Home | History | Annotate | Line # | Download | only in netinet
ip_icmp.c revision 1.36
      1 /*	$NetBSD: ip_icmp.c,v 1.36 1999/03/30 19:02:56 mycroft Exp $	*/
      2 
      3 /*-
      4  * Copyright (c) 1998 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  * 3. All advertising materials mentioning features or use of this software
     20  *    must display the following acknowledgement:
     21  *	This product includes software developed by the NetBSD
     22  *	Foundation, Inc. and its contributors.
     23  * 4. Neither the name of The NetBSD Foundation nor the names of its
     24  *    contributors may be used to endorse or promote products derived
     25  *    from this software without specific prior written permission.
     26  *
     27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     37  * POSSIBILITY OF SUCH DAMAGE.
     38  */
     39 
     40 /*
     41  * Copyright (c) 1982, 1986, 1988, 1993
     42  *	The Regents of the University of California.  All rights reserved.
     43  *
     44  * Redistribution and use in source and binary forms, with or without
     45  * modification, are permitted provided that the following conditions
     46  * are met:
     47  * 1. Redistributions of source code must retain the above copyright
     48  *    notice, this list of conditions and the following disclaimer.
     49  * 2. Redistributions in binary form must reproduce the above copyright
     50  *    notice, this list of conditions and the following disclaimer in the
     51  *    documentation and/or other materials provided with the distribution.
     52  * 3. All advertising materials mentioning features or use of this software
     53  *    must display the following acknowledgement:
     54  *	This product includes software developed by the University of
     55  *	California, Berkeley and its contributors.
     56  * 4. Neither the name of the University nor the names of its contributors
     57  *    may be used to endorse or promote products derived from this software
     58  *    without specific prior written permission.
     59  *
     60  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     61  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     62  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     63  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     64  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     65  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     66  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     67  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     68  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     69  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     70  * SUCH DAMAGE.
     71  *
     72  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
     73  */
     74 
     75 #include <sys/param.h>
     76 #include <sys/systm.h>
     77 #include <sys/malloc.h>
     78 #include <sys/mbuf.h>
     79 #include <sys/protosw.h>
     80 #include <sys/socket.h>
     81 #include <sys/time.h>
     82 #include <sys/kernel.h>
     83 #include <sys/proc.h>
     84 
     85 #include <vm/vm.h>
     86 #include <sys/sysctl.h>
     87 
     88 #include <net/if.h>
     89 #include <net/route.h>
     90 
     91 #include <netinet/in.h>
     92 #include <netinet/in_systm.h>
     93 #include <netinet/in_var.h>
     94 #include <netinet/ip.h>
     95 #include <netinet/ip_icmp.h>
     96 #include <netinet/ip_var.h>
     97 #include <netinet/icmp_var.h>
     98 
     99 #include <machine/stdarg.h>
    100 
    101 /*
    102  * ICMP routines: error generation, receive packet processing, and
    103  * routines to turnaround packets back to the originator, and
    104  * host table maintenance routines.
    105  */
    106 
    107 int	icmpmaskrepl = 0;
    108 #ifdef ICMPPRINTFS
    109 int	icmpprintfs = 0;
    110 #endif
    111 
    112 extern	struct protosw inetsw[];
    113 
    114 static void icmp_mtudisc __P((struct icmp *));
    115 static void icmp_mtudisc_timeout __P((struct rtentry *, struct rttimer *));
    116 
    117 /*
    118  * Generate an error packet of type error
    119  * in response to bad packet ip.
    120  */
    121 void
    122 icmp_error(n, type, code, dest, destifp)
    123 	struct mbuf *n;
    124 	int type, code;
    125 	n_long dest;
    126 	struct ifnet *destifp;
    127 {
    128 	register struct ip *oip = mtod(n, struct ip *), *nip;
    129 	register unsigned oiplen = oip->ip_hl << 2;
    130 	register struct icmp *icp;
    131 	register struct mbuf *m;
    132 	unsigned icmplen;
    133 
    134 #ifdef ICMPPRINTFS
    135 	if (icmpprintfs)
    136 		printf("icmp_error(%x, %d, %d)\n", oip, type, code);
    137 #endif
    138 	if (type != ICMP_REDIRECT)
    139 		icmpstat.icps_error++;
    140 	/*
    141 	 * Don't send error if not the first fragment of message.
    142 	 * Don't error if the old packet protocol was ICMP
    143 	 * error message, only known informational types.
    144 	 */
    145 	if (oip->ip_off &~ (IP_MF|IP_DF))
    146 		goto freeit;
    147 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
    148 	  n->m_len >= oiplen + ICMP_MINLEN &&
    149 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
    150 		icmpstat.icps_oldicmp++;
    151 		goto freeit;
    152 	}
    153 	/* Don't send error in response to a multicast or broadcast packet */
    154 	if (n->m_flags & (M_BCAST|M_MCAST))
    155 		goto freeit;
    156 	/*
    157 	 * First, formulate icmp message
    158 	 */
    159 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
    160 	if (m == NULL)
    161 		goto freeit;
    162 	icmplen = oiplen + min(8, oip->ip_len - oiplen);
    163 	m->m_len = icmplen + ICMP_MINLEN;
    164 	MH_ALIGN(m, m->m_len);
    165 	icp = mtod(m, struct icmp *);
    166 	if ((u_int)type > ICMP_MAXTYPE)
    167 		panic("icmp_error");
    168 	icmpstat.icps_outhist[type]++;
    169 	icp->icmp_type = type;
    170 	if (type == ICMP_REDIRECT)
    171 		icp->icmp_gwaddr.s_addr = dest;
    172 	else {
    173 		icp->icmp_void = 0;
    174 		/*
    175 		 * The following assignments assume an overlay with the
    176 		 * zeroed icmp_void field.
    177 		 */
    178 		if (type == ICMP_PARAMPROB) {
    179 			icp->icmp_pptr = code;
    180 			code = 0;
    181 		} else if (type == ICMP_UNREACH &&
    182 		    code == ICMP_UNREACH_NEEDFRAG && destifp)
    183 			icp->icmp_nextmtu = htons(destifp->if_mtu);
    184 	}
    185 
    186 	HTONS(oip->ip_off);
    187 	HTONS(oip->ip_len);
    188 	icp->icmp_code = code;
    189 	bcopy((caddr_t)oip, (caddr_t)&icp->icmp_ip, icmplen);
    190 	nip = &icp->icmp_ip;
    191 
    192 	/*
    193 	 * Now, copy old ip header (without options)
    194 	 * in front of icmp message.
    195 	 */
    196 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
    197 		panic("icmp len");
    198 	m->m_data -= sizeof(struct ip);
    199 	m->m_len += sizeof(struct ip);
    200 	m->m_pkthdr.len = m->m_len;
    201 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
    202 	nip = mtod(m, struct ip *);
    203 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
    204 	nip->ip_len = m->m_len;
    205 	nip->ip_hl = sizeof(struct ip) >> 2;
    206 	nip->ip_p = IPPROTO_ICMP;
    207 	nip->ip_tos = 0;
    208 	icmp_reflect(m);
    209 
    210 freeit:
    211 	m_freem(n);
    212 }
    213 
    214 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
    215 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
    216 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
    217 struct sockaddr_in icmpmask = { 8, 0 };
    218 
    219 /*
    220  * Process a received ICMP message.
    221  */
    222 void
    223 #if __STDC__
    224 icmp_input(struct mbuf *m, ...)
    225 #else
    226 icmp_input(m, va_alist)
    227 	struct mbuf *m;
    228 	va_dcl
    229 #endif
    230 {
    231 	register struct icmp *icp;
    232 	register struct ip *ip = mtod(m, struct ip *);
    233 	int icmplen;
    234 	register int i;
    235 	struct in_ifaddr *ia;
    236 	void *(*ctlfunc) __P((int, struct sockaddr *, void *));
    237 	int code;
    238 	extern u_char ip_protox[];
    239 	int hlen;
    240 	va_list ap;
    241 
    242 	va_start(ap, m);
    243 	hlen = va_arg(ap, int);
    244 	va_end(ap);
    245 
    246 	/*
    247 	 * Locate icmp structure in mbuf, and check
    248 	 * that not corrupted and of at least minimum length.
    249 	 */
    250 	icmplen = ip->ip_len - hlen;
    251 #ifdef ICMPPRINTFS
    252 	if (icmpprintfs)
    253 		printf("icmp_input from %x to %x, len %d\n",
    254 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr),
    255 		    icmplen);
    256 #endif
    257 	if (icmplen < ICMP_MINLEN) {
    258 		icmpstat.icps_tooshort++;
    259 		goto freeit;
    260 	}
    261 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
    262 	if (m->m_len < i && (m = m_pullup(m, i)) == 0) {
    263 		icmpstat.icps_tooshort++;
    264 		return;
    265 	}
    266 	ip = mtod(m, struct ip *);
    267 	m->m_len -= hlen;
    268 	m->m_data += hlen;
    269 	icp = mtod(m, struct icmp *);
    270 	if (in_cksum(m, icmplen)) {
    271 		icmpstat.icps_checksum++;
    272 		goto freeit;
    273 	}
    274 	m->m_len += hlen;
    275 	m->m_data -= hlen;
    276 
    277 #ifdef ICMPPRINTFS
    278 	/*
    279 	 * Message type specific processing.
    280 	 */
    281 	if (icmpprintfs)
    282 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
    283 		    icp->icmp_code);
    284 #endif
    285 	if (icp->icmp_type > ICMP_MAXTYPE)
    286 		goto raw;
    287 	icmpstat.icps_inhist[icp->icmp_type]++;
    288 	code = icp->icmp_code;
    289 	switch (icp->icmp_type) {
    290 
    291 	case ICMP_UNREACH:
    292 		switch (code) {
    293 			case ICMP_UNREACH_NET:
    294 			case ICMP_UNREACH_HOST:
    295 			case ICMP_UNREACH_PROTOCOL:
    296 			case ICMP_UNREACH_PORT:
    297 			case ICMP_UNREACH_SRCFAIL:
    298 				code += PRC_UNREACH_NET;
    299 				break;
    300 
    301 			case ICMP_UNREACH_NEEDFRAG:
    302 				code = PRC_MSGSIZE;
    303 				break;
    304 
    305 			case ICMP_UNREACH_NET_UNKNOWN:
    306 			case ICMP_UNREACH_NET_PROHIB:
    307 			case ICMP_UNREACH_TOSNET:
    308 				code = PRC_UNREACH_NET;
    309 				break;
    310 
    311 			case ICMP_UNREACH_HOST_UNKNOWN:
    312 			case ICMP_UNREACH_ISOLATED:
    313 			case ICMP_UNREACH_HOST_PROHIB:
    314 			case ICMP_UNREACH_TOSHOST:
    315 				code = PRC_UNREACH_HOST;
    316 				break;
    317 
    318 			default:
    319 				goto badcode;
    320 		}
    321 		goto deliver;
    322 
    323 	case ICMP_TIMXCEED:
    324 		if (code > 1)
    325 			goto badcode;
    326 		code += PRC_TIMXCEED_INTRANS;
    327 		goto deliver;
    328 
    329 	case ICMP_PARAMPROB:
    330 		if (code > 1)
    331 			goto badcode;
    332 		code = PRC_PARAMPROB;
    333 		goto deliver;
    334 
    335 	case ICMP_SOURCEQUENCH:
    336 		if (code)
    337 			goto badcode;
    338 		code = PRC_QUENCH;
    339 		goto deliver;
    340 
    341 	deliver:
    342 		/*
    343 		 * Problem with datagram; advise higher level routines.
    344 		 */
    345 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
    346 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
    347 			icmpstat.icps_badlen++;
    348 			goto freeit;
    349 		}
    350 		if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
    351 			goto badcode;
    352 		NTOHS(icp->icmp_ip.ip_len);
    353 #ifdef ICMPPRINTFS
    354 		if (icmpprintfs)
    355 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
    356 #endif
    357 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
    358 		if (code == PRC_MSGSIZE && ip_mtudisc)
    359 			icmp_mtudisc(icp);
    360 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
    361 		if (ctlfunc)
    362 			(*ctlfunc)(code, sintosa(&icmpsrc), &icp->icmp_ip);
    363 		break;
    364 
    365 	badcode:
    366 		icmpstat.icps_badcode++;
    367 		break;
    368 
    369 	case ICMP_ECHO:
    370 		icp->icmp_type = ICMP_ECHOREPLY;
    371 		goto reflect;
    372 
    373 	case ICMP_TSTAMP:
    374 		if (icmplen < ICMP_TSLEN) {
    375 			icmpstat.icps_badlen++;
    376 			break;
    377 		}
    378 		icp->icmp_type = ICMP_TSTAMPREPLY;
    379 		icp->icmp_rtime = iptime();
    380 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
    381 		goto reflect;
    382 
    383 	case ICMP_MASKREQ:
    384 		if (icmpmaskrepl == 0)
    385 			break;
    386 		/*
    387 		 * We are not able to respond with all ones broadcast
    388 		 * unless we receive it over a point-to-point interface.
    389 		 */
    390 		if (icmplen < ICMP_MASKLEN) {
    391 			icmpstat.icps_badlen++;
    392 			break;
    393 		}
    394 		if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
    395 		    in_nullhost(ip->ip_dst))
    396 			icmpdst.sin_addr = ip->ip_src;
    397 		else
    398 			icmpdst.sin_addr = ip->ip_dst;
    399 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
    400 		    m->m_pkthdr.rcvif));
    401 		if (ia == 0)
    402 			break;
    403 		icp->icmp_type = ICMP_MASKREPLY;
    404 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
    405 		if (in_nullhost(ip->ip_src)) {
    406 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
    407 				ip->ip_src = ia->ia_broadaddr.sin_addr;
    408 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
    409 				ip->ip_src = ia->ia_dstaddr.sin_addr;
    410 		}
    411 reflect:
    412 		icmpstat.icps_reflect++;
    413 		icmpstat.icps_outhist[icp->icmp_type]++;
    414 		icmp_reflect(m);
    415 		return;
    416 
    417 	case ICMP_REDIRECT:
    418 		if (code > 3)
    419 			goto badcode;
    420 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
    421 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
    422 			icmpstat.icps_badlen++;
    423 			break;
    424 		}
    425 		/*
    426 		 * Short circuit routing redirects to force
    427 		 * immediate change in the kernel's routing
    428 		 * tables.  The message is also handed to anyone
    429 		 * listening on a raw socket (e.g. the routing
    430 		 * daemon for use in updating its tables).
    431 		 */
    432 		icmpgw.sin_addr = ip->ip_src;
    433 		icmpdst.sin_addr = icp->icmp_gwaddr;
    434 #ifdef	ICMPPRINTFS
    435 		if (icmpprintfs)
    436 			printf("redirect dst %x to %x\n", icp->icmp_ip.ip_dst,
    437 			    icp->icmp_gwaddr);
    438 #endif
    439 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
    440 		rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst),
    441 		    (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
    442 		    sintosa(&icmpgw), (struct rtentry **)0);
    443 		pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc));
    444 		break;
    445 
    446 	/*
    447 	 * No kernel processing for the following;
    448 	 * just fall through to send to raw listener.
    449 	 */
    450 	case ICMP_ECHOREPLY:
    451 	case ICMP_ROUTERADVERT:
    452 	case ICMP_ROUTERSOLICIT:
    453 	case ICMP_TSTAMPREPLY:
    454 	case ICMP_IREQREPLY:
    455 	case ICMP_MASKREPLY:
    456 	default:
    457 		break;
    458 	}
    459 
    460 raw:
    461 	rip_input(m);
    462 	return;
    463 
    464 freeit:
    465 	m_freem(m);
    466 }
    467 
    468 /*
    469  * Reflect the ip packet back to the source
    470  */
    471 void
    472 icmp_reflect(m)
    473 	struct mbuf *m;
    474 {
    475 	register struct ip *ip = mtod(m, struct ip *);
    476 	register struct in_ifaddr *ia;
    477 	register struct ifaddr *ifa;
    478 	struct in_addr t;
    479 	struct mbuf *opts = 0;
    480 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
    481 
    482 	if (!in_canforward(ip->ip_src) &&
    483 	    ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
    484 	     htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
    485 		m_freem(m);	/* Bad return address */
    486 		goto done;	/* ip_output() will check for broadcast */
    487 	}
    488 	t = ip->ip_dst;
    489 	ip->ip_dst = ip->ip_src;
    490 	/*
    491 	 * If the incoming packet was addressed directly to us,
    492 	 * use dst as the src for the reply.  Otherwise (broadcast
    493 	 * or anonymous), use the address which corresponds
    494 	 * to the incoming interface.
    495 	 */
    496 	INADDR_TO_IA(t, ia);
    497 	if (ia == NULL && (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST)) {
    498 		for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
    499 		    ifa != NULL; ifa = ifa->ifa_list.tqe_next) {
    500 			if (ifa->ifa_addr->sa_family != AF_INET)
    501 				continue;
    502 			ia = ifatoia(ifa);
    503 			if (in_hosteq(t, ia->ia_broadaddr.sin_addr))
    504 				break;
    505 		}
    506 	}
    507 
    508 	icmpdst.sin_addr = t;
    509 	if (ia == (struct in_ifaddr *)0)
    510 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
    511 		    m->m_pkthdr.rcvif));
    512 	/*
    513 	 * The following happens if the packet was not addressed to us,
    514 	 * and was received on an interface with no IP address:
    515 	 * We find the first AF_INET address on the first non-loopback
    516 	 * interface.
    517 	 */
    518 	if (ia == (struct in_ifaddr *)0)
    519 		for (ia = in_ifaddr.tqh_first; ia != NULL;
    520 		    ia = ia->ia_list.tqe_next) {
    521 			if (ia->ia_ifp->if_flags & IFF_LOOPBACK)
    522 				continue;
    523 			break;
    524 		}
    525 	/*
    526 	 * If we still didn't find an address, punt.  We could have an
    527 	 * interface up (and receiving packets) with no address.
    528 	 */
    529 	if (ia == (struct in_ifaddr *)0) {
    530 		m_freem(m);
    531 		goto done;
    532 	}
    533 
    534 	t = ia->ia_addr.sin_addr;
    535 	ip->ip_src = t;
    536 	ip->ip_ttl = MAXTTL;
    537 
    538 	if (optlen > 0) {
    539 		register u_char *cp;
    540 		int opt, cnt;
    541 		u_int len;
    542 
    543 		/*
    544 		 * Retrieve any source routing from the incoming packet;
    545 		 * add on any record-route or timestamp options.
    546 		 */
    547 		cp = (u_char *) (ip + 1);
    548 		if ((opts = ip_srcroute()) == 0 &&
    549 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
    550 			opts->m_len = sizeof(struct in_addr);
    551 			*mtod(opts, struct in_addr *) = zeroin_addr;
    552 		}
    553 		if (opts) {
    554 #ifdef ICMPPRINTFS
    555 		    if (icmpprintfs)
    556 			    printf("icmp_reflect optlen %d rt %d => ",
    557 				optlen, opts->m_len);
    558 #endif
    559 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
    560 			    opt = cp[IPOPT_OPTVAL];
    561 			    if (opt == IPOPT_EOL)
    562 				    break;
    563 			    if (opt == IPOPT_NOP)
    564 				    len = 1;
    565 			    else {
    566 				    len = cp[IPOPT_OLEN];
    567 				    if (len <= 0 || len > cnt)
    568 					    break;
    569 			    }
    570 			    /*
    571 			     * Should check for overflow, but it "can't happen"
    572 			     */
    573 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
    574 				opt == IPOPT_SECURITY) {
    575 				    bcopy((caddr_t)cp,
    576 					mtod(opts, caddr_t) + opts->m_len, len);
    577 				    opts->m_len += len;
    578 			    }
    579 		    }
    580 		    /* Terminate & pad, if necessary */
    581 		    if ((cnt = opts->m_len % 4) != 0) {
    582 			    for (; cnt < 4; cnt++) {
    583 				    *(mtod(opts, caddr_t) + opts->m_len) =
    584 					IPOPT_EOL;
    585 				    opts->m_len++;
    586 			    }
    587 		    }
    588 #ifdef ICMPPRINTFS
    589 		    if (icmpprintfs)
    590 			    printf("%d\n", opts->m_len);
    591 #endif
    592 		}
    593 		/*
    594 		 * Now strip out original options by copying rest of first
    595 		 * mbuf's data back, and adjust the IP length.
    596 		 */
    597 		ip->ip_len -= optlen;
    598 		ip->ip_hl = sizeof(struct ip) >> 2;
    599 		m->m_len -= optlen;
    600 		if (m->m_flags & M_PKTHDR)
    601 			m->m_pkthdr.len -= optlen;
    602 		optlen += sizeof(struct ip);
    603 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
    604 			 (unsigned)(m->m_len - sizeof(struct ip)));
    605 	}
    606 	m->m_flags &= ~(M_BCAST|M_MCAST);
    607 	icmp_send(m, opts);
    608 done:
    609 	if (opts)
    610 		(void)m_free(opts);
    611 }
    612 
    613 /*
    614  * Send an icmp packet back to the ip level,
    615  * after supplying a checksum.
    616  */
    617 void
    618 icmp_send(m, opts)
    619 	register struct mbuf *m;
    620 	struct mbuf *opts;
    621 {
    622 	register struct ip *ip = mtod(m, struct ip *);
    623 	register int hlen;
    624 	register struct icmp *icp;
    625 
    626 	hlen = ip->ip_hl << 2;
    627 	m->m_data += hlen;
    628 	m->m_len -= hlen;
    629 	icp = mtod(m, struct icmp *);
    630 	icp->icmp_cksum = 0;
    631 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
    632 	m->m_data -= hlen;
    633 	m->m_len += hlen;
    634 #ifdef ICMPPRINTFS
    635 	if (icmpprintfs)
    636 		printf("icmp_send dst %x src %x\n", ip->ip_dst, ip->ip_src);
    637 #endif
    638 	(void) ip_output(m, opts, NULL, 0, NULL);
    639 }
    640 
    641 n_time
    642 iptime()
    643 {
    644 	struct timeval atv;
    645 	u_long t;
    646 
    647 	microtime(&atv);
    648 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
    649 	return (htonl(t));
    650 }
    651 
    652 int
    653 icmp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
    654 	int *name;
    655 	u_int namelen;
    656 	void *oldp;
    657 	size_t *oldlenp;
    658 	void *newp;
    659 	size_t newlen;
    660 {
    661 
    662 	/* All sysctl names at this level are terminal. */
    663 	if (namelen != 1)
    664 		return (ENOTDIR);
    665 
    666 	switch (name[0]) {
    667 	case ICMPCTL_MASKREPL:
    668 		return (sysctl_int(oldp, oldlenp, newp, newlen, &icmpmaskrepl));
    669 	default:
    670 		return (ENOPROTOOPT);
    671 	}
    672 	/* NOTREACHED */
    673 }
    674 
    675 static void
    676 icmp_mtudisc(icp)
    677 	struct icmp *icp;
    678 {
    679 	struct rtentry *rt;
    680 	struct sockaddr *dst = sintosa(&icmpsrc);
    681 	u_long mtu = ntohs(icp->icmp_nextmtu);  /* Why a long?  IPv6 */
    682 	int    error;
    683 
    684 	/* Table of common MTUs: */
    685 
    686 	static u_long mtu_table[] = {65535, 65280, 32000, 17914, 9180, 8166,
    687 				     4352, 2002, 1492, 1006, 508, 296, 68, 0};
    688 
    689 	rt = rtalloc1(dst, 1);
    690 	if (rt == 0)
    691 		return;
    692 
    693 	/* If we didn't get a host route, allocate one */
    694 
    695 	if ((rt->rt_flags & RTF_HOST) == 0) {
    696 		struct rtentry *nrt;
    697 
    698 		error = rtrequest((int) RTM_ADD, dst,
    699 		    (struct sockaddr *) rt->rt_gateway,
    700 		    (struct sockaddr *) 0,
    701 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
    702 		if (error) {
    703 			rtfree(rt);
    704 			rtfree(nrt);
    705 			return;
    706 		}
    707 		nrt->rt_rmx = rt->rt_rmx;
    708 		rtfree(rt);
    709 		rt = nrt;
    710 	}
    711 	error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q);
    712 	if (error) {
    713 		rtfree(rt);
    714 		return;
    715 	}
    716 
    717 	if (mtu == 0) {
    718 		int i = 0;
    719 
    720 		mtu = icp->icmp_ip.ip_len; /* NTOHS happened in deliver: */
    721 		/* Some 4.2BSD-based routers incorrectly adjust the ip_len */
    722 		if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0)
    723 			mtu -= (icp->icmp_ip.ip_hl << 2);
    724 
    725 		/* If we still can't guess a value, try the route */
    726 
    727 		if (mtu == 0) {
    728 			mtu = rt->rt_rmx.rmx_mtu;
    729 
    730 			/* If no route mtu, default to the interface mtu */
    731 
    732 			if (mtu == 0)
    733 				mtu = rt->rt_ifp->if_mtu;
    734 		}
    735 
    736 		for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++)
    737 			if (mtu > mtu_table[i]) {
    738 				mtu = mtu_table[i];
    739 				break;
    740 			}
    741 	}
    742 
    743 	/*
    744 	 * XXX:   RTV_MTU is overloaded, since the admin can set it
    745 	 *	  to turn off PMTU for a route, and the kernel can
    746 	 *	  set it to indicate a serious problem with PMTU
    747 	 *	  on a route.  We should be using a separate flag
    748 	 *	  for the kernel to indicate this.
    749 	 */
    750 
    751 	if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
    752 		if (mtu < 296 || mtu > rt->rt_ifp->if_mtu)
    753 			rt->rt_rmx.rmx_locks |= RTV_MTU;
    754 		else if (rt->rt_rmx.rmx_mtu > mtu ||
    755 			 rt->rt_rmx.rmx_mtu == 0)
    756 			rt->rt_rmx.rmx_mtu = mtu;
    757 	}
    758 
    759 	if (rt)
    760 		rtfree(rt);
    761 }
    762 
    763 
    764 static void
    765 icmp_mtudisc_timeout(rt, r)
    766 	struct rtentry *rt;
    767 	struct rttimer *r;
    768 {
    769 	if (rt == NULL)
    770 		panic("icmp_mtudisc_timeout:  bad route to timeout");
    771 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
    772 	    (RTF_DYNAMIC | RTF_HOST)) {
    773 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
    774 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
    775 	} else {
    776 		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
    777 			rt->rt_rmx.rmx_mtu = 0;
    778 		}
    779 	}
    780 }
    781