Home | History | Annotate | Line # | Download | only in net
if.c revision 1.264
      1 /*	$NetBSD: if.c,v 1.264 2013/06/20 13:56:29 roy 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 1.264 2013/06/20 13:56:29 roy 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 #include <sys/kmem.h>
    116 
    117 #include <net/if.h>
    118 #include <net/if_dl.h>
    119 #include <net/if_ether.h>
    120 #include <net/if_media.h>
    121 #include <net80211/ieee80211.h>
    122 #include <net80211/ieee80211_ioctl.h>
    123 #include <net/if_types.h>
    124 #include <net/radix.h>
    125 #include <net/route.h>
    126 #include <net/netisr.h>
    127 #include <sys/module.h>
    128 #ifdef NETATALK
    129 #include <netatalk/at_extern.h>
    130 #include <netatalk/at.h>
    131 #endif
    132 #include <net/pfil.h>
    133 
    134 #ifdef INET6
    135 #include <netinet/in.h>
    136 #include <netinet6/in6_var.h>
    137 #include <netinet6/nd6.h>
    138 #endif
    139 
    140 #include "carp.h"
    141 #if NCARP > 0
    142 #include <netinet/ip_carp.h>
    143 #endif
    144 
    145 #include <compat/sys/sockio.h>
    146 #include <compat/sys/socket.h>
    147 
    148 MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
    149 MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
    150 
    151 int	ifqmaxlen = IFQ_MAXLEN;
    152 callout_t if_slowtimo_ch;
    153 
    154 int netisr;			/* scheduling bits for network */
    155 
    156 static int	if_rt_walktree(struct rtentry *, void *);
    157 
    158 static struct if_clone *if_clone_lookup(const char *, int *);
    159 static int	if_clone_list(struct if_clonereq *);
    160 
    161 static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
    162 static int if_cloners_count;
    163 
    164 static uint64_t index_gen;
    165 static kmutex_t index_gen_mtx;
    166 
    167 #ifdef PFIL_HOOKS
    168 struct pfil_head if_pfil;	/* packet filtering hook for interfaces */
    169 #endif
    170 
    171 static kauth_listener_t if_listener;
    172 
    173 static int ifioctl_attach(struct ifnet *);
    174 static void ifioctl_detach(struct ifnet *);
    175 static void ifnet_lock_enter(struct ifnet_lock *);
    176 static void ifnet_lock_exit(struct ifnet_lock *);
    177 static void if_detach_queues(struct ifnet *, struct ifqueue *);
    178 static void sysctl_sndq_setup(struct sysctllog **, const char *,
    179     struct ifaltq *);
    180 
    181 #if defined(INET) || defined(INET6)
    182 static void sysctl_net_ifq_setup(struct sysctllog **, int, const char *,
    183 				 int, const char *, int, struct ifqueue *);
    184 #endif
    185 
    186 static int
    187 if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
    188     void *arg0, void *arg1, void *arg2, void *arg3)
    189 {
    190 	int result;
    191 	enum kauth_network_req req;
    192 
    193 	result = KAUTH_RESULT_DEFER;
    194 	req = (enum kauth_network_req)arg1;
    195 
    196 	if (action != KAUTH_NETWORK_INTERFACE)
    197 		return result;
    198 
    199 	if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
    200 	    (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
    201 		result = KAUTH_RESULT_ALLOW;
    202 
    203 	return result;
    204 }
    205 
    206 /*
    207  * Network interface utility routines.
    208  *
    209  * Routines with ifa_ifwith* names take sockaddr *'s as
    210  * parameters.
    211  */
    212 void
    213 ifinit(void)
    214 {
    215 #ifdef INET
    216 	{extern struct ifqueue ipintrq;
    217 	sysctl_net_ifq_setup(NULL, PF_INET, "inet", IPPROTO_IP, "ip",
    218 			     IPCTL_IFQ, &ipintrq);}
    219 #endif /* INET */
    220 #ifdef INET6
    221 	{extern struct ifqueue ip6intrq;
    222 	sysctl_net_ifq_setup(NULL, PF_INET6, "inet6", IPPROTO_IPV6, "ip6",
    223 			     IPV6CTL_IFQ, &ip6intrq);}
    224 #endif /* INET6 */
    225 
    226 	callout_init(&if_slowtimo_ch, 0);
    227 	if_slowtimo(NULL);
    228 
    229 	if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
    230 	    if_listener_cb, NULL);
    231 }
    232 
    233 /*
    234  * XXX Initialization before configure().
    235  * XXX hack to get pfil_add_hook working in autoconf.
    236  */
    237 void
    238 ifinit1(void)
    239 {
    240 
    241 	mutex_init(&index_gen_mtx, MUTEX_DEFAULT, IPL_NONE);
    242 
    243 #ifdef PFIL_HOOKS
    244 	if_pfil.ph_type = PFIL_TYPE_IFNET;
    245 	if_pfil.ph_ifnet = NULL;
    246 	if (pfil_head_register(&if_pfil) != 0)
    247 		printf("WARNING: unable to register pfil hook\n");
    248 #endif
    249 }
    250 
    251 struct ifnet *
    252 if_alloc(u_char type)
    253 {
    254 	return malloc(sizeof(struct ifnet), M_DEVBUF, M_WAITOK|M_ZERO);
    255 }
    256 
    257 void
    258 if_free(struct ifnet *ifp)
    259 {
    260 	free(ifp, M_DEVBUF);
    261 }
    262 
    263 void
    264 if_initname(struct ifnet *ifp, const char *name, int unit)
    265 {
    266 	(void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
    267 	    "%s%d", name, unit);
    268 }
    269 
    270 /*
    271  * Null routines used while an interface is going away.  These routines
    272  * just return an error.
    273  */
    274 
    275 int
    276 if_nulloutput(struct ifnet *ifp, struct mbuf *m,
    277     const struct sockaddr *so, struct rtentry *rt)
    278 {
    279 
    280 	return ENXIO;
    281 }
    282 
    283 void
    284 if_nullinput(struct ifnet *ifp, struct mbuf *m)
    285 {
    286 
    287 	/* Nothing. */
    288 }
    289 
    290 void
    291 if_nullstart(struct ifnet *ifp)
    292 {
    293 
    294 	/* Nothing. */
    295 }
    296 
    297 int
    298 if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
    299 {
    300 
    301 	/* Wake ifioctl_detach(), who may wait for all threads to
    302 	 * quit the critical section.
    303 	 */
    304 	cv_signal(&ifp->if_ioctl_lock->il_emptied);
    305 	return ENXIO;
    306 }
    307 
    308 int
    309 if_nullinit(struct ifnet *ifp)
    310 {
    311 
    312 	return ENXIO;
    313 }
    314 
    315 void
    316 if_nullstop(struct ifnet *ifp, int disable)
    317 {
    318 
    319 	/* Nothing. */
    320 }
    321 
    322 void
    323 if_nullwatchdog(struct ifnet *ifp)
    324 {
    325 
    326 	/* Nothing. */
    327 }
    328 
    329 void
    330 if_nulldrain(struct ifnet *ifp)
    331 {
    332 
    333 	/* Nothing. */
    334 }
    335 
    336 static u_int if_index = 1;
    337 struct ifnet_head ifnet;
    338 size_t if_indexlim = 0;
    339 struct ifaddr **ifnet_addrs = NULL;
    340 struct ifnet **ifindex2ifnet = NULL;
    341 struct ifnet *lo0ifp;
    342 
    343 void
    344 if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory)
    345 {
    346 	struct ifaddr *ifa;
    347 	struct sockaddr_dl *sdl;
    348 
    349 	ifp->if_addrlen = addrlen;
    350 	if_alloc_sadl(ifp);
    351 	ifa = ifp->if_dl;
    352 	sdl = satosdl(ifa->ifa_addr);
    353 
    354 	(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
    355 	if (factory) {
    356 		ifp->if_hwdl = ifp->if_dl;
    357 		IFAREF(ifp->if_hwdl);
    358 	}
    359 	/* TBD routing socket */
    360 }
    361 
    362 struct ifaddr *
    363 if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
    364 {
    365 	unsigned socksize, ifasize;
    366 	int addrlen, namelen;
    367 	struct sockaddr_dl *mask, *sdl;
    368 	struct ifaddr *ifa;
    369 
    370 	namelen = strlen(ifp->if_xname);
    371 	addrlen = ifp->if_addrlen;
    372 	socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
    373 	ifasize = sizeof(*ifa) + 2 * socksize;
    374 	ifa = (struct ifaddr *)malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
    375 
    376 	sdl = (struct sockaddr_dl *)(ifa + 1);
    377 	mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
    378 
    379 	sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
    380 	    ifp->if_xname, namelen, NULL, addrlen);
    381 	mask->sdl_len = sockaddr_dl_measure(namelen, 0);
    382 	memset(&mask->sdl_data[0], 0xff, namelen);
    383 	ifa->ifa_rtrequest = link_rtrequest;
    384 	ifa->ifa_addr = (struct sockaddr *)sdl;
    385 	ifa->ifa_netmask = (struct sockaddr *)mask;
    386 
    387 	*sdlp = sdl;
    388 
    389 	return ifa;
    390 }
    391 
    392 static void
    393 if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa)
    394 {
    395 	const struct sockaddr_dl *sdl;
    396 	ifnet_addrs[ifp->if_index] = ifa;
    397 	IFAREF(ifa);
    398 	ifp->if_dl = ifa;
    399 	IFAREF(ifa);
    400 	sdl = satosdl(ifa->ifa_addr);
    401 	ifp->if_sadl = sdl;
    402 }
    403 
    404 /*
    405  * Allocate the link level name for the specified interface.  This
    406  * is an attachment helper.  It must be called after ifp->if_addrlen
    407  * is initialized, which may not be the case when if_attach() is
    408  * called.
    409  */
    410 void
    411 if_alloc_sadl(struct ifnet *ifp)
    412 {
    413 	struct ifaddr *ifa;
    414 	const struct sockaddr_dl *sdl;
    415 
    416 	/*
    417 	 * If the interface already has a link name, release it
    418 	 * now.  This is useful for interfaces that can change
    419 	 * link types, and thus switch link names often.
    420 	 */
    421 	if (ifp->if_sadl != NULL)
    422 		if_free_sadl(ifp);
    423 
    424 	ifa = if_dl_create(ifp, &sdl);
    425 
    426 	ifa_insert(ifp, ifa);
    427 	if_sadl_setrefs(ifp, ifa);
    428 }
    429 
    430 static void
    431 if_deactivate_sadl(struct ifnet *ifp)
    432 {
    433 	struct ifaddr *ifa;
    434 
    435 	KASSERT(ifp->if_dl != NULL);
    436 
    437 	ifa = ifp->if_dl;
    438 
    439 	ifp->if_sadl = NULL;
    440 
    441 	ifnet_addrs[ifp->if_index] = NULL;
    442 	IFAFREE(ifa);
    443 	ifp->if_dl = NULL;
    444 	IFAFREE(ifa);
    445 }
    446 
    447 void
    448 if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa,
    449     const struct sockaddr_dl *sdl)
    450 {
    451 	int s;
    452 
    453 	s = splnet();
    454 
    455 	if_deactivate_sadl(ifp);
    456 
    457 	if_sadl_setrefs(ifp, ifa);
    458 	IFADDR_FOREACH(ifa, ifp)
    459 		rtinit(ifa, RTM_LLINFO_UPD, 0);
    460 	splx(s);
    461 }
    462 
    463 /*
    464  * Free the link level name for the specified interface.  This is
    465  * a detach helper.  This is called from if_detach() or from
    466  * link layer type specific detach functions.
    467  */
    468 void
    469 if_free_sadl(struct ifnet *ifp)
    470 {
    471 	struct ifaddr *ifa;
    472 	int s;
    473 
    474 	ifa = ifnet_addrs[ifp->if_index];
    475 	if (ifa == NULL) {
    476 		KASSERT(ifp->if_sadl == NULL);
    477 		KASSERT(ifp->if_dl == NULL);
    478 		return;
    479 	}
    480 
    481 	KASSERT(ifp->if_sadl != NULL);
    482 	KASSERT(ifp->if_dl != NULL);
    483 
    484 	s = splnet();
    485 	rtinit(ifa, RTM_DELETE, 0);
    486 	ifa_remove(ifp, ifa);
    487 	if_deactivate_sadl(ifp);
    488 	if (ifp->if_hwdl == ifa) {
    489 		IFAFREE(ifa);
    490 		ifp->if_hwdl = NULL;
    491 	}
    492 	splx(s);
    493 }
    494 
    495 /*
    496  * Attach an interface to the
    497  * list of "active" interfaces.
    498  */
    499 void
    500 if_attach(struct ifnet *ifp)
    501 {
    502 	int indexlim = 0;
    503 
    504 	if (if_indexlim == 0) {
    505 		TAILQ_INIT(&ifnet);
    506 		if_indexlim = 8;
    507 	}
    508 	TAILQ_INIT(&ifp->if_addrlist);
    509 	TAILQ_INSERT_TAIL(&ifnet, ifp, if_list);
    510 
    511 	if (ifioctl_attach(ifp) != 0)
    512 		panic("%s: ifioctl_attach() failed", __func__);
    513 
    514 	mutex_enter(&index_gen_mtx);
    515 	ifp->if_index_gen = index_gen++;
    516 	mutex_exit(&index_gen_mtx);
    517 
    518 	ifp->if_index = if_index;
    519 	if (ifindex2ifnet == NULL)
    520 		if_index++;
    521 	else
    522 		while (ifp->if_index < if_indexlim &&
    523 		    ifindex2ifnet[ifp->if_index] != NULL) {
    524 			++if_index;
    525 			if (if_index == 0)
    526 				if_index = 1;
    527 			/*
    528 			 * If we hit USHRT_MAX, we skip back to 0 since
    529 			 * there are a number of places where the value
    530 			 * of if_index or if_index itself is compared
    531 			 * to or stored in an unsigned short.  By
    532 			 * jumping back, we won't botch those assignments
    533 			 * or comparisons.
    534 			 */
    535 			else if (if_index == USHRT_MAX) {
    536 				/*
    537 				 * However, if we have to jump back to
    538 				 * zero *twice* without finding an empty
    539 				 * slot in ifindex2ifnet[], then there
    540 				 * there are too many (>65535) interfaces.
    541 				 */
    542 				if (indexlim++)
    543 					panic("too many interfaces");
    544 				else
    545 					if_index = 1;
    546 			}
    547 			ifp->if_index = if_index;
    548 		}
    549 
    550 	/*
    551 	 * We have some arrays that should be indexed by if_index.
    552 	 * since if_index will grow dynamically, they should grow too.
    553 	 *	struct ifadd **ifnet_addrs
    554 	 *	struct ifnet **ifindex2ifnet
    555 	 */
    556 	if (ifnet_addrs == NULL || ifindex2ifnet == NULL ||
    557 	    ifp->if_index >= if_indexlim) {
    558 		size_t m, n, oldlim;
    559 		void *q;
    560 
    561 		oldlim = if_indexlim;
    562 		while (ifp->if_index >= if_indexlim)
    563 			if_indexlim <<= 1;
    564 
    565 		/* grow ifnet_addrs */
    566 		m = oldlim * sizeof(struct ifaddr *);
    567 		n = if_indexlim * sizeof(struct ifaddr *);
    568 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
    569 		if (ifnet_addrs != NULL) {
    570 			memcpy(q, ifnet_addrs, m);
    571 			free(ifnet_addrs, M_IFADDR);
    572 		}
    573 		ifnet_addrs = (struct ifaddr **)q;
    574 
    575 		/* grow ifindex2ifnet */
    576 		m = oldlim * sizeof(struct ifnet *);
    577 		n = if_indexlim * sizeof(struct ifnet *);
    578 		q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
    579 		if (ifindex2ifnet != NULL) {
    580 			memcpy(q, ifindex2ifnet, m);
    581 			free(ifindex2ifnet, M_IFADDR);
    582 		}
    583 		ifindex2ifnet = (struct ifnet **)q;
    584 	}
    585 
    586 	ifindex2ifnet[ifp->if_index] = ifp;
    587 
    588 	/*
    589 	 * Link level name is allocated later by a separate call to
    590 	 * if_alloc_sadl().
    591 	 */
    592 
    593 	if (ifp->if_snd.ifq_maxlen == 0)
    594 		ifp->if_snd.ifq_maxlen = ifqmaxlen;
    595 
    596 	sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
    597 
    598 	ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
    599 
    600 	ifp->if_link_state = LINK_STATE_UNKNOWN;
    601 
    602 	ifp->if_capenable = 0;
    603 	ifp->if_csum_flags_tx = 0;
    604 	ifp->if_csum_flags_rx = 0;
    605 
    606 #ifdef ALTQ
    607 	ifp->if_snd.altq_type = 0;
    608 	ifp->if_snd.altq_disc = NULL;
    609 	ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
    610 	ifp->if_snd.altq_tbr  = NULL;
    611 	ifp->if_snd.altq_ifp  = ifp;
    612 #endif
    613 
    614 #ifdef PFIL_HOOKS
    615 	ifp->if_pfil.ph_type = PFIL_TYPE_IFNET;
    616 	ifp->if_pfil.ph_ifnet = ifp;
    617 	if (pfil_head_register(&ifp->if_pfil) != 0)
    618 		printf("%s: WARNING: unable to register pfil hook\n",
    619 		    ifp->if_xname);
    620 	(void)pfil_run_hooks(&if_pfil,
    621 	    (struct mbuf **)PFIL_IFNET_ATTACH, ifp, PFIL_IFNET);
    622 #endif
    623 
    624 	if (!STAILQ_EMPTY(&domains))
    625 		if_attachdomain1(ifp);
    626 
    627 	/* Announce the interface. */
    628 	rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
    629 }
    630 
    631 void
    632 if_attachdomain(void)
    633 {
    634 	struct ifnet *ifp;
    635 	int s;
    636 
    637 	s = splnet();
    638 	IFNET_FOREACH(ifp)
    639 		if_attachdomain1(ifp);
    640 	splx(s);
    641 }
    642 
    643 void
    644 if_attachdomain1(struct ifnet *ifp)
    645 {
    646 	struct domain *dp;
    647 	int s;
    648 
    649 	s = splnet();
    650 
    651 	/* address family dependent data region */
    652 	memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
    653 	DOMAIN_FOREACH(dp) {
    654 		if (dp->dom_ifattach != NULL)
    655 			ifp->if_afdata[dp->dom_family] =
    656 			    (*dp->dom_ifattach)(ifp);
    657 	}
    658 
    659 	splx(s);
    660 }
    661 
    662 /*
    663  * Deactivate an interface.  This points all of the procedure
    664  * handles at error stubs.  May be called from interrupt context.
    665  */
    666 void
    667 if_deactivate(struct ifnet *ifp)
    668 {
    669 	int s;
    670 
    671 	s = splnet();
    672 
    673 	ifp->if_output	 = if_nulloutput;
    674 	ifp->if_input	 = if_nullinput;
    675 	ifp->if_start	 = if_nullstart;
    676 	ifp->if_ioctl	 = if_nullioctl;
    677 	ifp->if_init	 = if_nullinit;
    678 	ifp->if_stop	 = if_nullstop;
    679 	ifp->if_watchdog = if_nullwatchdog;
    680 	ifp->if_drain	 = if_nulldrain;
    681 
    682 	/* No more packets may be enqueued. */
    683 	ifp->if_snd.ifq_maxlen = 0;
    684 
    685 	splx(s);
    686 }
    687 
    688 void
    689 if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
    690 {
    691 	struct ifaddr *ifa, *nifa;
    692 
    693 	for (ifa = IFADDR_FIRST(ifp); ifa != NULL; ifa = nifa) {
    694 		nifa = IFADDR_NEXT(ifa);
    695 		if (ifa->ifa_addr->sa_family != family)
    696 			continue;
    697 		(*purgeaddr)(ifa);
    698 	}
    699 }
    700 
    701 /*
    702  * Detach an interface from the list of "active" interfaces,
    703  * freeing any resources as we go along.
    704  *
    705  * NOTE: This routine must be called with a valid thread context,
    706  * as it may block.
    707  */
    708 void
    709 if_detach(struct ifnet *ifp)
    710 {
    711 	struct socket so;
    712 	struct ifaddr *ifa;
    713 #ifdef IFAREF_DEBUG
    714 	struct ifaddr *last_ifa = NULL;
    715 #endif
    716 	struct domain *dp;
    717 	const struct protosw *pr;
    718 	int s, i, family, purged;
    719 
    720 	/*
    721 	 * XXX It's kind of lame that we have to have the
    722 	 * XXX socket structure...
    723 	 */
    724 	memset(&so, 0, sizeof(so));
    725 
    726 	s = splnet();
    727 
    728 	/*
    729 	 * Do an if_down() to give protocols a chance to do something.
    730 	 */
    731 	if_down(ifp);
    732 
    733 #ifdef ALTQ
    734 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
    735 		altq_disable(&ifp->if_snd);
    736 	if (ALTQ_IS_ATTACHED(&ifp->if_snd))
    737 		altq_detach(&ifp->if_snd);
    738 #endif
    739 
    740 	sysctl_teardown(&ifp->if_sysctl_log);
    741 
    742 #if NCARP > 0
    743 	/* Remove the interface from any carp group it is a part of.  */
    744 	if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
    745 		carp_ifdetach(ifp);
    746 #endif
    747 
    748 	/*
    749 	 * Rip all the addresses off the interface.  This should make
    750 	 * all of the routes go away.
    751 	 *
    752 	 * pr_usrreq calls can remove an arbitrary number of ifaddrs
    753 	 * from the list, including our "cursor", ifa.  For safety,
    754 	 * and to honor the TAILQ abstraction, I just restart the
    755 	 * loop after each removal.  Note that the loop will exit
    756 	 * when all of the remaining ifaddrs belong to the AF_LINK
    757 	 * family.  I am counting on the historical fact that at
    758 	 * least one pr_usrreq in each address domain removes at
    759 	 * least one ifaddr.
    760 	 */
    761 again:
    762 	IFADDR_FOREACH(ifa, ifp) {
    763 		family = ifa->ifa_addr->sa_family;
    764 #ifdef IFAREF_DEBUG
    765 		printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
    766 		    ifa, family, ifa->ifa_refcnt);
    767 		if (last_ifa != NULL && ifa == last_ifa)
    768 			panic("if_detach: loop detected");
    769 		last_ifa = ifa;
    770 #endif
    771 		if (family == AF_LINK)
    772 			continue;
    773 		dp = pffinddomain(family);
    774 #ifdef DIAGNOSTIC
    775 		if (dp == NULL)
    776 			panic("if_detach: no domain for AF %d",
    777 			    family);
    778 #endif
    779 		/*
    780 		 * XXX These PURGEIF calls are redundant with the
    781 		 * purge-all-families calls below, but are left in for
    782 		 * now both to make a smaller change, and to avoid
    783 		 * unplanned interactions with clearing of
    784 		 * ifp->if_addrlist.
    785 		 */
    786 		purged = 0;
    787 		for (pr = dp->dom_protosw;
    788 		     pr < dp->dom_protoswNPROTOSW; pr++) {
    789 			so.so_proto = pr;
    790 			if (pr->pr_usrreq != NULL) {
    791 				(void) (*pr->pr_usrreq)(&so,
    792 				    PRU_PURGEIF, NULL, NULL,
    793 				    (struct mbuf *) ifp, curlwp);
    794 				purged = 1;
    795 			}
    796 		}
    797 		if (purged == 0) {
    798 			/*
    799 			 * XXX What's really the best thing to do
    800 			 * XXX here?  --thorpej (at) NetBSD.org
    801 			 */
    802 			printf("if_detach: WARNING: AF %d not purged\n",
    803 			    family);
    804 			ifa_remove(ifp, ifa);
    805 		}
    806 		goto again;
    807 	}
    808 
    809 	if_free_sadl(ifp);
    810 
    811 	/* Walk the routing table looking for stragglers. */
    812 	for (i = 0; i <= AF_MAX; i++) {
    813 		while (rt_walktree(i, if_rt_walktree, ifp) == ERESTART)
    814 			continue;
    815 	}
    816 
    817 	DOMAIN_FOREACH(dp) {
    818 		if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
    819 		{
    820 			void *p = ifp->if_afdata[dp->dom_family];
    821 			if (p) {
    822 				ifp->if_afdata[dp->dom_family] = NULL;
    823 				(*dp->dom_ifdetach)(ifp, p);
    824 			}
    825 		}
    826 
    827 		/*
    828 		 * One would expect multicast memberships (INET and
    829 		 * INET6) on UDP sockets to be purged by the PURGEIF
    830 		 * calls above, but if all addresses were removed from
    831 		 * the interface prior to destruction, the calls will
    832 		 * not be made (e.g. ppp, for which pppd(8) generally
    833 		 * removes addresses before destroying the interface).
    834 		 * Because there is no invariant that multicast
    835 		 * memberships only exist for interfaces with IPv4
    836 		 * addresses, we must call PURGEIF regardless of
    837 		 * addresses.  (Protocols which might store ifnet
    838 		 * pointers are marked with PR_PURGEIF.)
    839 		 */
    840 		for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
    841 			so.so_proto = pr;
    842 			if (pr->pr_usrreq != NULL && pr->pr_flags & PR_PURGEIF)
    843 				(void)(*pr->pr_usrreq)(&so, PRU_PURGEIF, NULL,
    844 				    NULL, (struct mbuf *)ifp, curlwp);
    845 		}
    846 	}
    847 
    848 #ifdef PFIL_HOOKS
    849 	(void)pfil_run_hooks(&if_pfil,
    850 	    (struct mbuf **)PFIL_IFNET_DETACH, ifp, PFIL_IFNET);
    851 	(void)pfil_head_unregister(&ifp->if_pfil);
    852 #endif
    853 
    854 	/* Announce that the interface is gone. */
    855 	rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
    856 
    857 	ifindex2ifnet[ifp->if_index] = NULL;
    858 
    859 	TAILQ_REMOVE(&ifnet, ifp, if_list);
    860 
    861 	ifioctl_detach(ifp);
    862 
    863 	/*
    864 	 * remove packets that came from ifp, from software interrupt queues.
    865 	 */
    866 	DOMAIN_FOREACH(dp) {
    867 		for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
    868 			struct ifqueue *iq = dp->dom_ifqueues[i];
    869 			if (iq == NULL)
    870 				break;
    871 			dp->dom_ifqueues[i] = NULL;
    872 			if_detach_queues(ifp, iq);
    873 		}
    874 	}
    875 
    876 	splx(s);
    877 }
    878 
    879 static void
    880 if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
    881 {
    882 	struct mbuf *m, *prev, *next;
    883 
    884 	prev = NULL;
    885 	for (m = q->ifq_head; m != NULL; m = next) {
    886 		next = m->m_nextpkt;
    887 #ifdef DIAGNOSTIC
    888 		if ((m->m_flags & M_PKTHDR) == 0) {
    889 			prev = m;
    890 			continue;
    891 		}
    892 #endif
    893 		if (m->m_pkthdr.rcvif != ifp) {
    894 			prev = m;
    895 			continue;
    896 		}
    897 
    898 		if (prev != NULL)
    899 			prev->m_nextpkt = m->m_nextpkt;
    900 		else
    901 			q->ifq_head = m->m_nextpkt;
    902 		if (q->ifq_tail == m)
    903 			q->ifq_tail = prev;
    904 		q->ifq_len--;
    905 
    906 		m->m_nextpkt = NULL;
    907 		m_freem(m);
    908 		IF_DROP(q);
    909 	}
    910 }
    911 
    912 /*
    913  * Callback for a radix tree walk to delete all references to an
    914  * ifnet.
    915  */
    916 static int
    917 if_rt_walktree(struct rtentry *rt, void *v)
    918 {
    919 	struct ifnet *ifp = (struct ifnet *)v;
    920 	int error;
    921 
    922 	if (rt->rt_ifp != ifp)
    923 		return 0;
    924 
    925 	/* Delete the entry. */
    926 	++rt->rt_refcnt;
    927 	error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
    928 	    rt_mask(rt), rt->rt_flags, NULL);
    929 	KASSERT((rt->rt_flags & RTF_UP) == 0);
    930 	rt->rt_ifp = NULL;
    931 	RTFREE(rt);
    932 	if (error != 0)
    933 		printf("%s: warning: unable to delete rtentry @ %p, "
    934 		    "error = %d\n", ifp->if_xname, rt, error);
    935 	return ERESTART;
    936 }
    937 
    938 /*
    939  * Create a clone network interface.
    940  */
    941 int
    942 if_clone_create(const char *name)
    943 {
    944 	struct if_clone *ifc;
    945 	int unit;
    946 
    947 	ifc = if_clone_lookup(name, &unit);
    948 	if (ifc == NULL)
    949 		return EINVAL;
    950 
    951 	if (ifunit(name) != NULL)
    952 		return EEXIST;
    953 
    954 	return (*ifc->ifc_create)(ifc, unit);
    955 }
    956 
    957 /*
    958  * Destroy a clone network interface.
    959  */
    960 int
    961 if_clone_destroy(const char *name)
    962 {
    963 	struct if_clone *ifc;
    964 	struct ifnet *ifp;
    965 
    966 	ifc = if_clone_lookup(name, NULL);
    967 	if (ifc == NULL)
    968 		return EINVAL;
    969 
    970 	ifp = ifunit(name);
    971 	if (ifp == NULL)
    972 		return ENXIO;
    973 
    974 	if (ifc->ifc_destroy == NULL)
    975 		return EOPNOTSUPP;
    976 
    977 	return (*ifc->ifc_destroy)(ifp);
    978 }
    979 
    980 /*
    981  * Look up a network interface cloner.
    982  */
    983 static struct if_clone *
    984 if_clone_lookup(const char *name, int *unitp)
    985 {
    986 	struct if_clone *ifc;
    987 	const char *cp;
    988 	char *dp, ifname[IFNAMSIZ + 3];
    989 	int unit;
    990 
    991 	strcpy(ifname, "if_");
    992 	/* separate interface name from unit */
    993 	for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ &&
    994 	    *cp && (*cp < '0' || *cp > '9');)
    995 		*dp++ = *cp++;
    996 
    997 	if (cp == name || cp - name == IFNAMSIZ || !*cp)
    998 		return NULL;	/* No name or unit number */
    999 	*dp++ = '\0';
   1000 
   1001 again:
   1002 	LIST_FOREACH(ifc, &if_cloners, ifc_list) {
   1003 		if (strcmp(ifname + 3, ifc->ifc_name) == 0)
   1004 			break;
   1005 	}
   1006 
   1007 	if (ifc == NULL) {
   1008 		if (*ifname == '\0' ||
   1009 		    module_autoload(ifname, MODULE_CLASS_DRIVER))
   1010 			return NULL;
   1011 		*ifname = '\0';
   1012 		goto again;
   1013 	}
   1014 
   1015 	unit = 0;
   1016 	while (cp - name < IFNAMSIZ && *cp) {
   1017 		if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) {
   1018 			/* Bogus unit number. */
   1019 			return NULL;
   1020 		}
   1021 		unit = (unit * 10) + (*cp++ - '0');
   1022 	}
   1023 
   1024 	if (unitp != NULL)
   1025 		*unitp = unit;
   1026 	return ifc;
   1027 }
   1028 
   1029 /*
   1030  * Register a network interface cloner.
   1031  */
   1032 void
   1033 if_clone_attach(struct if_clone *ifc)
   1034 {
   1035 
   1036 	LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
   1037 	if_cloners_count++;
   1038 }
   1039 
   1040 /*
   1041  * Unregister a network interface cloner.
   1042  */
   1043 void
   1044 if_clone_detach(struct if_clone *ifc)
   1045 {
   1046 
   1047 	LIST_REMOVE(ifc, ifc_list);
   1048 	if_cloners_count--;
   1049 }
   1050 
   1051 /*
   1052  * Provide list of interface cloners to userspace.
   1053  */
   1054 static int
   1055 if_clone_list(struct if_clonereq *ifcr)
   1056 {
   1057 	char outbuf[IFNAMSIZ], *dst;
   1058 	struct if_clone *ifc;
   1059 	int count, error = 0;
   1060 
   1061 	ifcr->ifcr_total = if_cloners_count;
   1062 	if ((dst = ifcr->ifcr_buffer) == NULL) {
   1063 		/* Just asking how many there are. */
   1064 		return 0;
   1065 	}
   1066 
   1067 	if (ifcr->ifcr_count < 0)
   1068 		return EINVAL;
   1069 
   1070 	count = (if_cloners_count < ifcr->ifcr_count) ?
   1071 	    if_cloners_count : ifcr->ifcr_count;
   1072 
   1073 	for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
   1074 	     ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
   1075 		(void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
   1076 		if (outbuf[sizeof(outbuf) - 1] != '\0')
   1077 			return ENAMETOOLONG;
   1078 		error = copyout(outbuf, dst, sizeof(outbuf));
   1079 		if (error != 0)
   1080 			break;
   1081 	}
   1082 
   1083 	return error;
   1084 }
   1085 
   1086 void
   1087 ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
   1088 {
   1089 	ifa->ifa_ifp = ifp;
   1090 	TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
   1091 	IFAREF(ifa);
   1092 }
   1093 
   1094 void
   1095 ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
   1096 {
   1097 	KASSERT(ifa->ifa_ifp == ifp);
   1098 	TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
   1099 	IFAFREE(ifa);
   1100 }
   1101 
   1102 static inline int
   1103 equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
   1104 {
   1105 	return sockaddr_cmp(sa1, sa2) == 0;
   1106 }
   1107 
   1108 /*
   1109  * Locate an interface based on a complete address.
   1110  */
   1111 /*ARGSUSED*/
   1112 struct ifaddr *
   1113 ifa_ifwithaddr(const struct sockaddr *addr)
   1114 {
   1115 	struct ifnet *ifp;
   1116 	struct ifaddr *ifa;
   1117 
   1118 	IFNET_FOREACH(ifp) {
   1119 		if (ifp->if_output == if_nulloutput)
   1120 			continue;
   1121 		IFADDR_FOREACH(ifa, ifp) {
   1122 			if (ifa->ifa_addr->sa_family != addr->sa_family)
   1123 				continue;
   1124 			if (equal(addr, ifa->ifa_addr))
   1125 				return ifa;
   1126 			if ((ifp->if_flags & IFF_BROADCAST) &&
   1127 			    ifa->ifa_broadaddr &&
   1128 			    /* IP6 doesn't have broadcast */
   1129 			    ifa->ifa_broadaddr->sa_len != 0 &&
   1130 			    equal(ifa->ifa_broadaddr, addr))
   1131 				return ifa;
   1132 		}
   1133 	}
   1134 	return NULL;
   1135 }
   1136 
   1137 /*
   1138  * Locate the point to point interface with a given destination address.
   1139  */
   1140 /*ARGSUSED*/
   1141 struct ifaddr *
   1142 ifa_ifwithdstaddr(const struct sockaddr *addr)
   1143 {
   1144 	struct ifnet *ifp;
   1145 	struct ifaddr *ifa;
   1146 
   1147 	IFNET_FOREACH(ifp) {
   1148 		if (ifp->if_output == if_nulloutput)
   1149 			continue;
   1150 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
   1151 			continue;
   1152 		IFADDR_FOREACH(ifa, ifp) {
   1153 			if (ifa->ifa_addr->sa_family != addr->sa_family ||
   1154 			    ifa->ifa_dstaddr == NULL)
   1155 				continue;
   1156 			if (equal(addr, ifa->ifa_dstaddr))
   1157 				return ifa;
   1158 		}
   1159 	}
   1160 	return NULL;
   1161 }
   1162 
   1163 /*
   1164  * Find an interface on a specific network.  If many, choice
   1165  * is most specific found.
   1166  */
   1167 struct ifaddr *
   1168 ifa_ifwithnet(const struct sockaddr *addr)
   1169 {
   1170 	struct ifnet *ifp;
   1171 	struct ifaddr *ifa;
   1172 	const struct sockaddr_dl *sdl;
   1173 	struct ifaddr *ifa_maybe = 0;
   1174 	u_int af = addr->sa_family;
   1175 	const char *addr_data = addr->sa_data, *cplim;
   1176 
   1177 	if (af == AF_LINK) {
   1178 		sdl = satocsdl(addr);
   1179 		if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
   1180 		    ifindex2ifnet[sdl->sdl_index] &&
   1181 		    ifindex2ifnet[sdl->sdl_index]->if_output != if_nulloutput)
   1182 			return ifnet_addrs[sdl->sdl_index];
   1183 	}
   1184 #ifdef NETATALK
   1185 	if (af == AF_APPLETALK) {
   1186 		const struct sockaddr_at *sat, *sat2;
   1187 		sat = (const struct sockaddr_at *)addr;
   1188 		IFNET_FOREACH(ifp) {
   1189 			if (ifp->if_output == if_nulloutput)
   1190 				continue;
   1191 			ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
   1192 			if (ifa == NULL)
   1193 				continue;
   1194 			sat2 = (struct sockaddr_at *)ifa->ifa_addr;
   1195 			if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
   1196 				return ifa; /* exact match */
   1197 			if (ifa_maybe == NULL) {
   1198 				/* else keep the if with the right range */
   1199 				ifa_maybe = ifa;
   1200 			}
   1201 		}
   1202 		return ifa_maybe;
   1203 	}
   1204 #endif
   1205 	IFNET_FOREACH(ifp) {
   1206 		if (ifp->if_output == if_nulloutput)
   1207 			continue;
   1208 		IFADDR_FOREACH(ifa, ifp) {
   1209 			const char *cp, *cp2, *cp3;
   1210 
   1211 			if (ifa->ifa_addr->sa_family != af ||
   1212 			    ifa->ifa_netmask == NULL)
   1213  next:				continue;
   1214 			cp = addr_data;
   1215 			cp2 = ifa->ifa_addr->sa_data;
   1216 			cp3 = ifa->ifa_netmask->sa_data;
   1217 			cplim = (const char *)ifa->ifa_netmask +
   1218 			    ifa->ifa_netmask->sa_len;
   1219 			while (cp3 < cplim) {
   1220 				if ((*cp++ ^ *cp2++) & *cp3++) {
   1221 					/* want to continue for() loop */
   1222 					goto next;
   1223 				}
   1224 			}
   1225 			if (ifa_maybe == NULL ||
   1226 			    rn_refines((void *)ifa->ifa_netmask,
   1227 			    (void *)ifa_maybe->ifa_netmask))
   1228 				ifa_maybe = ifa;
   1229 		}
   1230 	}
   1231 	return ifa_maybe;
   1232 }
   1233 
   1234 /*
   1235  * Find the interface of the addresss.
   1236  */
   1237 struct ifaddr *
   1238 ifa_ifwithladdr(const struct sockaddr *addr)
   1239 {
   1240 	struct ifaddr *ia;
   1241 
   1242 	if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
   1243 	    (ia = ifa_ifwithnet(addr)))
   1244 		return ia;
   1245 	return NULL;
   1246 }
   1247 
   1248 /*
   1249  * Find an interface using a specific address family
   1250  */
   1251 struct ifaddr *
   1252 ifa_ifwithaf(int af)
   1253 {
   1254 	struct ifnet *ifp;
   1255 	struct ifaddr *ifa;
   1256 
   1257 	IFNET_FOREACH(ifp) {
   1258 		if (ifp->if_output == if_nulloutput)
   1259 			continue;
   1260 		IFADDR_FOREACH(ifa, ifp) {
   1261 			if (ifa->ifa_addr->sa_family == af)
   1262 				return ifa;
   1263 		}
   1264 	}
   1265 	return NULL;
   1266 }
   1267 
   1268 /*
   1269  * Find an interface address specific to an interface best matching
   1270  * a given address.
   1271  */
   1272 struct ifaddr *
   1273 ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
   1274 {
   1275 	struct ifaddr *ifa;
   1276 	const char *cp, *cp2, *cp3;
   1277 	const char *cplim;
   1278 	struct ifaddr *ifa_maybe = 0;
   1279 	u_int af = addr->sa_family;
   1280 
   1281 	if (ifp->if_output == if_nulloutput)
   1282 		return NULL;
   1283 
   1284 	if (af >= AF_MAX)
   1285 		return NULL;
   1286 
   1287 	IFADDR_FOREACH(ifa, ifp) {
   1288 		if (ifa->ifa_addr->sa_family != af)
   1289 			continue;
   1290 		ifa_maybe = ifa;
   1291 		if (ifa->ifa_netmask == NULL) {
   1292 			if (equal(addr, ifa->ifa_addr) ||
   1293 			    (ifa->ifa_dstaddr &&
   1294 			     equal(addr, ifa->ifa_dstaddr)))
   1295 				return ifa;
   1296 			continue;
   1297 		}
   1298 		cp = addr->sa_data;
   1299 		cp2 = ifa->ifa_addr->sa_data;
   1300 		cp3 = ifa->ifa_netmask->sa_data;
   1301 		cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
   1302 		for (; cp3 < cplim; cp3++) {
   1303 			if ((*cp++ ^ *cp2++) & *cp3)
   1304 				break;
   1305 		}
   1306 		if (cp3 == cplim)
   1307 			return ifa;
   1308 	}
   1309 	return ifa_maybe;
   1310 }
   1311 
   1312 /*
   1313  * Default action when installing a route with a Link Level gateway.
   1314  * Lookup an appropriate real ifa to point to.
   1315  * This should be moved to /sys/net/link.c eventually.
   1316  */
   1317 void
   1318 link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
   1319 {
   1320 	struct ifaddr *ifa;
   1321 	const struct sockaddr *dst;
   1322 	struct ifnet *ifp;
   1323 
   1324 	if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
   1325 	    (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
   1326 		return;
   1327 	if ((ifa = ifaof_ifpforaddr(dst, ifp)) != NULL) {
   1328 		rt_replace_ifa(rt, ifa);
   1329 		if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
   1330 			ifa->ifa_rtrequest(cmd, rt, info);
   1331 	}
   1332 }
   1333 
   1334 /*
   1335  * Handle a change in the interface link state.
   1336  * XXX: We should listen to the routing socket in-kernel rather
   1337  * than calling in6_if_link_* functions directly from here.
   1338  */
   1339 void
   1340 if_link_state_change(struct ifnet *ifp, int link_state)
   1341 {
   1342 	int old_link_state, s;
   1343 
   1344 	s = splnet();
   1345 	if (ifp->if_link_state == link_state) {
   1346 		splx(s);
   1347 		return;
   1348 	}
   1349 
   1350 	old_link_state = ifp->if_link_state;
   1351 	ifp->if_link_state = link_state;
   1352 #ifdef DEBUG
   1353 	log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname,
   1354 		link_state == LINK_STATE_UP ? "UP" :
   1355 		link_state == LINK_STATE_DOWN ? "DOWN" :
   1356 		"UNKNOWN",
   1357 		 old_link_state == LINK_STATE_UP ? "UP" :
   1358 		old_link_state == LINK_STATE_DOWN ? "DOWN" :
   1359 		"UNKNOWN");
   1360 #endif
   1361 
   1362 #ifdef INET6
   1363 	/*
   1364 	 * When going from UNKNOWN to UP, we need to mark existing
   1365 	 * IPv6 addresses as tentative and restart DAD as we may have
   1366 	 * erroneously not found a duplicate.
   1367 	 *
   1368 	 * This needs to happen before rt_ifmsg to avoid a race where
   1369 	 * listeners would have an address and expect it to work right
   1370 	 * away.
   1371 	 */
   1372 	if (link_state == LINK_STATE_UP &&
   1373 	    old_link_state == LINK_STATE_UNKNOWN)
   1374 		in6_if_link_down(ifp);
   1375 #endif
   1376 
   1377 	/* Notify that the link state has changed. */
   1378 	rt_ifmsg(ifp);
   1379 
   1380 #if NCARP > 0
   1381 	if (ifp->if_carp)
   1382 		carp_carpdev_state(ifp);
   1383 #endif
   1384 
   1385 #ifdef INET6
   1386 	if (link_state == LINK_STATE_DOWN)
   1387 		in6_if_link_down(ifp);
   1388 	else if (link_state == LINK_STATE_UP)
   1389 		in6_if_link_up(ifp);
   1390 #endif
   1391 
   1392 	splx(s);
   1393 }
   1394 
   1395 /*
   1396  * Mark an interface down and notify protocols of
   1397  * the transition.
   1398  * NOTE: must be called at splsoftnet or equivalent.
   1399  */
   1400 void
   1401 if_down(struct ifnet *ifp)
   1402 {
   1403 	struct ifaddr *ifa;
   1404 
   1405 	ifp->if_flags &= ~IFF_UP;
   1406 	nanotime(&ifp->if_lastchange);
   1407 	IFADDR_FOREACH(ifa, ifp)
   1408 		pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
   1409 	IFQ_PURGE(&ifp->if_snd);
   1410 #if NCARP > 0
   1411 	if (ifp->if_carp)
   1412 		carp_carpdev_state(ifp);
   1413 #endif
   1414 	rt_ifmsg(ifp);
   1415 #ifdef INET6
   1416 	in6_if_down(ifp);
   1417 #endif
   1418 }
   1419 
   1420 /*
   1421  * Mark an interface up and notify protocols of
   1422  * the transition.
   1423  * NOTE: must be called at splsoftnet or equivalent.
   1424  */
   1425 void
   1426 if_up(struct ifnet *ifp)
   1427 {
   1428 #ifdef notyet
   1429 	struct ifaddr *ifa;
   1430 #endif
   1431 
   1432 	ifp->if_flags |= IFF_UP;
   1433 	nanotime(&ifp->if_lastchange);
   1434 #ifdef notyet
   1435 	/* this has no effect on IP, and will kill all ISO connections XXX */
   1436 	IFADDR_FOREACH(ifa, ifp)
   1437 		pfctlinput(PRC_IFUP, ifa->ifa_addr);
   1438 #endif
   1439 #if NCARP > 0
   1440 	if (ifp->if_carp)
   1441 		carp_carpdev_state(ifp);
   1442 #endif
   1443 	rt_ifmsg(ifp);
   1444 #ifdef INET6
   1445 	in6_if_up(ifp);
   1446 #endif
   1447 }
   1448 
   1449 /*
   1450  * Handle interface watchdog timer routines.  Called
   1451  * from softclock, we decrement timers (if set) and
   1452  * call the appropriate interface routine on expiration.
   1453  */
   1454 void
   1455 if_slowtimo(void *arg)
   1456 {
   1457 	struct ifnet *ifp;
   1458 	int s = splnet();
   1459 
   1460 	IFNET_FOREACH(ifp) {
   1461 		if (ifp->if_timer == 0 || --ifp->if_timer)
   1462 			continue;
   1463 		if (ifp->if_watchdog != NULL)
   1464 			(*ifp->if_watchdog)(ifp);
   1465 	}
   1466 	splx(s);
   1467 	callout_reset(&if_slowtimo_ch, hz / IFNET_SLOWHZ, if_slowtimo, NULL);
   1468 }
   1469 
   1470 /*
   1471  * Set/clear promiscuous mode on interface ifp based on the truth value
   1472  * of pswitch.  The calls are reference counted so that only the first
   1473  * "on" request actually has an effect, as does the final "off" request.
   1474  * Results are undefined if the "off" and "on" requests are not matched.
   1475  */
   1476 int
   1477 ifpromisc(struct ifnet *ifp, int pswitch)
   1478 {
   1479 	int pcount, ret;
   1480 	short nflags;
   1481 
   1482 	pcount = ifp->if_pcount;
   1483 	if (pswitch) {
   1484 		/*
   1485 		 * Allow the device to be "placed" into promiscuous
   1486 		 * mode even if it is not configured up.  It will
   1487 		 * consult IFF_PROMISC when it is brought up.
   1488 		 */
   1489 		if (ifp->if_pcount++ != 0)
   1490 			return 0;
   1491 		nflags = ifp->if_flags | IFF_PROMISC;
   1492 	} else {
   1493 		if (--ifp->if_pcount > 0)
   1494 			return 0;
   1495 		nflags = ifp->if_flags & ~IFF_PROMISC;
   1496 	}
   1497 	ret = if_flags_set(ifp, nflags);
   1498 	/* Restore interface state if not successful. */
   1499 	if (ret != 0) {
   1500 		ifp->if_pcount = pcount;
   1501 	}
   1502 	return ret;
   1503 }
   1504 
   1505 /*
   1506  * Map interface name to
   1507  * interface structure pointer.
   1508  */
   1509 struct ifnet *
   1510 ifunit(const char *name)
   1511 {
   1512 	struct ifnet *ifp;
   1513 	const char *cp = name;
   1514 	u_int unit = 0;
   1515 	u_int i;
   1516 
   1517 	/*
   1518 	 * If the entire name is a number, treat it as an ifindex.
   1519 	 */
   1520 	for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
   1521 		unit = unit * 10 + (*cp - '0');
   1522 	}
   1523 
   1524 	/*
   1525 	 * If the number took all of the name, then it's a valid ifindex.
   1526 	 */
   1527 	if (i == IFNAMSIZ || (cp != name && *cp == '\0')) {
   1528 		if (unit >= if_indexlim)
   1529 			return NULL;
   1530 		ifp = ifindex2ifnet[unit];
   1531 		if (ifp == NULL || ifp->if_output == if_nulloutput)
   1532 			return NULL;
   1533 		return ifp;
   1534 	}
   1535 
   1536 	IFNET_FOREACH(ifp) {
   1537 		if (ifp->if_output == if_nulloutput)
   1538 			continue;
   1539 	 	if (strcmp(ifp->if_xname, name) == 0)
   1540 			return ifp;
   1541 	}
   1542 	return NULL;
   1543 }
   1544 
   1545 ifnet_t *
   1546 if_byindex(u_int idx)
   1547 {
   1548 
   1549 	return (idx < if_indexlim) ? ifindex2ifnet[idx] : NULL;
   1550 }
   1551 
   1552 /* common */
   1553 int
   1554 ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
   1555 {
   1556 	int s;
   1557 	struct ifreq *ifr;
   1558 	struct ifcapreq *ifcr;
   1559 	struct ifdatareq *ifdr;
   1560 
   1561 	switch (cmd) {
   1562 	case SIOCSIFCAP:
   1563 		ifcr = data;
   1564 		if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
   1565 			return EINVAL;
   1566 
   1567 		if (ifcr->ifcr_capenable == ifp->if_capenable)
   1568 			return 0;
   1569 
   1570 		ifp->if_capenable = ifcr->ifcr_capenable;
   1571 
   1572 		/* Pre-compute the checksum flags mask. */
   1573 		ifp->if_csum_flags_tx = 0;
   1574 		ifp->if_csum_flags_rx = 0;
   1575 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
   1576 			ifp->if_csum_flags_tx |= M_CSUM_IPv4;
   1577 		}
   1578 		if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
   1579 			ifp->if_csum_flags_rx |= M_CSUM_IPv4;
   1580 		}
   1581 
   1582 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
   1583 			ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
   1584 		}
   1585 		if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
   1586 			ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
   1587 		}
   1588 
   1589 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
   1590 			ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
   1591 		}
   1592 		if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
   1593 			ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
   1594 		}
   1595 
   1596 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
   1597 			ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
   1598 		}
   1599 		if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
   1600 			ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
   1601 		}
   1602 
   1603 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
   1604 			ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
   1605 		}
   1606 		if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
   1607 			ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
   1608 		}
   1609 		if (ifp->if_flags & IFF_UP)
   1610 			return ENETRESET;
   1611 		return 0;
   1612 	case SIOCSIFFLAGS:
   1613 		ifr = data;
   1614 		if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
   1615 			s = splnet();
   1616 			if_down(ifp);
   1617 			splx(s);
   1618 		}
   1619 		if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
   1620 			s = splnet();
   1621 			if_up(ifp);
   1622 			splx(s);
   1623 		}
   1624 		ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
   1625 			(ifr->ifr_flags &~ IFF_CANTCHANGE);
   1626 		break;
   1627 	case SIOCGIFFLAGS:
   1628 		ifr = data;
   1629 		ifr->ifr_flags = ifp->if_flags;
   1630 		break;
   1631 
   1632 	case SIOCGIFMETRIC:
   1633 		ifr = data;
   1634 		ifr->ifr_metric = ifp->if_metric;
   1635 		break;
   1636 
   1637 	case SIOCGIFMTU:
   1638 		ifr = data;
   1639 		ifr->ifr_mtu = ifp->if_mtu;
   1640 		break;
   1641 
   1642 	case SIOCGIFDLT:
   1643 		ifr = data;
   1644 		ifr->ifr_dlt = ifp->if_dlt;
   1645 		break;
   1646 
   1647 	case SIOCGIFCAP:
   1648 		ifcr = data;
   1649 		ifcr->ifcr_capabilities = ifp->if_capabilities;
   1650 		ifcr->ifcr_capenable = ifp->if_capenable;
   1651 		break;
   1652 
   1653 	case SIOCSIFMETRIC:
   1654 		ifr = data;
   1655 		ifp->if_metric = ifr->ifr_metric;
   1656 		break;
   1657 
   1658 	case SIOCGIFDATA:
   1659 		ifdr = data;
   1660 		ifdr->ifdr_data = ifp->if_data;
   1661 		break;
   1662 
   1663 	case SIOCZIFDATA:
   1664 		ifdr = data;
   1665 		ifdr->ifdr_data = ifp->if_data;
   1666 		/*
   1667 		 * Assumes that the volatile counters that can be
   1668 		 * zero'ed are at the end of if_data.
   1669 		 */
   1670 		memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
   1671 		    offsetof(struct if_data, ifi_ipackets));
   1672 		/*
   1673 		 * The memset() clears to the bottm of if_data. In the area,
   1674 		 * if_lastchange is included. Please be careful if new entry
   1675 		 * will be added into if_data or rewite this.
   1676 		 *
   1677 		 * And also, update if_lastchnage.
   1678 		 */
   1679 		getnanotime(&ifp->if_lastchange);
   1680 		break;
   1681 	case SIOCSIFMTU:
   1682 		ifr = data;
   1683 		if (ifp->if_mtu == ifr->ifr_mtu)
   1684 			break;
   1685 		ifp->if_mtu = ifr->ifr_mtu;
   1686 		/*
   1687 		 * If the link MTU changed, do network layer specific procedure.
   1688 		 */
   1689 #ifdef INET6
   1690 		nd6_setmtu(ifp);
   1691 #endif
   1692 		return ENETRESET;
   1693 	default:
   1694 		return ENOTTY;
   1695 	}
   1696 	return 0;
   1697 }
   1698 
   1699 int
   1700 ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
   1701     lwp_t *l)
   1702 {
   1703 	struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
   1704 	struct ifaddr *ifa;
   1705 	const struct sockaddr *any, *sa;
   1706 	union {
   1707 		struct sockaddr sa;
   1708 		struct sockaddr_storage ss;
   1709 	} u, v;
   1710 
   1711 	switch (cmd) {
   1712 	case SIOCSIFADDRPREF:
   1713 		if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_INTERFACE,
   1714 		    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
   1715 		    NULL) != 0)
   1716 			return EPERM;
   1717 	case SIOCGIFADDRPREF:
   1718 		break;
   1719 	default:
   1720 		return EOPNOTSUPP;
   1721 	}
   1722 
   1723 	/* sanity checks */
   1724 	if (data == NULL || ifp == NULL) {
   1725 		panic("invalid argument to %s", __func__);
   1726 		/*NOTREACHED*/
   1727 	}
   1728 
   1729 	/* address must be specified on ADD and DELETE */
   1730 	sa = sstocsa(&ifap->ifap_addr);
   1731 	if (sa->sa_family != sofamily(so))
   1732 		return EINVAL;
   1733 	if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
   1734 		return EINVAL;
   1735 
   1736 	sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
   1737 
   1738 	IFADDR_FOREACH(ifa, ifp) {
   1739 		if (ifa->ifa_addr->sa_family != sa->sa_family)
   1740 			continue;
   1741 		sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
   1742 		if (sockaddr_cmp(&u.sa, &v.sa) == 0)
   1743 			break;
   1744 	}
   1745 	if (ifa == NULL)
   1746 		return EADDRNOTAVAIL;
   1747 
   1748 	switch (cmd) {
   1749 	case SIOCSIFADDRPREF:
   1750 		ifa->ifa_preference = ifap->ifap_preference;
   1751 		return 0;
   1752 	case SIOCGIFADDRPREF:
   1753 		/* fill in the if_laddrreq structure */
   1754 		(void)sockaddr_copy(sstosa(&ifap->ifap_addr),
   1755 		    sizeof(ifap->ifap_addr), ifa->ifa_addr);
   1756 		ifap->ifap_preference = ifa->ifa_preference;
   1757 		return 0;
   1758 	default:
   1759 		return EOPNOTSUPP;
   1760 	}
   1761 }
   1762 
   1763 static void
   1764 ifnet_lock_enter(struct ifnet_lock *il)
   1765 {
   1766 	uint64_t *nenter;
   1767 
   1768 	/* Before trying to acquire the mutex, increase the count of threads
   1769 	 * who have entered or who wait to enter the critical section.
   1770 	 * Avoid one costly locked memory transaction by keeping a count for
   1771 	 * each CPU.
   1772 	 */
   1773 	nenter = percpu_getref(il->il_nenter);
   1774 	(*nenter)++;
   1775 	percpu_putref(il->il_nenter);
   1776 	mutex_enter(&il->il_lock);
   1777 }
   1778 
   1779 static void
   1780 ifnet_lock_exit(struct ifnet_lock *il)
   1781 {
   1782 	/* Increase the count of threads who have exited the critical
   1783 	 * section.  Increase while we still hold the lock.
   1784 	 */
   1785 	il->il_nexit++;
   1786 	mutex_exit(&il->il_lock);
   1787 }
   1788 
   1789 /*
   1790  * Interface ioctls.
   1791  */
   1792 int
   1793 ifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
   1794 {
   1795 	struct ifnet *ifp;
   1796 	struct ifreq *ifr;
   1797 	int error = 0;
   1798 #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
   1799 	u_long ocmd = cmd;
   1800 #endif
   1801 	short oif_flags;
   1802 #ifdef COMPAT_OIFREQ
   1803 	struct ifreq ifrb;
   1804 	struct oifreq *oifr = NULL;
   1805 #endif
   1806 
   1807 	switch (cmd) {
   1808 #ifdef COMPAT_OIFREQ
   1809 	case OSIOCGIFCONF:
   1810 	case OOSIOCGIFCONF:
   1811 		return compat_ifconf(cmd, data);
   1812 #endif
   1813 #ifdef COMPAT_OIFDATA
   1814 	case OSIOCGIFDATA:
   1815 	case OSIOCZIFDATA:
   1816 		return compat_ifdatareq(l, cmd, data);
   1817 #endif
   1818 	case SIOCGIFCONF:
   1819 		return ifconf(cmd, data);
   1820 	case SIOCINITIFADDR:
   1821 		return EPERM;
   1822 	}
   1823 
   1824 #ifdef COMPAT_OIFREQ
   1825 	cmd = compat_cvtcmd(cmd);
   1826 	if (cmd != ocmd) {
   1827 		oifr = data;
   1828 		data = ifr = &ifrb;
   1829 		ifreqo2n(oifr, ifr);
   1830 	} else
   1831 #endif
   1832 		ifr = data;
   1833 
   1834 	ifp = ifunit(ifr->ifr_name);
   1835 
   1836 	switch (cmd) {
   1837 	case SIOCIFCREATE:
   1838 	case SIOCIFDESTROY:
   1839 		if (l != NULL) {
   1840 			error = kauth_authorize_network(l->l_cred,
   1841 			    KAUTH_NETWORK_INTERFACE,
   1842 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
   1843 			    (void *)cmd, NULL);
   1844 			if (error != 0)
   1845 				return error;
   1846 		}
   1847 		return (cmd == SIOCIFCREATE) ?
   1848 			if_clone_create(ifr->ifr_name) :
   1849 			if_clone_destroy(ifr->ifr_name);
   1850 
   1851 	case SIOCIFGCLONERS:
   1852 		return if_clone_list((struct if_clonereq *)data);
   1853 	}
   1854 
   1855 	if (ifp == NULL)
   1856 		return ENXIO;
   1857 
   1858 	switch (cmd) {
   1859 	case SIOCALIFADDR:
   1860 	case SIOCDLIFADDR:
   1861 	case SIOCSIFADDRPREF:
   1862 	case SIOCSIFFLAGS:
   1863 	case SIOCSIFCAP:
   1864 	case SIOCSIFMETRIC:
   1865 	case SIOCZIFDATA:
   1866 	case SIOCSIFMTU:
   1867 	case SIOCSIFPHYADDR:
   1868 	case SIOCDIFPHYADDR:
   1869 #ifdef INET6
   1870 	case SIOCSIFPHYADDR_IN6:
   1871 #endif
   1872 	case SIOCSLIFPHYADDR:
   1873 	case SIOCADDMULTI:
   1874 	case SIOCDELMULTI:
   1875 	case SIOCSIFMEDIA:
   1876 	case SIOCSDRVSPEC:
   1877 	case SIOCG80211:
   1878 	case SIOCS80211:
   1879 	case SIOCS80211NWID:
   1880 	case SIOCS80211NWKEY:
   1881 	case SIOCS80211POWER:
   1882 	case SIOCS80211BSSID:
   1883 	case SIOCS80211CHANNEL:
   1884 	case SIOCSLINKSTR:
   1885 		if (l != NULL) {
   1886 			error = kauth_authorize_network(l->l_cred,
   1887 			    KAUTH_NETWORK_INTERFACE,
   1888 			    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
   1889 			    (void *)cmd, NULL);
   1890 			if (error != 0)
   1891 				return error;
   1892 		}
   1893 	}
   1894 
   1895 	oif_flags = ifp->if_flags;
   1896 
   1897 	ifnet_lock_enter(ifp->if_ioctl_lock);
   1898 	error = (*ifp->if_ioctl)(ifp, cmd, data);
   1899 	if (error != ENOTTY)
   1900 		;
   1901 	else if (so->so_proto == NULL)
   1902 		error = EOPNOTSUPP;
   1903 	else {
   1904 #ifdef COMPAT_OSOCK
   1905 		error = compat_ifioctl(so, ocmd, cmd, data, l);
   1906 #else
   1907 		error = (*so->so_proto->pr_usrreq)(so, PRU_CONTROL,
   1908 		    (struct mbuf *)cmd, (struct mbuf *)data,
   1909 		    (struct mbuf *)ifp, l);
   1910 #endif
   1911 	}
   1912 
   1913 	if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
   1914 #ifdef INET6
   1915 		if ((ifp->if_flags & IFF_UP) != 0) {
   1916 			int s = splnet();
   1917 			in6_if_up(ifp);
   1918 			splx(s);
   1919 		}
   1920 #endif
   1921 	}
   1922 #ifdef COMPAT_OIFREQ
   1923 	if (cmd != ocmd)
   1924 		ifreqn2o(oifr, ifr);
   1925 #endif
   1926 
   1927 	ifnet_lock_exit(ifp->if_ioctl_lock);
   1928 	return error;
   1929 }
   1930 
   1931 /* This callback adds to the sum in `arg' the number of
   1932  * threads on `ci' who have entered or who wait to enter the
   1933  * critical section.
   1934  */
   1935 static void
   1936 ifnet_lock_sum(void *p, void *arg, struct cpu_info *ci)
   1937 {
   1938 	uint64_t *sum = arg, *nenter = p;
   1939 
   1940 	*sum += *nenter;
   1941 }
   1942 
   1943 /* Return the number of threads who have entered or who wait
   1944  * to enter the critical section on all CPUs.
   1945  */
   1946 static uint64_t
   1947 ifnet_lock_entrances(struct ifnet_lock *il)
   1948 {
   1949 	uint64_t sum = 0;
   1950 
   1951 	percpu_foreach(il->il_nenter, ifnet_lock_sum, &sum);
   1952 
   1953 	return sum;
   1954 }
   1955 
   1956 static int
   1957 ifioctl_attach(struct ifnet *ifp)
   1958 {
   1959 	struct ifnet_lock *il;
   1960 
   1961 	/* If the driver has not supplied its own if_ioctl, then
   1962 	 * supply the default.
   1963 	 */
   1964 	if (ifp->if_ioctl == NULL)
   1965 		ifp->if_ioctl = ifioctl_common;
   1966 
   1967 	/* Create an ifnet_lock for synchronizing ifioctls. */
   1968 	if ((il = kmem_zalloc(sizeof(*il), KM_SLEEP)) == NULL)
   1969 		return ENOMEM;
   1970 
   1971 	il->il_nenter = percpu_alloc(sizeof(uint64_t));
   1972 	if (il->il_nenter == NULL) {
   1973 		kmem_free(il, sizeof(*il));
   1974 		return ENOMEM;
   1975 	}
   1976 
   1977 	mutex_init(&il->il_lock, MUTEX_DEFAULT, IPL_NONE);
   1978 	cv_init(&il->il_emptied, ifp->if_xname);
   1979 
   1980 	ifp->if_ioctl_lock = il;
   1981 
   1982 	return 0;
   1983 }
   1984 
   1985 /*
   1986  * This must not be called until after `ifp' has been withdrawn from the
   1987  * ifnet tables so that ifioctl() cannot get a handle on it by calling
   1988  * ifunit().
   1989  */
   1990 static void
   1991 ifioctl_detach(struct ifnet *ifp)
   1992 {
   1993 	struct ifnet_lock *il;
   1994 
   1995 	il = ifp->if_ioctl_lock;
   1996 	mutex_enter(&il->il_lock);
   1997 	/* Install if_nullioctl to make sure that any thread that
   1998 	 * subsequently enters the critical section will quit it
   1999 	 * immediately and signal the condition variable that we
   2000 	 * wait on, below.
   2001 	 */
   2002 	ifp->if_ioctl = if_nullioctl;
   2003 	/* Sleep while threads are still in the critical section or
   2004 	 * wait to enter it.
   2005 	 */
   2006 	while (ifnet_lock_entrances(il) != il->il_nexit)
   2007 		cv_wait(&il->il_emptied, &il->il_lock);
   2008 	/* At this point, we are the only thread still in the critical
   2009 	 * section, and no new thread can get a handle on the ifioctl
   2010 	 * lock, so it is safe to free its memory.
   2011 	 */
   2012 	mutex_exit(&il->il_lock);
   2013 	ifp->if_ioctl_lock = NULL;
   2014 	percpu_free(il->il_nenter, sizeof(uint64_t));
   2015 	il->il_nenter = NULL;
   2016 	cv_destroy(&il->il_emptied);
   2017 	mutex_destroy(&il->il_lock);
   2018 	kmem_free(il, sizeof(*il));
   2019 }
   2020 
   2021 /*
   2022  * Return interface configuration
   2023  * of system.  List may be used
   2024  * in later ioctl's (above) to get
   2025  * other information.
   2026  *
   2027  * Each record is a struct ifreq.  Before the addition of
   2028  * sockaddr_storage, the API rule was that sockaddr flavors that did
   2029  * not fit would extend beyond the struct ifreq, with the next struct
   2030  * ifreq starting sa_len beyond the struct sockaddr.  Because the
   2031  * union in struct ifreq includes struct sockaddr_storage, every kind
   2032  * of sockaddr must fit.  Thus, there are no longer any overlength
   2033  * records.
   2034  *
   2035  * Records are added to the user buffer if they fit, and ifc_len is
   2036  * adjusted to the length that was written.  Thus, the user is only
   2037  * assured of getting the complete list if ifc_len on return is at
   2038  * least sizeof(struct ifreq) less than it was on entry.
   2039  *
   2040  * If the user buffer pointer is NULL, this routine copies no data and
   2041  * returns the amount of space that would be needed.
   2042  *
   2043  * Invariants:
   2044  * ifrp points to the next part of the user's buffer to be used.  If
   2045  * ifrp != NULL, space holds the number of bytes remaining that we may
   2046  * write at ifrp.  Otherwise, space holds the number of bytes that
   2047  * would have been written had there been adequate space.
   2048  */
   2049 /*ARGSUSED*/
   2050 int
   2051 ifconf(u_long cmd, void *data)
   2052 {
   2053 	struct ifconf *ifc = (struct ifconf *)data;
   2054 	struct ifnet *ifp;
   2055 	struct ifaddr *ifa;
   2056 	struct ifreq ifr, *ifrp;
   2057 	int space, error = 0;
   2058 	const int sz = (int)sizeof(struct ifreq);
   2059 
   2060 	if ((ifrp = ifc->ifc_req) == NULL)
   2061 		space = 0;
   2062 	else
   2063 		space = ifc->ifc_len;
   2064 	IFNET_FOREACH(ifp) {
   2065 		(void)strncpy(ifr.ifr_name, ifp->if_xname,
   2066 		    sizeof(ifr.ifr_name));
   2067 		if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0')
   2068 			return ENAMETOOLONG;
   2069 		if (IFADDR_EMPTY(ifp)) {
   2070 			/* Interface with no addresses - send zero sockaddr. */
   2071 			memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
   2072 			if (ifrp == NULL) {
   2073 				space += sz;
   2074 				continue;
   2075 			}
   2076 			if (space >= sz) {
   2077 				error = copyout(&ifr, ifrp, sz);
   2078 				if (error != 0)
   2079 					return error;
   2080 				ifrp++;
   2081 				space -= sz;
   2082 			}
   2083 		}
   2084 
   2085 		IFADDR_FOREACH(ifa, ifp) {
   2086 			struct sockaddr *sa = ifa->ifa_addr;
   2087 			/* all sockaddrs must fit in sockaddr_storage */
   2088 			KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
   2089 
   2090 			if (ifrp == NULL) {
   2091 				space += sz;
   2092 				continue;
   2093 			}
   2094 			memcpy(&ifr.ifr_space, sa, sa->sa_len);
   2095 			if (space >= sz) {
   2096 				error = copyout(&ifr, ifrp, sz);
   2097 				if (error != 0)
   2098 					return (error);
   2099 				ifrp++; space -= sz;
   2100 			}
   2101 		}
   2102 	}
   2103 	if (ifrp != NULL) {
   2104 		KASSERT(0 <= space && space <= ifc->ifc_len);
   2105 		ifc->ifc_len -= space;
   2106 	} else {
   2107 		KASSERT(space >= 0);
   2108 		ifc->ifc_len = space;
   2109 	}
   2110 	return (0);
   2111 }
   2112 
   2113 int
   2114 ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
   2115 {
   2116 	uint8_t len;
   2117 #ifdef COMPAT_OIFREQ
   2118 	struct ifreq ifrb;
   2119 	struct oifreq *oifr = NULL;
   2120 	u_long ocmd = cmd;
   2121 	cmd = compat_cvtcmd(cmd);
   2122 	if (cmd != ocmd) {
   2123 		oifr = (struct oifreq *)(void *)ifr;
   2124 		ifr = &ifrb;
   2125 		ifreqo2n(oifr, ifr);
   2126 		len = sizeof(oifr->ifr_addr);
   2127 	} else
   2128 #endif
   2129 		len = sizeof(ifr->ifr_ifru.ifru_space);
   2130 
   2131 	if (len < sa->sa_len)
   2132 		return EFBIG;
   2133 
   2134 	memset(&ifr->ifr_addr, 0, len);
   2135 	sockaddr_copy(&ifr->ifr_addr, len, sa);
   2136 
   2137 #ifdef COMPAT_OIFREQ
   2138 	if (cmd != ocmd)
   2139 		ifreqn2o(oifr, ifr);
   2140 #endif
   2141 	return 0;
   2142 }
   2143 
   2144 /*
   2145  * Queue message on interface, and start output if interface
   2146  * not yet active.
   2147  */
   2148 int
   2149 ifq_enqueue(struct ifnet *ifp, struct mbuf *m
   2150     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
   2151 {
   2152 	int len = m->m_pkthdr.len;
   2153 	int mflags = m->m_flags;
   2154 	int s = splnet();
   2155 	int error;
   2156 
   2157 	IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
   2158 	if (error != 0)
   2159 		goto out;
   2160 	ifp->if_obytes += len;
   2161 	if (mflags & M_MCAST)
   2162 		ifp->if_omcasts++;
   2163 	if ((ifp->if_flags & IFF_OACTIVE) == 0)
   2164 		(*ifp->if_start)(ifp);
   2165 out:
   2166 	splx(s);
   2167 	return error;
   2168 }
   2169 
   2170 /*
   2171  * Queue message on interface, possibly using a second fast queue
   2172  */
   2173 int
   2174 ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m
   2175     ALTQ_COMMA ALTQ_DECL(struct altq_pktattr *pktattr))
   2176 {
   2177 	int error = 0;
   2178 
   2179 	if (ifq != NULL
   2180 #ifdef ALTQ
   2181 	    && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
   2182 #endif
   2183 	    ) {
   2184 		if (IF_QFULL(ifq)) {
   2185 			IF_DROP(&ifp->if_snd);
   2186 			m_freem(m);
   2187 			if (error == 0)
   2188 				error = ENOBUFS;
   2189 		} else
   2190 			IF_ENQUEUE(ifq, m);
   2191 	} else
   2192 		IFQ_ENQUEUE(&ifp->if_snd, m, pktattr, error);
   2193 	if (error != 0) {
   2194 		++ifp->if_oerrors;
   2195 		return error;
   2196 	}
   2197 	return 0;
   2198 }
   2199 
   2200 int
   2201 if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src)
   2202 {
   2203 	int rc;
   2204 
   2205 	if (ifp->if_initaddr != NULL)
   2206 		rc = (*ifp->if_initaddr)(ifp, ifa, src);
   2207 	else if (src ||
   2208 	         (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY)
   2209 		rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa);
   2210 
   2211 	return rc;
   2212 }
   2213 
   2214 int
   2215 if_flags_set(ifnet_t *ifp, const short flags)
   2216 {
   2217 	int rc;
   2218 
   2219 	if (ifp->if_setflags != NULL)
   2220 		rc = (*ifp->if_setflags)(ifp, flags);
   2221 	else {
   2222 		short cantflags, chgdflags;
   2223 		struct ifreq ifr;
   2224 
   2225 		chgdflags = ifp->if_flags ^ flags;
   2226 		cantflags = chgdflags & IFF_CANTCHANGE;
   2227 
   2228 		if (cantflags != 0)
   2229 			ifp->if_flags ^= cantflags;
   2230 
   2231                 /* Traditionally, we do not call if_ioctl after
   2232                  * setting/clearing only IFF_PROMISC if the interface
   2233                  * isn't IFF_UP.  Uphold that tradition.
   2234 		 */
   2235 		if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0)
   2236 			return 0;
   2237 
   2238 		memset(&ifr, 0, sizeof(ifr));
   2239 
   2240 		ifr.ifr_flags = flags & ~IFF_CANTCHANGE;
   2241 		rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
   2242 
   2243 		if (rc != 0 && cantflags != 0)
   2244 			ifp->if_flags ^= cantflags;
   2245 	}
   2246 
   2247 	return rc;
   2248 }
   2249 
   2250 int
   2251 if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa)
   2252 {
   2253 	int rc;
   2254 	struct ifreq ifr;
   2255 
   2256 	if (ifp->if_mcastop != NULL)
   2257 		rc = (*ifp->if_mcastop)(ifp, cmd, sa);
   2258 	else {
   2259 		ifreq_setaddr(cmd, &ifr, sa);
   2260 		rc = (*ifp->if_ioctl)(ifp, cmd, &ifr);
   2261 	}
   2262 
   2263 	return rc;
   2264 }
   2265 
   2266 static void
   2267 sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
   2268     struct ifaltq *ifq)
   2269 {
   2270 	const struct sysctlnode *cnode, *rnode;
   2271 
   2272 	if (sysctl_createv(clog, 0, NULL, &rnode,
   2273 		       CTLFLAG_PERMANENT,
   2274 		       CTLTYPE_NODE, "net", NULL,
   2275 		       NULL, 0, NULL, 0,
   2276 		       CTL_NET, CTL_EOL) != 0)
   2277 		goto bad;
   2278 
   2279 	if (sysctl_createv(clog, 0, &rnode, &rnode,
   2280 		       CTLFLAG_PERMANENT,
   2281 		       CTLTYPE_NODE, "interfaces",
   2282 		       SYSCTL_DESCR("Per-interface controls"),
   2283 		       NULL, 0, NULL, 0,
   2284 		       CTL_CREATE, CTL_EOL) != 0)
   2285 		goto bad;
   2286 
   2287 	if (sysctl_createv(clog, 0, &rnode, &rnode,
   2288 		       CTLFLAG_PERMANENT,
   2289 		       CTLTYPE_NODE, ifname,
   2290 		       SYSCTL_DESCR("Interface controls"),
   2291 		       NULL, 0, NULL, 0,
   2292 		       CTL_CREATE, CTL_EOL) != 0)
   2293 		goto bad;
   2294 
   2295 	if (sysctl_createv(clog, 0, &rnode, &rnode,
   2296 		       CTLFLAG_PERMANENT,
   2297 		       CTLTYPE_NODE, "sndq",
   2298 		       SYSCTL_DESCR("Interface output queue controls"),
   2299 		       NULL, 0, NULL, 0,
   2300 		       CTL_CREATE, CTL_EOL) != 0)
   2301 		goto bad;
   2302 
   2303 	if (sysctl_createv(clog, 0, &rnode, &cnode,
   2304 		       CTLFLAG_PERMANENT,
   2305 		       CTLTYPE_INT, "len",
   2306 		       SYSCTL_DESCR("Current output queue length"),
   2307 		       NULL, 0, &ifq->ifq_len, 0,
   2308 		       CTL_CREATE, CTL_EOL) != 0)
   2309 		goto bad;
   2310 
   2311 	if (sysctl_createv(clog, 0, &rnode, &cnode,
   2312 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2313 		       CTLTYPE_INT, "maxlen",
   2314 		       SYSCTL_DESCR("Maximum allowed output queue length"),
   2315 		       NULL, 0, &ifq->ifq_maxlen, 0,
   2316 		       CTL_CREATE, CTL_EOL) != 0)
   2317 		goto bad;
   2318 
   2319 	if (sysctl_createv(clog, 0, &rnode, &cnode,
   2320 		       CTLFLAG_PERMANENT,
   2321 		       CTLTYPE_INT, "drops",
   2322 		       SYSCTL_DESCR("Packets dropped due to full output queue"),
   2323 		       NULL, 0, &ifq->ifq_drops, 0,
   2324 		       CTL_CREATE, CTL_EOL) != 0)
   2325 		goto bad;
   2326 
   2327 	return;
   2328 bad:
   2329 	printf("%s: could not attach sysctl nodes\n", ifname);
   2330 	return;
   2331 }
   2332 
   2333 #if defined(INET) || defined(INET6)
   2334 static void
   2335 sysctl_net_ifq_setup(struct sysctllog **clog,
   2336 		     int pf, const char *pfname,
   2337 		     int ipn, const char *ipname,
   2338 		     int qid, struct ifqueue *ifq)
   2339 {
   2340 
   2341 	sysctl_createv(clog, 0, NULL, NULL,
   2342 		       CTLFLAG_PERMANENT,
   2343 		       CTLTYPE_NODE, "net", NULL,
   2344 		       NULL, 0, NULL, 0,
   2345 		       CTL_NET, CTL_EOL);
   2346 	sysctl_createv(clog, 0, NULL, NULL,
   2347 		       CTLFLAG_PERMANENT,
   2348 		       CTLTYPE_NODE, pfname, NULL,
   2349 		       NULL, 0, NULL, 0,
   2350 		       CTL_NET, pf, CTL_EOL);
   2351 	sysctl_createv(clog, 0, NULL, NULL,
   2352 		       CTLFLAG_PERMANENT,
   2353 		       CTLTYPE_NODE, ipname, NULL,
   2354 		       NULL, 0, NULL, 0,
   2355 		       CTL_NET, pf, ipn, CTL_EOL);
   2356 	sysctl_createv(clog, 0, NULL, NULL,
   2357 		       CTLFLAG_PERMANENT,
   2358 		       CTLTYPE_NODE, "ifq",
   2359 		       SYSCTL_DESCR("Protocol input queue controls"),
   2360 		       NULL, 0, NULL, 0,
   2361 		       CTL_NET, pf, ipn, qid, CTL_EOL);
   2362 
   2363 	sysctl_createv(clog, 0, NULL, NULL,
   2364 		       CTLFLAG_PERMANENT,
   2365 		       CTLTYPE_INT, "len",
   2366 		       SYSCTL_DESCR("Current input queue length"),
   2367 		       NULL, 0, &ifq->ifq_len, 0,
   2368 		       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
   2369 	sysctl_createv(clog, 0, NULL, NULL,
   2370 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2371 		       CTLTYPE_INT, "maxlen",
   2372 		       SYSCTL_DESCR("Maximum allowed input queue length"),
   2373 		       NULL, 0, &ifq->ifq_maxlen, 0,
   2374 		       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
   2375 #ifdef notyet
   2376 	sysctl_createv(clog, 0, NULL, NULL,
   2377 		       CTLFLAG_PERMANENT,
   2378 		       CTLTYPE_INT, "peak",
   2379 		       SYSCTL_DESCR("Highest input queue length"),
   2380 		       NULL, 0, &ifq->ifq_peak, 0,
   2381 		       CTL_NET, pf, ipn, qid, IFQCTL_PEAK, CTL_EOL);
   2382 #endif
   2383 	sysctl_createv(clog, 0, NULL, NULL,
   2384 		       CTLFLAG_PERMANENT,
   2385 		       CTLTYPE_INT, "drops",
   2386 		       SYSCTL_DESCR("Packets dropped due to full input queue"),
   2387 		       NULL, 0, &ifq->ifq_drops, 0,
   2388 		       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
   2389 }
   2390 #endif /* INET || INET6 */
   2391