Home | History | Annotate | Line # | Graph | Download | only in netinet
      1 /*	$NetBSD: if_arp.c,v 1.216 2016/07/06 06:30:08 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 1.216 2016/07/06 06:30:08 ozaki-r Exp $");
     72 
     73 #ifdef _KERNEL_OPT
     74 #include "opt_ddb.h"
     75 #include "opt_inet.h"
     76 #endif
     77 
     78 #ifdef INET
     79 
     80 #include "bridge.h"
     81 
     82 #include <sys/param.h>
     83 #include <sys/systm.h>
     84 #include <sys/callout.h>
     85 #include <sys/malloc.h>
     86 #include <sys/mbuf.h>
     87 #include <sys/socket.h>
     88 #include <sys/time.h>
     89 #include <sys/timetc.h>
     90 #include <sys/kernel.h>
     91 #include <sys/errno.h>
     92 #include <sys/ioctl.h>
     93 #include <sys/syslog.h>
     94 #include <sys/proc.h>
     95 #include <sys/protosw.h>
     96 #include <sys/domain.h>
     97 #include <sys/sysctl.h>
     98 #include <sys/socketvar.h>
     99 #include <sys/percpu.h>
    100 #include <sys/cprng.h>
    101 #include <sys/kmem.h>
    102 
    103 #include <net/ethertypes.h>
    104 #include <net/if.h>
    105 #include <net/if_dl.h>
    106 #include <net/if_token.h>
    107 #include <net/if_types.h>
    108 #include <net/if_ether.h>
    109 #include <net/if_llatbl.h>
    110 #include <net/net_osdep.h>
    111 #include <net/route.h>
    112 #include <net/net_stats.h>
    113 
    114 #include <netinet/in.h>
    115 #include <netinet/in_systm.h>
    116 #include <netinet/in_var.h>
    117 #include <netinet/ip.h>
    118 #include <netinet/if_inarp.h>
    119 
    120 #include "arcnet.h"
    121 #if NARCNET > 0
    122 #include <net/if_arc.h>
    123 #endif
    124 #include "fddi.h"
    125 #if NFDDI > 0
    126 #include <net/if_fddi.h>
    127 #endif
    128 #include "token.h"
    129 #include "carp.h"
    130 #if NCARP > 0
    131 #include <netinet/ip_carp.h>
    132 #endif
    133 
    134 #define SIN(s) ((struct sockaddr_in *)s)
    135 #define SRP(s) ((struct sockaddr_inarp *)s)
    136 
    137 /*
    138  * ARP trailer negotiation.  Trailer protocol is not IP specific,
    139  * but ARP request/response use IP addresses.
    140  */
    141 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL
    142 
    143 /* timer values */
    144 static int	arpt_keep = (20*60);	/* once resolved, good for 20 more minutes */
    145 static int	arpt_down = 20;		/* once declared down, don't send for 20 secs */
    146 static int	arp_maxhold = 1;	/* number of packets to hold per ARP entry */
    147 #define	rt_expire rt_rmx.rmx_expire
    148 #define	rt_pksent rt_rmx.rmx_pksent
    149 
    150 int		ip_dad_count = PROBE_NUM;
    151 #ifdef ARP_DEBUG
    152 static int	arp_debug = 1;
    153 #else
    154 static int	arp_debug = 0;
    155 #endif
    156 #define arplog(x)	do { if (arp_debug) log x; } while (/*CONSTCOND*/ 0)
    157 
    158 static	void arp_init(void);
    159 
    160 static	struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *,
    161 	    const struct sockaddr *);
    162 static	void arptimer(void *);
    163 static	void arp_settimer(struct llentry *, int);
    164 static	struct llentry *arplookup(struct ifnet *, struct mbuf *,
    165 	    const struct in_addr *, const struct sockaddr *, int);
    166 static	struct llentry *arpcreate(struct ifnet *, struct mbuf *,
    167 	    const struct in_addr *, const struct sockaddr *, int);
    168 static	void in_arpinput(struct mbuf *);
    169 static	void in_revarpinput(struct mbuf *);
    170 static	void revarprequest(struct ifnet *);
    171 
    172 static	void arp_drainstub(void);
    173 
    174 static void arp_dad_timer(struct ifaddr *);
    175 static void arp_dad_start(struct ifaddr *);
    176 static void arp_dad_stop(struct ifaddr *);
    177 static void arp_dad_duplicated(struct ifaddr *);
    178 
    179 static void arp_init_llentry(struct ifnet *, struct llentry *);
    180 #if NTOKEN > 0
    181 static void arp_free_llentry_tokenring(struct llentry *);
    182 #endif
    183 
    184 struct	ifqueue arpintrq = {
    185 	.ifq_head = NULL,
    186 	.ifq_tail = NULL,
    187 	.ifq_len = 0,
    188 	.ifq_maxlen = 50,
    189 	.ifq_drops = 0,
    190 };
    191 static int	arp_maxtries = 5;
    192 static int	useloopback = 1;	/* use loopback interface for local traffic */
    193 
    194 static percpu_t *arpstat_percpu;
    195 
    196 #define	ARP_STAT_GETREF()	_NET_STAT_GETREF(arpstat_percpu)
    197 #define	ARP_STAT_PUTREF()	_NET_STAT_PUTREF(arpstat_percpu)
    198 
    199 #define	ARP_STATINC(x)		_NET_STATINC(arpstat_percpu, x)
    200 #define	ARP_STATADD(x, v)	_NET_STATADD(arpstat_percpu, x, v)
    201 
    202 /* revarp state */
    203 static struct	in_addr myip, srv_ip;
    204 static int	myip_initialized = 0;
    205 static int	revarp_in_progress = 0;
    206 static struct	ifnet *myip_ifp = NULL;
    207 
    208 static int arp_drainwanted;
    209 
    210 static int log_movements = 1;
    211 static int log_permanent_modify = 1;
    212 static int log_wrong_iface = 1;
    213 static int log_unknown_network = 1;
    214 
    215 /*
    216  * this should be elsewhere.
    217  */
    218 
    219 static char *
    220 lla_snprintf(u_int8_t *, int);
    221 
    222 static char *
    223 lla_snprintf(u_int8_t *adrp, int len)
    224 {
    225 #define NUMBUFS 3
    226 	static char buf[NUMBUFS][16*3];
    227 	static int bnum = 0;
    228 
    229 	int i;
    230 	char *p;
    231 
    232 	p = buf[bnum];
    233 
    234 	*p++ = hexdigits[(*adrp)>>4];
    235 	*p++ = hexdigits[(*adrp++)&0xf];
    236 
    237 	for (i=1; i<len && i<16; i++) {
    238 		*p++ = ':';
    239 		*p++ = hexdigits[(*adrp)>>4];
    240 		*p++ = hexdigits[(*adrp++)&0xf];
    241 	}
    242 
    243 	*p = 0;
    244 	p = buf[bnum];
    245 	bnum = (bnum + 1) % NUMBUFS;
    246 	return p;
    247 }
    248 
    249 DOMAIN_DEFINE(arpdomain);	/* forward declare and add to link set */
    250 
    251 static void
    252 arp_fasttimo(void)
    253 {
    254 	if (arp_drainwanted) {
    255 		arp_drain();
    256 		arp_drainwanted = 0;
    257 	}
    258 }
    259 
    260 const struct protosw arpsw[] = {
    261 	{ .pr_type = 0,
    262 	  .pr_domain = &arpdomain,
    263 	  .pr_protocol = 0,
    264 	  .pr_flags = 0,
    265 	  .pr_input = 0,
    266 	  .pr_ctlinput = 0,
    267 	  .pr_ctloutput = 0,
    268 	  .pr_usrreqs = 0,
    269 	  .pr_init = arp_init,
    270 	  .pr_fasttimo = arp_fasttimo,
    271 	  .pr_slowtimo = 0,
    272 	  .pr_drain = arp_drainstub,
    273 	}
    274 };
    275 
    276 struct domain arpdomain = {
    277 	.dom_family = PF_ARP,
    278 	.dom_name = "arp",
    279 	.dom_protosw = arpsw,
    280 	.dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)],
    281 };
    282 
    283 static void sysctl_net_inet_arp_setup(struct sysctllog **);
    284 
    285 void
    286 arp_init(void)
    287 {
    288 
    289 	sysctl_net_inet_arp_setup(NULL);
    290 	arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS);
    291 }
    292 
    293 static void
    294 arp_drainstub(void)
    295 {
    296 	arp_drainwanted = 1;
    297 }
    298 
    299 /*
    300  * ARP protocol drain routine.  Called when memory is in short supply.
    301  * Called at splvm();  don't acquire softnet_lock as can be called from
    302  * hardware interrupt handlers.
    303  */
    304 void
    305 arp_drain(void)
    306 {
    307 
    308 	lltable_drain(AF_INET);
    309 }
    310 
    311 static void
    312 arptimer(void *arg)
    313 {
    314 	struct llentry *lle = arg;
    315 	struct ifnet *ifp;
    316 
    317 	if (lle == NULL)
    318 		return;
    319 
    320 	if (lle->la_flags & LLE_STATIC)
    321 		return;
    322 
    323 	LLE_WLOCK(lle);
    324 	if (callout_pending(&lle->la_timer)) {
    325 		/*
    326 		 * Here we are a bit odd here in the treatment of
    327 		 * active/pending. If the pending bit is set, it got
    328 		 * rescheduled before I ran. The active
    329 		 * bit we ignore, since if it was stopped
    330 		 * in ll_tablefree() and was currently running
    331 		 * it would have return 0 so the code would
    332 		 * not have deleted it since the callout could
    333 		 * not be stopped so we want to go through
    334 		 * with the delete here now. If the callout
    335 		 * was restarted, the pending bit will be back on and
    336 		 * we just want to bail since the callout_reset would
    337 		 * return 1 and our reference would have been removed
    338 		 * by arpresolve() below.
    339 		 */
    340 		LLE_WUNLOCK(lle);
    341 		return;
    342 	}
    343 	ifp = lle->lle_tbl->llt_ifp;
    344 
    345 	callout_stop(&lle->la_timer);
    346 
    347 	/* XXX: LOR avoidance. We still have ref on lle. */
    348 	LLE_WUNLOCK(lle);
    349 
    350 	IF_AFDATA_LOCK(ifp);
    351 	LLE_WLOCK(lle);
    352 
    353 	/* Guard against race with other llentry_free(). */
    354 	if (lle->la_flags & LLE_LINKED) {
    355 		size_t pkts_dropped;
    356 
    357 		LLE_REMREF(lle);
    358 		pkts_dropped = llentry_free(lle);
    359 		ARP_STATADD(ARP_STAT_DFRDROPPED, pkts_dropped);
    360 	} else {
    361 		LLE_FREE_LOCKED(lle);
    362 	}
    363 
    364 	IF_AFDATA_UNLOCK(ifp);
    365 }
    366 
    367 static void
    368 arp_settimer(struct llentry *la, int sec)
    369 {
    370 
    371 	LLE_WLOCK_ASSERT(la);
    372 	LLE_ADDREF(la);
    373 	callout_reset(&la->la_timer, hz * sec, arptimer, la);
    374 }
    375 
    376 /*
    377  * We set the gateway for RTF_CLONING routes to a "prototype"
    378  * link-layer sockaddr whose interface type (if_type) and interface
    379  * index (if_index) fields are prepared.
    380  */
    381 static struct sockaddr *
    382 arp_setgate(struct rtentry *rt, struct sockaddr *gate,
    383     const struct sockaddr *netmask)
    384 {
    385 	const struct ifnet *ifp = rt->rt_ifp;
    386 	uint8_t namelen = strlen(ifp->if_xname);
    387 	uint8_t addrlen = ifp->if_addrlen;
    388 
    389 	/*
    390 	 * XXX: If this is a manually added route to interface
    391 	 * such as older version of routed or gated might provide,
    392 	 * restore cloning bit.
    393 	 */
    394 	if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL &&
    395 	    satocsin(netmask)->sin_addr.s_addr != 0xffffffff)
    396 		rt->rt_flags |= RTF_CONNECTED;
    397 
    398 	if ((rt->rt_flags & (RTF_CONNECTED | RTF_LOCAL))) {
    399 		union {
    400 			struct sockaddr sa;
    401 			struct sockaddr_storage ss;
    402 			struct sockaddr_dl sdl;
    403 		} u;
    404 		/*
    405 		 * Case 1: This route should come from a route to iface.
    406 		 */
    407 		sockaddr_dl_init(&u.sdl, sizeof(u.ss),
    408 		    ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen);
    409 		rt_setgate(rt, &u.sa);
    410 		gate = rt->rt_gateway;
    411 	}
    412 	return gate;
    413 }
    414 
    415 static void
    416 arp_init_llentry(struct ifnet *ifp, struct llentry *lle)
    417 {
    418 
    419 	switch (ifp->if_type) {
    420 #if NTOKEN > 0
    421 	case IFT_ISO88025:
    422 		lle->la_opaque = kmem_intr_alloc(sizeof(struct token_rif),
    423 		    KM_NOSLEEP);
    424 		lle->lle_ll_free = arp_free_llentry_tokenring;
    425 		break;
    426 #endif
    427 	}
    428 }
    429 
    430 #if NTOKEN > 0
    431 static void
    432 arp_free_llentry_tokenring(struct llentry *lle)
    433 {
    434 
    435 	kmem_intr_free(lle->la_opaque, sizeof(struct token_rif));
    436 }
    437 #endif
    438 
    439 /*
    440  * Parallel to llc_rtrequest.
    441  */
    442 void
    443 arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info)
    444 {
    445 	struct sockaddr *gate = rt->rt_gateway;
    446 	struct in_ifaddr *ia;
    447 	struct ifaddr *ifa;
    448 	struct ifnet *ifp = rt->rt_ifp;
    449 
    450 	if (req == RTM_LLINFO_UPD) {
    451 		struct in_addr *in;
    452 
    453 		if ((ifa = info->rti_ifa) == NULL)
    454 			return;
    455 
    456 		in = &ifatoia(ifa)->ia_addr.sin_addr;
    457 
    458 		if (ifatoia(ifa)->ia4_flags &
    459 		    (IN_IFF_NOTREADY | IN_IFF_DETACHED))
    460 		{
    461 			arplog((LOG_DEBUG, "arp_request: %s not ready\n",
    462 			   in_fmtaddr(*in)));
    463 			return;
    464 		}
    465 
    466 		arprequest(ifa->ifa_ifp, in, in,
    467 		    CLLADDR(ifa->ifa_ifp->if_sadl));
    468 		return;
    469 	}
    470 
    471 	if ((rt->rt_flags & RTF_GATEWAY) != 0) {
    472 		if (req != RTM_ADD)
    473 			return;
    474 
    475 		/*
    476 		 * linklayers with particular link MTU limitation.
    477 		 */
    478 		switch(ifp->if_type) {
    479 #if NFDDI > 0
    480 		case IFT_FDDI:
    481 			if (ifp->if_mtu > FDDIIPMTU)
    482 				rt->rt_rmx.rmx_mtu = FDDIIPMTU;
    483 			break;
    484 #endif
    485 #if NARCNET > 0
    486 		case IFT_ARCNET:
    487 		    {
    488 			int arcipifmtu;
    489 
    490 			if (ifp->if_flags & IFF_LINK0)
    491 				arcipifmtu = arc_ipmtu;
    492 			else
    493 				arcipifmtu = ARCMTU;
    494 			if (ifp->if_mtu > arcipifmtu)
    495 				rt->rt_rmx.rmx_mtu = arcipifmtu;
    496 			break;
    497 		    }
    498 #endif
    499 		}
    500 		return;
    501 	}
    502 
    503 	switch (req) {
    504 	case RTM_SETGATE:
    505 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
    506 		break;
    507 	case RTM_ADD:
    508 		gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]);
    509 		if (gate == NULL) {
    510 			log(LOG_ERR, "%s: arp_setgate failed\n", __func__);
    511 			break;
    512 		}
    513 		if ((rt->rt_flags & RTF_CONNECTED) ||
    514 		    (rt->rt_flags & RTF_LOCAL)) {
    515 			/*
    516 			 * Give this route an expiration time, even though
    517 			 * it's a "permanent" route, so that routes cloned
    518 			 * from it do not need their expiration time set.
    519 			 */
    520 			KASSERT(time_uptime != 0);
    521 			rt->rt_expire = time_uptime;
    522 			/*
    523 			 * linklayers with particular link MTU limitation.
    524 			 */
    525 			switch (ifp->if_type) {
    526 #if NFDDI > 0
    527 			case IFT_FDDI:
    528 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
    529 				    (rt->rt_rmx.rmx_mtu > FDDIIPMTU ||
    530 				     (rt->rt_rmx.rmx_mtu == 0 &&
    531 				      ifp->if_mtu > FDDIIPMTU)))
    532 					rt->rt_rmx.rmx_mtu = FDDIIPMTU;
    533 				break;
    534 #endif
    535 #if NARCNET > 0
    536 			case IFT_ARCNET:
    537 			    {
    538 				int arcipifmtu;
    539 				if (ifp->if_flags & IFF_LINK0)
    540 					arcipifmtu = arc_ipmtu;
    541 				else
    542 					arcipifmtu = ARCMTU;
    543 
    544 				if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
    545 				    (rt->rt_rmx.rmx_mtu > arcipifmtu ||
    546 				     (rt->rt_rmx.rmx_mtu == 0 &&
    547 				      ifp->if_mtu > arcipifmtu)))
    548 					rt->rt_rmx.rmx_mtu = arcipifmtu;
    549 				break;
    550 			    }
    551 #endif
    552 			}
    553 			if (rt->rt_flags & RTF_CONNECTED)
    554 				break;
    555 		}
    556 		/* Announce a new entry if requested. */
    557 		if (rt->rt_flags & RTF_ANNOUNCE) {
    558 			ia = in_get_ia_on_iface(
    559 			    satocsin(rt_getkey(rt))->sin_addr, ifp);
    560 			if (ia == NULL ||
    561 			    ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
    562 				;
    563 			else
    564 				arprequest(ifp,
    565 				    &satocsin(rt_getkey(rt))->sin_addr,
    566 				    &satocsin(rt_getkey(rt))->sin_addr,
    567 				    CLLADDR(satocsdl(gate)));
    568 		}
    569 
    570 		if (gate->sa_family != AF_LINK ||
    571 		    gate->sa_len < sockaddr_dl_measure(0, ifp->if_addrlen)) {
    572 			log(LOG_DEBUG, "%s: bad gateway value\n", __func__);
    573 			break;
    574 		}
    575 
    576 		satosdl(gate)->sdl_type = ifp->if_type;
    577 		satosdl(gate)->sdl_index = ifp->if_index;
    578 
    579 		/* If the route is for a broadcast address mark it as such.
    580 		 * This way we can avoid an expensive call to in_broadcast()
    581 		 * in ip_output() most of the time (because the route passed
    582 		 * to ip_output() is almost always a host route). */
    583 		if (rt->rt_flags & RTF_HOST &&
    584 		    !(rt->rt_flags & RTF_BROADCAST) &&
    585 		    in_broadcast(satocsin(rt_getkey(rt))->sin_addr, rt->rt_ifp))
    586 			rt->rt_flags |= RTF_BROADCAST;
    587 		/* There is little point in resolving the broadcast address */
    588 		if (rt->rt_flags & RTF_BROADCAST)
    589 			break;
    590 
    591 		/*
    592 		 * When called from rt_ifa_addlocal, we cannot depend on that
    593 		 * the address (rt_getkey(rt)) exits in the address list of the
    594 		 * interface. So check RTF_LOCAL instead.
    595 		 */
    596 		if (rt->rt_flags & RTF_LOCAL) {
    597 			rt->rt_expire = 0;
    598 			if (useloopback) {
    599 				rt->rt_ifp = lo0ifp;
    600 				rt->rt_rmx.rmx_mtu = 0;
    601 			}
    602 			break;
    603 		}
    604 
    605 		ia = in_get_ia_on_iface(satocsin(rt_getkey(rt))->sin_addr, ifp);
    606 		if (ia == NULL)
    607 			break;
    608 
    609 		rt->rt_expire = 0;
    610 		if (useloopback) {
    611 			rt->rt_ifp = lo0ifp;
    612 			rt->rt_rmx.rmx_mtu = 0;
    613 		}
    614 		rt->rt_flags |= RTF_LOCAL;
    615 		/*
    616 		 * make sure to set rt->rt_ifa to the interface
    617 		 * address we are using, otherwise we will have trouble
    618 		 * with source address selection.
    619 		 */
    620 		ifa = &ia->ia_ifa;
    621 		if (ifa != rt->rt_ifa)
    622 			rt_replace_ifa(rt, ifa);
    623 		break;
    624 	}
    625 }
    626 
    627 /*
    628  * Broadcast an ARP request. Caller specifies:
    629  *	- arp header source ip address
    630  *	- arp header target ip address
    631  *	- arp header source ethernet address
    632  */
    633 void
    634 arprequest(struct ifnet *ifp,
    635     const struct in_addr *sip, const struct in_addr *tip,
    636     const u_int8_t *enaddr)
    637 {
    638 	struct mbuf *m;
    639 	struct arphdr *ah;
    640 	struct sockaddr sa;
    641 	uint64_t *arps;
    642 
    643 	KASSERT(sip != NULL);
    644 	KASSERT(tip != NULL);
    645 	KASSERT(enaddr != NULL);
    646 
    647 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
    648 		return;
    649 	MCLAIM(m, &arpdomain.dom_mowner);
    650 	switch (ifp->if_type) {
    651 	case IFT_IEEE1394:
    652 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
    653 		    ifp->if_addrlen;
    654 		break;
    655 	default:
    656 		m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
    657 		    2 * ifp->if_addrlen;
    658 		break;
    659 	}
    660 	m->m_pkthdr.len = m->m_len;
    661 	MH_ALIGN(m, m->m_len);
    662 	ah = mtod(m, struct arphdr *);
    663 	memset(ah, 0, m->m_len);
    664 	switch (ifp->if_type) {
    665 	case IFT_IEEE1394:	/* RFC2734 */
    666 		/* fill it now for ar_tpa computation */
    667 		ah->ar_hrd = htons(ARPHRD_IEEE1394);
    668 		break;
    669 	default:
    670 		/* ifp->if_output will fill ar_hrd */
    671 		break;
    672 	}
    673 	ah->ar_pro = htons(ETHERTYPE_IP);
    674 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
    675 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
    676 	ah->ar_op = htons(ARPOP_REQUEST);
    677 	memcpy(ar_sha(ah), enaddr, ah->ar_hln);
    678 	memcpy(ar_spa(ah), sip, ah->ar_pln);
    679 	memcpy(ar_tpa(ah), tip, ah->ar_pln);
    680 	sa.sa_family = AF_ARP;
    681 	sa.sa_len = 2;
    682 	m->m_flags |= M_BCAST;
    683 	arps = ARP_STAT_GETREF();
    684 	arps[ARP_STAT_SNDTOTAL]++;
    685 	arps[ARP_STAT_SENDREQUEST]++;
    686 	ARP_STAT_PUTREF();
    687 	if_output_lock(ifp, ifp, m, &sa, NULL);
    688 }
    689 
    690 /*
    691  * Resolve an IP address into an ethernet address.  If success,
    692  * desten is filled in.  If there is no entry in arptab,
    693  * set one up and broadcast a request for the IP address.
    694  * Hold onto this mbuf and resend it once the address
    695  * is finally resolved.  A return value of 0 indicates
    696  * that desten has been filled in and the packet should be sent
    697  * normally; a return value of EWOULDBLOCK indicates that the packet has been
    698  * held pending resolution.
    699  * Any other value indicates an error.
    700  */
    701 int
    702 arpresolve(struct ifnet *ifp, const struct rtentry *rt, struct mbuf *m,
    703     const struct sockaddr *dst, void *desten, size_t destlen)
    704 {
    705 	struct llentry *la;
    706 	const char *create_lookup;
    707 	bool renew;
    708 	int error;
    709 
    710 	KASSERT(m != NULL);
    711 
    712 	la = arplookup(ifp, m, NULL, dst, 0);
    713 	if (la == NULL)
    714 		goto notfound;
    715 
    716 	if ((la->la_flags & LLE_VALID) &&
    717 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
    718 		KASSERT(destlen >= ifp->if_addrlen);
    719 		memcpy(desten, &la->ll_addr, ifp->if_addrlen);
    720 		LLE_RUNLOCK(la);
    721 		return 0;
    722 	}
    723 
    724 notfound:
    725 #ifdef IFF_STATICARP /* FreeBSD */
    726 #define _IFF_NOARP (IFF_NOARP | IFF_STATICARP)
    727 #else
    728 #define _IFF_NOARP IFF_NOARP
    729 #endif
    730 	if (ifp->if_flags & _IFF_NOARP) {
    731 		if (la != NULL)
    732 			LLE_RUNLOCK(la);
    733 		error = ENOTSUP;
    734 		goto bad;
    735 	}
    736 #undef _IFF_NOARP
    737 	if (la == NULL) {
    738 		create_lookup = "create";
    739 		IF_AFDATA_WLOCK(ifp);
    740 		la = lla_create(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
    741 		IF_AFDATA_WUNLOCK(ifp);
    742 		if (la == NULL)
    743 			ARP_STATINC(ARP_STAT_ALLOCFAIL);
    744 		else
    745 			arp_init_llentry(ifp, la);
    746 	} else if (LLE_TRY_UPGRADE(la) == 0) {
    747 		create_lookup = "lookup";
    748 		LLE_RUNLOCK(la);
    749 		IF_AFDATA_RLOCK(ifp);
    750 		la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
    751 		IF_AFDATA_RUNLOCK(ifp);
    752 	}
    753 
    754 	error = EINVAL;
    755 	if (la == NULL) {
    756 		log(LOG_DEBUG,
    757 		    "%s: failed to %s llentry for %s on %s\n",
    758 		    __func__, create_lookup, inet_ntoa(satocsin(dst)->sin_addr),
    759 		    ifp->if_xname);
    760 		goto bad;
    761 	}
    762 
    763 	if ((la->la_flags & LLE_VALID) &&
    764 	    ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime))
    765 	{
    766 		KASSERT(destlen >= ifp->if_addrlen);
    767 		memcpy(desten, &la->ll_addr, ifp->if_addrlen);
    768 		renew = false;
    769 		/*
    770 		 * If entry has an expiry time and it is approaching,
    771 		 * see if we need to send an ARP request within this
    772 		 * arpt_down interval.
    773 		 */
    774 		if (!(la->la_flags & LLE_STATIC) &&
    775 		    time_uptime + la->la_preempt > la->la_expire)
    776 		{
    777 			renew = true;
    778 			la->la_preempt--;
    779 		}
    780 
    781 		LLE_WUNLOCK(la);
    782 
    783 		if (renew) {
    784 			const u_int8_t *enaddr =
    785 #if NCARP > 0
    786 			    (ifp->if_type == IFT_CARP) ?
    787 			    CLLADDR(ifp->if_sadl):
    788 #endif
    789 			    CLLADDR(ifp->if_sadl);
    790 			arprequest(ifp,
    791 			    &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
    792 			    &satocsin(dst)->sin_addr, enaddr);
    793 		}
    794 
    795 		return 0;
    796 	}
    797 
    798 	if (la->la_flags & LLE_STATIC) {   /* should not happen! */
    799 		LLE_RUNLOCK(la);
    800 		log(LOG_DEBUG, "%s: ouch, empty static llinfo for %s\n",
    801 		    __func__, inet_ntoa(satocsin(dst)->sin_addr));
    802 		error = EINVAL;
    803 		goto bad;
    804 	}
    805 
    806 	renew = (la->la_asked == 0 || la->la_expire != time_uptime);
    807 
    808 	/*
    809 	 * There is an arptab entry, but no ethernet address
    810 	 * response yet.  Add the mbuf to the list, dropping
    811 	 * the oldest packet if we have exceeded the system
    812 	 * setting.
    813 	 */
    814 	LLE_WLOCK_ASSERT(la);
    815 	if (la->la_numheld >= arp_maxhold) {
    816 		if (la->la_hold != NULL) {
    817 			struct mbuf *next = la->la_hold->m_nextpkt;
    818 			m_freem(la->la_hold);
    819 			la->la_hold = next;
    820 			la->la_numheld--;
    821 			ARP_STATINC(ARP_STAT_DFRDROPPED);
    822 		}
    823 	}
    824 	if (la->la_hold != NULL) {
    825 		struct mbuf *curr = la->la_hold;
    826 		while (curr->m_nextpkt != NULL)
    827 			curr = curr->m_nextpkt;
    828 		curr->m_nextpkt = m;
    829 	} else
    830 		la->la_hold = m;
    831 	la->la_numheld++;
    832 	if (!renew)
    833 		LLE_DOWNGRADE(la);
    834 
    835 	/*
    836 	 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
    837 	 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
    838 	 * if we have already sent arp_maxtries ARP requests. Retransmit the
    839 	 * ARP request, but not faster than one request per second.
    840 	 */
    841 	if (la->la_asked < arp_maxtries)
    842 		error = EWOULDBLOCK;	/* First request. */
    843 	else
    844 		error = (rt != NULL && rt->rt_flags & RTF_GATEWAY) ?
    845 		    EHOSTUNREACH : EHOSTDOWN;
    846 
    847 	if (renew) {
    848 		const u_int8_t *enaddr =
    849 #if NCARP > 0
    850 		    (rt != NULL && rt->rt_ifp->if_type == IFT_CARP) ?
    851 		    CLLADDR(rt->rt_ifp->if_sadl):
    852 #endif
    853 		    CLLADDR(ifp->if_sadl);
    854 		la->la_expire = time_uptime;
    855 		arp_settimer(la, arpt_down);
    856 		la->la_asked++;
    857 		LLE_WUNLOCK(la);
    858 
    859 		if (rt != NULL) {
    860 			arprequest(ifp, &satocsin(rt->rt_ifa->ifa_addr)->sin_addr,
    861 			    &satocsin(dst)->sin_addr, enaddr);
    862 		} else {
    863 			struct sockaddr_in sin;
    864 			struct rtentry *_rt;
    865 
    866 			sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
    867 
    868 			/* XXX */
    869 			_rt = rtalloc1((struct sockaddr *)&sin, 0);
    870 			if (_rt == NULL)
    871 				goto bad;
    872 			arprequest(ifp,
    873 			    &satocsin(_rt->rt_ifa->ifa_addr)->sin_addr,
    874 			    &satocsin(dst)->sin_addr, enaddr);
    875 			rtfree(_rt);
    876 		}
    877 		return error;
    878 	}
    879 
    880 	LLE_RUNLOCK(la);
    881 	return error;
    882 
    883 bad:
    884 	m_freem(m);
    885 	return error;
    886 }
    887 
    888 /*
    889  * Common length and type checks are done here,
    890  * then the protocol-specific routine is called.
    891  */
    892 void
    893 arpintr(void)
    894 {
    895 	struct mbuf *m;
    896 	struct arphdr *ar;
    897 	int s;
    898 	int arplen;
    899 
    900 	mutex_enter(softnet_lock);
    901 	KERNEL_LOCK(1, NULL);
    902 	while (arpintrq.ifq_head) {
    903 		struct ifnet *rcvif;
    904 
    905 		s = splnet();
    906 		IF_DEQUEUE(&arpintrq, m);
    907 		splx(s);
    908 		if (m == NULL || (m->m_flags & M_PKTHDR) == 0)
    909 			panic("arpintr");
    910 
    911 		MCLAIM(m, &arpdomain.dom_mowner);
    912 		ARP_STATINC(ARP_STAT_RCVTOTAL);
    913 
    914 		/*
    915 		 * First, make sure we have at least struct arphdr.
    916 		 */
    917 		if (m->m_len < sizeof(struct arphdr) ||
    918 		    (ar = mtod(m, struct arphdr *)) == NULL)
    919 			goto badlen;
    920 
    921 		rcvif = m_get_rcvif(m, &s);
    922 		switch (rcvif->if_type) {
    923 		case IFT_IEEE1394:
    924 			arplen = sizeof(struct arphdr) +
    925 			    ar->ar_hln + 2 * ar->ar_pln;
    926 			break;
    927 		default:
    928 			arplen = sizeof(struct arphdr) +
    929 			    2 * ar->ar_hln + 2 * ar->ar_pln;
    930 			break;
    931 		}
    932 		m_put_rcvif(rcvif, &s);
    933 
    934 		if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */
    935 		    m->m_len >= arplen)
    936 			switch (ntohs(ar->ar_pro)) {
    937 			case ETHERTYPE_IP:
    938 			case ETHERTYPE_IPTRAILERS:
    939 				in_arpinput(m);
    940 				continue;
    941 			default:
    942 				ARP_STATINC(ARP_STAT_RCVBADPROTO);
    943 			}
    944 		else {
    945 badlen:
    946 			ARP_STATINC(ARP_STAT_RCVBADLEN);
    947 		}
    948 		m_freem(m);
    949 	}
    950 	KERNEL_UNLOCK_ONE(NULL);
    951 	mutex_exit(softnet_lock);
    952 }
    953 
    954 /*
    955  * ARP for Internet protocols on 10 Mb/s Ethernet.
    956  * Algorithm is that given in RFC 826.
    957  * In addition, a sanity check is performed on the sender
    958  * protocol address, to catch impersonators.
    959  * We no longer handle negotiations for use of trailer protocol:
    960  * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
    961  * along with IP replies if we wanted trailers sent to us,
    962  * and also sent them in response to IP replies.
    963  * This allowed either end to announce the desire to receive
    964  * trailer packets.
    965  * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
    966  * but formerly didn't normally send requests.
    967  */
    968 static void
    969 in_arpinput(struct mbuf *m)
    970 {
    971 	struct arphdr *ah;
    972 	struct ifnet *ifp, *rcvif = NULL;
    973 	struct llentry *la = NULL;
    974 	struct in_ifaddr *ia;
    975 #if NBRIDGE > 0
    976 	struct in_ifaddr *bridge_ia = NULL;
    977 #endif
    978 #if NCARP > 0
    979 	u_int32_t count = 0, index = 0;
    980 #endif
    981 	struct sockaddr sa;
    982 	struct in_addr isaddr, itaddr, myaddr;
    983 	int op;
    984 	void *tha;
    985 	uint64_t *arps;
    986 	struct psref psref;
    987 
    988 	if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT)))
    989 		goto out;
    990 	ah = mtod(m, struct arphdr *);
    991 	op = ntohs(ah->ar_op);
    992 
    993 	rcvif = ifp = m_get_rcvif_psref(m, &psref);
    994 	if (__predict_false(rcvif == NULL))
    995 		goto drop;
    996 	/*
    997 	 * Fix up ah->ar_hrd if necessary, before using ar_tha() or
    998 	 * ar_tpa().
    999 	 */
   1000 	switch (ifp->if_type) {
   1001 	case IFT_IEEE1394:
   1002 		if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394)
   1003 			;
   1004 		else {
   1005 			/* XXX this is to make sure we compute ar_tha right */
   1006 			/* XXX check ar_hrd more strictly? */
   1007 			ah->ar_hrd = htons(ARPHRD_IEEE1394);
   1008 		}
   1009 		break;
   1010 	default:
   1011 		/* XXX check ar_hrd? */
   1012 		break;
   1013 	}
   1014 
   1015 	memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
   1016 	memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
   1017 
   1018 	if (m->m_flags & (M_BCAST|M_MCAST))
   1019 		ARP_STATINC(ARP_STAT_RCVMCAST);
   1020 
   1021 
   1022 	/*
   1023 	 * Search for a matching interface address
   1024 	 * or any address on the interface to use
   1025 	 * as a dummy address in the rest of this function
   1026 	 */
   1027 	IN_ADDRHASH_READER_FOREACH(ia, itaddr.s_addr) {
   1028 		if (!in_hosteq(ia->ia_addr.sin_addr, itaddr))
   1029 			continue;
   1030 #if NCARP > 0
   1031 		if (ia->ia_ifp->if_type == IFT_CARP &&
   1032 		    ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) ==
   1033 		    (IFF_UP|IFF_RUNNING))) {
   1034 			index++;
   1035 			if (ia->ia_ifp == rcvif &&
   1036 			    carp_iamatch(ia, ar_sha(ah),
   1037 			    &count, index)) {
   1038 				break;
   1039 			}
   1040 		} else
   1041 #endif
   1042 		if (ia->ia_ifp == rcvif)
   1043 			break;
   1044 #if NBRIDGE > 0
   1045 		/*
   1046 		 * If the interface we received the packet on
   1047 		 * is part of a bridge, check to see if we need
   1048 		 * to "bridge" the packet to ourselves at this
   1049 		 * layer.  Note we still prefer a perfect match,
   1050 		 * but allow this weaker match if necessary.
   1051 		 */
   1052 		if (rcvif->if_bridge != NULL &&
   1053 		    rcvif->if_bridge == ia->ia_ifp->if_bridge)
   1054 			bridge_ia = ia;
   1055 #endif /* NBRIDGE > 0 */
   1056 	}
   1057 
   1058 #if NBRIDGE > 0
   1059 	if (ia == NULL && bridge_ia != NULL) {
   1060 		ia = bridge_ia;
   1061 		m_put_rcvif_psref(rcvif, &psref);
   1062 		rcvif = NULL;
   1063 		/* FIXME */
   1064 		ifp = bridge_ia->ia_ifp;
   1065 	}
   1066 #endif
   1067 
   1068 	if (ia == NULL) {
   1069 		ia = in_get_ia_on_iface(isaddr, rcvif);
   1070 		if (ia == NULL) {
   1071 			ia = in_get_ia_from_ifp(ifp);
   1072 			if (ia == NULL) {
   1073 				ARP_STATINC(ARP_STAT_RCVNOINT);
   1074 				goto out;
   1075 			}
   1076 		}
   1077 	}
   1078 
   1079 	myaddr = ia->ia_addr.sin_addr;
   1080 
   1081 	/* XXX checks for bridge case? */
   1082 	if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) {
   1083 		ARP_STATINC(ARP_STAT_RCVLOCALSHA);
   1084 		goto out;	/* it's from me, ignore it. */
   1085 	}
   1086 
   1087 	/* XXX checks for bridge case? */
   1088 	if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
   1089 		ARP_STATINC(ARP_STAT_RCVBCASTSHA);
   1090 		log(LOG_ERR,
   1091 		    "%s: arp: link address is broadcast for IP address %s!\n",
   1092 		    ifp->if_xname, in_fmtaddr(isaddr));
   1093 		goto out;
   1094 	}
   1095 
   1096 	/*
   1097 	 * If the source IP address is zero, this is an RFC 5227 ARP probe
   1098 	 */
   1099 	if (in_nullhost(isaddr))
   1100 		ARP_STATINC(ARP_STAT_RCVZEROSPA);
   1101 	else if (in_hosteq(isaddr, myaddr))
   1102 		ARP_STATINC(ARP_STAT_RCVLOCALSPA);
   1103 
   1104 	if (in_nullhost(itaddr))
   1105 		ARP_STATINC(ARP_STAT_RCVZEROTPA);
   1106 
   1107 	/* DAD check, RFC 5227 2.1.1, Probe Details */
   1108 	if (in_hosteq(isaddr, myaddr) ||
   1109 	    (in_nullhost(isaddr) && in_hosteq(itaddr, myaddr)))
   1110 	{
   1111 		/* If our address is tentative, mark it as duplicated */
   1112 		if (ia->ia4_flags & IN_IFF_TENTATIVE)
   1113 			arp_dad_duplicated((struct ifaddr *)ia);
   1114 		/* If our address is unuseable, don't reply */
   1115 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
   1116 			goto out;
   1117 	}
   1118 
   1119 	/*
   1120 	 * If the target IP address is zero, ignore the packet.
   1121 	 * This prevents the code below from tring to answer
   1122 	 * when we are using IP address zero (booting).
   1123 	 */
   1124 	if (in_nullhost(itaddr))
   1125 		goto out;
   1126 
   1127 	if (in_nullhost(isaddr))
   1128 		goto reply;
   1129 
   1130 	if (in_hosteq(isaddr, myaddr)) {
   1131 		log(LOG_ERR,
   1132 		   "duplicate IP address %s sent from link address %s\n",
   1133 		   in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln));
   1134 		itaddr = myaddr;
   1135 		goto reply;
   1136 	}
   1137 
   1138 	if (in_hosteq(itaddr, myaddr))
   1139 		la = arpcreate(ifp, m, &isaddr, NULL, 1);
   1140 	else
   1141 		la = arplookup(ifp, m, &isaddr, NULL, 1);
   1142 	if (la == NULL)
   1143 		goto reply;
   1144 
   1145 	if ((la->la_flags & LLE_VALID) &&
   1146 	    memcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
   1147 		if (la->la_flags & LLE_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 (la->lle_tbl->llt_ifp != ifp) {
   1158 			/* XXX should not happen? */
   1159 			ARP_STATINC(ARP_STAT_RCVOVERINT);
   1160 			if (!log_wrong_iface)
   1161 				goto out;
   1162 			log(LOG_INFO,
   1163 			    "%s on %s tried to overwrite "
   1164 			    "arp info for %s on %s\n",
   1165 			    lla_snprintf(ar_sha(ah), ah->ar_hln),
   1166 			    ifp->if_xname, in_fmtaddr(isaddr),
   1167 			    la->lle_tbl->llt_ifp->if_xname);
   1168 				goto out;
   1169 		} else {
   1170 			ARP_STATINC(ARP_STAT_RCVOVER);
   1171 			if (log_movements)
   1172 				log(LOG_INFO, "arp info overwritten "
   1173 				    "for %s by %s\n",
   1174 				    in_fmtaddr(isaddr),
   1175 				    lla_snprintf(ar_sha(ah),
   1176 				    ah->ar_hln));
   1177 		}
   1178 	}
   1179 
   1180 	/* XXX llentry should have addrlen? */
   1181 #if 0
   1182 	/*
   1183 	 * sanity check for the address length.
   1184 	 * XXX this does not work for protocols with variable address
   1185 	 * length. -is
   1186 	 */
   1187 	if (sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) {
   1188 		ARP_STATINC(ARP_STAT_RCVLENCHG);
   1189 		log(LOG_WARNING,
   1190 		    "arp from %s: new addr len %d, was %d\n",
   1191 		    in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen);
   1192 	}
   1193 #endif
   1194 
   1195 	if (ifp->if_addrlen != ah->ar_hln) {
   1196 		ARP_STATINC(ARP_STAT_RCVBADLEN);
   1197 		log(LOG_WARNING,
   1198 		    "arp from %s: addr len: new %d, i/f %d (ignored)\n",
   1199 		    in_fmtaddr(isaddr), ah->ar_hln,
   1200 		    ifp->if_addrlen);
   1201 		goto reply;
   1202 	}
   1203 
   1204 #if NTOKEN > 0
   1205 	/*
   1206 	 * XXX uses m_data and assumes the complete answer including
   1207 	 * XXX token-ring headers is in the same buf
   1208 	 */
   1209 	if (ifp->if_type == IFT_ISO88025) {
   1210 		struct token_header *trh;
   1211 
   1212 		trh = (struct token_header *)M_TRHSTART(m);
   1213 		if (trh->token_shost[0] & TOKEN_RI_PRESENT) {
   1214 			struct token_rif *rif;
   1215 			size_t riflen;
   1216 
   1217 			rif = TOKEN_RIF(trh);
   1218 			riflen = (ntohs(rif->tr_rcf) &
   1219 			    TOKEN_RCF_LEN_MASK) >> 8;
   1220 
   1221 			if (riflen > 2 &&
   1222 			    riflen < sizeof(struct token_rif) &&
   1223 			    (riflen & 1) == 0) {
   1224 				rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION);
   1225 				rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK);
   1226 				memcpy(TOKEN_RIF_LLE(la), rif, riflen);
   1227 			}
   1228 		}
   1229 	}
   1230 #endif /* NTOKEN > 0 */
   1231 
   1232 	KASSERT(sizeof(la->ll_addr) >= ifp->if_addrlen);
   1233 	(void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
   1234 	la->la_flags |= LLE_VALID;
   1235 	if ((la->la_flags & LLE_STATIC) == 0) {
   1236 		la->la_expire = time_uptime + arpt_keep;
   1237 		arp_settimer(la, arpt_keep);
   1238 	}
   1239 	la->la_asked = 0;
   1240 	/* rt->rt_flags &= ~RTF_REJECT; */
   1241 
   1242 	if (la->la_hold != NULL) {
   1243 		int n = la->la_numheld;
   1244 		struct mbuf *m_hold, *m_hold_next;
   1245 		struct sockaddr_in sin;
   1246 
   1247 		sockaddr_in_init(&sin, &la->r_l3addr.addr4, 0);
   1248 
   1249 		m_hold = la->la_hold;
   1250 		la->la_hold = NULL;
   1251 		la->la_numheld = 0;
   1252 		/*
   1253 		 * We have to unlock here because if_output would call
   1254 		 * arpresolve
   1255 		 */
   1256 		LLE_WUNLOCK(la);
   1257 		ARP_STATADD(ARP_STAT_DFRSENT, n);
   1258 		for (; m_hold != NULL; m_hold = m_hold_next) {
   1259 			m_hold_next = m_hold->m_nextpkt;
   1260 			m_hold->m_nextpkt = NULL;
   1261 			if_output_lock(ifp, ifp, m_hold, sintosa(&sin), NULL);
   1262 		}
   1263 	} else
   1264 		LLE_WUNLOCK(la);
   1265 	la = NULL;
   1266 
   1267 reply:
   1268 	if (la != NULL) {
   1269 		LLE_WUNLOCK(la);
   1270 		la = NULL;
   1271 	}
   1272 	if (op != ARPOP_REQUEST) {
   1273 		if (op == ARPOP_REPLY)
   1274 			ARP_STATINC(ARP_STAT_RCVREPLY);
   1275 		goto out;
   1276 	}
   1277 	ARP_STATINC(ARP_STAT_RCVREQUEST);
   1278 	if (in_hosteq(itaddr, myaddr)) {
   1279 		/* If our address is unuseable, don't reply */
   1280 		if (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED))
   1281 			goto out;
   1282 		/* I am the target */
   1283 		tha = ar_tha(ah);
   1284 		if (tha)
   1285 			memcpy(tha, ar_sha(ah), ah->ar_hln);
   1286 		memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
   1287 	} else {
   1288 		/* Proxy ARP */
   1289 		struct llentry *lle = NULL;
   1290 		struct sockaddr_in sin;
   1291 #if NCARP > 0
   1292 		int s;
   1293 		struct ifnet *_rcvif = m_get_rcvif(m, &s);
   1294 		if (ifp->if_type == IFT_CARP && _rcvif->if_type != IFT_CARP)
   1295 			goto out;
   1296 		m_put_rcvif(_rcvif, &s);
   1297 #endif
   1298 
   1299 		tha = ar_tha(ah);
   1300 
   1301 		sockaddr_in_init(&sin, &itaddr, 0);
   1302 
   1303 		IF_AFDATA_RLOCK(ifp);
   1304 		lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
   1305 		IF_AFDATA_RUNLOCK(ifp);
   1306 
   1307 		if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
   1308 			(void)memcpy(tha, ar_sha(ah), ah->ar_hln);
   1309 			(void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
   1310 			LLE_RUNLOCK(lle);
   1311 		} else {
   1312 			if (lle != NULL)
   1313 				LLE_RUNLOCK(lle);
   1314 			goto drop;
   1315 		}
   1316 	}
   1317 
   1318 	memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
   1319 	memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
   1320 	ah->ar_op = htons(ARPOP_REPLY);
   1321 	ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
   1322 	switch (ifp->if_type) {
   1323 	case IFT_IEEE1394:
   1324 		/*
   1325 		 * ieee1394 arp reply is broadcast
   1326 		 */
   1327 		m->m_flags &= ~M_MCAST;
   1328 		m->m_flags |= M_BCAST;
   1329 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln;
   1330 		break;
   1331 
   1332 	default:
   1333 		m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */
   1334 		m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
   1335 		break;
   1336 	}
   1337 	m->m_pkthdr.len = m->m_len;
   1338 	sa.sa_family = AF_ARP;
   1339 	sa.sa_len = 2;
   1340 	arps = ARP_STAT_GETREF();
   1341 	arps[ARP_STAT_SNDTOTAL]++;
   1342 	arps[ARP_STAT_SNDREPLY]++;
   1343 	ARP_STAT_PUTREF();
   1344 	if_output_lock(ifp, ifp, m, &sa, NULL);
   1345 	if (rcvif != NULL)
   1346 		m_put_rcvif_psref(rcvif, &psref);
   1347 	return;
   1348 
   1349 out:
   1350 	if (la != NULL)
   1351 		LLE_WUNLOCK(la);
   1352 drop:
   1353 	if (rcvif != NULL)
   1354 		m_put_rcvif_psref(rcvif, &psref);
   1355 	m_freem(m);
   1356 }
   1357 
   1358 /*
   1359  * Lookup or a new address in arptab.
   1360  */
   1361 static struct llentry *
   1362 arplookup(struct ifnet *ifp, struct mbuf *m, const struct in_addr *addr,
   1363     const struct sockaddr *sa, int wlock)
   1364 {
   1365 	struct sockaddr_in sin;
   1366 	struct llentry *la;
   1367 	int flags = wlock ? LLE_EXCLUSIVE : 0;
   1368 
   1369 
   1370 	if (sa == NULL) {
   1371 		KASSERT(addr != NULL);
   1372 		sockaddr_in_init(&sin, addr, 0);
   1373 		sa = sintocsa(&sin);
   1374 	}
   1375 
   1376 	IF_AFDATA_RLOCK(ifp);
   1377 	la = lla_lookup(LLTABLE(ifp), flags, sa);
   1378 	IF_AFDATA_RUNLOCK(ifp);
   1379 
   1380 	return la;
   1381 }
   1382 
   1383 static struct llentry *
   1384 arpcreate(struct ifnet *ifp, struct mbuf *m, const struct in_addr *addr,
   1385     const struct sockaddr *sa, int wlock)
   1386 {
   1387 	struct sockaddr_in sin;
   1388 	struct llentry *la;
   1389 	int flags = wlock ? LLE_EXCLUSIVE : 0;
   1390 
   1391 	if (sa == NULL) {
   1392 		KASSERT(addr != NULL);
   1393 		sockaddr_in_init(&sin, addr, 0);
   1394 		sa = sintocsa(&sin);
   1395 	}
   1396 
   1397 	la = arplookup(ifp, m, addr, sa, wlock);
   1398 
   1399 	if (la == NULL) {
   1400 		IF_AFDATA_WLOCK(ifp);
   1401 		la = lla_create(LLTABLE(ifp), flags, sa);
   1402 		IF_AFDATA_WUNLOCK(ifp);
   1403 
   1404 		if (la != NULL)
   1405 			arp_init_llentry(ifp, la);
   1406 	}
   1407 
   1408 	return la;
   1409 }
   1410 
   1411 int
   1412 arpioctl(u_long cmd, void *data)
   1413 {
   1414 
   1415 	return EOPNOTSUPP;
   1416 }
   1417 
   1418 void
   1419 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
   1420 {
   1421 	struct in_addr *ip;
   1422 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1423 
   1424 	/*
   1425 	 * Warn the user if another station has this IP address,
   1426 	 * but only if the interface IP address is not zero.
   1427 	 */
   1428 	ip = &IA_SIN(ifa)->sin_addr;
   1429 	if (!in_nullhost(*ip) &&
   1430 	    (ia->ia4_flags & (IN_IFF_NOTREADY | IN_IFF_DETACHED)) == 0) {
   1431 		struct llentry *lle;
   1432 
   1433 		arprequest(ifp, ip, ip, CLLADDR(ifp->if_sadl));
   1434 
   1435 		/*
   1436 		 * interface address is considered static entry
   1437 		 * because the output of the arp utility shows
   1438 		 * that L2 entry as permanent
   1439 		 */
   1440 		IF_AFDATA_WLOCK(ifp);
   1441 		lle = lla_create(LLTABLE(ifp), (LLE_IFADDR | LLE_STATIC),
   1442 				 (struct sockaddr *)IA_SIN(ifa));
   1443 		IF_AFDATA_WUNLOCK(ifp);
   1444 		if (lle == NULL)
   1445 			log(LOG_INFO, "%s: cannot create arp entry for"
   1446 			    " interface address\n", __func__);
   1447 		else {
   1448 			arp_init_llentry(ifp, lle);
   1449 			LLE_RUNLOCK(lle);
   1450 		}
   1451 	}
   1452 
   1453 	ifa->ifa_rtrequest = arp_rtrequest;
   1454 	ifa->ifa_flags |= RTF_CONNECTED;
   1455 
   1456 	/* ARP will handle DAD for this address. */
   1457 	if (ia->ia4_flags & IN_IFF_TRYTENTATIVE) {
   1458 		ia->ia4_flags |= IN_IFF_TENTATIVE;
   1459 		ia->ia_dad_start = arp_dad_start;
   1460 		ia->ia_dad_stop = arp_dad_stop;
   1461 	}
   1462 }
   1463 
   1464 TAILQ_HEAD(dadq_head, dadq);
   1465 struct dadq {
   1466 	TAILQ_ENTRY(dadq) dad_list;
   1467 	struct ifaddr *dad_ifa;
   1468 	int dad_count;		/* max ARP to send */
   1469 	int dad_arp_tcount;	/* # of trials to send ARP */
   1470 	int dad_arp_ocount;	/* ARP sent so far */
   1471 	int dad_arp_announce;	/* max ARP announcements */
   1472 	int dad_arp_acount;	/* # of announcements */
   1473 	struct callout dad_timer_ch;
   1474 };
   1475 MALLOC_JUSTDEFINE(M_IPARP, "ARP DAD", "ARP DAD Structure");
   1476 
   1477 static struct dadq_head dadq;
   1478 static int dad_init = 0;
   1479 static int dad_maxtry = 15;     /* max # of *tries* to transmit DAD packet */
   1480 
   1481 static struct dadq *
   1482 arp_dad_find(struct ifaddr *ifa)
   1483 {
   1484 	struct dadq *dp;
   1485 
   1486 	TAILQ_FOREACH(dp, &dadq, dad_list) {
   1487 		if (dp->dad_ifa == ifa)
   1488 			return dp;
   1489 	}
   1490 	return NULL;
   1491 }
   1492 
   1493 static void
   1494 arp_dad_starttimer(struct dadq *dp, int ticks)
   1495 {
   1496 
   1497 	callout_reset(&dp->dad_timer_ch, ticks,
   1498 	    (void (*)(void *))arp_dad_timer, (void *)dp->dad_ifa);
   1499 }
   1500 
   1501 static void
   1502 arp_dad_stoptimer(struct dadq *dp)
   1503 {
   1504 
   1505 	callout_halt(&dp->dad_timer_ch, NULL);
   1506 }
   1507 
   1508 static void
   1509 arp_dad_output(struct dadq *dp, struct ifaddr *ifa)
   1510 {
   1511 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1512 	struct ifnet *ifp = ifa->ifa_ifp;
   1513 	struct in_addr sip;
   1514 
   1515 	dp->dad_arp_tcount++;
   1516 	if ((ifp->if_flags & IFF_UP) == 0)
   1517 		return;
   1518 	if ((ifp->if_flags & IFF_RUNNING) == 0)
   1519 		return;
   1520 
   1521 	dp->dad_arp_tcount = 0;
   1522 	dp->dad_arp_ocount++;
   1523 
   1524 	memset(&sip, 0, sizeof(sip));
   1525 	arprequest(ifa->ifa_ifp, &sip, &ia->ia_addr.sin_addr,
   1526 	    CLLADDR(ifa->ifa_ifp->if_sadl));
   1527 }
   1528 
   1529 /*
   1530  * Start Duplicate Address Detection (DAD) for specified interface address.
   1531  */
   1532 static void
   1533 arp_dad_start(struct ifaddr *ifa)
   1534 {
   1535 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1536 	struct dadq *dp;
   1537 
   1538 	if (!dad_init) {
   1539 		TAILQ_INIT(&dadq);
   1540 		dad_init++;
   1541 	}
   1542 
   1543 	/*
   1544 	 * If we don't need DAD, don't do it.
   1545 	 * - DAD is disabled (ip_dad_count == 0)
   1546 	 */
   1547 	if (!(ia->ia4_flags & IN_IFF_TENTATIVE)) {
   1548 		log(LOG_DEBUG,
   1549 		    "%s: called with non-tentative address %s(%s)\n", __func__,
   1550 		    in_fmtaddr(ia->ia_addr.sin_addr),
   1551 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
   1552 		return;
   1553 	}
   1554 	if (!ip_dad_count) {
   1555 		struct in_addr *ip = &IA_SIN(ifa)->sin_addr;
   1556 
   1557 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
   1558 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1559 		arprequest(ifa->ifa_ifp, ip, ip,
   1560 		    CLLADDR(ifa->ifa_ifp->if_sadl));
   1561 		return;
   1562 	}
   1563 	if (ifa->ifa_ifp == NULL)
   1564 		panic("arp_dad_start: ifa->ifa_ifp == NULL");
   1565 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
   1566 		return;
   1567 	if (arp_dad_find(ifa) != NULL) {
   1568 		/* DAD already in progress */
   1569 		return;
   1570 	}
   1571 
   1572 	dp = malloc(sizeof(*dp), M_IPARP, M_NOWAIT);
   1573 	if (dp == NULL) {
   1574 		log(LOG_ERR, "%s: memory allocation failed for %s(%s)\n",
   1575 		    __func__, in_fmtaddr(ia->ia_addr.sin_addr),
   1576 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
   1577 		return;
   1578 	}
   1579 	memset(dp, 0, sizeof(*dp));
   1580 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
   1581 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
   1582 
   1583 	arplog((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
   1584 	    in_fmtaddr(ia->ia_addr.sin_addr)));
   1585 
   1586 	/*
   1587 	 * Send ARP packet for DAD, ip_dad_count times.
   1588 	 * Note that we must delay the first transmission.
   1589 	 */
   1590 	dp->dad_ifa = ifa;
   1591 	ifaref(ifa);	/* just for safety */
   1592 	dp->dad_count = ip_dad_count;
   1593 	dp->dad_arp_announce = 0; /* Will be set when starting to announce */
   1594 	dp->dad_arp_acount = dp->dad_arp_ocount = dp->dad_arp_tcount = 0;
   1595 
   1596 	arp_dad_starttimer(dp, cprng_fast32() % (PROBE_WAIT * hz));
   1597 }
   1598 
   1599 /*
   1600  * terminate DAD unconditionally.  used for address removals.
   1601  */
   1602 static void
   1603 arp_dad_stop(struct ifaddr *ifa)
   1604 {
   1605 	struct dadq *dp;
   1606 
   1607 	if (!dad_init)
   1608 		return;
   1609 	dp = arp_dad_find(ifa);
   1610 	if (dp == NULL) {
   1611 		/* DAD wasn't started yet */
   1612 		return;
   1613 	}
   1614 
   1615 	arp_dad_stoptimer(dp);
   1616 
   1617 	TAILQ_REMOVE(&dadq, dp, dad_list);
   1618 	free(dp, M_IPARP);
   1619 	dp = NULL;
   1620 	ifafree(ifa);
   1621 }
   1622 
   1623 static void
   1624 arp_dad_timer(struct ifaddr *ifa)
   1625 {
   1626 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1627 	struct dadq *dp;
   1628 	struct in_addr *ip;
   1629 
   1630 	mutex_enter(softnet_lock);
   1631 	KERNEL_LOCK(1, NULL);
   1632 
   1633 	/* Sanity check */
   1634 	if (ia == NULL) {
   1635 		log(LOG_ERR, "%s: called with null parameter\n", __func__);
   1636 		goto done;
   1637 	}
   1638 	dp = arp_dad_find(ifa);
   1639 	if (dp == NULL) {
   1640 		log(LOG_ERR, "%s: DAD structure not found\n", __func__);
   1641 		goto done;
   1642 	}
   1643 	if (ia->ia4_flags & IN_IFF_DUPLICATED) {
   1644 		log(LOG_ERR, "%s: called with duplicate address %s(%s)\n",
   1645 		    __func__, in_fmtaddr(ia->ia_addr.sin_addr),
   1646 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
   1647 		goto done;
   1648 	}
   1649 	if ((ia->ia4_flags & IN_IFF_TENTATIVE) == 0 && dp->dad_arp_acount == 0)
   1650 	{
   1651 		log(LOG_ERR, "%s: called with non-tentative address %s(%s)\n",
   1652 		    __func__, in_fmtaddr(ia->ia_addr.sin_addr),
   1653 		    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
   1654 		goto done;
   1655 	}
   1656 
   1657 	/* timeouted with IFF_{RUNNING,UP} check */
   1658 	if (dp->dad_arp_tcount > dad_maxtry) {
   1659 		arplog((LOG_INFO, "%s: could not run DAD, driver problem?\n",
   1660 		    if_name(ifa->ifa_ifp)));
   1661 
   1662 		TAILQ_REMOVE(&dadq, dp, dad_list);
   1663 		free(dp, M_IPARP);
   1664 		dp = NULL;
   1665 		ifafree(ifa);
   1666 		goto done;
   1667 	}
   1668 
   1669 	/* Need more checks? */
   1670 	if (dp->dad_arp_ocount < dp->dad_count) {
   1671 		int adelay;
   1672 
   1673 		/*
   1674 		 * We have more ARP to go.  Send ARP packet for DAD.
   1675 		 */
   1676 		arp_dad_output(dp, ifa);
   1677 		if (dp->dad_arp_ocount < dp->dad_count)
   1678 			adelay = (PROBE_MIN * hz) +
   1679 			    (cprng_fast32() %
   1680 			    ((PROBE_MAX * hz) - (PROBE_MIN * hz)));
   1681 		else
   1682 			adelay = ANNOUNCE_WAIT * hz;
   1683 		arp_dad_starttimer(dp, adelay);
   1684 		goto done;
   1685 	} else if (dp->dad_arp_acount == 0) {
   1686 		/*
   1687 		 * We are done with DAD.
   1688 		 * No duplicate address found.
   1689 		 */
   1690 		ia->ia4_flags &= ~IN_IFF_TENTATIVE;
   1691 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1692 		arplog((LOG_DEBUG,
   1693 		    "%s: DAD complete for %s - no duplicates found\n",
   1694 		    if_name(ifa->ifa_ifp),
   1695 		    in_fmtaddr(ia->ia_addr.sin_addr)));
   1696 		dp->dad_arp_announce = ANNOUNCE_NUM;
   1697 		goto announce;
   1698 	} else if (dp->dad_arp_acount < dp->dad_arp_announce) {
   1699 announce:
   1700 		/*
   1701 		 * Announce the address.
   1702 		 */
   1703 		ip = &IA_SIN(ifa)->sin_addr;
   1704 		arprequest(ifa->ifa_ifp, ip, ip,
   1705 		    CLLADDR(ifa->ifa_ifp->if_sadl));
   1706 		dp->dad_arp_acount++;
   1707 		if (dp->dad_arp_acount < dp->dad_arp_announce) {
   1708 			arp_dad_starttimer(dp, ANNOUNCE_INTERVAL * hz);
   1709 			goto done;
   1710 		}
   1711 		arplog((LOG_DEBUG,
   1712 		    "%s: ARP announcement complete for %s\n",
   1713 		    if_name(ifa->ifa_ifp),
   1714 		    in_fmtaddr(ia->ia_addr.sin_addr)));
   1715 	}
   1716 
   1717 	TAILQ_REMOVE(&dadq, dp, dad_list);
   1718 	free(dp, M_IPARP);
   1719 	dp = NULL;
   1720 	ifafree(ifa);
   1721 
   1722 done:
   1723 	KERNEL_UNLOCK_ONE(NULL);
   1724 	mutex_exit(softnet_lock);
   1725 }
   1726 
   1727 static void
   1728 arp_dad_duplicated(struct ifaddr *ifa)
   1729 {
   1730 	struct in_ifaddr *ia = (struct in_ifaddr *)ifa;
   1731 	struct ifnet *ifp;
   1732 	struct dadq *dp;
   1733 
   1734 	dp = arp_dad_find(ifa);
   1735 	if (dp == NULL) {
   1736 		log(LOG_ERR, "%s: DAD structure not found\n", __func__);
   1737 		return;
   1738 	}
   1739 
   1740 	ifp = ifa->ifa_ifp;
   1741 	log(LOG_ERR,
   1742 	    "%s: DAD detected duplicate IPv4 address %s: ARP out=%d\n",
   1743 	    if_name(ifp), in_fmtaddr(ia->ia_addr.sin_addr),
   1744 	    dp->dad_arp_ocount);
   1745 
   1746 	ia->ia4_flags &= ~IN_IFF_TENTATIVE;
   1747 	ia->ia4_flags |= IN_IFF_DUPLICATED;
   1748 
   1749 	/* We are done with DAD, with duplicated address found. (failure) */
   1750 	arp_dad_stoptimer(dp);
   1751 
   1752 	/* Inform the routing socket that DAD has completed */
   1753 	rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   1754 
   1755 	TAILQ_REMOVE(&dadq, dp, dad_list);
   1756 	free(dp, M_IPARP);
   1757 	dp = NULL;
   1758 	ifafree(ifa);
   1759 }
   1760 
   1761 /*
   1762  * Called from 10 Mb/s Ethernet interrupt handlers
   1763  * when ether packet type ETHERTYPE_REVARP
   1764  * is received.  Common length and type checks are done here,
   1765  * then the protocol-specific routine is called.
   1766  */
   1767 void
   1768 revarpinput(struct mbuf *m)
   1769 {
   1770 	struct arphdr *ar;
   1771 
   1772 	if (m->m_len < sizeof(struct arphdr))
   1773 		goto out;
   1774 	ar = mtod(m, struct arphdr *);
   1775 #if 0 /* XXX I don't think we need this... and it will prevent other LL */
   1776 	if (ntohs(ar->ar_hrd) != ARPHRD_ETHER)
   1777 		goto out;
   1778 #endif
   1779 	if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln))
   1780 		goto out;
   1781 	switch (ntohs(ar->ar_pro)) {
   1782 	case ETHERTYPE_IP:
   1783 	case ETHERTYPE_IPTRAILERS:
   1784 		in_revarpinput(m);
   1785 		return;
   1786 
   1787 	default:
   1788 		break;
   1789 	}
   1790 out:
   1791 	m_freem(m);
   1792 }
   1793 
   1794 /*
   1795  * RARP for Internet protocols on 10 Mb/s Ethernet.
   1796  * Algorithm is that given in RFC 903.
   1797  * We are only using for bootstrap purposes to get an ip address for one of
   1798  * our interfaces.  Thus we support no user-interface.
   1799  *
   1800  * Since the contents of the RARP reply are specific to the interface that
   1801  * sent the request, this code must ensure that they are properly associated.
   1802  *
   1803  * Note: also supports ARP via RARP packets, per the RFC.
   1804  */
   1805 void
   1806 in_revarpinput(struct mbuf *m)
   1807 {
   1808 	struct arphdr *ah;
   1809 	void *tha;
   1810 	int op;
   1811 	struct ifnet *rcvif;
   1812 	int s;
   1813 
   1814 	ah = mtod(m, struct arphdr *);
   1815 	op = ntohs(ah->ar_op);
   1816 
   1817 	rcvif = m_get_rcvif(m, &s);
   1818 	switch (rcvif->if_type) {
   1819 	case IFT_IEEE1394:
   1820 		/* ARP without target hardware address is not supported */
   1821 		goto out;
   1822 	default:
   1823 		break;
   1824 	}
   1825 
   1826 	switch (op) {
   1827 	case ARPOP_REQUEST:
   1828 	case ARPOP_REPLY:	/* per RFC */
   1829 		m_put_rcvif(rcvif, &s);
   1830 		in_arpinput(m);
   1831 		return;
   1832 	case ARPOP_REVREPLY:
   1833 		break;
   1834 	case ARPOP_REVREQUEST:	/* handled by rarpd(8) */
   1835 	default:
   1836 		goto out;
   1837 	}
   1838 	if (!revarp_in_progress)
   1839 		goto out;
   1840 	if (rcvif != myip_ifp) /* !same interface */
   1841 		goto out;
   1842 	if (myip_initialized)
   1843 		goto wake;
   1844 	tha = ar_tha(ah);
   1845 	if (tha == NULL)
   1846 		goto out;
   1847 	if (memcmp(tha, CLLADDR(rcvif->if_sadl), rcvif->if_sadl->sdl_alen))
   1848 		goto out;
   1849 	memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip));
   1850 	memcpy(&myip, ar_tpa(ah), sizeof(myip));
   1851 	myip_initialized = 1;
   1852 wake:	/* Do wakeup every time in case it was missed. */
   1853 	wakeup((void *)&myip);
   1854 
   1855 out:
   1856 	m_put_rcvif(rcvif, &s);
   1857 	m_freem(m);
   1858 }
   1859 
   1860 /*
   1861  * Send a RARP request for the ip address of the specified interface.
   1862  * The request should be RFC 903-compliant.
   1863  */
   1864 static void
   1865 revarprequest(struct ifnet *ifp)
   1866 {
   1867 	struct sockaddr sa;
   1868 	struct mbuf *m;
   1869 	struct arphdr *ah;
   1870 	void *tha;
   1871 
   1872 	if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL)
   1873 		return;
   1874 	MCLAIM(m, &arpdomain.dom_mowner);
   1875 	m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
   1876 	    2*ifp->if_addrlen;
   1877 	m->m_pkthdr.len = m->m_len;
   1878 	MH_ALIGN(m, m->m_len);
   1879 	ah = mtod(m, struct arphdr *);
   1880 	memset(ah, 0, m->m_len);
   1881 	ah->ar_pro = htons(ETHERTYPE_IP);
   1882 	ah->ar_hln = ifp->if_addrlen;		/* hardware address length */
   1883 	ah->ar_pln = sizeof(struct in_addr);	/* protocol address length */
   1884 	ah->ar_op = htons(ARPOP_REVREQUEST);
   1885 
   1886 	memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln);
   1887 	tha = ar_tha(ah);
   1888 	if (tha == NULL) {
   1889 		m_free(m);
   1890 		return;
   1891 	}
   1892 	memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln);
   1893 
   1894 	sa.sa_family = AF_ARP;
   1895 	sa.sa_len = 2;
   1896 	m->m_flags |= M_BCAST;
   1897 
   1898 	if_output_lock(ifp, ifp, m, &sa, NULL);
   1899 }
   1900 
   1901 /*
   1902  * RARP for the ip address of the specified interface, but also
   1903  * save the ip address of the server that sent the answer.
   1904  * Timeout if no response is received.
   1905  */
   1906 int
   1907 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in,
   1908     struct in_addr *clnt_in)
   1909 {
   1910 	int result, count = 20;
   1911 
   1912 	myip_initialized = 0;
   1913 	myip_ifp = ifp;
   1914 
   1915 	revarp_in_progress = 1;
   1916 	while (count--) {
   1917 		revarprequest(ifp);
   1918 		result = tsleep((void *)&myip, PSOCK, "revarp", hz/2);
   1919 		if (result != EWOULDBLOCK)
   1920 			break;
   1921 	}
   1922 	revarp_in_progress = 0;
   1923 
   1924 	if (!myip_initialized)
   1925 		return ENETUNREACH;
   1926 
   1927 	memcpy(serv_in, &srv_ip, sizeof(*serv_in));
   1928 	memcpy(clnt_in, &myip, sizeof(*clnt_in));
   1929 	return 0;
   1930 }
   1931 
   1932 void
   1933 arp_stat_add(int type, uint64_t count)
   1934 {
   1935 	ARP_STATADD(type, count);
   1936 }
   1937 
   1938 static int
   1939 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS)
   1940 {
   1941 
   1942 	return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS);
   1943 }
   1944 
   1945 static void
   1946 sysctl_net_inet_arp_setup(struct sysctllog **clog)
   1947 {
   1948 	const struct sysctlnode *node;
   1949 
   1950 	sysctl_createv(clog, 0, NULL, NULL,
   1951 			CTLFLAG_PERMANENT,
   1952 			CTLTYPE_NODE, "inet", NULL,
   1953 			NULL, 0, NULL, 0,
   1954 			CTL_NET, PF_INET, CTL_EOL);
   1955 	sysctl_createv(clog, 0, NULL, &node,
   1956 			CTLFLAG_PERMANENT,
   1957 			CTLTYPE_NODE, "arp",
   1958 			SYSCTL_DESCR("Address Resolution Protocol"),
   1959 			NULL, 0, NULL, 0,
   1960 			CTL_NET, PF_INET, CTL_CREATE, CTL_EOL);
   1961 
   1962 	sysctl_createv(clog, 0, NULL, NULL,
   1963 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1964 			CTLTYPE_INT, "keep",
   1965 			SYSCTL_DESCR("Valid ARP entry lifetime in seconds"),
   1966 			NULL, 0, &arpt_keep, 0,
   1967 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   1968 
   1969 	sysctl_createv(clog, 0, NULL, NULL,
   1970 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1971 			CTLTYPE_INT, "down",
   1972 			SYSCTL_DESCR("Failed ARP entry lifetime in seconds"),
   1973 			NULL, 0, &arpt_down, 0,
   1974 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   1975 
   1976 	sysctl_createv(clog, 0, NULL, NULL,
   1977 			CTLFLAG_PERMANENT,
   1978 			CTLTYPE_STRUCT, "stats",
   1979 			SYSCTL_DESCR("ARP statistics"),
   1980 			sysctl_net_inet_arp_stats, 0, NULL, 0,
   1981 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   1982 
   1983 	sysctl_createv(clog, 0, NULL, NULL,
   1984 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1985 			CTLTYPE_INT, "log_movements",
   1986 			SYSCTL_DESCR("log ARP replies from MACs different than"
   1987 			    " the one in the cache"),
   1988 			NULL, 0, &log_movements, 0,
   1989 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   1990 
   1991 	sysctl_createv(clog, 0, NULL, NULL,
   1992 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   1993 			CTLTYPE_INT, "log_permanent_modify",
   1994 			SYSCTL_DESCR("log ARP replies from MACs different than"
   1995 			    " the one in the permanent arp entry"),
   1996 			NULL, 0, &log_permanent_modify, 0,
   1997 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   1998 
   1999 	sysctl_createv(clog, 0, NULL, NULL,
   2000 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2001 			CTLTYPE_INT, "log_wrong_iface",
   2002 			SYSCTL_DESCR("log ARP packets arriving on the wrong"
   2003 			    " interface"),
   2004 			NULL, 0, &log_wrong_iface, 0,
   2005 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2006 
   2007 	sysctl_createv(clog, 0, NULL, NULL,
   2008 			CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2009 			CTLTYPE_INT, "log_unknown_network",
   2010 			SYSCTL_DESCR("log ARP packets from non-local network"),
   2011 			NULL, 0, &log_unknown_network, 0,
   2012 			CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2013 
   2014 	sysctl_createv(clog, 0, NULL, NULL,
   2015 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   2016 		       CTLTYPE_INT, "debug",
   2017 		       SYSCTL_DESCR("Enable ARP DAD debug output"),
   2018 		       NULL, 0, &arp_debug, 0,
   2019 		       CTL_NET, PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL);
   2020 }
   2021 
   2022 #endif /* INET */
   2023