFF_BROADCAST) &&
    880 			    ifa->ifa_broadaddr &&
    881 			    /* IP6 doesn't have broadcast */
    882 			    ifa->ifa_broadaddr->sa_len != 0 &&
    883 			    equal(ifa->ifa_broadaddr, addr))
    884 				return (ifa);
    885 		}
    886 	}
    887 	return (NULL);
    888 }
    889 
    890 /*
    891  * Locate the point to point interface with a given destination address.
    892  */
    893 /*ARGSUSED*/
    894 struct ifaddr *
    895 ifa_ifwithdstaddr(addr)
    896 	struct sockaddr *addr;
    897 {
    898 	struct ifnet *ifp;
    899 	struct ifaddr *ifa;
    900 
    901 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
    902 	     ifp = TAILQ_NEXT(ifp, if_list)) {
    903 		if (ifp->if_output == if_nulloutput)
    904 			continue;
    905 		if (ifp->if_flags & IFF_POINTOPOINT) {
    906 			for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
    907 			     ifa = TAILQ_NEXT(ifa, ifa_list)) {
    908 				if (ifa->ifa_addr->sa_family !=
    909 				      addr->sa_family ||
    910 				    ifa->ifa_dstaddr == NULL)
    911 					continue;
    912 				if (equal(addr, ifa->ifa_dstaddr))
    913 					return (ifa);
    914 			}
    915 		}
    916 	}
    917 	return (NULL);
    918 }
    919 
    920 /*
    921  * Find an interface on a specific network.  If many, choice
    922  * is most specific found.
    923  */
    924 struct ifaddr *
    925 ifa_ifwithnet(addr)
    926 	struct sockaddr *addr;
    927 {
    928 	struct ifnet *ifp;
    929 	struct ifaddr *ifa;
    930 	struct sockaddr_dl *sdl;
    931 	struct ifaddr *ifa_maybe = 0;
    932 	u_int af = addr->sa_family;
    933 	char *addr_data = addr->sa_data, *cplim;
    934 
    935 	if (af == AF_LINK) {
    936 		sdl = (struct sockaddr_dl *)addr;
    937 		if (sdl->sdl_index && sdl->sdl_index <= if_index &&
    938 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
    939 			return (ifnet_addrs[sdl->sdl_index]);
    940 	}
    941 #ifdef NETATALK
    942 	if (af == AF_APPLETALK) {
    943 		struct sockaddr_at *sat, *sat2;
    944 		sat = (struct sockaddr_at *)addr;
    945 		for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
    946 		     ifp = TAILQ_NEXT(ifp, if_list)) {
    947 			if (ifp->if_output == if_nulloutput)
    948 				continue;
    949 			ifa = at_ifawithnet((struct sockaddr_at *)addr, ifp);
    950 			if (ifa == NULL)
    951 				continue;
    952 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
    953 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
    954 				return (ifa); /* exact match */
    955 			if (ifa_maybe == NULL) {
    956 				/* else keep the if with the rigth range */
    957 				ifa_maybe = ifa;
    958 			}
    959 		}
    960 		return (ifa_maybe);
    961 	}
    962 #endif
    963 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
    964 	     ifp = TAILQ_NEXT(ifp, if_list)) {
    965 		if (ifp->if_output == if_nulloutput)
    966 			continue;
    967 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
    968 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
    969 			char *cp, *cp2, *cp3;
    970 
    971 			if (ifa->ifa_addr->sa_family != af ||
    972 			    ifa->ifa_netmask == 0)
    973  next:				continue;
    974 			cp = addr_data;
    975 			cp2 = ifa->ifa_addr->sa_data;
    976 			cp3 = ifa->ifa_netmask->sa_data;
    977 			cplim = (char *)ifa->ifa_netmask +
    978 			    ifa->ifa_netmask->sa_len;
    979 			while (cp3 < cplim) {
    980 				if ((*cp++ ^ *cp2++) & *cp3++) {
    981 					/* want to continue for() loop */
    982 					goto next;
    983 				}
    984 			}
    985 			if (ifa_maybe == 0 ||
    986 			    rn_refines((caddr_t)ifa->ifa_netmask,
    987 			    (caddr_t)ifa_maybe->ifa_netmask))
    988 				ifa_maybe = ifa;
    989 		}
    990 	}
    991 	return (ifa_maybe);
    992 }
    993 
    994 /*
    995  * Find the interface of the addresss.
    996  */
    997 struct ifaddr *
    998 ifa_ifwithladdr(addr)
    999 	struct sockaddr *addr;
   1000 {
   1001 	struct ifaddr *ia;
   1002 
   1003 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
   1004 	    (ia = ifa_ifwithnet(addr)))
   1005 		return (ia);
   1006 	return (NULL);
   1007 }
   1008 
   1009 /*
   1010  * Find an interface using a specific address family
   1011  */
   1012 struct ifaddr *
   1013 ifa_ifwithaf(af)
   1014 	int af;
   1015 {
   1016 	struct ifnet *ifp;
   1017 	struct ifaddr *ifa;
   1018 
   1019 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
   1020 	     ifp = TAILQ_NEXT(ifp, if_list)) {
   1021 		if (ifp->if_output == if_nulloutput)
   1022 			continue;
   1023 		for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
   1024 		     ifa = TAILQ_NEXT(ifa, ifa_list)) {
   1025 			if (ifa->ifa_addr->sa_family == af)
   1026 				return (ifa);
   1027 		}
   1028 	}
   1029 	return (NULL);
   1030 }
   1031 
   1032 /*
   1033  * Find an interface address specific to an interface best matching
   1034  * a given address.
   1035  */
   1036 struct ifaddr *
   1037 ifaof_ifpforaddr(addr, ifp)
   1038 	struct sockaddr *addr;
   1039 	struct ifnet *ifp;
   1040 {
   1041 	struct ifaddr *ifa;
   1042 	char *cp, *cp2, *cp3;
   1043 	char *cplim;
   1044 	struct ifaddr *ifa_maybe = 0;
   1045 	u_int af = addr->sa_family;
   1046 
   1047 	if (ifp->if_output == if_nulloutput)
   1048 		return (NULL);
   1049 
   1050 	if (af >= AF_MAX)
   1051 		return (NULL);
   1052 
   1053 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
   1054 	     ifa = TAILQ_NEXT(ifa, ifa_list)) {
   1055 		if (ifa->ifa_addr->sa_family != af)
   1056 			continue;
   1057 		ifa_maybe = ifa;
   1058 		if (ifa->ifa_netmask == 0) {
   1059 			if (equal(addr, ifa->ifa_addr) ||
   1060 			    (ifa->ifa_dstaddr &&
   1061 			     equal(addr, ifa->ifa_dstaddr)))
   1062 				return (ifa);
   1063 			continue;
   1064 		}
   1065 		cp = addr->sa_data;
   1066 		cp2 = ifa->ifa_addr->sa_data;
   1067 		cp3 = ifa->ifa_netmask->sa_data;
   1068 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
   1069 		for (; cp3 < cplim; cp3++) {
   1070 			if ((*cp++ ^ *cp2++) & *cp3)
   1071 				break;
   1072 		}
   1073 		if (cp3 == cplim)
   1074 			return (ifa);
   1075 	}
   1076 	return (ifa_maybe);
   1077 }
   1078 
   1079 /*
   1080  * Default action when installing a route with a Link Level gateway.
   1081  * Lookup an appropriate real ifa to point to.
   1082  * This should be moved to /sys/net/link.c eventually.
   1083  */
   1084 void
   1085 link_rtrequest(cmd, rt, info)
   1086 	int cmd;
   1087 	struct rtentry *rt;
   1088 	struct rt_addrinfo *info;
   1089 {
   1090 	struct ifaddr *ifa;
   1091 	struct sockaddr *dst;
   1092 	struct ifnet *ifp;
   1093 
   1094 	if (cmd != RTM_ADD || ((ifa = rt->rt_ifa) == 0) ||
   1095 	    ((ifp = ifa->ifa_ifp) == 0) || ((dst = rt_key(rt)) == 0))
   1096 		return;
   1097 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
   1098 		IFAFREE(rt->rt_ifa);
   1099 		rt->rt_ifa = ifa;
   1100 		IFAREF(ifa);
   1101 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
   1102 			ifa->ifa_rtrequest(cmd, rt, info);
   1103 	}
   1104 }
   1105 
   1106 /*
   1107  * Mark an interface down and notify protocols of
   1108  * the transition.
   1109  * NOTE: must be called at splsoftnet or equivalent.
   1110  */
   1111 void
   1112 if_down(ifp)
   1113 	struct ifnet *ifp;
   1114 {
   1115 	struct ifaddr *ifa;
   1116 
   1117 	ifp->if_flags &= ~IFF_UP;
   1118 	microtime(&ifp->if_lastchange);
   1119 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
   1120 	     ifa = TAILQ_NEXT(ifa, ifa_list))
   1121 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
   1122 	IFQ_PURGE(&ifp->if_snd);
   1123 	rt_ifmsg(ifp);
   1124 }
   1125 
   1126 /*
   1127  * Mark an interface up and notify protocols of
   1128  * the transition.
   1129  * NOTE: must be called at splsoftnet or equivalent.
   1130  */
   1131 void
   1132 if_up(ifp)
   1133 	struct ifnet *ifp;
   1134 {
   1135 #ifdef notyet
   1136 	struct ifaddr *ifa;
   1137 #endif
   1138 
   1139 	ifp->if_flags |= IFF_UP;
   1140 	microtime(&ifp->if_lastchange);
   1141 #ifdef notyet
   1142 	/* this has no effect on IP, and will kill all ISO connections XXX */
   1143 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL;
   1144 	     ifa = TAILQ_NEXT(ifa, ifa_list))
   1145 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
   1146 #endif
   1147 	rt_ifmsg(ifp);
   1148 #ifdef INET6
   1149 	in6_if_up(ifp);
   1150 #endif
   1151 }
   1152 
   1153 /*
   1154  * Handle interface watchdog timer routines.  Called
   1155  * from softclock, we decrement timers (if set) and
   1156  * call the appropriate interface routine on expiration.
   1157  */
   1158 void
   1159 if_slowtimo(arg)
   1160 	void *arg;
   1161 {
   1162 	struct ifnet *ifp;
   1163 	int s = splnet();
   1164 
   1165 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
   1166 	     ifp = TAILQ_NEXT(ifp, if_list)) {
   1167 		if (ifp->if_timer == 0 || --ifp->if_timer)
   1168 			continue;
   1169 		if (ifp->if_watchdog)
   1170 			(*ifp->if_watchdog)(ifp);
   1171 	}
   1172 	splx(s);
   1173 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ,
   1174 	    if_slowtimo, NULL);
   1175 }
   1176 
   1177 /*
   1178  * Set/clear promiscuous mode on interface ifp based on the truth value
   1179  * of pswitch.  The calls are reference counted so that only the first
   1180  * "on" request actually has an effect, as does the final "off" request.
   1181  * Results are undefined if the "off" and "on" requests are not matched.
   1182  */
   1183 int
   1184 ifpromisc(ifp, pswitch)
   1185 	struct ifnet *ifp;
   1186 	int pswitch;
   1187 {
   1188 	int pcount, ret;
   1189 	short flags;
   1190 	struct ifreq ifr;
   1191 
   1192 	pcount = ifp->if_pcount;
   1193 	flags = ifp->if_flags;
   1194 	if (pswitch) {
   1195 		/*
   1196 		 * Allow the device to be "placed" into promiscuous
   1197 		 * mode even if it is not configured up.  It will
   1198 		 * consult IFF_PROMISC when it is is brought up.
   1199 		 */
   1200 		if (ifp->if_pcount++ != 0)
   1201 			return (0);
   1202 		ifp->if_flags |= IFF_PROMISC;
   1203 		if ((ifp->if_flags & IFF_UP) == 0)
   1204 			return (0);
   1205 	} else {
   1206 		if (--ifp->if_pcount > 0)
   1207 			return (0);
   1208 		ifp->if_flags &= ~IFF_PROMISC;
   1209 		/*
   1210 		 * If the device is not configured up, we should not need to
   1211 		 * turn off promiscuous mode (device should have turned it
   1212 		 * off when interface went down; and will look at IFF_PROMISC
   1213 		 * again next time interface comes up).
   1214 		 */
   1215 		if ((ifp->if_flags & IFF_UP) == 0)
   1216 			return (0);
   1217 	}
   1218 	memset(&ifr, 0, sizeof(ifr));
   1219 	ifr.ifr_flags = ifp->if_flags;
   1220 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t) &ifr);
   1221 	/* Restore interface state if not successful. */
   1222 	if (ret != 0) {
   1223 		ifp->if_pcount = pcount;
   1224 		ifp->if_flags = flags;
   1225 	}
   1226 	return (ret);
   1227 }
   1228 
   1229 /*
   1230  * Map interface name to
   1231  * interface structure pointer.
   1232  */
   1233 struct ifnet *
   1234 ifunit(name)
   1235 	const char *name;
   1236 {
   1237 	struct ifnet *ifp;
   1238 
   1239 	for (ifp = TAILQ_FIRST(&ifnet); ifp != NULL;
   1240 	     ifp = TAILQ_NEXT(ifp, if_list)) {
   1241 		if (ifp->if_output == if_nulloutput)
   1242 			continue;
   1243 	 	if (strcmp(ifp->if_xname, name) == 0)
   1244 			return (ifp);
   1245 	}
   1246 	return (NULL);
   1247 }
   1248 
   1249 /*
   1250  * Interface ioctls.
   1251  */
   1252 int
   1253 ifioctl(so, cmd, data, p)
   1254 	struct socket *so;
   1255 	u_long cmd;
   1256 	caddr_t data;
   1257 	struct proc *p;
   1258 {
   1259 	struct ifnet *ifp;
   1260 	struct ifreq *ifr;
   1261 	struct ifcapreq *ifcr;
   1262 	int s, error = 0;
   1263 	short oif_flags;
   1264 
   1265 	switch (cmd) {
   1266 
   1267 	case SIOCGIFCONF:
   1268 	case OSIOCGIFCONF:
   1269 		return (ifconf(cmd, data));
   1270 	}
   1271 	ifr = (struct ifreq *)data;
   1272 	ifcr = (struct ifcapreq *)data;
   1273 
   1274 	switch (cmd) {
   1275 	case SIOCIFCREATE:
   1276 	case SIOCIFDESTROY:
   1277 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1278 			return (error);
   1279 		return ((cmd == SIOCIFCREATE) ?
   1280 			if_clone_create(ifr->ifr_name) :
   1281 			if_clone_destroy(ifr->ifr_name));
   1282 
   1283 	case SIOCIFGCLONERS:
   1284 		return (if_clone_list((struct if_clonereq *)data));
   1285 	}
   1286 
   1287 	ifp = ifunit(ifr->ifr_name);
   1288 	if (ifp == 0)
   1289 		return (ENXIO);
   1290 	oif_flags = ifp->if_flags;
   1291 	switch (cmd) {
   1292 
   1293 	case SIOCGIFFLAGS:
   1294 		ifr->ifr_flags = ifp->if_flags;
   1295 		break;
   1296 
   1297 	case SIOCGIFMETRIC:
   1298 		ifr->ifr_metric = ifp->if_metric;
   1299 		break;
   1300 
   1301 	case SIOCGIFMTU:
   1302 		ifr->ifr_mtu = ifp->if_mtu;
   1303 		break;
   1304 
   1305 	case SIOCGIFDLT:
   1306 		ifr->ifr_dlt = ifp->if_dlt;
   1307 		break;
   1308 
   1309 	case SIOCSIFFLAGS:
   1310 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1311 			return (error);
   1312 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
   1313 			s = splnet();
   1314 			if_down(ifp);
   1315 			splx(s);
   1316 		}
   1317 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
   1318 			s = splnet();
   1319 			if_up(ifp);
   1320 			splx(s);
   1321 		}
   1322 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
   1323 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
   1324 		if (ifp->if_ioctl)
   1325 			(void) (*ifp->if_ioctl)(ifp, cmd, data);
   1326 		break;
   1327 
   1328 	case SIOCGIFCAP:
   1329 		ifcr->ifcr_capabilities = ifp->if_capabilities;
   1330 		ifcr->ifcr_capenable = ifp->if_capenable;
   1331 		break;
   1332 
   1333 	case SIOCSIFCAP:
   1334 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1335 			return (error);
   1336 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
   1337 			return (EINVAL);
   1338 		if (ifp->if_ioctl == NULL)
   1339 			return (EOPNOTSUPP);
   1340 
   1341 		/* Must prevent race with packet reception here. */
   1342 		s = splnet();
   1343 		if (ifcr->ifcr_capenable != ifp->if_capenable) {
   1344 			struct ifreq ifrq;
   1345 
   1346 			ifrq.ifr_flags = ifp->if_flags;
   1347 			ifp->if_capenable = ifcr->ifcr_capenable;
   1348 
   1349 			/* Pre-compute the checksum flags mask. */
   1350 			ifp->if_csum_flags_tx = 0;
   1351 			ifp->if_csum_flags_rx = 0;
   1352 			if (ifp->if_capenable & IFCAP_CSUM_IPv4) {
   1353 				ifp->if_csum_flags_tx |= M_CSUM_IPv4;
   1354 				ifp->if_csum_flags_rx |= M_CSUM_IPv4;
   1355 			}
   1356 
   1357 			if (ifp->if_capenable & IFCAP_CSUM_TCPv4) {
   1358 				ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
   1359 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
   1360 			} else if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx)
   1361 				ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
   1362 
   1363 			if (ifp->if_capenable & IFCAP_CSUM_UDPv4) {
   1364 				ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
   1365 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
   1366 			} else if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx)
   1367 				ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
   1368 
   1369 			if (ifp->if_capenable & IFCAP_CSUM_TCPv6) {
   1370 				ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
   1371 				ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
   1372 			}
   1373 
   1374 			if (ifp->if_capenable & IFCAP_CSUM_UDPv6) {
   1375 				ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
   1376 				ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
   1377 			}
   1378 
   1379 			/*
   1380 			 * Only kick the interface if it's up.  If it's
   1381 			 * not up now, it will notice the cap enables
   1382 			 * when it is brought up later.
   1383 			 */
   1384 			if (ifp->if_flags & IFF_UP)
   1385 				(void) (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS,
   1386 				    (caddr_t) &ifrq);
   1387 		}
   1388 		splx(s);
   1389 		break;
   1390 
   1391 	case SIOCSIFMETRIC:
   1392 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1393 			return (error);
   1394 		ifp->if_metric = ifr->ifr_metric;
   1395 		break;
   1396 
   1397 	case SIOCSIFMTU:
   1398 	{
   1399 		u_long oldmtu = ifp->if_mtu;
   1400 
   1401 		error = suser(p->p_ucred, &p->p_acflag);
   1402 		if (error)
   1403 			return (error);
   1404 		if (ifp->if_ioctl == NULL)
   1405 			return (EOPNOTSUPP);
   1406 		error = (*ifp->if_ioctl)(ifp, cmd, data);
   1407 
   1408 		/*
   1409 		 * If the link MTU changed, do network layer specific procedure.
   1410 		 */
   1411 		if (ifp->if_mtu != oldmtu) {
   1412 #ifdef INET6
   1413 			nd6_setmtu(ifp);
   1414 #endif
   1415 		}
   1416 		break;
   1417 	}
   1418 	case SIOCSIFPHYADDR:
   1419 	case SIOCDIFPHYADDR:
   1420 #ifdef INET6
   1421 	case SIOCSIFPHYADDR_IN6:
   1422 #endif
   1423 	case SIOCSLIFPHYADDR:
   1424 	case SIOCADDMULTI:
   1425 	case SIOCDELMULTI:
   1426 	case SIOCSIFMEDIA:
   1427 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1428 			return (error);
   1429 		/* FALLTHROUGH */
   1430 	case SIOCGIFPSRCADDR:
   1431 	case SIOCGIFPDSTADDR:
   1432 	case SIOCGLIFPHYADDR:
   1433 	case SIOCGIFMEDIA:
   1434 		if (ifp->if_ioctl == 0)
   1435 			return (EOPNOTSUPP);
   1436 		error = (*ifp->if_ioctl)(ifp, cmd, data);
   1437 		break;
   1438 
   1439 	case SIOCSDRVSPEC:
   1440 	case SIOCS80211NWID:
   1441 	case SIOCS80211NWKEY:
   1442 	case SIOCS80211POWER:
   1443 		/* XXX:  need to pass proc pointer through to driver... */
   1444 		if ((error = suser(p->p_ucred, &p->p_acflag)) != 0)
   1445 			return (error);
   1446 	/* FALLTHROUGH */
   1447 	default:
   1448 		if (so->so_proto == 0)
   1449 			return (EOPNOTSUPP);
   1450 #if !defined(COMPAT_43) && !defined(COMPAT_LINUX) && !defined(COMPAT_SVR4) && !defined(LKM)
   1451 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
   1452 		    (struct mbuf *)cmd, (struct mbuf *)data,
   1453 		    (struct mbuf *)ifp, p));
   1454 #else
   1455 	    {
   1456 		int ocmd = cmd;
   1457 
   1458 		switch (cmd) {
   1459 
   1460 		case SIOCSIFADDR:
   1461 		case SIOCSIFDSTADDR:
   1462 		case SIOCSIFBRDADDR:
   1463 		case SIOCSIFNETMASK:
   1464 #if BYTE_ORDER != BIG_ENDIAN
   1465 			if (ifr->ifr_addr.sa_family == 0 &&
   1466 			    ifr->ifr_addr.sa_len < 16) {
   1467 				ifr->ifr_addr.sa_family = ifr->ifr_addr.sa_len;
   1468 				ifr->ifr_addr.sa_len = 16;
   1469 			}
   1470 #else
   1471 			if (ifr->ifr_addr.sa_len == 0)
   1472 				ifr->ifr_addr.sa_len = 16;
   1473 #endif
   1474 			break;
   1475 
   1476 		case OSIOCGIFADDR:
   1477 			cmd = SIOCGIFADDR;
   1478 			break;
   1479 
   1480 		case OSIOCGIFDSTADDR:
   1481 			cmd = SIOCGIFDSTADDR;
   1482 			break;
   1483 
   1484 		case OSIOCGIFBRDADDR:
   1485 			cmd = SIOCGIFBRDADDR;
   1486 			break;
   1487 
   1488 		case OSIOCGIFNETMASK:
   1489 			cmd = SIOCGIFNETMASK;
   1490 		}
   1491 
   1492 		error = ((*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
   1493 		    (struct mbuf *)cmd, (struct mbuf *)data,
   1494 		    (struct mbuf *)ifp, p));
   1495 
   1496 		switch (ocmd) {
   1497 		case OSIOCGIFADDR:
   1498 		case OSIOCGIFDSTADDR:
   1499 		case OSIOCGIFBRDADDR:
   1500 		case OSIOCGIFNETMASK:
   1501 			*(u_int16_t *)&ifr->ifr_addr = ifr->ifr_addr.sa_family;
   1502 		}
   1503 	    }
   1504 #endif /* COMPAT_43 */
   1505 		break;
   1506 	}
   1507 
   1508 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
   1509 #ifdef INET6
   1510 		if ((ifp->if_flags & IFF_UP) != 0) {
   1511 			s = splnet();
   1512 			in6_if_up(ifp);
   1513 			splx(s);
   1514 		}
   1515 #endif
   1516 	}
   1517 
   1518 	return (error);
   1519 }
   1520 
   1521 /*
   1522  * Return interface configuration
   1523  * of system.  List may be used
   1524  * in later ioctl's (above) to get
   1525  * other information.
   1526  */
   1527 /*ARGSUSED*/
   1528 int
   1529 ifconf(cmd, data)
   1530 	u_long cmd;
   1531 	caddr_t data;
   1532 {
   1533 	struct ifconf *ifc = (struct ifconf *)data;
   1534 	struct ifnet *ifp;
   1535 	struct ifaddr *ifa;
   1536 	struct ifreq ifr, *ifrp;
   1537 	int space = ifc->ifc_len, error = 0;
   1538 
   1539 	ifrp = ifc->ifc_req;
   1540 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
   1541 		bcopy(ifp->if_xname, ifr.ifr_name, IFNAMSIZ);
   1542 		if ((ifa = TAILQ_FIRST(&ifp->if_addrlist)) == 0) {
   1543 			memset((caddr_t)&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
   1544 			if (space >= (int)sizeof (ifr)) {
   1545 				error = copyout((caddr_t)&ifr, (caddr_t)ifrp,
   1546 						sizeof(ifr));
   1547 				if (error)
   1548 					break;
   1549 			}
   1550 			space -= sizeof (ifr), ifrp++;
   1551 		} else
   1552 		    for (; ifa != 0; ifa = TAILQ_NEXT(ifa, ifa_list)) {
   1553 			struct sockaddr *sa = ifa->ifa_addr;
   1554 #if defined(COMPAT_43) || defined(COMPAT_LINUX) || defined(COMPAT_SVR4)
   1555 			if (cmd == OSIOCGIFCONF) {
   1556 				struct osockaddr *osa =
   1557 					 (struct osockaddr *)&ifr.ifr_addr;
   1558 				ifr.ifr_addr = *sa;
   1559 				osa->sa_family = sa->sa_family;
   1560 				if (space >= (int)sizeof (ifr)) {
   1561 					error = copyout((caddr_t)&ifr,
   1562 							(caddr_t)ifrp,
   1563 							sizeof (ifr));
   1564 					ifrp++;
   1565 				}
   1566 			} else
   1567 #endif
   1568 			if (sa->sa_len <= sizeof(*sa)) {
   1569 				ifr.ifr_addr = *sa;
   1570 				if (space >= (int)sizeof (ifr)) {
   1571 					error = copyout((caddr_t)&ifr,
   1572 							(caddr_t)ifrp,
   1573 							sizeof (ifr));
   1574 					ifrp++;
   1575 				}
   1576 			} else {
   1577 				space -= sa->sa_len - sizeof(*sa);
   1578 				if (space >= (int)sizeof (ifr)) {
   1579 					error = copyout((caddr_t)&ifr,
   1580 							(caddr_t)ifrp,
   1581 							sizeof (ifr.ifr_name));
   1582 					if (error == 0) {
   1583 						error = copyout((caddr_t)sa,
   1584 						  (caddr_t)&ifrp->ifr_addr,
   1585 						  sa->sa_len);
   1586 					}
   1587 					ifrp = (struct ifreq *)
   1588 						(sa->sa_len +
   1589 						 (caddr_t)&ifrp->ifr_addr);
   1590 				}
   1591 			}
   1592 			if (error)
   1593 				break;
   1594 			space -= sizeof (ifr);
   1595 		}
   1596 	}
   1597 	ifc->ifc_len -= space;
   1598 	return (error);
   1599 }
   1600