Home | History | Annotate | Line # | Graph | Download | only in netinet
      1 /*	$NetBSD: if_arp.c,v 1.171 2015/08/07 08:11:33 ozaki-r Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1998, 2000, 2008 The NetBSD Foundation, Inc.
      5  * All rights reserved.
      6  *
      7  * This code is derived from software contributed to The NetBSD Foundation
      8  * by Public Access Networks Corporation ("Panix").  It was developed under
      9  * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
     10  *
     11  * Redistribution and use in source and binary forms, with or without
     12  * modification, are permitted provided that the following conditions
     13  * are met:
     14  * 1. Redistributions of source code must retain the above copyright
     15  *    notice, this list of conditions and the following disclaimer.
     16  * 2. Redistributions in binary form must reproduce the above copyright
     17  *    notice, this list of conditions and the following disclaimer in the
     18  *    documentation and/or other materials provided with the distribution.
     19  *
     20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     30  * POSSIBILITY OF SUCH DAMAGE.
     31  */
     32 
     33 /*
     34  * Copyright (c) 1982, 1986, 1988, 1993
     35  *	The Regents of the University of California.  All rights reserved.
     36  *
     37  * Redistribution and use in source and binary forms, with or without
     38  * modification, are permitted provided that the following conditions
     39  * are met:
     40  * 1. Redistributions of source code must retain the above copyright
     41  *    notice, this list of conditions and the following disclaimer.
     42  * 2. Redistributions in binary form must reproduce the above copyright
     43  *    notice, this list of conditions and the following disclaimer in the
     44  *    documentation and/or other materials provided with the distribution.
     45  * 3. Neither the name of the University nor the names of its contributors
     46  *    may be used to endorse or promote products derived from this software
     47  *    without specific prior written permission.
     48  *
     49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     59  * SUCH DAMAGE.
     60  *
     61  *	@(#)if_ether.c	8.2 (Berkeley) 9/26/94
     62  */
     63 
     64 /*
     65  * Ethernet address resolution protocol.
     66  * TODO:
     67  *	add "inuse/lock" bit (or ref. count) along with valid bit
     68  */
     69 
     70 #include <sys/cdefs.h>
     71 __KERNEL_RCSID(0, "$NetBSD: if_arp.c,v 1.171 2015/08/07 08:11:33 ozaki-r Exp $");
     72 
     73 #include "opt_ddb.h"
     74 #include "opt_inet.h"
     75 
     76 #ifdef INET
     77 
     78 #include "bridge.h"
     79 
     80 #include <sys/param.h>
     81 #include <sys/systm.h>
     82 #include <sys/callout.h>
     83 #include <sys/malloc.h>
     84 #include <sys/mbuf.h>
     85 #include <sys/socket.h>
     86 #include <sys/time.h>
     87 #include <sys/timetc.h>
     88 #include <sys/kernel.h>
     89 #include <sys/errno.h>
     90 #include <sys/ioctl.h>
     91 #include <sys/syslog.h>
     92 #include <sys/proc.h>
     93 #include <sys/protosw.h>
     94 #include <sys/domain.h>
     95 #include <sys/sysctl.h>
     96 #include <sys/socketvar.h>
     97 #include <sys/percpu.h>
     98 #include <sys/cprng.h>
     99 
    100 #include <net/ethertypes.h>
    101 #include <net/if.h>
    102 #include <net/if_dl.h>
    103 #include <net/if_token.h>
    104 #include <net/if_types.h>
    105 #include <net/if_ether.h>
    106 #include <net/net_osdep.h>
    107 #include <net/route.h>
    108 #include <net/net_stats.h>
    109 
    110 #include <netinet/in.h>
    111 #include <netinet/in_systm.h>
    112 #include <netinet/in_var.h>
    113 #include <netinet/ip.h>
    114 #include <netinet/if_inarp.h>
    115 
    116 #include "arcnet.h"
    117 #if NARCNET > 0
    118 #include <net/if_arc.h>
    119 #endif
    120 #include "fddi.h"
    121 #if NFDDI > 0
    122 #include <net/if_fddi.h>
    123 #endif
    124 #include "token.h"
    125 #include "carp.h"
    126 #if NCARP > 0
    127 #include <netinet/ip_carp.h>
    128 #endif
    129 
    130 #define SIN(s) ((struct sockaddr_in *)s)
    131 #define SRP(s) ((struct sockaddr_inarp *)s)
    132 
    133 /*
    134  * ARP trailer negotiation.  Trailer protocol is not IP specific,
    135  * but ARP request/response use IP addresses.
    136  */
    137 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
    138 
    139 /* timer values */
    140 static int	arpt_prune = (5*60*1);	/* walk list every 5 minutes */
    141 static int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
    142 static int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
    143 static int	arpt_refresh = (5*60);	/* time left before refreshing */
    144 #define	rt_expire rt_rmx.rmx_expire
    145 #define	rt_pksent rt_rmx.rmx_pksent
    146 
    147 int		ip_dad_count = PROBE_NUM;
    148 #ifdef ARP_DEBUG
    149 static int	arp_debug = 1;
    150 #else
    151 static int	arp_debug = 0;
    152 #endif
    153 #define arplog(x)	do { if (arp_debug) log x; } while (/*CONSTCOND*/ 0)
    154 
    155 static	void arp_init(void);
    156 
    157 static	struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *,
    158 	    const struct sockaddr *);
    159 static	void arptfree(struct llinfo_arp *);
    160 static	void arptimer(void *);
    161 static	struct llinfo_arp *arplookup1(struct mbuf *, const struct in_addr *,
    162 				      int, int, struct rtentry *);
    163 static	struct llinfo_arp *arplookup(struct mbuf *, const struct in_addr *,
    164 					  int, int);
    165 static	void in_arpinput(struct mbuf *);
    166 static	void in_revarpinput(struct mbuf *);
    167 static	void revarprequest(struct ifnet *);
    168 
    169 static	void arp_drainstub(void);
    170 
    171 static void arp_dad_timer(struct ifaddr *);
    172 static void arp_dad_start(struct ifaddr *);
    173 static void arp_dad_stop(struct ifaddr *);
    174 static void arp_dad_duplicated(struct ifaddr *);
    175 
    176 LIST_HEAD(llinfo_arpq, llinfo_arp) llinfo_arp;
    177 struct	ifqueue arpintrq = {
    178 	.ifq_head = NULL,
    179 	.ifq_tail = NULL,
    180 	.ifq_len = 0,
    181 	.ifq_maxlen = 50,
    182 	.ifq_drops = 0,
    183 };
    184 static int	arp_inuse, arp_allocated;
    185 static int	arp_maxtries = 5;
    186 static int	useloopback = 1;	/* use loopback interface for local traffic */
    187 static int	arpinit_done = 0;
    188 
    189 static percpu_t *arpstat_percpu;
    190 
    191 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
    192 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
    193 
    194 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
    195 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
    196 
    197 struct	callout arptimer_ch;
    198 
    199 /* revarp state */
    200 static struct	in_addr myip, srv_ip;
    201 static int	myip_initialized = 0;
    202 static int	revarp_in_progress = 0;
    203 static struct	ifnet *myip_ifp = NULL;
    204 
    205 #ifdef DDB
    206 static void db_print_sa(const struct sockaddr *);
    207 static void db_print_ifa(struct ifaddr *);
    208 static void db_print_llinfo(void *);
    209 static int db_show_rtentry(struct rtentry *, void *);
    210 #endif
    211 
    212 static int arp_drainwanted;
    213 
    214 static int log_movements = 1;
    215 static int log_permanent_modify = 1;
    216 static int log_wrong_iface = 1;
    217 static int log_unknown_network = 1;
    218 
    219 /*
    220  * this should be elsewhere.
    221  */
    222 
    223 static char *
    224 lla_snprintf(u_int8_t *, int);
    225 
    226 static char *
    227 lla_snprintf(u_int8_t *adrp, int len)
    228 {
    229 #define NUMBUFS 3
    230 	static char buf[NUMBUFS][16*3];
    231 	static int bnum = 0;
    232 
    233 	int i;
    234 	char *p;
    235 
    236 	p = buf[bnum];
    237 
    238 	*p++ = hexdigits[(*adrp)>>4];
    239 	*p++ = hexdigits[(*adrp++)&0xf];
    240 
    241 	for (i=1; i<len && i<16; i++) {
    242 		*p++ = ':';
    243 		*p++ = hexdigits[(*adrp)>>4];
    244 		*p++ = hexdigits[(*adrp++)&0xf];
    245 	}
    246 
    247 	*p = 0;
    248 	p = buf[bnum];
    249 	bnum = (bnum + 1) % NUMBUFS;
    250 	return p;
    251 }
    252 
    253 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
    254 
    255 static void
    256 arp_fasttimo(void)
    257 {
    258 	if (arp_drainwanted) {
    259 		arp_drain();
    260 		arp_drainwanted = 0;
    261 	}
    262 }
    263 
    264 const struct protosw arpsw[] = {
    265 	{ .pr_type = 0,
    266 	  .pr_domain = &arpdomain,
    267 	  .pr_protocol = 0,
    268 	  .pr_flags = 0,
    269 	  .pr_input = 0,
    270 	  .pr_output = 0,
    271 	  .pr_ctlinput = 0,
    272 	  .pr_ctloutput = 0,
    273 	  .pr_usrreqs = 0,
    274 	  .pr_init = arp_init,
    275 	  .pr_fasttimo = arp_fasttimo,
    276 	  .pr_slowtimo = 0,
    277 	  .pr_drain = arp_drainstub,
    278 	}
    279 };
    280 
    281 struct domain arpdomain = {
    282 	.dom_family = PF_ARP,
    283 	.dom_name = "arp",
    284 	.dom_protosw = arpsw,
    285 	.dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)],
    286 };
    287 
    288 /*
    289  * ARP table locking.
    290  *
    291  * to prevent lossage vs. the arp_drain routine (which may be called at
    292  * any time, including in a device driver context), we do two things:
    293  *
    294  * 1) manipulation of la->la_hold is done at splnet() (for all of
    295  * about two instructions).
    296  *
    297  * 2) manipulation of the arp table's linked list is done under the
    298  * protection of the ARP_LOCK; if arp_drain() or arptimer is called
    299  * while the arp table is locked, we punt and try again later.
    300  */
    301 
    302 static int	arp_locked;
    303 static inline int arp_lock_try(int);
    304 static inline void arp_unlock(void);
    305 
    306 static inline int
    307 arp_lock_try(int recurse)
    308 {
    309 	int s;
    310 
    311 	/*
    312 	 * Use splvm() -- we're blocking things that would cause
    313 	 * mbuf allocation.
    314 	 */
    315 	s = splvm();
    316 	if (!recurse && arp_locked) {
    317 		splx(s);
    318 		return 0;
    319 	}
    320 	arp_locked++;
    321 	splx(s);
    322 	return 1;
    323 }
    324 
    325 static inline void
    326 arp_unlock(void)
    327 {
    328 	int s;
    329 
    330 	s = splvm();
    331 	arp_locked--;
    332 	splx(s);
    333 }
    334 
    335 #ifdef DIAGNOSTIC
    336 #define	ARP_LOCK(recurse)						\
    337 do {									\
    338 	if (arp_lock_try(recurse) == 0) {				\
    339 		printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \
    340 		panic("arp_lock");					\
    341 	}								\
    342 } while (/*CONSTCOND*/ 0)
    343 #define	ARP_LOCK_CHECK()						\
    344 do {									\
    345 	if (arp_locked == 0) {						\
    346 		printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \
    347 		panic("arp lock check");				\
    348 	}								\
    349 } while (/*CONSTCOND*/ 0)
    350 #else
    351 #define	ARP_LOCK(x)		(void) arp_lock_try(x)
    352 #define	ARP_LOCK_CHECK()	/* nothing */
    353 #endif
    354 
    355 #define	ARP_UNLOCK()		arp_unlock()
    356 
    357 static void sysctl_net_inet_arp_setup(struct sysctllog **);
    358 
    359 void
    360 arp_init(void)
    361 {
    362 
    363 	sysctl_net_inet_arp_setup(NULL);
    364 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
    365 }
    366 
    367 static void
    368 arp_drainstub(void)
    369 {
    370 	arp_drainwanted = 1;
    371 }
    372 
    373 /*
    374  * ARP protocol drain routine.  Called when memory is in short supply.
    375  * Called at splvm();  don't acquire softnet_lock as can be called from
    376  * hardware interrupt handlers.
    377  */
    378 void
    379 arp_drain(void)
    380 {
    381 	struct llinfo_arp *la;
    382 	int count = 0;
    383 	struct mbuf *mold;
    384 
    385 	KERNEL_LOCK(1, NULL);
    386 
    387 	if (arp_lock_try(0) == 0) {
    388 		KERNEL_UNLOCK_ONE(NULL);
    389 		return;
    390 	}
    391 
    392 	LIST_FOREACH(la, &llinfo_arp, la_list) {
    393 		mold = la->la_hold;
    394 		la->la_hold = NULL;
    395 
    396 		if (mold) {
    397 			m_freem(mold);
    398 			count++;
    399 		}
    400 	}
    401 	ARP_UNLOCK();
    402 	ARP_STATADD(ARP_STAT_DFRDROPPED, count);
    403 	KERNEL_UNLOCK_ONE(NULL);
    404 }
    405 
    406 
    407 /*
    408  * Timeout routine.  Age arp_tab entries periodically.
    409  */
    410 /* ARGSUSED */
    411 static void
    412 arptimer(void *arg)
    413 {
    414 	struct llinfo_arp *la, *nla;
    415 
    416 	mutex_enter(softnet_lock);
    417 	KERNEL_LOCK(1, NULL);
    418 
    419 	if (arp_lock_try(0) == 0) {
    420 		/* get it later.. */
    421 		KERNEL_UNLOCK_ONE(NULL);
    422 		mutex_exit(softnet_lock);
    423 		return;
    424 	}
    425 
    426 	callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL);
    427 	LIST_FOREACH_SAFE(la, &llinfo_arp, la_list, nla) {
    428 		struct rtentry *rt = la->la_rt;
    429 
    430 		if (rt->rt_expire == 0)
    431 			continue;
    432 		if ((rt->rt_expire - time_uptime) < arpt_refresh &&
    433 		    rt->rt_pksent > (time_uptime - arpt_keep)) {
    434 			/*
    435 			 * If the entry has been used during since last
    436 			 * refresh, try to renew it before deleting.
    437 			 */
    438 			arprequest(rt->rt_ifp,
    439 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
    440 			    &satocsin(rt_getkey(rt))->sin_addr,
    441 			    CLLADDR(rt->rt_ifp->if_sadl));
    442 		} else if (rt->rt_expire <= time_uptime)
    443 			arptfree(la); /* timer has expired; clear */
    444 	}
    445 
    446 	ARP_UNLOCK();
    447 
    448 	KERNEL_UNLOCK_ONE(NULL);
    449 	mutex_exit(softnet_lock);
    450 }
    451 
    452 /*
    453  * We set the gateway for RTF_CLONING routes to a "prototype"
    454  * link-layer sockaddr whose interface type (if_type) and interface
    455  * index (if_index) fields are prepared.
    456  */
    457 static struct sockaddr *
    458 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
    459     const struct sockaddr *netmask)
    460 {
    461 	const struct ifnet *ifp = rt->rt_ifp;
    462 	uint8_t namelen = strlen(ifp->if_xname);
    463 	uint8_t addrlen = ifp->if_addrlen;
    464 
    465 	/*
    466 	 * XXX: If this is a manually added route to interface
    467 	 * such as older version of routed or gated might provide,
    468 	 * restore cloning bit.
    469 	 */
    470 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
    471 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
    472 		rt->rt_flags |= RTF_CLONING;
    473 	if (rt->rt_flags & RTF_CLONING ||
    474 	    ((rt->rt_flags & (RTF_LLINFO | RTF_LOCAL)) && !rt->rt_llinfo))
    475 	{
    476 		union {
    477 			struct sockaddr sa;
    478 			struct sockaddr_storage ss;
    479 			struct sockaddr_dl sdl;
    480 		} u;
    481 		/*
    482 		 * Case 1: This route should come from a route to iface.
    483 		 */
    484 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
    485 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
    486 		rt_setgate(rt, &u.sa);
    487 		gate = rt->rt_gateway;
    488 	}
    489 	return gate;
    490 }
    491 
    492 /*
    493  * Parallel to llc_rtrequest.
    494  */
    495 void
    496 arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info)
    497 {
    498 	struct sockaddr *gate = rt->rt_gateway;
    499 	struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo;
    500 	size_t allocsize;
    501 	struct mbuf *mold;
    502 	int s;
    503 	struct in_ifaddr *ia;
    504 	struct ifaddr *ifa;
    505 	struct ifnet *ifp = rt->rt_ifp;
    506 
    507 	if (!arpinit_done) {
    508 		arpinit_done = 1;
    509 		callout_init(&arptimer_ch, CALLOUT_MPSAFE);
    510 		callout_reset(&arptimer_ch, hz, arptimer, NULL);
    511 	}
    512 
    513 	if (req == RTM_LLINFO_UPD) {
    514 		struct in_addr *in;
    515 
    516 		if ((ifa = info->rti_ifa) == NULL)
    517 			return;
    518 
    519 		in = &ifatoia(ifa)->ia_addr.sin_addr;
    520 
    521 		if (ifatoia(ifa)->ia4_flags &
    522 		    (IN_IFF_NOTREADY | IN_IFF_DETACHED))
    523 		{
    524 			arplog((LOG_DEBUG, "arp_request: %s not ready\n",
    525 			   in_fmtaddr(*in)));
    526 			return;
    527 		}
    528 
    529 		arprequest(ifa->ifa_ifp, in, in,
    530 		    CLLADDR(ifa->ifa_ifp->if_sadl));
    531 		return;
    532 	}
    533 
    534 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
    535 		if (req != RTM_ADD)
    536 			return;
    537 
    538 		/*
    539 		 * linklayers with particular link MTU limitation.
    540 		 */
    541 		switch(ifp->if_type) {
    542 #if NFDDI > 0
    543 		case IFT_FDDI:
    544 			if (ifp->if_mtu > FDDIIPMTU)
    545 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
    546 			break;
    547 #endif
    548 #if NARCNET > 0
    549 		case IFT_ARCNET:
    550 		    {
    551 			int arcipifmtu;
    552 
    553 			if (ifp->if_flags & IFF_LINK0)
    554 				arcipifmtu = arc_ipmtu;
    555 			else
    556 				arcipifmtu = ARCMTU;
    557 			if (ifp->if_mtu > arcipifmtu)
    558 				rt->rt_rmx.rmx_mtu = arcipifmtu;
    559 			break;
    560 		    }
    561 #endif
    562 		}
    563 		return;
    564 	}
    565 
    566 	ARP_LOCK(1);		/* we may already be locked here. */
    567 
    568 	switch (req) {
    569 	case RTM_SETGATE:
    570 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
    571 		break;
    572 	case RTM_ADD:
    573 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
    574 		if (rt->rt_flags & RTF_CLONING ||
    575 		    ((rt->rt_flags & (RTF_LLINFO | RTF_LOCAL)) && !la))
    576 		{
    577 			/*
    578 			 * Give this route an expiration time, even though
    579 			 * it's a "permanent" route, so that routes cloned
    580 			 * from it do not need their expiration time set.
    581 			 */
    582 			KASSERT(time_uptime != 0);
    583 			rt->rt_expire = time_uptime;
    584 			/*
    585 			 * linklayers with particular link MTU limitation.
    586 			 */
    587 			switch (ifp->if_type) {
    588 #if NFDDI > 0
    589 			case IFT_FDDI:
    590 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
    591 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
    592 				     (rt->rt_rmx.rmx_mtu == 0 &&
    593 				      ifp->if_mtu > FDDIIPMTU)))
    594 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
    595 				break;
    596 #endif
    597 #if NARCNET > 0
    598 			case IFT_ARCNET:
    599 			    {
    600 				int arcipifmtu;
    601 				if (ifp->if_flags & IFF_LINK0)
    602 					arcipifmtu = arc_ipmtu;
    603 				else
    604 					arcipifmtu = ARCMTU;
    605 
    606 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
    607 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
    608 				     (rt->rt_rmx.rmx_mtu == 0 &&
    609 				      ifp->if_mtu > arcipifmtu)))
    610 					rt->rt_rmx.rmx_mtu = arcipifmtu;
    611 				break;
    612 			    }
    613 #endif
    614 			}
    615 			if (rt->rt_flags & RTF_CLONING)
    616 				break;
    617 		}
    618 		/* Announce a new entry if requested. */
    619 		if (rt->rt_flags & RTF_ANNOUNCE) {
    620 			INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia);
    621 			while (ia && ia->ia_ifp != ifp)
    622 				NEXT_IA_WITH_SAME_ADDR(ia);
    623 			if (ia == NULL ||
    624 			    ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
    625 				;
    626 			else
    627 				arprequest(ifp,
    628 				    &satocsin(rt_getkey(rt))->sin_addr,
    629 				    &satocsin(rt_getkey(rt))->sin_addr,
    630 				    CLLADDR(satocsdl(gate)));
    631 		}
    632 		/*FALLTHROUGH*/
    633 	case RTM_RESOLVE:
    634 		if (gate->sa_family != AF_LINK ||
    635 		    gate->sa_len < sockaddr_dl_measure(0, ifp->if_addrlen)) {
    636 			log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n");
    637 			break;
    638 		}
    639 
    640 		satosdl(gate)->sdl_type = ifp->if_type;
    641 		satosdl(gate)->sdl_index = ifp->if_index;
    642 		if (la != NULL)
    643 			break; /* This happens on a route change */
    644 
    645 		/* If the route is for a broadcast address mark it as such.
    646 		 * This way we can avoid an expensive call to in_broadcast()
    647 		 * in ip_output() most of the time (because the route passed
    648 		 * to ip_output() is almost always a host route). */
    649 		if (rt->rt_flags & RTF_HOST &&
    650 		    !(rt->rt_flags & RTF_BROADCAST) &&
    651 		    in_broadcast(satocsin(rt_getkey(rt))->sin_addr, rt->rt_ifp))
    652 			rt->rt_flags |= RTF_BROADCAST;
    653 		/* There is little point in resolving the broadcast address */
    654 		if (rt->rt_flags & RTF_BROADCAST)
    655 			break;
    656 
    657 		/*
    658 		 * Case 2:  This route may come from cloning, or a manual route
    659 		 * add with a LL address.
    660 		 */
    661 		switch (ifp->if_type) {
    662 #if NTOKEN > 0
    663 		case IFT_ISO88025:
    664 			allocsize = sizeof(*la) + sizeof(struct token_rif);
    665 			break;
    666 #endif /* NTOKEN > 0 */
    667 		default:
    668 			allocsize = sizeof(*la);
    669 		}
    670 		R_Malloc(la, struct llinfo_arp *, allocsize);
    671 		rt->rt_llinfo = (void *)la;
    672 		if (la == NULL) {
    673 			log(LOG_DEBUG, "arp_rtrequest: malloc failed\n");
    674 			break;
    675 		}
    676 		arp_inuse++, arp_allocated++;
    677 		memset(la, 0, allocsize);
    678 		la->la_rt = rt;
    679 		rt->rt_flags |= RTF_LLINFO;
    680 		LIST_INSERT_HEAD(&llinfo_arp, la, la_list);
    681 
    682 		INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia);
    683 		while (ia && ia->ia_ifp != ifp)
    684 			NEXT_IA_WITH_SAME_ADDR(ia);
    685 		if (ia) {
    686 			/*
    687 			 * This test used to be
    688 			 *	if (lo0ifp->if_flags & IFF_UP)
    689 			 * It allowed local traffic to be forced through
    690 			 * the hardware by configuring the loopback down.
    691 			 * However, it causes problems during network
    692 			 * configuration for boards that can't receive
    693 			 * packets they send.  It is now necessary to clear
    694 			 * "useloopback" and remove the route to force
    695 			 * traffic out to the hardware.
    696 			 *
    697 			 * In 4.4BSD, the above "if" statement checked
    698 			 * rt->rt_ifa against rt_getkey(rt).  It was changed
    699 			 * to the current form so that we can provide a
    700 			 * better support for multiple IPv4 addresses on a
    701 			 * interface.
    702 			 */
    703 			rt->rt_expire = 0;
    704 			if (sockaddr_dl_init(satosdl(gate), gate->sa_len,
    705 			    ifp->if_index, ifp->if_type, NULL, 0,
    706 			    CLLADDR(ifp->if_sadl), ifp->if_addrlen) == NULL) {
    707 				panic("%s(%s): sockaddr_dl_init cannot fail",
    708 				    __func__, ifp->if_xname);
    709 			}
    710 			if (useloopback)
    711 				ifp = rt->rt_ifp = lo0ifp;
    712 			/*
    713 			 * make sure to set rt->rt_ifa to the interface
    714 			 * address we are using, otherwise we will have trouble
    715 			 * with source address selection.
    716 			 */
    717 			ifa = &ia->ia_ifa;
    718 			if (ifa != rt->rt_ifa)
    719 				rt_replace_ifa(rt, ifa);
    720 		}
    721 		break;
    722 
    723 	case RTM_DELETE:
    724 		if (la == NULL)
    725 			break;
    726 		arp_inuse--;
    727 		LIST_REMOVE(la, la_list);
    728 		rt->rt_llinfo = NULL;
    729 		rt->rt_flags &= ~RTF_LLINFO;
    730 
    731 		s = splnet();
    732 		mold = la->la_hold;
    733 		la->la_hold = NULL;
    734 		splx(s);
    735 
    736 		if (mold)
    737 			m_freem(mold);
    738 
    739 		Free((void *)la);
    740 	}
    741 	ARP_UNLOCK();
    742 }
    743 
    744 /*
    745  * Broadcast an ARP request. Caller specifies:
    746  *	- arp header source ip address
    747  *	- arp header target ip address
    748  *	- arp header source ethernet address
    749  */
    750 void
    751 arprequest(struct ifnet *ifp,
    752     const struct in_addr *sip, const struct in_addr *tip,
    753     const u_int8_t *enaddr)
    754 {
    755 	struct mbuf *m;
    756 	struct arphdr *ah;
    757 	struct sockaddr sa;
    758 	uint64_t *arps;
    759 
    760 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
    761 		return;
    762 	MCLAIM(m, &arpdomain.dom_mowner);
    763 	switch (ifp->if_type) {
    764 	case IFT_IEEE1394:
    765 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
    766 		    ifp->if_addrlen;
    767 		break;
    768 	default:
    769 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
    770 		    2 * ifp->if_addrlen;
    771 		break;
    772 	}
    773 	m->m_pkthdr.len = m->m_len;
    774 	MH_ALIGN(m, m->m_len);
    775 	ah = mtod(m, struct arphdr *);
    776 	memset(ah, 0, m->m_len);
    777 	switch (ifp->if_type) {
    778 	case IFT_IEEE1394:	/* RFC2734 */
    779 		/* fill it now for ar_tpa computation */
    780 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
    781 		break;
    782 	default:
    783 		/* ifp->if_output will fill ar_hrd */
    784 		break;
    785 	}
    786 	ah->ar_pro = htons(ETHERTYPE_IP);
    787 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
    788 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
    789 	ah->ar_op = htons(ARPOP_REQUEST);
    790 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
    791 	memcpy(ar_spa(ah), sip, ah->ar_pln);
    792 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
    793 	sa.sa_family = AF_ARP;
    794 	sa.sa_len = 2;
    795 	m->m_flags |= M_BCAST;
    796 	arps = ARP_STAT_GETREF();
    797 	arps[ARP_STAT_SNDTOTAL]++;
    798 	arps[ARP_STAT_SENDREQUEST]++;
    799 	ARP_STAT_PUTREF();
    800 	(*ifp->if_output)(ifp, m, &sa, NULL);
    801 }
    802 
    803 /*
    804  * Resolve an IP address into an ethernet address.  If success,
    805  * desten is filled in.  If there is no entry in arptab,
    806  * set one up and broadcast a request for the IP address.
    807  * Hold onto this mbuf and resend it once the address
    808  * is finally resolved.  A return value of 1 indicates
    809  * that desten has been filled in and the packet should be sent
    810  * normally; a 0 return indicates that the packet has been
    811  * taken over here, either now or for later transmission.
    812  */
    813 int
    814 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m,
    815     const struct sockaddr *dst, u_char *desten)
    816 {
    817 	struct llinfo_arp *la;
    818 	const struct sockaddr_dl *sdl;
    819 	struct mbuf *mold;
    820 	int s;
    821 
    822 	if ((la = arplookup1(m, &satocsin(dst)->sin_addr, 1, 0, rt)) != NULL)
    823 		rt = la->la_rt;
    824 
    825 	if (la == NULL || rt == NULL) {
    826 		ARP_STATINC(ARP_STAT_ALLOCFAIL);
    827 		log(LOG_DEBUG,
    828 		    "arpresolve: can't allocate llinfo on %s for %s\n",
    829 		    ifp->if_xname, in_fmtaddr(satocsin(dst)->sin_addr));
    830 		m_freem(m);
    831 		return 0;
    832 	}
    833 	sdl = satocsdl(rt->rt_gateway);
    834 	/*
    835 	 * Check the address family and length is valid, the address
    836 	 * is resolved; otherwise, try to resolve.
    837 	 */
    838 	if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) &&
    839 	    sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) {
    840 		memcpy(desten, CLLADDR(sdl),
    841 		    min(sdl->sdl_alen, ifp->if_addrlen));
    842 		rt->rt_pksent = time_uptime; /* Time for last pkt sent */
    843 		return 1;
    844 	}
    845 	/*
    846 	 * There is an arptab entry, but no ethernet address
    847 	 * response yet.  Replace the held mbuf with this
    848 	 * latest one.
    849 	 */
    850 
    851 	ARP_STATINC(ARP_STAT_DFRTOTAL);
    852 	s = splnet();
    853 	mold = la->la_hold;
    854 	la->la_hold = m;
    855 	splx(s);
    856 
    857 	if (mold) {
    858 		ARP_STATINC(ARP_STAT_DFRDROPPED);
    859 		m_freem(mold);
    860 	}
    861 
    862 	/*
    863 	 * Re-send the ARP request when appropriate.
    864 	 */
    865 #ifdef	DIAGNOSTIC
    866 	if (rt->rt_expire == 0) {
    867 		/* This should never happen. (Should it? -gwr) */
    868 		printf("arpresolve: unresolved and rt_expire == 0\n");
    869 		/* Set expiration time to now (expired). */
    870 		rt->rt_expire = time_uptime;
    871 	}
    872 #endif
    873 	if (rt->rt_expire) {
    874 		rt->rt_flags &= ~RTF_REJECT;
    875 		if (la->la_asked == 0 || rt->rt_expire != time_uptime) {
    876 			rt->rt_expire = time_uptime;
    877 			if (la->la_asked++ < arp_maxtries) {
    878 				arprequest(ifp,
    879 				    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
    880 				    &satocsin(dst)->sin_addr,
    881 #if NCARP > 0
    882 				    (rt->rt_ifp->if_type == IFT_CARP) ?
    883 				    CLLADDR(rt->rt_ifp->if_sadl):
    884 #endif
    885 				    CLLADDR(ifp->if_sadl));
    886 			} else {
    887 				rt->rt_flags |= RTF_REJECT;
    888 				rt->rt_expire += arpt_down;
    889 				la->la_asked = 0;
    890 			}
    891 		}
    892 	}
    893 	return 0;
    894 }
    895 
    896 /*
    897  * Common length and type checks are done here,
    898  * then the protocol-specific routine is called.
    899  */
    900 void
    901 arpintr(void)
    902 {
    903 	struct mbuf *m;
    904 	struct arphdr *ar;
    905 	int s;
    906 	int arplen;
    907 
    908 	mutex_enter(softnet_lock);
    909 	KERNEL_LOCK(1, NULL);
    910 	while (arpintrq.ifq_head) {
    911 		s = splnet();
    912 		IF_DEQUEUE(&arpintrq, m);
    913 		splx(s);
    914 		if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
    915 			panic("arpintr");
    916 
    917 		MCLAIM(m, &arpdomain.dom_mowner);
    918 		ARP_STATINC(ARP_STAT_RCVTOTAL);
    919 
    920 		/*
    921 		 * First, make sure we have at least struct arphdr.
    922 		 */
    923 		if (m->m_len < sizeof(struct arphdr) ||
    924 		    (ar = mtod(m, struct arphdr *)) == NULL)
    925 			goto badlen;
    926 
    927 		switch (m->m_pkthdr.rcvif->if_type) {
    928 		case IFT_IEEE1394:
    929 			arplen = sizeof(struct arphdr) +
    930 			    ar->ar_hln + 2 * ar->ar_pln;
    931 			break;
    932 		default:
    933 			arplen = sizeof(struct arphdr) +
    934 			    2 * ar->ar_hln + 2 * ar->ar_pln;
    935 			break;
    936 		}
    937 
    938 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
    939 		    m->m_len >= arplen)
    940 			switch (ntohs(ar->ar_pro)) {
    941 			case ETHERTYPE_IP:
    942 			case ETHERTYPE_IPTRAILERS:
    943 				in_arpinput(m);
    944 				continue;
    945 			default:
    946 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
    947 			}
    948 		else {
    949 badlen:
    950 			ARP_STATINC(ARP_STAT_RCVBADLEN);
    951 		}
    952 		m_freem(m);
    953 	}
    954 	KERNEL_UNLOCK_ONE(NULL);
    955 	mutex_exit(softnet_lock);
    956 }
    957 
    958 /*
    959  * ARP for Internet protocols on 10 Mb/s Ethernet.
    960  * Algorithm is that given in RFC 826.
    961  * In addition, a sanity check is performed on the sender
    962  * protocol address, to catch impersonators.
    963  * We no longer handle negotiations for use of trailer protocol:
    964  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
    965  * along with IP replies if we wanted trailers sent to us,
    966  * and also sent them in response to IP replies.
    967  * This allowed either end to announce the desire to receive
    968  * trailer packets.
    969  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
    970  * but formerly didn't normally send requests.
    971  */
    972 static void
    973 in_arpinput(struct mbuf *m)
    974 {
    975 	struct arphdr *ah;
    976 	struct ifnet *ifp = m->m_pkthdr.rcvif;
    977 	struct llinfo_arp *la = NULL;
    978 	struct rtentry  *rt;
    979 	struct in_ifaddr *ia;
    980 #if NBRIDGE > 0
    981 	struct in_ifaddr *bridge_ia = NULL;
    982 #endif
    983 #if NCARP > 0
    984 	u_int32_t count = 0, index = 0;
    985 #endif
    986 	struct sockaddr_dl *sdl;
    987 	struct sockaddr sa;
    988 	struct in_addr isaddr, itaddr, myaddr;
    989 	int op;
    990 	struct mbuf *mold;
    991 	void *tha;
    992 	int s;
    993 	uint64_t *arps;
    994 
    995 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
    996 		goto out;
    997 	ah = mtod(m, struct arphdr *);
    998 	op = ntohs(ah->ar_op);
    999 
   1000 	/*
   1001 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
   1002 	 * ar_tpa().
   1003 	 */
   1004 	switch (ifp->if_type) {
   1005 	case IFT_IEEE1394:
   1006 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
   1007 			;
   1008 		else {
   1009 			/* XXX this is to make sure we compute ar_tha right */
   1010 			/* XXX check ar_hrd more strictly? */
   1011 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
   1012 		}
   1013 		break;
   1014 	default:
   1015 		/* XXX check ar_hrd? */
   1016 		break;
   1017 	}
   1018 
   1019 	memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
   1020 	memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
   1021 
   1022 	if (m->m_flags & (M_BCAST|M_MCAST))
   1023 		ARP_STATINC(ARP_STAT_RCVMCAST);
   1024 
   1025 
   1026 	/*
   1027 	 * Search for a matching interface address
   1028 	 * or any address on the interface to use
   1029 	 * as a dummy address in the rest of this function
   1030 	 */
   1031 
   1032 	INADDR_TO_IA(itaddr, ia);
   1033 	while (ia != NULL) {
   1034 #if NCARP > 0
   1035 		if (ia->ia_ifp->if_type == IFT_CARP &&
   1036 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
   1037 		    (IFF_UP|IFF_RUNNING))) {
   1038 			index++;
   1039 			if (ia->ia_ifp == m->m_pkthdr.rcvif &&
   1040 			    carp_iamatch(ia, ar_sha(ah),
   1041 			    &count, index)) {
   1042 				break;
   1043 				}
   1044 		} else
   1045 #endif
   1046 			    if (ia->ia_ifp == m->m_pkthdr.rcvif)
   1047 				break;
   1048 #if NBRIDGE > 0
   1049 		/*
   1050 		 * If the interface we received the packet on
   1051 		 * is part of a bridge, check to see if we need
   1052 		 * to "bridge" the packet to ourselves at this
   1053 		 * layer.  Note we still prefer a perfect match,
   1054 		 * but allow this weaker match if necessary.
   1055 		 */
   1056 		if (m->m_pkthdr.rcvif->if_bridge != NULL &&
   1057 		    m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge)
   1058 			bridge_ia = ia;
   1059 #endif /* NBRIDGE > 0 */
   1060 
   1061 		NEXT_IA_WITH_SAME_ADDR(ia);
   1062 	}
   1063 
   1064 #if NBRIDGE > 0
   1065 	if (ia == NULL && bridge_ia != NULL) {
   1066 		ia = bridge_ia;
   1067 		ifp = bridge_ia->ia_ifp;
   1068 	}
   1069 #endif
   1070 
   1071 	if (ia == NULL) {
   1072 		INADDR_TO_IA(isaddr, ia);
   1073 		while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif)
   1074 			NEXT_IA_WITH_SAME_ADDR(ia);
   1075 
   1076 		if (ia == NULL) {
   1077 			IFP_TO_IA(ifp, ia);
   1078 			if (ia == NULL) {
   1079 				ARP_STATINC(ARP_STAT_RCVNOINT);
   1080 				goto out;
   1081 			}
   1082 		}
   1083 	}
   1084 
   1085 	myaddr = ia->ia_addr.sin_addr;
   1086 
   1087 	/* XXX checks for bridge case? */
   1088 	if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) {
   1089 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
   1090 		goto out;	/* it's from me, ignore it. */
   1091 	}
   1092 
   1093 	/* XXX checks for bridge case? */
   1094 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
   1095 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
   1096 		log(LOG_ERR,
   1097 		    "%s: arp: link address is broadcast for IP address %s!\n",
   1098 		    ifp->if_xname, in_fmtaddr(isaddr));
   1099 		goto out;
   1100 	}
   1101 
   1102 	/*
   1103 	 * If the source IP address is zero, this is an RFC 5227 ARP probe
   1104 	 */
   1105 	if (in_nullhost(isaddr))
   1106 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
   1107 	else if (in_hosteq(isaddr, myaddr))
   1108 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
   1109 
   1110 	if (in_nullhost(itaddr))
   1111 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
   1112 
   1113 	/* DAD check, RFC 5227 2.1.1, Probe Details */
   1114 	if (in_hosteq(isaddr, myaddr) ||
   1115 	    (in_nullhost(isaddr) && in_hosteq(itaddr, myaddr)))
   1116 	{
   1117 		/* If our address is tentative, mark it as duplicated */
   1118 		if (ia->ia4_flags & IN_IFF_TENTATIVE)
   1119 			arp_dad_duplicated((struct ifaddr *)ia);
   1120 		/* If our address is unuseable, don't reply */
   1121 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
   1122 			goto out;
   1123 	}
   1124 
   1125 	/*
   1126 	 * If the target IP address is zero, ignore the packet.
   1127 	 * This prevents the code below from tring to answer
   1128 	 * when we are using IP address zero (booting).
   1129 	 */
   1130 	if (in_nullhost(itaddr))
   1131 		goto out;
   1132 
   1133 	if (in_nullhost(isaddr))
   1134 		goto reply;
   1135 
   1136 	if (in_hosteq(isaddr, myaddr)) {
   1137 		log(LOG_ERR,
   1138 		   "duplicate IP address %s sent from link address %s\n",
   1139 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
   1140 		itaddr = myaddr;
   1141 		goto reply;
   1142 	}
   1143 	la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0);
   1144 	if (la != NULL && (rt = la->la_rt) && (sdl = satosdl(rt->rt_gateway))) {
   1145 		if (sdl->sdl_alen &&
   1146 		    memcmp(ar_sha(ah), CLLADDR(sdl), sdl->sdl_alen)) {
   1147 			if (rt->rt_flags & RTF_STATIC) {
   1148 				ARP_STATINC(ARP_STAT_RCVOVERPERM);
   1149 				if (!log_permanent_modify)
   1150 					goto out;
   1151 				log(LOG_INFO,
   1152 				    "%s tried to overwrite permanent arp info"
   1153 				    " for %s\n",
   1154 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
   1155 				    in_fmtaddr(isaddr));
   1156 				goto out;
   1157 			} else if (rt->rt_ifp != ifp) {
   1158 				ARP_STATINC(ARP_STAT_RCVOVERINT);
   1159 				if (!log_wrong_iface)
   1160 					goto out;
   1161 				log(LOG_INFO,
   1162 				    "%s on %s tried to overwrite "
   1163 				    "arp info for %s on %s\n",
   1164 				    lla_snprintf(ar_sha(ah), ah->ar_hln),
   1165 				    ifp->if_xname, in_fmtaddr(isaddr),
   1166 				    rt->rt_ifp->if_xname);
   1167 				    goto out;
   1168 			} else {
   1169 				ARP_STATINC(ARP_STAT_RCVOVER);
   1170 				if (log_movements)
   1171 					log(LOG_INFO, "arp info overwritten "
   1172 					    "for %s by %s\n",
   1173 					    in_fmtaddr(isaddr),
   1174 					    lla_snprintf(ar_sha(ah),
   1175 					    ah->ar_hln));
   1176 			}
   1177 		}
   1178 		/*
   1179 		 * sanity check for the address length.
   1180 		 * XXX this does not work for protocols with variable address
   1181 		 * length. -is
   1182 		 */
   1183 		if (sdl->sdl_alen &&
   1184 		    sdl->sdl_alen != ah->ar_hln) {
   1185 			ARP_STATINC(ARP_STAT_RCVLENCHG);
   1186 			log(LOG_WARNING,
   1187 			    "arp from %s: new addr len %d, was %d\n",
   1188 			    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
   1189 		}
   1190 		if (ifp->if_addrlen != ah->ar_hln) {
   1191 			ARP_STATINC(ARP_STAT_RCVBADLEN);
   1192 			log(LOG_WARNING,
   1193 			    "arp from %s: addr len: new %d, i/f %d (ignored)\n",
   1194 			    in_fmtaddr(isaddr), ah->ar_hln,
   1195 			    ifp->if_addrlen);
   1196 			goto reply;
   1197 		}
   1198 #if NTOKEN > 0
   1199 		/*
   1200 		 * XXX uses m_data and assumes the complete answer including
   1201 		 * XXX token-ring headers is in the same buf
   1202 		 */
   1203 		if (ifp->if_type == IFT_ISO88025) {
   1204 			struct token_header *trh;
   1205 
   1206 			trh = (struct token_header *)M_TRHSTART(m);
   1207 			if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
   1208 				struct token_rif	*rif;
   1209 				size_t	riflen;
   1210 
   1211 				rif = TOKEN_RIF(trh);
   1212 				riflen = (ntohs(rif->tr_rcf) &
   1213 				    TOKEN_RCF_LEN_MASK) >> 8;
   1214 
   1215 				if (riflen > 2 &&
   1216 				    riflen < sizeof(struct token_rif) &&
   1217 				    (riflen & 1) == 0) {
   1218 					rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
   1219 					rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
   1220 					memcpy(TOKEN_RIF(la), rif, riflen);
   1221 				}
   1222 			}
   1223 		}
   1224 #endif /* NTOKEN > 0 */
   1225 		(void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, ar_sha(ah),
   1226 		    ah->ar_hln);
   1227 		if (rt->rt_expire)
   1228 			rt->rt_expire = time_uptime + arpt_keep;
   1229 		rt->rt_flags &= ~RTF_REJECT;
   1230 		la->la_asked = 0;
   1231 
   1232 		s = splnet();
   1233 		mold = la->la_hold;
   1234 		la->la_hold = NULL;
   1235 		splx(s);
   1236 
   1237 		if (mold) {
   1238 			ARP_STATINC(ARP_STAT_DFRSENT);
   1239 			(*ifp->if_output)(ifp, mold, rt_getkey(rt), rt);
   1240 		}
   1241 	}
   1242 reply:
   1243 	if (op != ARPOP_REQUEST) {
   1244 		if (op == ARPOP_REPLY)
   1245 			ARP_STATINC(ARP_STAT_RCVREPLY);
   1246 		goto out;
   1247 	}
   1248 	ARP_STATINC(ARP_STAT_RCVREQUEST);
   1249 	if (in_hosteq(itaddr, myaddr)) {
   1250 		/* If our address is unuseable, don't reply */
   1251 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
   1252 			goto out;
   1253 		/* I am the target */
   1254 		tha = ar_tha(ah);
   1255 		if (tha)
   1256 			memcpy(tha, ar_sha(ah), ah->ar_hln);
   1257 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
   1258 	} else {
   1259 		la = arplookup(m, &itaddr, 0, SIN_PROXY);
   1260 		if (la == NULL)
   1261 			goto out;
   1262 		rt = la->la_rt;
   1263 		if (rt->rt_ifp->if_type == IFT_CARP &&
   1264 		    m->m_pkthdr.rcvif->if_type != IFT_CARP)
   1265 			goto out;
   1266 		tha = ar_tha(ah);
   1267 		if (tha)
   1268 			memcpy(tha, ar_sha(ah), ah->ar_hln);
   1269 		sdl = satosdl(rt->rt_gateway);
   1270 		memcpy(ar_sha(ah), CLLADDR(sdl), ah->ar_hln);
   1271 	}
   1272 
   1273 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
   1274 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
   1275 	ah->ar_op = htons(ARPOP_REPLY);
   1276 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
   1277 	switch (ifp->if_type) {
   1278 	case IFT_IEEE1394:
   1279 		/*
   1280 		 * ieee1394 arp reply is broadcast
   1281 		 */
   1282 		m->m_flags &= ~M_MCAST;
   1283 		m->m_flags |= M_BCAST;
   1284 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
   1285 		break;
   1286 
   1287 	default:
   1288 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
   1289 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
   1290 		break;
   1291 	}
   1292 	m->m_pkthdr.len = m->m_len;
   1293 	sa.sa_family = AF_ARP;
   1294 	sa.sa_len = 2;
   1295 	arps = ARP_STAT_GETREF();
   1296 	arps[ARP_STAT_SNDTOTAL]++;
   1297 	arps[ARP_STAT_SNDREPLY]++;
   1298 	ARP_STAT_PUTREF();
   1299 	(*ifp->if_output)(ifp, m, &sa, NULL);
   1300 	return;
   1301 out:
   1302 	m_freem(m);
   1303 }
   1304 
   1305 /*
   1306  * Free an arp entry.
   1307  */
   1308 static void arptfree(struct llinfo_arp *la)
   1309 {
   1310 	struct rtentry *rt = la->la_rt;
   1311 	struct sockaddr_dl *sdl;
   1312 
   1313 	ARP_LOCK_CHECK();
   1314 
   1315 	if (rt == NULL)
   1316 		panic("arptfree");
   1317 	if (rt->rt_refcnt > 0 && (sdl = satosdl(rt->rt_gateway)) &&
   1318 	    sdl->sdl_family == AF_LINK) {
   1319 		sdl->sdl_alen = 0;
   1320 		la->la_asked = 0;
   1321 		rt->rt_flags &= ~RTF_REJECT;
   1322 		return;
   1323 	}
   1324 	rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL);
   1325 }
   1326 
   1327 static struct llinfo_arp *
   1328 arplookup(struct mbuf *m, const struct in_addr *addr, int create, int proxy)
   1329 {
   1330 	return arplookup1(m, addr, create, proxy, NULL);
   1331 }
   1332 
   1333 /*
   1334  * Lookup or enter a new address in arptab.
   1335  */
   1336 static struct llinfo_arp *
   1337 arplookup1(struct mbuf *m, const struct in_addr *addr, int create, int proxy,
   1338     struct rtentry *rt0)
   1339 {
   1340 	struct arphdr *ah;
   1341 	struct ifnet *ifp = m->m_pkthdr.rcvif;
   1342 	struct rtentry *rt;
   1343 	struct sockaddr_inarp sin;
   1344 	const char *why = NULL;
   1345 
   1346 	ah = mtod(m, struct arphdr *);
   1347 	if (rt0 == NULL) {
   1348 		memset(&sin, 0, sizeof(sin));
   1349 		sin.sin_len = sizeof(sin);
   1350 		sin.sin_family = AF_INET;
   1351 		sin.sin_addr = *addr;
   1352 		sin.sin_other = proxy ? SIN_PROXY : 0;
   1353 		rt = rtalloc1(sintosa(&sin), create);
   1354 		if (rt == NULL)
   1355 			return NULL;
   1356 		rt->rt_refcnt--;
   1357 	} else
   1358 		rt = rt0;
   1359 
   1360 #define	IS_LLINFO(__rt)							  \
   1361 	(((__rt)->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO && \
   1362 	 (__rt)->rt_gateway->sa_family == AF_LINK)
   1363 
   1364 
   1365 	if (IS_LLINFO(rt))
   1366 		return (struct llinfo_arp *)rt->rt_llinfo;
   1367 
   1368 	if (create) {
   1369 		if (rt->rt_flags & RTF_GATEWAY) {
   1370 			if (log_unknown_network)
   1371 				why = "host is not on local network";
   1372 		} else if ((rt->rt_flags & RTF_LLINFO) == 0) {
   1373 			ARP_STATINC(ARP_STAT_ALLOCFAIL);
   1374 			why = "could not allocate llinfo";
   1375 		} else
   1376 			why = "gateway route is not ours";
   1377 		if (why) {
   1378 			log(LOG_DEBUG, "arplookup: unable to enter address"
   1379 			    " for %s@%s on %s (%s)\n", in_fmtaddr(*addr),
   1380 			    lla_snprintf(ar_sha(ah), ah->ar_hln),
   1381 			    (ifp) ? ifp->if_xname : "null", why);
   1382 		}
   1383 		if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) {
   1384 			rtrequest(RTM_DELETE, rt_getkey(rt),
   1385 		    	    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
   1386 		}
   1387 	}
   1388 	return NULL;
   1389 }
   1390 
   1391 int
   1392 arpioctl(u_long cmd, void *data)
   1393 {
   1394 
   1395 	return EOPNOTSUPP;
   1396 }
   1397 
   1398 void
   1399 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
   1400 {
   1401 	struct in_addr *ip;
   1402 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1403 
   1404 	/*
   1405 	 * Warn the user if another station has this IP address,
   1406 	 * but only if the interface IP address is not zero.
   1407 	 */
   1408 	ip = &IA_SIN(ifa)->sin_addr;
   1409 	if (!in_nullhost(*ip) &&
   1410 	    (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) == 0)
   1411 		arprequest(ifp, ip, ip, CLLADDR(ifp->if_sadl));
   1412 
   1413 	ifa->ifa_rtrequest = arp_rtrequest;
   1414 	ifa->ifa_flags |= RTF_CLONING;
   1415 
   1416 	/* ARP will handle DAD for this address. */
   1417 	if (ia->ia4_flags & IN_IFF_TRYTENTATIVE) {
   1418 		ia->ia4_flags |= IN_IFF_TENTATIVE;
   1419 		ia->ia_dad_start = arp_dad_start;
   1420 		ia->ia_dad_stop = arp_dad_stop;
   1421 	}
   1422 }
   1423 
   1424 TAILQ_HEAD(dadq_head, dadq);
   1425 struct dadq {
   1426 	TAILQ_ENTRY(dadq) dad_list;
   1427 	struct ifaddr *dad_ifa;
   1428 	int dad_count;		/* max ARP to send */
   1429 	int dad_arp_tcount;	/* # of trials to send ARP */
   1430 	int dad_arp_ocount;	/* ARP sent so far */
   1431 	int dad_arp_announce;	/* max ARP announcements */
   1432 	int dad_arp_acount;	/* # of announcements */
   1433 	struct callout dad_timer_ch;
   1434 };
   1435 MALLOC_JUSTDEFINE(M_IPARP, "ARP DAD", "ARP DAD Structure");
   1436 
   1437 static struct dadq_head dadq;
   1438 static int dad_init = 0;
   1439 static int dad_maxtry = 15;     /* max # of *tries* to transmit DAD packet */
   1440 
   1441 static struct dadq *
   1442 arp_dad_find(struct ifaddr *ifa)
   1443 {
   1444 	struct dadq *dp;
   1445 
   1446 	TAILQ_FOREACH(dp, &dadq, dad_list) {
   1447 		if (dp->dad_ifa == ifa)
   1448 			return dp;
   1449 	}
   1450 	return NULL;
   1451 }
   1452 
   1453 static void
   1454 arp_dad_starttimer(struct dadq *dp, int ticks)
   1455 {
   1456 
   1457 	callout_reset(&dp->dad_timer_ch, ticks,
   1458 	    (void (*)(void *))arp_dad_timer, (void *)dp->dad_ifa);
   1459 }
   1460 
   1461 static void
   1462 arp_dad_stoptimer(struct dadq *dp)
   1463 {
   1464 
   1465 	callout_stop(&dp->dad_timer_ch);
   1466 }
   1467 
   1468 static void
   1469 arp_dad_output(struct dadq *dp, struct ifaddr *ifa)
   1470 {
   1471 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1472 	struct ifnet *ifp = ifa->ifa_ifp;
   1473 	struct in_addr sip;
   1474 
   1475 	dp->dad_arp_tcount++;
   1476 	if ((ifp->if_flags & IFF_UP) == 0)
   1477 		return;
   1478 	if ((ifp->if_flags & IFF_RUNNING) == 0)
   1479 		return;
   1480 
   1481 	dp->dad_arp_tcount = 0;
   1482 	dp->dad_arp_ocount++;
   1483 
   1484 	memset(&sip, 0, sizeof(sip));
   1485 	arprequest(ifa->ifa_ifp, &sip, &ia->ia_addr.sin_addr,
   1486 	    CLLADDR(ifa->ifa_ifp->if_sadl));
   1487 }
   1488 
   1489 /*
   1490  * Start Duplicate Address Detection (DAD) for specified interface address.
   1491  */
   1492 static void
   1493 arp_dad_start(struct ifaddr *ifa)
   1494 {
   1495 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1496 	struct dadq *dp;
   1497 
   1498 	if (!dad_init) {
   1499 		TAILQ_INIT(&dadq);
   1500 		dad_init++;
   1501 	}
   1502 
   1503 	/*
   1504 	 * If we don't need DAD, don't do it.
   1505 	 * - DAD is disabled (ip_dad_count == 0)
   1506 	 */
   1507 	if (!(ia->ia4_flags & IN_IFF_TENTATIVE)) {
   1508 		log(LOG_DEBUG,
   1509 			"arp_dad_start: called with non-tentative address "
   1510 			"%s(%s)\n",
   1511 			in_fmtaddr(ia->ia_addr.sin_addr),
   1512 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
   1513 		return;
   1514 	}
   1515 	if (!ip_dad_count) {
   1516 		struct in_addr *ip = &IA_SIN(ifa)->sin_addr;
   1517 
   1518 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
   1519 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1520 		arprequest(ifa->ifa_ifp, ip, ip,
   1521 		    CLLADDR(ifa->ifa_ifp->if_sadl));
   1522 		return;
   1523 	}
   1524 	if (ifa->ifa_ifp == NULL)
   1525 		panic("arp_dad_start: ifa->ifa_ifp == NULL");
   1526 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
   1527 		return;
   1528 	if (arp_dad_find(ifa) != NULL) {
   1529 		/* DAD already in progress */
   1530 		return;
   1531 	}
   1532 
   1533 	dp = malloc(sizeof(*dp), M_IPARP, M_NOWAIT);
   1534 	if (dp == NULL) {
   1535 		log(LOG_ERR, "arp_dad_start: memory allocation failed for "
   1536 			"%s(%s)\n",
   1537 			in_fmtaddr(ia->ia_addr.sin_addr),
   1538 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
   1539 		return;
   1540 	}
   1541 	memset(dp, 0, sizeof(*dp));
   1542 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
   1543 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
   1544 
   1545 	arplog((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
   1546 	    in_fmtaddr(ia->ia_addr.sin_addr)));
   1547 
   1548 	/*
   1549 	 * Send ARP packet for DAD, ip_dad_count times.
   1550 	 * Note that we must delay the first transmission.
   1551 	 */
   1552 	dp->dad_ifa = ifa;
   1553 	ifaref(ifa);	/* just for safety */
   1554 	dp->dad_count = ip_dad_count;
   1555 	dp->dad_arp_announce = 0; /* Will be set when starting to announce */
   1556 	dp->dad_arp_acount = dp->dad_arp_ocount = dp->dad_arp_tcount = 0;
   1557 
   1558 	arp_dad_starttimer(dp, cprng_fast32() % (PROBE_WAIT * hz));
   1559 }
   1560 
   1561 /*
   1562  * terminate DAD unconditionally.  used for address removals.
   1563  */
   1564 static void
   1565 arp_dad_stop(struct ifaddr *ifa)
   1566 {
   1567 	struct dadq *dp;
   1568 
   1569 	if (!dad_init)
   1570 		return;
   1571 	dp = arp_dad_find(ifa);
   1572 	if (dp == NULL) {
   1573 		/* DAD wasn't started yet */
   1574 		return;
   1575 	}
   1576 
   1577 	arp_dad_stoptimer(dp);
   1578 
   1579 	TAILQ_REMOVE(&dadq, dp, dad_list);
   1580 	free(dp, M_IPARP);
   1581 	dp = NULL;
   1582 	ifafree(ifa);
   1583 }
   1584 
   1585 static void
   1586 arp_dad_timer(struct ifaddr *ifa)
   1587 {
   1588 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1589 	struct dadq *dp;
   1590 	struct in_addr *ip;
   1591 
   1592 	mutex_enter(softnet_lock);
   1593 	KERNEL_LOCK(1, NULL);
   1594 
   1595 	/* Sanity check */
   1596 	if (ia == NULL) {
   1597 		log(LOG_ERR, "arp_dad_timer: called with null parameter\n");
   1598 		goto done;
   1599 	}
   1600 	dp = arp_dad_find(ifa);
   1601 	if (dp == NULL) {
   1602 		log(LOG_ERR, "arp_dad_timer: DAD structure not found\n");
   1603 		goto done;
   1604 	}
   1605 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
   1606 		log(LOG_ERR, "nd4_dad_timer: called with duplicate address "
   1607 			"%s(%s)\n",
   1608 			in_fmtaddr(ia->ia_addr.sin_addr),
   1609 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
   1610 		goto done;
   1611 	}
   1612 	if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0 && dp->dad_arp_acount == 0){
   1613 		log(LOG_ERR, "arp_dad_timer: called with non-tentative address "
   1614 			"%s(%s)\n",
   1615 			in_fmtaddr(ia->ia_addr.sin_addr),
   1616 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
   1617 		goto done;
   1618 	}
   1619 
   1620 	/* timeouted with IFF_{RUNNING,UP} check */
   1621 	if (dp->dad_arp_tcount > dad_maxtry) {
   1622 		arplog((LOG_INFO, "%s: could not run DAD, driver problem?\n",
   1623 			if_name(ifa->ifa_ifp)));
   1624 
   1625 		TAILQ_REMOVE(&dadq, dp, dad_list);
   1626 		free(dp, M_IPARP);
   1627 		dp = NULL;
   1628 		ifafree(ifa);
   1629 		goto done;
   1630 	}
   1631 
   1632 	/* Need more checks? */
   1633 	if (dp->dad_arp_ocount < dp->dad_count) {
   1634 		int adelay;
   1635 
   1636 		/*
   1637 		 * We have more ARP to go.  Send ARP packet for DAD.
   1638 		 */
   1639 		arp_dad_output(dp, ifa);
   1640 		if (dp->dad_arp_ocount < dp->dad_count)
   1641 			adelay = (PROBE_MIN * hz) +
   1642 			    (cprng_fast32() %
   1643 			    ((PROBE_MAX * hz) - (PROBE_MIN * hz)));
   1644 		else
   1645 			adelay = ANNOUNCE_WAIT * hz;
   1646 		arp_dad_starttimer(dp, adelay);
   1647 		goto done;
   1648 	} else if (dp->dad_arp_acount == 0) {
   1649 		/*
   1650 		 * We are done with DAD.
   1651 		 * No duplicate address found.
   1652 		 */
   1653 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
   1654 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1655 		arplog((LOG_DEBUG,
   1656 		    "%s: DAD complete for %s - no duplicates found\n",
   1657 		    if_name(ifa->ifa_ifp),
   1658 		    in_fmtaddr(ia->ia_addr.sin_addr)));
   1659 		dp->dad_arp_announce = ANNOUNCE_NUM;
   1660 		goto announce;
   1661 	} else if (dp->dad_arp_acount < dp->dad_arp_announce) {
   1662 announce:
   1663 		/*
   1664 		 * Announce the address.
   1665 		 */
   1666 		ip = &IA_SIN(ifa)->sin_addr;
   1667 		arprequest(ifa->ifa_ifp, ip, ip,
   1668 		    CLLADDR(ifa->ifa_ifp->if_sadl));
   1669 		dp->dad_arp_acount++;
   1670 		if (dp->dad_arp_acount < dp->dad_arp_announce) {
   1671 			arp_dad_starttimer(dp, ANNOUNCE_INTERVAL * hz);
   1672 			goto done;
   1673 		}
   1674 		arplog((LOG_DEBUG,
   1675 		    "%s: ARP announcement complete for %s\n",
   1676 		    if_name(ifa->ifa_ifp),
   1677 		    in_fmtaddr(ia->ia_addr.sin_addr)));
   1678 	}
   1679 
   1680 	TAILQ_REMOVE(&dadq, dp, dad_list);
   1681 	free(dp, M_IPARP);
   1682 	dp = NULL;
   1683 	ifafree(ifa);
   1684 
   1685 done:
   1686 	KERNEL_UNLOCK_ONE(NULL);
   1687 	mutex_exit(softnet_lock);
   1688 }
   1689 
   1690 static void
   1691 arp_dad_duplicated(struct ifaddr *ifa)
   1692 {
   1693 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1694 	struct ifnet *ifp;
   1695 	struct dadq *dp;
   1696 
   1697 	dp = arp_dad_find(ifa);
   1698 	if (dp == NULL) {
   1699 		log(LOG_ERR, "arp_dad_duplicated: DAD structure not found\n");
   1700 		return;
   1701 	}
   1702 
   1703 	ifp = ifa->ifa_ifp;
   1704 	log(LOG_ERR, "%s: DAD detected duplicate IPv4 address %s: "
   1705 	    "ARP out=%d\n",
   1706 	    if_name(ifp), in_fmtaddr(ia->ia_addr.sin_addr),
   1707 	    dp->dad_arp_ocount);
   1708 
   1709 	ia->ia4_flags &= ~IN_IFF_TENTATIVE;
   1710 	ia->ia4_flags |= IN_IFF_DUPLICATED;
   1711 
   1712 	/* We are done with DAD, with duplicated address found. (failure) */
   1713 	arp_dad_stoptimer(dp);
   1714 
   1715 	/* Inform the routing socket that DAD has completed */
   1716 	rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1717 
   1718 	TAILQ_REMOVE(&dadq, dp, dad_list);
   1719 	free(dp, M_IPARP);
   1720 	dp = NULL;
   1721 	ifafree(ifa);
   1722 }
   1723 
   1724 /*
   1725  * Called from 10 Mb/s Ethernet interrupt handlers
   1726  * when ether packet type ETHERTYPE_REVARP
   1727  * is received.  Common length and type checks are done here,
   1728  * then the protocol-specific routine is called.
   1729  */
   1730 void
   1731 revarpinput(struct mbuf *m)
   1732 {
   1733 	struct arphdr *ar;
   1734 
   1735 	if (m->m_len < sizeof(struct arphdr))
   1736 		goto out;
   1737 	ar = mtod(m, struct arphdr *);
   1738 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
   1739 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
   1740 		goto out;
   1741 #endif
   1742 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
   1743 		goto out;
   1744 	switch (ntohs(ar->ar_pro)) {
   1745 	case ETHERTYPE_IP:
   1746 	case ETHERTYPE_IPTRAILERS:
   1747 		in_revarpinput(m);
   1748 		return;
   1749 
   1750 	default:
   1751 		break;
   1752 	}
   1753 out:
   1754 	m_freem(m);
   1755 }
   1756 
   1757 /*
   1758  * RARP for Internet protocols on 10 Mb/s Ethernet.
   1759  * Algorithm is that given in RFC 903.
   1760  * We are only using for bootstrap purposes to get an ip address for one of
   1761  * our interfaces.  Thus we support no user-interface.
   1762  *
   1763  * Since the contents of the RARP reply are specific to the interface that
   1764  * sent the request, this code must ensure that they are properly associated.
   1765  *
   1766  * Note: also supports ARP via RARP packets, per the RFC.
   1767  */
   1768 void
   1769 in_revarpinput(struct mbuf *m)
   1770 {
   1771 	struct ifnet *ifp;
   1772 	struct arphdr *ah;
   1773 	void *tha;
   1774 	int op;
   1775 
   1776 	ah = mtod(m, struct arphdr *);
   1777 	op = ntohs(ah->ar_op);
   1778 
   1779 	switch (m->m_pkthdr.rcvif->if_type) {
   1780 	case IFT_IEEE1394:
   1781 		/* ARP without target hardware address is not supported */
   1782 		goto out;
   1783 	default:
   1784 		break;
   1785 	}
   1786 
   1787 	switch (op) {
   1788 	case ARPOP_REQUEST:
   1789 	case ARPOP_REPLY:	/* per RFC */
   1790 		in_arpinput(m);
   1791 		return;
   1792 	case ARPOP_REVREPLY:
   1793 		break;
   1794 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
   1795 	default:
   1796 		goto out;
   1797 	}
   1798 	if (!revarp_in_progress)
   1799 		goto out;
   1800 	ifp = m->m_pkthdr.rcvif;
   1801 	if (ifp != myip_ifp) /* !same interface */
   1802 		goto out;
   1803 	if (myip_initialized)
   1804 		goto wake;
   1805 	tha = ar_tha(ah);
   1806 	if (tha == NULL)
   1807 		goto out;
   1808 	if (memcmp(tha, CLLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen))
   1809 		goto out;
   1810 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
   1811 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
   1812 	myip_initialized = 1;
   1813 wake:	/* Do wakeup every time in case it was missed. */
   1814 	wakeup((void *)&myip);
   1815 
   1816 out:
   1817 	m_freem(m);
   1818 }
   1819 
   1820 /*
   1821  * Send a RARP request for the ip address of the specified interface.
   1822  * The request should be RFC 903-compliant.
   1823  */
   1824 void
   1825 revarprequest(struct ifnet *ifp)
   1826 {
   1827 	struct sockaddr sa;
   1828 	struct mbuf *m;
   1829 	struct arphdr *ah;
   1830 	void *tha;
   1831 
   1832 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
   1833 		return;
   1834 	MCLAIM(m, &arpdomain.dom_mowner);
   1835 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
   1836 	    2*ifp->if_addrlen;
   1837 	m->m_pkthdr.len = m->m_len;
   1838 	MH_ALIGN(m, m->m_len);
   1839 	ah = mtod(m, struct arphdr *);
   1840 	memset(ah, 0, m->m_len);
   1841 	ah->ar_pro = htons(ETHERTYPE_IP);
   1842 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
   1843 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
   1844 	ah->ar_op = htons(ARPOP_REVREQUEST);
   1845 
   1846 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
   1847 	tha = ar_tha(ah);
   1848 	if (tha == NULL) {
   1849 		m_free(m);
   1850 		return;
   1851 	}
   1852 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
   1853 
   1854 	sa.sa_family = AF_ARP;
   1855 	sa.sa_len = 2;
   1856 	m->m_flags |= M_BCAST;
   1857 
   1858 	KERNEL_LOCK(1, NULL);
   1859 	(*ifp->if_output)(ifp, m, &sa, NULL);
   1860 	KERNEL_UNLOCK_ONE(NULL);
   1861 }
   1862 
   1863 /*
   1864  * RARP for the ip address of the specified interface, but also
   1865  * save the ip address of the server that sent the answer.
   1866  * Timeout if no response is received.
   1867  */
   1868 int
   1869 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
   1870     struct in_addr *clnt_in)
   1871 {
   1872 	int result, count = 20;
   1873 
   1874 	myip_initialized = 0;
   1875 	myip_ifp = ifp;
   1876 
   1877 	revarp_in_progress = 1;
   1878 	while (count--) {
   1879 		revarprequest(ifp);
   1880 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
   1881 		if (result != EWOULDBLOCK)
   1882 			break;
   1883 	}
   1884 	revarp_in_progress = 0;
   1885 
   1886 	if (!myip_initialized)
   1887 		return ENETUNREACH;
   1888 
   1889 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
   1890 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
   1891 	return 0;
   1892 }
   1893 
   1894 
   1895 
   1897 #ifdef DDB
   1898 
   1899 #include <machine/db_machdep.h>
   1900 #include <ddb/db_interface.h>
   1901 #include <ddb/db_output.h>
   1902 
   1903 static void
   1904 db_print_sa(const struct sockaddr *sa)
   1905 {
   1906 	int len;
   1907 	const u_char *p;
   1908 
   1909 	if (sa == NULL) {
   1910 		db_printf("[NULL]");
   1911 		return;
   1912 	}
   1913 
   1914 	p = (const u_char *)sa;
   1915 	len = sa->sa_len;
   1916 	db_printf("[");
   1917 	while (len > 0) {
   1918 		db_printf("%d", *p);
   1919 		p++; len--;
   1920 		if (len) db_printf(",");
   1921 	}
   1922 	db_printf("]\n");
   1923 }
   1924 
   1925 static void
   1926 db_print_ifa(struct ifaddr *ifa)
   1927 {
   1928 	if (ifa == NULL)
   1929 		return;
   1930 	db_printf("  ifa_addr=");
   1931 	db_print_sa(ifa->ifa_addr);
   1932 	db_printf("  ifa_dsta=");
   1933 	db_print_sa(ifa->ifa_dstaddr);
   1934 	db_printf("  ifa_mask=");
   1935 	db_print_sa(ifa->ifa_netmask);
   1936 	db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
   1937 			  ifa->ifa_flags,
   1938 			  ifa->ifa_refcnt,
   1939 			  ifa->ifa_metric);
   1940 }
   1941 
   1942 static void
   1943 db_print_llinfo(void *li)
   1944 {
   1945 	struct llinfo_arp *la;
   1946 
   1947 	if (li == NULL)
   1948 		return;
   1949 	la = (struct llinfo_arp *)li;
   1950 	db_printf("  la_rt=%p la_hold=%p, la_asked=0x%lx\n",
   1951 			  la->la_rt, la->la_hold, la->la_asked);
   1952 }
   1953 
   1954 /*
   1955  * Function to pass to rt_walktree().
   1956  * Return non-zero error to abort walk.
   1957  */
   1958 static int
   1959 db_show_rtentry(struct rtentry *rt, void *w)
   1960 {
   1961 	db_printf("rtentry=%p", rt);
   1962 
   1963 	db_printf(" flags=0x%x refcnt=%d use=%"PRId64" expire=%"PRId64"\n",
   1964 			  rt->rt_flags, rt->rt_refcnt,
   1965 			  rt->rt_use, (uint64_t)rt->rt_expire);
   1966 
   1967 	db_printf(" key="); db_print_sa(rt_getkey(rt));
   1968 	db_printf(" mask="); db_print_sa(rt_mask(rt));
   1969 	db_printf(" gw="); db_print_sa(rt->rt_gateway);
   1970 
   1971 	db_printf(" ifp=%p ", rt->rt_ifp);
   1972 	if (rt->rt_ifp)
   1973 		db_printf("(%s)", rt->rt_ifp->if_xname);
   1974 	else
   1975 		db_printf("(NULL)");
   1976 
   1977 	db_printf(" ifa=%p\n", rt->rt_ifa);
   1978 	db_print_ifa(rt->rt_ifa);
   1979 
   1980 	db_printf(" gwroute=%p llinfo=%p\n",
   1981 			  rt->rt_gwroute, rt->rt_llinfo);
   1982 	db_print_llinfo(rt->rt_llinfo);
   1983 
   1984 	return 0;
   1985 }
   1986 
   1987 /*
   1988  * Function to print all the route trees.
   1989  * Use this from ddb:  "show arptab"
   1990  */
   1991 void
   1992 db_show_arptab(db_expr_t addr, bool have_addr,
   1993     db_expr_t count, const char *modif)
   1994 {
   1995 	rt_walktree(AF_INET, db_show_rtentry, NULL);
   1996 }
   1997 #endif
   1998 
   1999 static int
   2000 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
   2001 {
   2002 
   2003 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
   2004 }
   2005 
   2006 static void
   2007 sysctl_net_inet_arp_setup(struct sysctllog **clog)
   2008 {
   2009 	const struct sysctlnode *node;
   2010 
   2011 	sysctl_createv(clog, 0, NULL, NULL,
   2012 			CTLFLAG_PERMANENT,
   2013 			CTLTYPE_NODE, "inet", NULL,
   2014 			NULL, 0, NULL, 0,
   2015 			CTL_NET, PF_INET, CTL_EOL);
   2016 	sysctl_createv(clog, 0, NULL, &node,
   2017 			CTLFLAG_PERMANENT,
   2018 			CTLTYPE_NODE, "arp",
   2019 			SYSCTL_DESCR("Address Resolution Protocol"),
   2020 			NULL, 0, NULL, 0,
   2021 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
   2022 
   2023 	sysctl_createv(clog, 0, NULL, NULL,
   2024 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2025 			CTLTYPE_INT, "prune",
   2026 			SYSCTL_DESCR("ARP cache pruning interval in seconds"),
   2027 			NULL, 0, &arpt_prune, 0,
   2028 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2029 
   2030 	sysctl_createv(clog, 0, NULL, NULL,
   2031 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2032 			CTLTYPE_INT, "keep",
   2033 			SYSCTL_DESCR("Valid ARP entry lifetime in seconds"),
   2034 			NULL, 0, &arpt_keep, 0,
   2035 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2036 
   2037 	sysctl_createv(clog, 0, NULL, NULL,
   2038 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2039 			CTLTYPE_INT, "down",
   2040 			SYSCTL_DESCR("Failed ARP entry lifetime in seconds"),
   2041 			NULL, 0, &arpt_down, 0,
   2042 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2043 
   2044 	sysctl_createv(clog, 0, NULL, NULL,
   2045 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2046 			CTLTYPE_INT, "refresh",
   2047 			SYSCTL_DESCR("ARP entry refresh interval"),
   2048 			NULL, 0, &arpt_refresh, 0,
   2049 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2050 
   2051 	sysctl_createv(clog, 0, NULL, NULL,
   2052 			CTLFLAG_PERMANENT,
   2053 			CTLTYPE_STRUCT, "stats",
   2054 			SYSCTL_DESCR("ARP statistics"),
   2055 			sysctl_net_inet_arp_stats, 0, NULL, 0,
   2056 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2057 
   2058 	sysctl_createv(clog, 0, NULL, NULL,
   2059 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2060 			CTLTYPE_INT, "log_movements",
   2061 			SYSCTL_DESCR("log ARP replies from MACs different than"
   2062 			    " the one in the cache"),
   2063 			NULL, 0, &log_movements, 0,
   2064 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2065 
   2066 	sysctl_createv(clog, 0, NULL, NULL,
   2067 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2068 			CTLTYPE_INT, "log_permanent_modify",
   2069 			SYSCTL_DESCR("log ARP replies from MACs different than"
   2070 			    " the one in the permanent arp entry"),
   2071 			NULL, 0, &log_permanent_modify, 0,
   2072 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2073 
   2074 	sysctl_createv(clog, 0, NULL, NULL,
   2075 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2076 			CTLTYPE_INT, "log_wrong_iface",
   2077 			SYSCTL_DESCR("log ARP packets arriving on the wrong"
   2078 			    " interface"),
   2079 			NULL, 0, &log_wrong_iface, 0,
   2080 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2081 
   2082 	sysctl_createv(clog, 0, NULL, NULL,
   2083 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2084 			CTLTYPE_INT, "log_unknown_network",
   2085 			SYSCTL_DESCR("log ARP packets from non-local network"),
   2086 			NULL, 0, &log_unknown_network, 0,
   2087 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2088 
   2089 	sysctl_createv(clog, 0, NULL, NULL,
   2090 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2091 		       CTLTYPE_INT, "debug",
   2092 		       SYSCTL_DESCR("Enable ARP DAD debug output"),
   2093 		       NULL, 0, &arp_debug, 0,
   2094 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2095 }
   2096 
   2097 #endif /* INET */
   2098