Home | History | Annotate | Line # | Download | only in net
if.c revision 1.241
      1 /*	$NetBSD: if.c,v 1.241 2009/11/13 23:11:08 joerg Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1999, 2000, 2001, 2008 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by William Studenmund 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  *
     19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     29  * POSSIBILITY OF SUCH DAMAGE.
     30  */
     31 
     32 /*
     33  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
     34  * All rights reserved.
     35  *
     36  * Redistribution and use in source and binary forms, with or without
     37  * modification, are permitted provided that the following conditions
     38  * are met:
     39  * 1. Redistributions of source code must retain the above copyright
     40  *    notice, this list of conditions and the following disclaimer.
     41  * 2. Redistributions in binary form must reproduce the above copyright
     42  *    notice, this list of conditions and the following disclaimer in the
     43  *    documentation and/or other materials provided with the distribution.
     44  * 3. Neither the name of the project nor the names of its contributors
     45  *    may be used to endorse or promote products derived from this software
     46  *    without specific prior written permission.
     47  *
     48  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     58  * SUCH DAMAGE.
     59  */
     60 
     61 /*
     62  * Copyright (c) 1980, 1986, 1993
     63  *	The Regents of the University of California.  All rights reserved.
     64  *
     65  * Redistribution and use in source and binary forms, with or without
     66  * modification, are permitted provided that the following conditions
     67  * are met:
     68  * 1. Redistributions of source code must retain the above copyright
     69  *    notice, this list of conditions and the following disclaimer.
     70  * 2. Redistributions in binary form must reproduce the above copyright
     71  *    notice, this list of conditions and the following disclaimer in the
     72  *    documentation and/or other materials provided with the distribution.
     73  * 3. Neither the name of the University nor the names of its contributors
     74  *    may be used to endorse or promote products derived from this software
     75  *    without specific prior written permission.
     76  *
     77  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     78  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     79  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     80  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     81  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     82  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     83  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     84  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     85  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     86  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     87  * SUCH DAMAGE.
     88  *
     89  *	@(#)if.c	8.5 (Berkeley) 1/9/95
     90  */
     91 
     92 #include <sys/cdefs.h>
     93 __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.241 2009/11/13 23:11:08 joerg Exp $");
     94 
     95 #include "opt_inet.h"
     96 
     97 #include "opt_atalk.h"
     98 #include "opt_natm.h"
     99 #include "opt_pfil_hooks.h"
    100 
    101 #include <sys/param.h>
    102 #include <sys/mbuf.h>
    103 #include <sys/systm.h>
    104 #include <sys/callout.h>
    105 #include <sys/proc.h>
    106 #include <sys/socket.h>
    107 #include <sys/socketvar.h>
    108 #include <sys/domain.h>
    109 #include <sys/protosw.h>
    110 #include <sys/kernel.h>
    111 #include <sys/ioctl.h>
    112 #include <sys/sysctl.h>
    113 #include <sys/syslog.h>
    114 #include <sys/kauth.h>
    115 
    116 #include <net/if.h>
    117 #include <net/if_dl.h>
    118 #include <net/if_ether.h>
    119 #include <net/if_media.h>
    120 #include <net80211/ieee80211.h>
    121 #include <net80211/ieee80211_ioctl.h>
    122 #include <net/if_types.h>
    123 #include <net/radix.h>
    124 #include <net/route.h>
    125 #include <net/netisr.h>
    126 #ifdef NETATALK
    127 #include <netatalk/at_extern.h>
    128 #include <netatalk/at.h>
    129 #endif
    130 #include <net/pfil.h>
    131 
    132 #ifdef INET6
    133 #include <netinet/in.h>
    134 #include <netinet6/in6_var.h>
    135 #include <netinet6/nd6.h>
    136 #endif
    137 
    138 #include "carp.h"
    139 #if NCARP > 0
    140 #include <netinet/ip_carp.h>
    141 #endif
    142 
    143 #include <compat/sys/sockio.h>
    144 #include <compat/sys/socket.h>
    145 
    146 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
    147 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
    148 
    149 int	ifqmaxlen = IFQ_MAXLEN;
    150 callout_t if_slowtimo_ch;
    151 
    152 int netisr;			/* scheduling bits for network */
    153 
    154 static int	if_rt_walktree(struct rtentry *, void *);
    155 
    156 static struct if_clone *if_clone_lookup(const char *, int *);
    157 static int	if_clone_list(struct if_clonereq *);
    158 
    159 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
    160 static int if_cloners_count;
    161 
    162 static uint64_t index_gen;
    163 static kmutex_t index_gen_mtx;
    164 
    165 #ifdef PFIL_HOOKS
    166 struct pfil_head if_pfil;	/* packet filtering hook for interfaces */
    167 #endif
    168 
    169 static kauth_listener_t if_listener;
    170 
    171 static void if_detach_queues(struct ifnet *, struct ifqueue *);
    172 static void sysctl_sndq_setup(struct sysctllog **, const char *,
    173     struct ifaltq *);
    174 
    175 #if defined(INET) || defined(INET6)
    176 static void sysctl_net_ifq_setup(struct sysctllog **, int, const char *,
    177 				 int, const char *, int, struct ifqueue *);
    178 #endif
    179 
    180 static int
    181 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
    182     void *arg0, void *arg1, void *arg2, void *arg3)
    183 {
    184 	int result;
    185 	enum kauth_network_req req;
    186 
    187 	result = KAUTH_RESULT_DEFER;
    188 	req = (enum kauth_network_req)arg1;
    189 
    190 	if (action != KAUTH_NETWORK_INTERFACE)
    191 		return result;
    192 
    193 	if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
    194 	    (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
    195 		result = KAUTH_RESULT_ALLOW;
    196 
    197 	return result;
    198 }
    199 
    200 /*
    201  * Network interface utility routines.
    202  *
    203  * Routines with ifa_ifwith* names take sockaddr *'s as
    204  * parameters.
    205  */
    206 void
    207 ifinit(void)
    208 {
    209 #ifdef INET
    210 	{extern struct ifqueue ipintrq;
    211 	sysctl_net_ifq_setup(NULL, PF_INET, "inet", IPPROTO_IP, "ip",
    212 			     IPCTL_IFQ, &ipintrq);}
    213 #endif /* INET */
    214 #ifdef INET6
    215 	{extern struct ifqueue ip6intrq;
    216 	sysctl_net_ifq_setup(NULL, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
    217 			     IPV6CTL_IFQ, &ip6intrq);}
    218 #endif /* INET6 */
    219 
    220 	callout_init(&if_slowtimo_ch, 0);
    221 	if_slowtimo(NULL);
    222 
    223 	if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
    224 	    if_listener_cb, NULL);
    225 }
    226 
    227 /*
    228  * XXX Initialization before configure().
    229  * XXX hack to get pfil_add_hook working in autoconf.
    230  */
    231 void
    232 ifinit1(void)
    233 {
    234 
    235 	mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE);
    236 
    237 #ifdef PFIL_HOOKS
    238 	if_pfil.ph_type = PFIL_TYPE_IFNET;
    239 	if_pfil.ph_ifnet = NULL;
    240 	if (pfil_head_register(&if_pfil) != 0)
    241 		printf("WARNING: unable to register pfil hook\n");
    242 #endif
    243 }
    244 
    245 struct ifnet *
    246 if_alloc(u_char type)
    247 {
    248 	return malloc(sizeof(struct ifnet), M_DEVBUF, M_WAITOK|M_ZERO);
    249 }
    250 
    251 void
    252 if_initname(struct ifnet *ifp, const char *name, int unit)
    253 {
    254 	(void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
    255 	    "%s%d", name, unit);
    256 }
    257 
    258 /*
    259  * Null routines used while an interface is going away.  These routines
    260  * just return an error.
    261  */
    262 
    263 int
    264 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
    265     const struct sockaddr *so, struct rtentry *rt)
    266 {
    267 
    268 	return ENXIO;
    269 }
    270 
    271 void
    272 if_nullinput(struct ifnet *ifp, struct mbuf *m)
    273 {
    274 
    275 	/* Nothing. */
    276 }
    277 
    278 void
    279 if_nullstart(struct ifnet *ifp)
    280 {
    281 
    282 	/* Nothing. */
    283 }
    284 
    285 int
    286 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
    287 {
    288 
    289 	return ENXIO;
    290 }
    291 
    292 int
    293 if_nullinit(struct ifnet *ifp)
    294 {
    295 
    296 	return ENXIO;
    297 }
    298 
    299 void
    300 if_nullstop(struct ifnet *ifp, int disable)
    301 {
    302 
    303 	/* Nothing. */
    304 }
    305 
    306 void
    307 if_nullwatchdog(struct ifnet *ifp)
    308 {
    309 
    310 	/* Nothing. */
    311 }
    312 
    313 void
    314 if_nulldrain(struct ifnet *ifp)
    315 {
    316 
    317 	/* Nothing. */
    318 }
    319 
    320 static u_int if_index = 1;
    321 struct ifnet_head ifnet;
    322 size_t if_indexlim = 0;
    323 struct ifaddr **ifnet_addrs = NULL;
    324 struct ifnet **ifindex2ifnet = NULL;
    325 struct ifnet *lo0ifp;
    326 
    327 void
    328 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory)
    329 {
    330 	struct ifaddr *ifa;
    331 	struct sockaddr_dl *sdl;
    332 
    333 	ifp->if_addrlen = addrlen;
    334 	if_alloc_sadl(ifp);
    335 	ifa = ifp->if_dl;
    336 	sdl = satosdl(ifa->ifa_addr);
    337 
    338 	(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
    339 	if (factory) {
    340 		ifp->if_hwdl = ifp->if_dl;
    341 		IFAREF(ifp->if_hwdl);
    342 	}
    343 	/* TBD routing socket */
    344 }
    345 
    346 struct ifaddr *
    347 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
    348 {
    349 	unsigned socksize, ifasize;
    350 	int addrlen, namelen;
    351 	struct sockaddr_dl *mask, *sdl;
    352 	struct ifaddr *ifa;
    353 
    354 	namelen = strlen(ifp->if_xname);
    355 	addrlen = ifp->if_addrlen;
    356 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
    357 	ifasize = sizeof(*ifa) + 2 * socksize;
    358 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
    359 
    360 	sdl = (struct sockaddr_dl *)(ifa + 1);
    361 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
    362 
    363 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
    364 	    ifp->if_xname, namelen, NULL, addrlen);
    365 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
    366 	memset(&mask->sdl_data[0], 0xff, namelen);
    367 	ifa->ifa_rtrequest = link_rtrequest;
    368 	ifa->ifa_addr = (struct sockaddr *)sdl;
    369 	ifa->ifa_netmask = (struct sockaddr *)mask;
    370 
    371 	*sdlp = sdl;
    372 
    373 	return ifa;
    374 }
    375 
    376 static void
    377 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa)
    378 {
    379 	const struct sockaddr_dl *sdl;
    380 	ifnet_addrs[ifp->if_index] = ifa;
    381 	IFAREF(ifa);
    382 	ifp->if_dl = ifa;
    383 	IFAREF(ifa);
    384 	sdl = satosdl(ifa->ifa_addr);
    385 	ifp->if_sadl = sdl;
    386 }
    387 
    388 /*
    389  * Allocate the link level name for the specified interface.  This
    390  * is an attachment helper.  It must be called after ifp->if_addrlen
    391  * is initialized, which may not be the case when if_attach() is
    392  * called.
    393  */
    394 void
    395 if_alloc_sadl(struct ifnet *ifp)
    396 {
    397 	struct ifaddr *ifa;
    398 	const struct sockaddr_dl *sdl;
    399 
    400 	/*
    401 	 * If the interface already has a link name, release it
    402 	 * now.  This is useful for interfaces that can change
    403 	 * link types, and thus switch link names often.
    404 	 */
    405 	if (ifp->if_sadl != NULL)
    406 		if_free_sadl(ifp);
    407 
    408 	ifa = if_dl_create(ifp, &sdl);
    409 
    410 	ifa_insert(ifp, ifa);
    411 	if_sadl_setrefs(ifp, ifa);
    412 }
    413 
    414 static void
    415 if_deactivate_sadl(struct ifnet *ifp)
    416 {
    417 	struct ifaddr *ifa;
    418 
    419 	KASSERT(ifp->if_dl != NULL);
    420 
    421 	ifa = ifp->if_dl;
    422 
    423 	ifp->if_sadl = NULL;
    424 
    425 	ifnet_addrs[ifp->if_index] = NULL;
    426 	IFAFREE(ifa);
    427 	ifp->if_dl = NULL;
    428 	IFAFREE(ifa);
    429 }
    430 
    431 void
    432 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa,
    433     const struct sockaddr_dl *sdl)
    434 {
    435 	int s;
    436 
    437 	s = splnet();
    438 
    439 	if_deactivate_sadl(ifp);
    440 
    441 	if_sadl_setrefs(ifp, ifa);
    442 	IFADDR_FOREACH(ifa, ifp)
    443 		rtinit(ifa, RTM_LLINFO_UPD, 0);
    444 	splx(s);
    445 }
    446 
    447 /*
    448  * Free the link level name for the specified interface.  This is
    449  * a detach helper.  This is called from if_detach() or from
    450  * link layer type specific detach functions.
    451  */
    452 void
    453 if_free_sadl(struct ifnet *ifp)
    454 {
    455 	struct ifaddr *ifa;
    456 	int s;
    457 
    458 	ifa = ifnet_addrs[ifp->if_index];
    459 	if (ifa == NULL) {
    460 		KASSERT(ifp->if_sadl == NULL);
    461 		KASSERT(ifp->if_dl == NULL);
    462 		return;
    463 	}
    464 
    465 	KASSERT(ifp->if_sadl != NULL);
    466 	KASSERT(ifp->if_dl != NULL);
    467 
    468 	s = splnet();
    469 	rtinit(ifa, RTM_DELETE, 0);
    470 	ifa_remove(ifp, ifa);
    471 	if_deactivate_sadl(ifp);
    472 	if (ifp->if_hwdl == ifa) {
    473 		IFAFREE(ifa);
    474 		ifp->if_hwdl = NULL;
    475 	}
    476 	splx(s);
    477 }
    478 
    479 /*
    480  * Attach an interface to the
    481  * list of "active" interfaces.
    482  */
    483 void
    484 if_attach(struct ifnet *ifp)
    485 {
    486 	int indexlim = 0;
    487 
    488 	if (if_indexlim == 0) {
    489 		TAILQ_INIT(&ifnet);
    490 		if_indexlim = 8;
    491 	}
    492 	TAILQ_INIT(&ifp->if_addrlist);
    493 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
    494 	if (ifp->if_ioctl == NULL)
    495 		ifp->if_ioctl = ifioctl_common;
    496 
    497 	mutex_enter(&index_gen_mtx);
    498 	ifp->if_index_gen = index_gen++;
    499 	mutex_exit(&index_gen_mtx);
    500 
    501 	ifp->if_index = if_index;
    502 	if (ifindex2ifnet == NULL)
    503 		if_index++;
    504 	else
    505 		while (ifp->if_index < if_indexlim &&
    506 		    ifindex2ifnet[ifp->if_index] != NULL) {
    507 			++if_index;
    508 			if (if_index == 0)
    509 				if_index = 1;
    510 			/*
    511 			 * If we hit USHRT_MAX, we skip back to 0 since
    512 			 * there are a number of places where the value
    513 			 * of if_index or if_index itself is compared
    514 			 * to or stored in an unsigned short.  By
    515 			 * jumping back, we won't botch those assignments
    516 			 * or comparisons.
    517 			 */
    518 			else if (if_index == USHRT_MAX) {
    519 				/*
    520 				 * However, if we have to jump back to
    521 				 * zero *twice* without finding an empty
    522 				 * slot in ifindex2ifnet[], then there
    523 				 * there are too many (>65535) interfaces.
    524 				 */
    525 				if (indexlim++)
    526 					panic("too many interfaces");
    527 				else
    528 					if_index = 1;
    529 			}
    530 			ifp->if_index = if_index;
    531 		}
    532 
    533 	/*
    534 	 * We have some arrays that should be indexed by if_index.
    535 	 * since if_index will grow dynamically, they should grow too.
    536 	 *	struct ifadd **ifnet_addrs
    537 	 *	struct ifnet **ifindex2ifnet
    538 	 */
    539 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
    540 	    ifp->if_index >= if_indexlim) {
    541 		size_t m, n, oldlim;
    542 		void *q;
    543 
    544 		oldlim = if_indexlim;
    545 		while (ifp->if_index >= if_indexlim)
    546 			if_indexlim <<= 1;
    547 
    548 		/* grow ifnet_addrs */
    549 		m = oldlim * sizeof(struct ifaddr *);
    550 		n = if_indexlim * sizeof(struct ifaddr *);
    551 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
    552 		if (ifnet_addrs != NULL) {
    553 			memcpy(q, ifnet_addrs, m);
    554 			free(ifnet_addrs, M_IFADDR);
    555 		}
    556 		ifnet_addrs = (struct ifaddr **)q;
    557 
    558 		/* grow ifindex2ifnet */
    559 		m = oldlim * sizeof(struct ifnet *);
    560 		n = if_indexlim * sizeof(struct ifnet *);
    561 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
    562 		if (ifindex2ifnet != NULL) {
    563 			memcpy(q, ifindex2ifnet, m);
    564 			free(ifindex2ifnet, M_IFADDR);
    565 		}
    566 		ifindex2ifnet = (struct ifnet **)q;
    567 	}
    568 
    569 	ifindex2ifnet[ifp->if_index] = ifp;
    570 
    571 	/*
    572 	 * Link level name is allocated later by a separate call to
    573 	 * if_alloc_sadl().
    574 	 */
    575 
    576 	if (ifp->if_snd.ifq_maxlen == 0)
    577 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
    578 
    579 	sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
    580 
    581 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
    582 
    583 	ifp->if_link_state = LINK_STATE_UNKNOWN;
    584 
    585 	ifp->if_capenable = 0;
    586 	ifp->if_csum_flags_tx = 0;
    587 	ifp->if_csum_flags_rx = 0;
    588 
    589 #ifdef ALTQ
    590 	ifp->if_snd.altq_type = 0;
    591 	ifp->if_snd.altq_disc = NULL;
    592 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
    593 	ifp->if_snd.altq_tbr  = NULL;
    594 	ifp->if_snd.altq_ifp  = ifp;
    595 #endif
    596 
    597 #ifdef PFIL_HOOKS
    598 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
    599 	ifp->if_pfil.ph_ifnet = ifp;
    600 	if (pfil_head_register(&ifp->if_pfil) != 0)
    601 		printf("%s: WARNING: unable to register pfil hook\n",
    602 		    ifp->if_xname);
    603 	(void)pfil_run_hooks(&if_pfil,
    604 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
    605 #endif
    606 
    607 	if (!STAILQ_EMPTY(&domains))
    608 		if_attachdomain1(ifp);
    609 
    610 	/* Announce the interface. */
    611 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
    612 }
    613 
    614 void
    615 if_attachdomain(void)
    616 {
    617 	struct ifnet *ifp;
    618 	int s;
    619 
    620 	s = splnet();
    621 	IFNET_FOREACH(ifp)
    622 		if_attachdomain1(ifp);
    623 	splx(s);
    624 }
    625 
    626 void
    627 if_attachdomain1(struct ifnet *ifp)
    628 {
    629 	struct domain *dp;
    630 	int s;
    631 
    632 	s = splnet();
    633 
    634 	/* address family dependent data region */
    635 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
    636 	DOMAIN_FOREACH(dp) {
    637 		if (dp->dom_ifattach != NULL)
    638 			ifp->if_afdata[dp->dom_family] =
    639 			    (*dp->dom_ifattach)(ifp);
    640 	}
    641 
    642 	splx(s);
    643 }
    644 
    645 /*
    646  * Deactivate an interface.  This points all of the procedure
    647  * handles at error stubs.  May be called from interrupt context.
    648  */
    649 void
    650 if_deactivate(struct ifnet *ifp)
    651 {
    652 	int s;
    653 
    654 	s = splnet();
    655 
    656 	ifp->if_output	 = if_nulloutput;
    657 	ifp->if_input	 = if_nullinput;
    658 	ifp->if_start	 = if_nullstart;
    659 	ifp->if_ioctl	 = if_nullioctl;
    660 	ifp->if_init	 = if_nullinit;
    661 	ifp->if_stop	 = if_nullstop;
    662 	ifp->if_watchdog = if_nullwatchdog;
    663 	ifp->if_drain	 = if_nulldrain;
    664 
    665 	/* No more packets may be enqueued. */
    666 	ifp->if_snd.ifq_maxlen = 0;
    667 
    668 	splx(s);
    669 }
    670 
    671 void
    672 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
    673 {
    674 	struct ifaddr *ifa, *nifa;
    675 
    676 	for (ifa = IFADDR_FIRST(ifp); ifa != NULL; ifa = nifa) {
    677 		nifa = IFADDR_NEXT(ifa);
    678 		if (ifa->ifa_addr->sa_family != family)
    679 			continue;
    680 		(*purgeaddr)(ifa);
    681 	}
    682 }
    683 
    684 /*
    685  * Detach an interface from the list of "active" interfaces,
    686  * freeing any resources as we go along.
    687  *
    688  * NOTE: This routine must be called with a valid thread context,
    689  * as it may block.
    690  */
    691 void
    692 if_detach(struct ifnet *ifp)
    693 {
    694 	struct socket so;
    695 	struct ifaddr *ifa;
    696 #ifdef IFAREF_DEBUG
    697 	struct ifaddr *last_ifa = NULL;
    698 #endif
    699 	struct domain *dp;
    700 	const struct protosw *pr;
    701 	int s, i, family, purged;
    702 
    703 	/*
    704 	 * XXX It's kind of lame that we have to have the
    705 	 * XXX socket structure...
    706 	 */
    707 	memset(&so, 0, sizeof(so));
    708 
    709 	s = splnet();
    710 
    711 	/*
    712 	 * Do an if_down() to give protocols a chance to do something.
    713 	 */
    714 	if_down(ifp);
    715 
    716 #ifdef ALTQ
    717 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
    718 		altq_disable(&ifp->if_snd);
    719 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
    720 		altq_detach(&ifp->if_snd);
    721 #endif
    722 
    723 	sysctl_teardown(&ifp->if_sysctl_log);
    724 
    725 #if NCARP > 0
    726 	/* Remove the interface from any carp group it is a part of.  */
    727 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
    728 		carp_ifdetach(ifp);
    729 #endif
    730 
    731 	/*
    732 	 * Rip all the addresses off the interface.  This should make
    733 	 * all of the routes go away.
    734 	 *
    735 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
    736 	 * from the list, including our "cursor", ifa.  For safety,
    737 	 * and to honor the TAILQ abstraction, I just restart the
    738 	 * loop after each removal.  Note that the loop will exit
    739 	 * when all of the remaining ifaddrs belong to the AF_LINK
    740 	 * family.  I am counting on the historical fact that at
    741 	 * least one pr_usrreq in each address domain removes at
    742 	 * least one ifaddr.
    743 	 */
    744 again:
    745 	IFADDR_FOREACH(ifa, ifp) {
    746 		family = ifa->ifa_addr->sa_family;
    747 #ifdef IFAREF_DEBUG
    748 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
    749 		    ifa, family, ifa->ifa_refcnt);
    750 		if (last_ifa != NULL && ifa == last_ifa)
    751 			panic("if_detach: loop detected");
    752 		last_ifa = ifa;
    753 #endif
    754 		if (family == AF_LINK)
    755 			continue;
    756 		dp = pffinddomain(family);
    757 #ifdef DIAGNOSTIC
    758 		if (dp == NULL)
    759 			panic("if_detach: no domain for AF %d",
    760 			    family);
    761 #endif
    762 		/*
    763 		 * XXX These PURGEIF calls are redundant with the
    764 		 * purge-all-families calls below, but are left in for
    765 		 * now both to make a smaller change, and to avoid
    766 		 * unplanned interactions with clearing of
    767 		 * ifp->if_addrlist.
    768 		 */
    769 		purged = 0;
    770 		for (pr = dp->dom_protosw;
    771 		     pr < dp->dom_protoswNPROTOSW; pr++) {
    772 			so.so_proto = pr;
    773 			if (pr->pr_usrreq != NULL) {
    774 				(void) (*pr->pr_usrreq)(&so,
    775 				    PRU_PURGEIF, NULL, NULL,
    776 				    (struct mbuf *) ifp, curlwp);
    777 				purged = 1;
    778 			}
    779 		}
    780 		if (purged == 0) {
    781 			/*
    782 			 * XXX What's really the best thing to do
    783 			 * XXX here?  --thorpej (at) NetBSD.org
    784 			 */
    785 			printf("if_detach: WARNING: AF %d not purged\n",
    786 			    family);
    787 			ifa_remove(ifp, ifa);
    788 		}
    789 		goto again;
    790 	}
    791 
    792 	if_free_sadl(ifp);
    793 
    794 	/* Walk the routing table looking for stragglers. */
    795 	for (i = 0; i <= AF_MAX; i++)
    796 		(void)rt_walktree(i, if_rt_walktree, ifp);
    797 
    798 	DOMAIN_FOREACH(dp) {
    799 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
    800 			(*dp->dom_ifdetach)(ifp,
    801 			    ifp->if_afdata[dp->dom_family]);
    802 
    803 		/*
    804 		 * One would expect multicast memberships (INET and
    805 		 * INET6) on UDP sockets to be purged by the PURGEIF
    806 		 * calls above, but if all addresses were removed from
    807 		 * the interface prior to destruction, the calls will
    808 		 * not be made (e.g. ppp, for which pppd(8) generally
    809 		 * removes addresses before destroying the interface).
    810 		 * Because there is no invariant that multicast
    811 		 * memberships only exist for interfaces with IPv4
    812 		 * addresses, we must call PURGEIF regardless of
    813 		 * addresses.  (Protocols which might store ifnet
    814 		 * pointers are marked with PR_PURGEIF.)
    815 		 */
    816 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
    817 			so.so_proto = pr;
    818 			if (pr->pr_usrreq != NULL && pr->pr_flags & PR_PURGEIF)
    819 				(void)(*pr->pr_usrreq)(&so, PRU_PURGEIF, NULL,
    820 				    NULL, (struct mbuf *)ifp, curlwp);
    821 		}
    822 	}
    823 
    824 #ifdef PFIL_HOOKS
    825 	(void)pfil_run_hooks(&if_pfil,
    826 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
    827 	(void)pfil_head_unregister(&ifp->if_pfil);
    828 #endif
    829 
    830 	/* Announce that the interface is gone. */
    831 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
    832 
    833 	ifindex2ifnet[ifp->if_index] = NULL;
    834 
    835 	TAILQ_REMOVE(&ifnet, ifp, if_list);
    836 
    837 	/*
    838 	 * remove packets that came from ifp, from software interrupt queues.
    839 	 */
    840 	DOMAIN_FOREACH(dp) {
    841 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
    842 			if (dp->dom_ifqueues[i] == NULL)
    843 				break;
    844 			if_detach_queues(ifp, dp->dom_ifqueues[i]);
    845 		}
    846 	}
    847 
    848 	splx(s);
    849 }
    850 
    851 static void
    852 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
    853 {
    854 	struct mbuf *m, *prev, *next;
    855 
    856 	prev = NULL;
    857 	for (m = q->ifq_head; m != NULL; m = next) {
    858 		next = m->m_nextpkt;
    859 #ifdef DIAGNOSTIC
    860 		if ((m->m_flags & M_PKTHDR) == 0) {
    861 			prev = m;
    862 			continue;
    863 		}
    864 #endif
    865 		if (m->m_pkthdr.rcvif != ifp) {
    866 			prev = m;
    867 			continue;
    868 		}
    869 
    870 		if (prev != NULL)
    871 			prev->m_nextpkt = m->m_nextpkt;
    872 		else
    873 			q->ifq_head = m->m_nextpkt;
    874 		if (q->ifq_tail == m)
    875 			q->ifq_tail = prev;
    876 		q->ifq_len--;
    877 
    878 		m->m_nextpkt = NULL;
    879 		m_freem(m);
    880 		IF_DROP(q);
    881 	}
    882 }
    883 
    884 /*
    885  * Callback for a radix tree walk to delete all references to an
    886  * ifnet.
    887  */
    888 static int
    889 if_rt_walktree(struct rtentry *rt, void *v)
    890 {
    891 	struct ifnet *ifp = (struct ifnet *)v;
    892 	int error;
    893 
    894 	if (rt->rt_ifp != ifp)
    895 		return 0;
    896 
    897 	/* Delete the entry. */
    898 	++rt->rt_refcnt;
    899 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
    900 	    rt_mask(rt), rt->rt_flags, NULL);
    901 	KASSERT((rt->rt_flags & RTF_UP) == 0);
    902 	rt->rt_ifp = NULL;
    903 	RTFREE(rt);
    904 	if (error != 0)
    905 		printf("%s: warning: unable to delete rtentry @ %p, "
    906 		    "error = %d\n", ifp->if_xname, rt, error);
    907 	return 0;
    908 }
    909 
    910 /*
    911  * Create a clone network interface.
    912  */
    913 int
    914 if_clone_create(const char *name)
    915 {
    916 	struct if_clone *ifc;
    917 	int unit;
    918 
    919 	ifc = if_clone_lookup(name, &unit);
    920 	if (ifc == NULL)
    921 		return EINVAL;
    922 
    923 	if (ifunit(name) != NULL)
    924 		return EEXIST;
    925 
    926 	return (*ifc->ifc_create)(ifc, unit);
    927 }
    928 
    929 /*
    930  * Destroy a clone network interface.
    931  */
    932 int
    933 if_clone_destroy(const char *name)
    934 {
    935 	struct if_clone *ifc;
    936 	struct ifnet *ifp;
    937 
    938 	ifc = if_clone_lookup(name, NULL);
    939 	if (ifc == NULL)
    940 		return EINVAL;
    941 
    942 	ifp = ifunit(name);
    943 	if (ifp == NULL)
    944 		return ENXIO;
    945 
    946 	if (ifc->ifc_destroy == NULL)
    947 		return EOPNOTSUPP;
    948 
    949 	return (*ifc->ifc_destroy)(ifp);
    950 }
    951 
    952 /*
    953  * Look up a network interface cloner.
    954  */
    955 static struct if_clone *
    956 if_clone_lookup(const char *name, int *unitp)
    957 {
    958 	struct if_clone *ifc;
    959 	const char *cp;
    960 	int unit;
    961 
    962 	/* separate interface name from unit */
    963 	for (cp = name;
    964 	    cp - name < IFNAMSIZ && *cp && (*cp < '0' || *cp > '9');
    965 	    cp++)
    966 		continue;
    967 
    968 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
    969 		return NULL;	/* No name or unit number */
    970 
    971 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
    972 		if (strlen(ifc->ifc_name) == cp - name &&
    973 		    strncmp(name, ifc->ifc_name, cp - name) == 0)
    974 			break;
    975 	}
    976 
    977 	if (ifc == NULL)
    978 		return NULL;
    979 
    980 	unit = 0;
    981 	while (cp - name < IFNAMSIZ && *cp) {
    982 		if (*cp < '0' || *cp > '9' || unit > INT_MAX / 10) {
    983 			/* Bogus unit number. */
    984 			return NULL;
    985 		}
    986 		unit = (unit * 10) + (*cp++ - '0');
    987 	}
    988 
    989 	if (unitp != NULL)
    990 		*unitp = unit;
    991 	return ifc;
    992 }
    993 
    994 /*
    995  * Register a network interface cloner.
    996  */
    997 void
    998 if_clone_attach(struct if_clone *ifc)
    999 {
   1000 
   1001 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
   1002 	if_cloners_count++;
   1003 }
   1004 
   1005 /*
   1006  * Unregister a network interface cloner.
   1007  */
   1008 void
   1009 if_clone_detach(struct if_clone *ifc)
   1010 {
   1011 
   1012 	LIST_REMOVE(ifc, ifc_list);
   1013 	if_cloners_count--;
   1014 }
   1015 
   1016 /*
   1017  * Provide list of interface cloners to userspace.
   1018  */
   1019 static int
   1020 if_clone_list(struct if_clonereq *ifcr)
   1021 {
   1022 	char outbuf[IFNAMSIZ], *dst;
   1023 	struct if_clone *ifc;
   1024 	int count, error = 0;
   1025 
   1026 	ifcr->ifcr_total = if_cloners_count;
   1027 	if ((dst = ifcr->ifcr_buffer) == NULL) {
   1028 		/* Just asking how many there are. */
   1029 		return 0;
   1030 	}
   1031 
   1032 	if (ifcr->ifcr_count < 0)
   1033 		return EINVAL;
   1034 
   1035 	count = (if_cloners_count < ifcr->ifcr_count) ?
   1036 	    if_cloners_count : ifcr->ifcr_count;
   1037 
   1038 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
   1039 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
   1040 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
   1041 		if (outbuf[sizeof(outbuf) - 1] != '\0')
   1042 			return ENAMETOOLONG;
   1043 		error = copyout(outbuf, dst, sizeof(outbuf));
   1044 		if (error != 0)
   1045 			break;
   1046 	}
   1047 
   1048 	return error;
   1049 }
   1050 
   1051 void
   1052 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
   1053 {
   1054 	ifa->ifa_ifp = ifp;
   1055 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
   1056 	IFAREF(ifa);
   1057 }
   1058 
   1059 void
   1060 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
   1061 {
   1062 	KASSERT(ifa->ifa_ifp == ifp);
   1063 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
   1064 	IFAFREE(ifa);
   1065 }
   1066 
   1067 static inline int
   1068 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
   1069 {
   1070 	return sockaddr_cmp(sa1, sa2) == 0;
   1071 }
   1072 
   1073 /*
   1074  * Locate an interface based on a complete address.
   1075  */
   1076 /*ARGSUSED*/
   1077 struct ifaddr *
   1078 ifa_ifwithaddr(const struct sockaddr *addr)
   1079 {
   1080 	struct ifnet *ifp;
   1081 	struct ifaddr *ifa;
   1082 
   1083 	IFNET_FOREACH(ifp) {
   1084 		if (ifp->if_output == if_nulloutput)
   1085 			continue;
   1086 		IFADDR_FOREACH(ifa, ifp) {
   1087 			if (ifa->ifa_addr->sa_family != addr->sa_family)
   1088 				continue;
   1089 			if (equal(addr, ifa->ifa_addr))
   1090 				return ifa;
   1091 			if ((ifp->if_flags & IFF_BROADCAST) &&
   1092 			    ifa->ifa_broadaddr &&
   1093 			    /* IP6 doesn't have broadcast */
   1094 			    ifa->ifa_broadaddr->sa_len != 0 &&
   1095 			    equal(ifa->ifa_broadaddr, addr))
   1096 				return ifa;
   1097 		}
   1098 	}
   1099 	return NULL;
   1100 }
   1101 
   1102 /*
   1103  * Locate the point to point interface with a given destination address.
   1104  */
   1105 /*ARGSUSED*/
   1106 struct ifaddr *
   1107 ifa_ifwithdstaddr(const struct sockaddr *addr)
   1108 {
   1109 	struct ifnet *ifp;
   1110 	struct ifaddr *ifa;
   1111 
   1112 	IFNET_FOREACH(ifp) {
   1113 		if (ifp->if_output == if_nulloutput)
   1114 			continue;
   1115 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
   1116 			continue;
   1117 		IFADDR_FOREACH(ifa, ifp) {
   1118 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
   1119 			    ifa->ifa_dstaddr == NULL)
   1120 				continue;
   1121 			if (equal(addr, ifa->ifa_dstaddr))
   1122 				return ifa;
   1123 		}
   1124 	}
   1125 	return NULL;
   1126 }
   1127 
   1128 /*
   1129  * Find an interface on a specific network.  If many, choice
   1130  * is most specific found.
   1131  */
   1132 struct ifaddr *
   1133 ifa_ifwithnet(const struct sockaddr *addr)
   1134 {
   1135 	struct ifnet *ifp;
   1136 	struct ifaddr *ifa;
   1137 	const struct sockaddr_dl *sdl;
   1138 	struct ifaddr *ifa_maybe = 0;
   1139 	u_int af = addr->sa_family;
   1140 	const char *addr_data = addr->sa_data, *cplim;
   1141 
   1142 	if (af == AF_LINK) {
   1143 		sdl = satocsdl(addr);
   1144 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
   1145 		    ifindex2ifnet[sdl->sdl_index] &&
   1146 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
   1147 			return ifnet_addrs[sdl->sdl_index];
   1148 	}
   1149 #ifdef NETATALK
   1150 	if (af == AF_APPLETALK) {
   1151 		const struct sockaddr_at *sat, *sat2;
   1152 		sat = (const struct sockaddr_at *)addr;
   1153 		IFNET_FOREACH(ifp) {
   1154 			if (ifp->if_output == if_nulloutput)
   1155 				continue;
   1156 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
   1157 			if (ifa == NULL)
   1158 				continue;
   1159 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
   1160 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
   1161 				return ifa; /* exact match */
   1162 			if (ifa_maybe == NULL) {
   1163 				/* else keep the if with the right range */
   1164 				ifa_maybe = ifa;
   1165 			}
   1166 		}
   1167 		return ifa_maybe;
   1168 	}
   1169 #endif
   1170 	IFNET_FOREACH(ifp) {
   1171 		if (ifp->if_output == if_nulloutput)
   1172 			continue;
   1173 		IFADDR_FOREACH(ifa, ifp) {
   1174 			const char *cp, *cp2, *cp3;
   1175 
   1176 			if (ifa->ifa_addr->sa_family != af ||
   1177 			    ifa->ifa_netmask == NULL)
   1178  next:				continue;
   1179 			cp = addr_data;
   1180 			cp2 = ifa->ifa_addr->sa_data;
   1181 			cp3 = ifa->ifa_netmask->sa_data;
   1182 			cplim = (const char *)ifa->ifa_netmask +
   1183 			    ifa->ifa_netmask->sa_len;
   1184 			while (cp3 < cplim) {
   1185 				if ((*cp++ ^ *cp2++) & *cp3++) {
   1186 					/* want to continue for() loop */
   1187 					goto next;
   1188 				}
   1189 			}
   1190 			if (ifa_maybe == NULL ||
   1191 			    rn_refines((void *)ifa->ifa_netmask,
   1192 			    (void *)ifa_maybe->ifa_netmask))
   1193 				ifa_maybe = ifa;
   1194 		}
   1195 	}
   1196 	return ifa_maybe;
   1197 }
   1198 
   1199 /*
   1200  * Find the interface of the addresss.
   1201  */
   1202 struct ifaddr *
   1203 ifa_ifwithladdr(const struct sockaddr *addr)
   1204 {
   1205 	struct ifaddr *ia;
   1206 
   1207 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
   1208 	    (ia = ifa_ifwithnet(addr)))
   1209 		return ia;
   1210 	return NULL;
   1211 }
   1212 
   1213 /*
   1214  * Find an interface using a specific address family
   1215  */
   1216 struct ifaddr *
   1217 ifa_ifwithaf(int af)
   1218 {
   1219 	struct ifnet *ifp;
   1220 	struct ifaddr *ifa;
   1221 
   1222 	IFNET_FOREACH(ifp) {
   1223 		if (ifp->if_output == if_nulloutput)
   1224 			continue;
   1225 		IFADDR_FOREACH(ifa, ifp) {
   1226 			if (ifa->ifa_addr->sa_family == af)
   1227 				return ifa;
   1228 		}
   1229 	}
   1230 	return NULL;
   1231 }
   1232 
   1233 /*
   1234  * Find an interface address specific to an interface best matching
   1235  * a given address.
   1236  */
   1237 struct ifaddr *
   1238 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
   1239 {
   1240 	struct ifaddr *ifa;
   1241 	const char *cp, *cp2, *cp3;
   1242 	const char *cplim;
   1243 	struct ifaddr *ifa_maybe = 0;
   1244 	u_int af = addr->sa_family;
   1245 
   1246 	if (ifp->if_output == if_nulloutput)
   1247 		return NULL;
   1248 
   1249 	if (af >= AF_MAX)
   1250 		return NULL;
   1251 
   1252 	IFADDR_FOREACH(ifa, ifp) {
   1253 		if (ifa->ifa_addr->sa_family != af)
   1254 			continue;
   1255 		ifa_maybe = ifa;
   1256 		if (ifa->ifa_netmask == NULL) {
   1257 			if (equal(addr, ifa->ifa_addr) ||
   1258 			    (ifa->ifa_dstaddr &&
   1259 			     equal(addr, ifa->ifa_dstaddr)))
   1260 				return ifa;
   1261 			continue;
   1262 		}
   1263 		cp = addr->sa_data;
   1264 		cp2 = ifa->ifa_addr->sa_data;
   1265 		cp3 = ifa->ifa_netmask->sa_data;
   1266 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
   1267 		for (; cp3 < cplim; cp3++) {
   1268 			if ((*cp++ ^ *cp2++) & *cp3)
   1269 				break;
   1270 		}
   1271 		if (cp3 == cplim)
   1272 			return ifa;
   1273 	}
   1274 	return ifa_maybe;
   1275 }
   1276 
   1277 /*
   1278  * Default action when installing a route with a Link Level gateway.
   1279  * Lookup an appropriate real ifa to point to.
   1280  * This should be moved to /sys/net/link.c eventually.
   1281  */
   1282 void
   1283 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
   1284 {
   1285 	struct ifaddr *ifa;
   1286 	const struct sockaddr *dst;
   1287 	struct ifnet *ifp;
   1288 
   1289 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
   1290 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
   1291 		return;
   1292 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
   1293 		rt_replace_ifa(rt, ifa);
   1294 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
   1295 			ifa->ifa_rtrequest(cmd, rt, info);
   1296 	}
   1297 }
   1298 
   1299 /*
   1300  * Handle a change in the interface link state.
   1301  */
   1302 void
   1303 if_link_state_change(struct ifnet *ifp, int link_state)
   1304 {
   1305 	if (ifp->if_link_state == link_state)
   1306 		return;
   1307 	ifp->if_link_state = link_state;
   1308 	/* Notify that the link state has changed. */
   1309 	rt_ifmsg(ifp);
   1310 #if NCARP > 0
   1311 	if (ifp->if_carp)
   1312 		carp_carpdev_state(ifp);
   1313 #endif
   1314 }
   1315 
   1316 /*
   1317  * Mark an interface down and notify protocols of
   1318  * the transition.
   1319  * NOTE: must be called at splsoftnet or equivalent.
   1320  */
   1321 void
   1322 if_down(struct ifnet *ifp)
   1323 {
   1324 	struct ifaddr *ifa;
   1325 
   1326 	ifp->if_flags &= ~IFF_UP;
   1327 	nanotime(&ifp->if_lastchange);
   1328 	IFADDR_FOREACH(ifa, ifp)
   1329 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
   1330 	IFQ_PURGE(&ifp->if_snd);
   1331 #if NCARP > 0
   1332 	if (ifp->if_carp)
   1333 		carp_carpdev_state(ifp);
   1334 #endif
   1335 	rt_ifmsg(ifp);
   1336 }
   1337 
   1338 /*
   1339  * Mark an interface up and notify protocols of
   1340  * the transition.
   1341  * NOTE: must be called at splsoftnet or equivalent.
   1342  */
   1343 void
   1344 if_up(struct ifnet *ifp)
   1345 {
   1346 #ifdef notyet
   1347 	struct ifaddr *ifa;
   1348 #endif
   1349 
   1350 	ifp->if_flags |= IFF_UP;
   1351 	nanotime(&ifp->if_lastchange);
   1352 #ifdef notyet
   1353 	/* this has no effect on IP, and will kill all ISO connections XXX */
   1354 	IFADDR_FOREACH(ifa, ifp)
   1355 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
   1356 #endif
   1357 #if NCARP > 0
   1358 	if (ifp->if_carp)
   1359 		carp_carpdev_state(ifp);
   1360 #endif
   1361 	rt_ifmsg(ifp);
   1362 #ifdef INET6
   1363 	in6_if_up(ifp);
   1364 #endif
   1365 }
   1366 
   1367 /*
   1368  * Handle interface watchdog timer routines.  Called
   1369  * from softclock, we decrement timers (if set) and
   1370  * call the appropriate interface routine on expiration.
   1371  */
   1372 void
   1373 if_slowtimo(void *arg)
   1374 {
   1375 	struct ifnet *ifp;
   1376 	int s = splnet();
   1377 
   1378 	IFNET_FOREACH(ifp) {
   1379 		if (ifp->if_timer == 0 || --ifp->if_timer)
   1380 			continue;
   1381 		if (ifp->if_watchdog != NULL)
   1382 			(*ifp->if_watchdog)(ifp);
   1383 	}
   1384 	splx(s);
   1385 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
   1386 }
   1387 
   1388 /*
   1389  * Set/clear promiscuous mode on interface ifp based on the truth value
   1390  * of pswitch.  The calls are reference counted so that only the first
   1391  * "on" request actually has an effect, as does the final "off" request.
   1392  * Results are undefined if the "off" and "on" requests are not matched.
   1393  */
   1394 int
   1395 ifpromisc(struct ifnet *ifp, int pswitch)
   1396 {
   1397 	int pcount, ret;
   1398 	short flags;
   1399 	struct ifreq ifr;
   1400 
   1401 	pcount = ifp->if_pcount;
   1402 	flags = ifp->if_flags;
   1403 	if (pswitch) {
   1404 		/*
   1405 		 * Allow the device to be "placed" into promiscuous
   1406 		 * mode even if it is not configured up.  It will
   1407 		 * consult IFF_PROMISC when it is is brought up.
   1408 		 */
   1409 		if (ifp->if_pcount++ != 0)
   1410 			return 0;
   1411 		ifp->if_flags |= IFF_PROMISC;
   1412 		if ((ifp->if_flags & IFF_UP) == 0)
   1413 			return 0;
   1414 	} else {
   1415 		if (--ifp->if_pcount > 0)
   1416 			return 0;
   1417 		ifp->if_flags &= ~IFF_PROMISC;
   1418 		/*
   1419 		 * If the device is not configured up, we should not need to
   1420 		 * turn off promiscuous mode (device should have turned it
   1421 		 * off when interface went down; and will look at IFF_PROMISC
   1422 		 * again next time interface comes up).
   1423 		 */
   1424 		if ((ifp->if_flags & IFF_UP) == 0)
   1425 			return 0;
   1426 	}
   1427 	memset(&ifr, 0, sizeof(ifr));
   1428 	ifr.ifr_flags = ifp->if_flags;
   1429 	ret = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
   1430 	/* Restore interface state if not successful. */
   1431 	if (ret != 0) {
   1432 		ifp->if_pcount = pcount;
   1433 		ifp->if_flags = flags;
   1434 	}
   1435 	return ret;
   1436 }
   1437 
   1438 /*
   1439  * Map interface name to
   1440  * interface structure pointer.
   1441  */
   1442 struct ifnet *
   1443 ifunit(const char *name)
   1444 {
   1445 	struct ifnet *ifp;
   1446 	const char *cp = name;
   1447 	u_int unit = 0;
   1448 	u_int i;
   1449 
   1450 	/*
   1451 	 * If the entire name is a number, treat it as an ifindex.
   1452 	 */
   1453 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
   1454 		unit = unit * 10 + (*cp - '0');
   1455 	}
   1456 
   1457 	/*
   1458 	 * If the number took all of the name, then it's a valid ifindex.
   1459 	 */
   1460 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
   1461 		if (unit >= if_indexlim)
   1462 			return NULL;
   1463 		ifp = ifindex2ifnet[unit];
   1464 		if (ifp == NULL || ifp->if_output == if_nulloutput)
   1465 			return NULL;
   1466 		return ifp;
   1467 	}
   1468 
   1469 	IFNET_FOREACH(ifp) {
   1470 		if (ifp->if_output == if_nulloutput)
   1471 			continue;
   1472 	 	if (strcmp(ifp->if_xname, name) == 0)
   1473 			return ifp;
   1474 	}
   1475 	return NULL;
   1476 }
   1477 
   1478 /* common */
   1479 int
   1480 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
   1481 {
   1482 	int s;
   1483 	struct ifreq *ifr;
   1484 	struct ifcapreq *ifcr;
   1485 	struct ifdatareq *ifdr;
   1486 
   1487 	switch (cmd) {
   1488 	case SIOCSIFCAP:
   1489 		ifcr = data;
   1490 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
   1491 			return EINVAL;
   1492 
   1493 		if (ifcr->ifcr_capenable == ifp->if_capenable)
   1494 			return 0;
   1495 
   1496 		ifp->if_capenable = ifcr->ifcr_capenable;
   1497 
   1498 		/* Pre-compute the checksum flags mask. */
   1499 		ifp->if_csum_flags_tx = 0;
   1500 		ifp->if_csum_flags_rx = 0;
   1501 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
   1502 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
   1503 		}
   1504 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
   1505 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
   1506 		}
   1507 
   1508 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
   1509 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
   1510 		}
   1511 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
   1512 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
   1513 		}
   1514 
   1515 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
   1516 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
   1517 		}
   1518 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
   1519 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
   1520 		}
   1521 
   1522 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
   1523 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
   1524 		}
   1525 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
   1526 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
   1527 		}
   1528 
   1529 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
   1530 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
   1531 		}
   1532 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
   1533 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
   1534 		}
   1535 		if (ifp->if_flags & IFF_UP)
   1536 			return ENETRESET;
   1537 		return 0;
   1538 	case SIOCSIFFLAGS:
   1539 		ifr = data;
   1540 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
   1541 			s = splnet();
   1542 			if_down(ifp);
   1543 			splx(s);
   1544 		}
   1545 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
   1546 			s = splnet();
   1547 			if_up(ifp);
   1548 			splx(s);
   1549 		}
   1550 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
   1551 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
   1552 		break;
   1553 	case SIOCGIFFLAGS:
   1554 		ifr = data;
   1555 		ifr->ifr_flags = ifp->if_flags;
   1556 		break;
   1557 
   1558 	case SIOCGIFMETRIC:
   1559 		ifr = data;
   1560 		ifr->ifr_metric = ifp->if_metric;
   1561 		break;
   1562 
   1563 	case SIOCGIFMTU:
   1564 		ifr = data;
   1565 		ifr->ifr_mtu = ifp->if_mtu;
   1566 		break;
   1567 
   1568 	case SIOCGIFDLT:
   1569 		ifr = data;
   1570 		ifr->ifr_dlt = ifp->if_dlt;
   1571 		break;
   1572 
   1573 	case SIOCGIFCAP:
   1574 		ifcr = data;
   1575 		ifcr->ifcr_capabilities = ifp->if_capabilities;
   1576 		ifcr->ifcr_capenable = ifp->if_capenable;
   1577 		break;
   1578 
   1579 	case SIOCSIFMETRIC:
   1580 		ifr = data;
   1581 		ifp->if_metric = ifr->ifr_metric;
   1582 		break;
   1583 
   1584 	case SIOCGIFDATA:
   1585 		ifdr = data;
   1586 		ifdr->ifdr_data = ifp->if_data;
   1587 		break;
   1588 
   1589 	case SIOCZIFDATA:
   1590 		ifdr = data;
   1591 		ifdr->ifdr_data = ifp->if_data;
   1592 		/*
   1593 		 * Assumes that the volatile counters that can be
   1594 		 * zero'ed are at the end of if_data.
   1595 		 */
   1596 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
   1597 		    offsetof(struct if_data, ifi_ipackets));
   1598 		break;
   1599 	case SIOCSIFMTU:
   1600 		ifr = data;
   1601 		if (ifp->if_mtu == ifr->ifr_mtu)
   1602 			break;
   1603 		ifp->if_mtu = ifr->ifr_mtu;
   1604 		/*
   1605 		 * If the link MTU changed, do network layer specific procedure.
   1606 		 */
   1607 #ifdef INET6
   1608 		nd6_setmtu(ifp);
   1609 #endif
   1610 		return ENETRESET;
   1611 	default:
   1612 		return ENOTTY;
   1613 	}
   1614 	return 0;
   1615 }
   1616 
   1617 int
   1618 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
   1619     lwp_t *l)
   1620 {
   1621 	struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
   1622 	struct ifaddr *ifa;
   1623 	const struct sockaddr *any, *sa;
   1624 	union {
   1625 		struct sockaddr sa;
   1626 		struct sockaddr_storage ss;
   1627 	} u, v;
   1628 
   1629 	switch (cmd) {
   1630 	case SIOCSIFADDRPREF:
   1631 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_INTERFACE,
   1632 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
   1633 		    NULL) != 0)
   1634 			return EPERM;
   1635 	case SIOCGIFADDRPREF:
   1636 		break;
   1637 	default:
   1638 		return EOPNOTSUPP;
   1639 	}
   1640 
   1641 	/* sanity checks */
   1642 	if (data == NULL || ifp == NULL) {
   1643 		panic("invalid argument to %s", __func__);
   1644 		/*NOTREACHED*/
   1645 	}
   1646 
   1647 	/* address must be specified on ADD and DELETE */
   1648 	sa = sstocsa(&ifap->ifap_addr);
   1649 	if (sa->sa_family != sofamily(so))
   1650 		return EINVAL;
   1651 	if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
   1652 		return EINVAL;
   1653 
   1654 	sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
   1655 
   1656 	IFADDR_FOREACH(ifa, ifp) {
   1657 		if (ifa->ifa_addr->sa_family != sa->sa_family)
   1658 			continue;
   1659 		sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
   1660 		if (sockaddr_cmp(&u.sa, &v.sa) == 0)
   1661 			break;
   1662 	}
   1663 	if (ifa == NULL)
   1664 		return EADDRNOTAVAIL;
   1665 
   1666 	switch (cmd) {
   1667 	case SIOCSIFADDRPREF:
   1668 		ifa->ifa_preference = ifap->ifap_preference;
   1669 		return 0;
   1670 	case SIOCGIFADDRPREF:
   1671 		/* fill in the if_laddrreq structure */
   1672 		(void)sockaddr_copy(sstosa(&ifap->ifap_addr),
   1673 		    sizeof(ifap->ifap_addr), ifa->ifa_addr);
   1674 		ifap->ifap_preference = ifa->ifa_preference;
   1675 		return 0;
   1676 	default:
   1677 		return EOPNOTSUPP;
   1678 	}
   1679 }
   1680 
   1681 /*
   1682  * Interface ioctls.
   1683  */
   1684 int
   1685 ifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
   1686 {
   1687 	struct ifnet *ifp;
   1688 	struct ifreq *ifr;
   1689 	struct ifcapreq *ifcr;
   1690 	struct ifdatareq *ifdr;
   1691 	int error = 0;
   1692 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
   1693 	u_long ocmd = cmd;
   1694 #endif
   1695 	short oif_flags;
   1696 #ifdef COMPAT_OIFREQ
   1697 	struct ifreq ifrb;
   1698 	struct oifreq *oifr = NULL;
   1699 #endif
   1700 
   1701 	switch (cmd) {
   1702 #ifdef COMPAT_OIFREQ
   1703 	case OSIOCGIFCONF:
   1704 	case OOSIOCGIFCONF:
   1705 		return compat_ifconf(cmd, data);
   1706 #endif
   1707 #ifdef COMPAT_OIFDATA
   1708 	case OSIOCGIFDATA:
   1709 	case OSIOCZIFDATA:
   1710 		return compat_ifdatareq(l, cmd, data);
   1711 #endif
   1712 	case SIOCGIFCONF:
   1713 		return ifconf(cmd, data);
   1714 	case SIOCINITIFADDR:
   1715 		return EPERM;
   1716 	}
   1717 
   1718 #ifdef COMPAT_OIFREQ
   1719 	cmd = compat_cvtcmd(cmd);
   1720 	if (cmd != ocmd) {
   1721 		oifr = data;
   1722 		data = ifr = &ifrb;
   1723 		ifreqo2n(oifr, ifr);
   1724 	} else
   1725 #endif
   1726 		ifr = data;
   1727 	ifcr = data;
   1728 	ifdr = data;
   1729 
   1730 	ifp = ifunit(ifr->ifr_name);
   1731 
   1732 	switch (cmd) {
   1733 	case SIOCIFCREATE:
   1734 	case SIOCIFDESTROY:
   1735 		if (l != NULL) {
   1736 			error = kauth_authorize_network(l->l_cred,
   1737 			    KAUTH_NETWORK_INTERFACE,
   1738 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
   1739 			    (void *)cmd, NULL);
   1740 			if (error != 0)
   1741 				return error;
   1742 		}
   1743 		return (cmd == SIOCIFCREATE) ?
   1744 			if_clone_create(ifr->ifr_name) :
   1745 			if_clone_destroy(ifr->ifr_name);
   1746 
   1747 	case SIOCIFGCLONERS:
   1748 		return if_clone_list((struct if_clonereq *)data);
   1749 	}
   1750 
   1751 	if (ifp == NULL)
   1752 		return ENXIO;
   1753 
   1754 	switch (cmd) {
   1755 	case SIOCALIFADDR:
   1756 	case SIOCDLIFADDR:
   1757 	case SIOCSIFADDRPREF:
   1758 	case SIOCSIFFLAGS:
   1759 	case SIOCSIFCAP:
   1760 	case SIOCSIFMETRIC:
   1761 	case SIOCZIFDATA:
   1762 	case SIOCSIFMTU:
   1763 	case SIOCSIFPHYADDR:
   1764 	case SIOCDIFPHYADDR:
   1765 #ifdef INET6
   1766 	case SIOCSIFPHYADDR_IN6:
   1767 #endif
   1768 	case SIOCSLIFPHYADDR:
   1769 	case SIOCADDMULTI:
   1770 	case SIOCDELMULTI:
   1771 	case SIOCSIFMEDIA:
   1772 	case SIOCSDRVSPEC:
   1773 	case SIOCG80211:
   1774 	case SIOCS80211:
   1775 	case SIOCS80211NWID:
   1776 	case SIOCS80211NWKEY:
   1777 	case SIOCS80211POWER:
   1778 	case SIOCS80211BSSID:
   1779 	case SIOCS80211CHANNEL:
   1780 		if (l != NULL) {
   1781 			error = kauth_authorize_network(l->l_cred,
   1782 			    KAUTH_NETWORK_INTERFACE,
   1783 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
   1784 			    (void *)cmd, NULL);
   1785 			if (error != 0)
   1786 				return error;
   1787 		}
   1788 	}
   1789 
   1790 	oif_flags = ifp->if_flags;
   1791 
   1792 	error = (*ifp->if_ioctl)(ifp, cmd, data);
   1793 	if (error != ENOTTY)
   1794 		;
   1795 	else if (so->so_proto == NULL)
   1796 		return EOPNOTSUPP;
   1797 	else {
   1798 #ifdef COMPAT_OSOCK
   1799 		error = compat_ifioctl(so, ocmd, cmd, data, l);
   1800 #else
   1801 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
   1802 		    (struct mbuf *)cmd, (struct mbuf *)data,
   1803 		    (struct mbuf *)ifp, l);
   1804 #endif
   1805 	}
   1806 
   1807 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
   1808 #ifdef INET6
   1809 		if ((ifp->if_flags & IFF_UP) != 0) {
   1810 			int s = splnet();
   1811 			in6_if_up(ifp);
   1812 			splx(s);
   1813 		}
   1814 #endif
   1815 	}
   1816 #ifdef COMPAT_OIFREQ
   1817 	if (cmd != ocmd)
   1818 		ifreqn2o(oifr, ifr);
   1819 #endif
   1820 
   1821 	return error;
   1822 }
   1823 
   1824 /*
   1825  * Return interface configuration
   1826  * of system.  List may be used
   1827  * in later ioctl's (above) to get
   1828  * other information.
   1829  *
   1830  * Each record is a struct ifreq.  Before the addition of
   1831  * sockaddr_storage, the API rule was that sockaddr flavors that did
   1832  * not fit would extend beyond the struct ifreq, with the next struct
   1833  * ifreq starting sa_len beyond the struct sockaddr.  Because the
   1834  * union in struct ifreq includes struct sockaddr_storage, every kind
   1835  * of sockaddr must fit.  Thus, there are no longer any overlength
   1836  * records.
   1837  *
   1838  * Records are added to the user buffer if they fit, and ifc_len is
   1839  * adjusted to the length that was written.  Thus, the user is only
   1840  * assured of getting the complete list if ifc_len on return is at
   1841  * least sizeof(struct ifreq) less than it was on entry.
   1842  *
   1843  * If the user buffer pointer is NULL, this routine copies no data and
   1844  * returns the amount of space that would be needed.
   1845  *
   1846  * Invariants:
   1847  * ifrp points to the next part of the user's buffer to be used.  If
   1848  * ifrp != NULL, space holds the number of bytes remaining that we may
   1849  * write at ifrp.  Otherwise, space holds the number of bytes that
   1850  * would have been written had there been adequate space.
   1851  */
   1852 /*ARGSUSED*/
   1853 int
   1854 ifconf(u_long cmd, void *data)
   1855 {
   1856 	struct ifconf *ifc = (struct ifconf *)data;
   1857 	struct ifnet *ifp;
   1858 	struct ifaddr *ifa;
   1859 	struct ifreq ifr, *ifrp;
   1860 	int space, error = 0;
   1861 	const int sz = (int)sizeof(struct ifreq);
   1862 
   1863 	if ((ifrp = ifc->ifc_req) == NULL)
   1864 		space = 0;
   1865 	else
   1866 		space = ifc->ifc_len;
   1867 	IFNET_FOREACH(ifp) {
   1868 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
   1869 		    sizeof(ifr.ifr_name));
   1870 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
   1871 			return ENAMETOOLONG;
   1872 		if (IFADDR_EMPTY(ifp)) {
   1873 			/* Interface with no addresses - send zero sockaddr. */
   1874 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
   1875 			if (ifrp == NULL) {
   1876 				space += sz;
   1877 				continue;
   1878 			}
   1879 			if (space >= sz) {
   1880 				error = copyout(&ifr, ifrp, sz);
   1881 				if (error != 0)
   1882 					return error;
   1883 				ifrp++;
   1884 				space -= sz;
   1885 			}
   1886 		}
   1887 
   1888 		IFADDR_FOREACH(ifa, ifp) {
   1889 			struct sockaddr *sa = ifa->ifa_addr;
   1890 			/* all sockaddrs must fit in sockaddr_storage */
   1891 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
   1892 
   1893 			if (ifrp == NULL) {
   1894 				space += sz;
   1895 				continue;
   1896 			}
   1897 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
   1898 			if (space >= sz) {
   1899 				error = copyout(&ifr, ifrp, sz);
   1900 				if (error != 0)
   1901 					return (error);
   1902 				ifrp++; space -= sz;
   1903 			}
   1904 		}
   1905 	}
   1906 	if (ifrp != NULL) {
   1907 		KASSERT(0 <= space && space <= ifc->ifc_len);
   1908 		ifc->ifc_len -= space;
   1909 	} else {
   1910 		KASSERT(space >= 0);
   1911 		ifc->ifc_len = space;
   1912 	}
   1913 	return (0);
   1914 }
   1915 
   1916 int
   1917 ifreq_setaddr(const u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
   1918 {
   1919 	uint8_t len;
   1920 	u_long ncmd;
   1921 
   1922 	if ((ncmd = compat_cvtcmd(cmd)) != cmd)
   1923 		len = sizeof(ifr->ifr_addr);
   1924 	else
   1925 		len = sizeof(ifr->ifr_ifru.ifru_space);
   1926 	if (len < sa->sa_len)
   1927 		return EFBIG;
   1928 	memset(&ifr->ifr_addr, 0, len);
   1929 	sockaddr_copy(&ifr->ifr_addr, len, sa);
   1930 	return 0;
   1931 }
   1932 
   1933 /*
   1934  * Queue message on interface, and start output if interface
   1935  * not yet active.
   1936  */
   1937 int
   1938 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
   1939     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
   1940 {
   1941 	int len = m->m_pkthdr.len;
   1942 	int mflags = m->m_flags;
   1943 	int s = splnet();
   1944 	int error;
   1945 
   1946 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
   1947 	if (error != 0)
   1948 		goto out;
   1949 	ifp->if_obytes += len;
   1950 	if (mflags & M_MCAST)
   1951 		ifp->if_omcasts++;
   1952 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
   1953 		(*ifp->if_start)(ifp);
   1954 out:
   1955 	splx(s);
   1956 	return error;
   1957 }
   1958 
   1959 /*
   1960  * Queue message on interface, possibly using a second fast queue
   1961  */
   1962 int
   1963 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
   1964     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
   1965 {
   1966 	int error = 0;
   1967 
   1968 	if (ifq != NULL
   1969 #ifdef ALTQ
   1970 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
   1971 #endif
   1972 	    ) {
   1973 		if (IF_QFULL(ifq)) {
   1974 			IF_DROP(&ifp->if_snd);
   1975 			m_freem(m);
   1976 			if (error == 0)
   1977 				error = ENOBUFS;
   1978 		} else
   1979 			IF_ENQUEUE(ifq, m);
   1980 	} else
   1981 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
   1982 	if (error != 0) {
   1983 		++ifp->if_oerrors;
   1984 		return error;
   1985 	}
   1986 	return 0;
   1987 }
   1988 
   1989 
   1990 static void
   1991 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
   1992     struct ifaltq *ifq)
   1993 {
   1994 	const struct sysctlnode *cnode, *rnode;
   1995 
   1996 	if (sysctl_createv(clog, 0, NULL, &rnode,
   1997 		       CTLFLAG_PERMANENT,
   1998 		       CTLTYPE_NODE, "net", NULL,
   1999 		       NULL, 0, NULL, 0,
   2000 		       CTL_NET, CTL_EOL) != 0)
   2001 		goto bad;
   2002 
   2003 	if (sysctl_createv(clog, 0, &rnode, &rnode,
   2004 		       CTLFLAG_PERMANENT,
   2005 		       CTLTYPE_NODE, "interfaces",
   2006 		       SYSCTL_DESCR("Per-interface controls"),
   2007 		       NULL, 0, NULL, 0,
   2008 		       CTL_CREATE, CTL_EOL) != 0)
   2009 		goto bad;
   2010 
   2011 	if (sysctl_createv(clog, 0, &rnode, &rnode,
   2012 		       CTLFLAG_PERMANENT,
   2013 		       CTLTYPE_NODE, ifname,
   2014 		       SYSCTL_DESCR("Interface controls"),
   2015 		       NULL, 0, NULL, 0,
   2016 		       CTL_CREATE, CTL_EOL) != 0)
   2017 		goto bad;
   2018 
   2019 	if (sysctl_createv(clog, 0, &rnode, &rnode,
   2020 		       CTLFLAG_PERMANENT,
   2021 		       CTLTYPE_NODE, "sndq",
   2022 		       SYSCTL_DESCR("Interface output queue controls"),
   2023 		       NULL, 0, NULL, 0,
   2024 		       CTL_CREATE, CTL_EOL) != 0)
   2025 		goto bad;
   2026 
   2027 	if (sysctl_createv(clog, 0, &rnode, &cnode,
   2028 		       CTLFLAG_PERMANENT,
   2029 		       CTLTYPE_INT, "len",
   2030 		       SYSCTL_DESCR("Current output queue length"),
   2031 		       NULL, 0, &ifq->ifq_len, 0,
   2032 		       CTL_CREATE, CTL_EOL) != 0)
   2033 		goto bad;
   2034 
   2035 	if (sysctl_createv(clog, 0, &rnode, &cnode,
   2036 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2037 		       CTLTYPE_INT, "maxlen",
   2038 		       SYSCTL_DESCR("Maximum allowed output queue length"),
   2039 		       NULL, 0, &ifq->ifq_maxlen, 0,
   2040 		       CTL_CREATE, CTL_EOL) != 0)
   2041 		goto bad;
   2042 
   2043 	if (sysctl_createv(clog, 0, &rnode, &cnode,
   2044 		       CTLFLAG_PERMANENT,
   2045 		       CTLTYPE_INT, "drops",
   2046 		       SYSCTL_DESCR("Packets dropped due to full output queue"),
   2047 		       NULL, 0, &ifq->ifq_drops, 0,
   2048 		       CTL_CREATE, CTL_EOL) != 0)
   2049 		goto bad;
   2050 
   2051 	return;
   2052 bad:
   2053 	printf("%s: could not attach sysctl nodes\n", ifname);
   2054 	return;
   2055 }
   2056 
   2057 #if defined(INET) || defined(INET6)
   2058 static void
   2059 sysctl_net_ifq_setup(struct sysctllog **clog,
   2060 		     int pf, const char *pfname,
   2061 		     int ipn, const char *ipname,
   2062 		     int qid, struct ifqueue *ifq)
   2063 {
   2064 
   2065 	sysctl_createv(clog, 0, NULL, NULL,
   2066 		       CTLFLAG_PERMANENT,
   2067 		       CTLTYPE_NODE, "net", NULL,
   2068 		       NULL, 0, NULL, 0,
   2069 		       CTL_NET, CTL_EOL);
   2070 	sysctl_createv(clog, 0, NULL, NULL,
   2071 		       CTLFLAG_PERMANENT,
   2072 		       CTLTYPE_NODE, pfname, NULL,
   2073 		       NULL, 0, NULL, 0,
   2074 		       CTL_NET, pf, CTL_EOL);
   2075 	sysctl_createv(clog, 0, NULL, NULL,
   2076 		       CTLFLAG_PERMANENT,
   2077 		       CTLTYPE_NODE, ipname, NULL,
   2078 		       NULL, 0, NULL, 0,
   2079 		       CTL_NET, pf, ipn, CTL_EOL);
   2080 	sysctl_createv(clog, 0, NULL, NULL,
   2081 		       CTLFLAG_PERMANENT,
   2082 		       CTLTYPE_NODE, "ifq",
   2083 		       SYSCTL_DESCR("Protocol input queue controls"),
   2084 		       NULL, 0, NULL, 0,
   2085 		       CTL_NET, pf, ipn, qid, CTL_EOL);
   2086 
   2087 	sysctl_createv(clog, 0, NULL, NULL,
   2088 		       CTLFLAG_PERMANENT,
   2089 		       CTLTYPE_INT, "len",
   2090 		       SYSCTL_DESCR("Current input queue length"),
   2091 		       NULL, 0, &ifq->ifq_len, 0,
   2092 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
   2093 	sysctl_createv(clog, 0, NULL, NULL,
   2094 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2095 		       CTLTYPE_INT, "maxlen",
   2096 		       SYSCTL_DESCR("Maximum allowed input queue length"),
   2097 		       NULL, 0, &ifq->ifq_maxlen, 0,
   2098 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
   2099 #ifdef notyet
   2100 	sysctl_createv(clog, 0, NULL, NULL,
   2101 		       CTLFLAG_PERMANENT,
   2102 		       CTLTYPE_INT, "peak",
   2103 		       SYSCTL_DESCR("Highest input queue length"),
   2104 		       NULL, 0, &ifq->ifq_peak, 0,
   2105 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
   2106 #endif
   2107 	sysctl_createv(clog, 0, NULL, NULL,
   2108 		       CTLFLAG_PERMANENT,
   2109 		       CTLTYPE_INT, "drops",
   2110 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
   2111 		       NULL, 0, &ifq->ifq_drops, 0,
   2112 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
   2113 }
   2114 #endif /* INET || INET6 */
   2115