Home | History | Annotate | Line # | Download | only in netinet
ip_icmp.c revision 1.29
      1 /*	$NetBSD: ip_icmp.c,v 1.29 1998/04/29 03:44:11 kml 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);
    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 	icp->icmp_code = code;
    187 	bcopy((caddr_t)oip, (caddr_t)&icp->icmp_ip, icmplen);
    188 	nip = &icp->icmp_ip;
    189 	nip->ip_len = htons((u_int16_t)(nip->ip_len + oiplen));
    190 
    191 	/*
    192 	 * Now, copy old ip header (without options)
    193 	 * in front of icmp message.
    194 	 */
    195 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
    196 		panic("icmp len");
    197 	m->m_data -= sizeof(struct ip);
    198 	m->m_len += sizeof(struct ip);
    199 	m->m_pkthdr.len = m->m_len;
    200 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
    201 	nip = mtod(m, struct ip *);
    202 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
    203 	nip->ip_len = m->m_len;
    204 	nip->ip_hl = sizeof(struct ip) >> 2;
    205 	nip->ip_p = IPPROTO_ICMP;
    206 	nip->ip_tos = 0;
    207 	icmp_reflect(m);
    208 
    209 freeit:
    210 	m_freem(n);
    211 }
    212 
    213 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
    214 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
    215 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
    216 struct sockaddr_in icmpmask = { 8, 0 };
    217 
    218 /*
    219  * Process a received ICMP message.
    220  */
    221 void
    222 #if __STDC__
    223 icmp_input(struct mbuf *m, ...)
    224 #else
    225 icmp_input(m, va_alist)
    226 	struct mbuf *m;
    227 	va_dcl
    228 #endif
    229 {
    230 	register struct icmp *icp;
    231 	register struct ip *ip = mtod(m, struct ip *);
    232 	int icmplen = ip->ip_len;
    233 	register int i;
    234 	struct in_ifaddr *ia;
    235 	void *(*ctlfunc) __P((int, struct sockaddr *, void *));
    236 	int code;
    237 	extern u_char ip_protox[];
    238 	int hlen;
    239 	va_list ap;
    240 
    241 	va_start(ap, m);
    242 	hlen = va_arg(ap, int);
    243 	va_end(ap);
    244 
    245 	/*
    246 	 * Locate icmp structure in mbuf, and check
    247 	 * that not corrupted and of at least minimum length.
    248 	 */
    249 #ifdef ICMPPRINTFS
    250 	if (icmpprintfs)
    251 		printf("icmp_input from %x to %x, len %d\n",
    252 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr),
    253 		    icmplen);
    254 #endif
    255 	if (icmplen < ICMP_MINLEN) {
    256 		icmpstat.icps_tooshort++;
    257 		goto freeit;
    258 	}
    259 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
    260 	if (m->m_len < i && (m = m_pullup(m, i)) == 0) {
    261 		icmpstat.icps_tooshort++;
    262 		return;
    263 	}
    264 	ip = mtod(m, struct ip *);
    265 	m->m_len -= hlen;
    266 	m->m_data += hlen;
    267 	icp = mtod(m, struct icmp *);
    268 	if (in_cksum(m, icmplen)) {
    269 		icmpstat.icps_checksum++;
    270 		goto freeit;
    271 	}
    272 	m->m_len += hlen;
    273 	m->m_data -= hlen;
    274 
    275 #ifdef ICMPPRINTFS
    276 	/*
    277 	 * Message type specific processing.
    278 	 */
    279 	if (icmpprintfs)
    280 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
    281 		    icp->icmp_code);
    282 #endif
    283 	if (icp->icmp_type > ICMP_MAXTYPE)
    284 		goto raw;
    285 	icmpstat.icps_inhist[icp->icmp_type]++;
    286 	code = icp->icmp_code;
    287 	switch (icp->icmp_type) {
    288 
    289 	case ICMP_UNREACH:
    290 		switch (code) {
    291 			case ICMP_UNREACH_NET:
    292 			case ICMP_UNREACH_HOST:
    293 			case ICMP_UNREACH_PROTOCOL:
    294 			case ICMP_UNREACH_PORT:
    295 			case ICMP_UNREACH_SRCFAIL:
    296 				code += PRC_UNREACH_NET;
    297 				break;
    298 
    299 			case ICMP_UNREACH_NEEDFRAG:
    300 				code = PRC_MSGSIZE;
    301 				break;
    302 
    303 			case ICMP_UNREACH_NET_UNKNOWN:
    304 			case ICMP_UNREACH_NET_PROHIB:
    305 			case ICMP_UNREACH_TOSNET:
    306 				code = PRC_UNREACH_NET;
    307 				break;
    308 
    309 			case ICMP_UNREACH_HOST_UNKNOWN:
    310 			case ICMP_UNREACH_ISOLATED:
    311 			case ICMP_UNREACH_HOST_PROHIB:
    312 			case ICMP_UNREACH_TOSHOST:
    313 				code = PRC_UNREACH_HOST;
    314 				break;
    315 
    316 			default:
    317 				goto badcode;
    318 		}
    319 		goto deliver;
    320 
    321 	case ICMP_TIMXCEED:
    322 		if (code > 1)
    323 			goto badcode;
    324 		code += PRC_TIMXCEED_INTRANS;
    325 		goto deliver;
    326 
    327 	case ICMP_PARAMPROB:
    328 		if (code > 1)
    329 			goto badcode;
    330 		code = PRC_PARAMPROB;
    331 		goto deliver;
    332 
    333 	case ICMP_SOURCEQUENCH:
    334 		if (code)
    335 			goto badcode;
    336 		code = PRC_QUENCH;
    337 		goto deliver;
    338 
    339 	deliver:
    340 		/*
    341 		 * Problem with datagram; advise higher level routines.
    342 		 */
    343 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
    344 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
    345 			icmpstat.icps_badlen++;
    346 			goto freeit;
    347 		}
    348 		if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
    349 			goto badcode;
    350 		NTOHS(icp->icmp_ip.ip_len);
    351 #ifdef ICMPPRINTFS
    352 		if (icmpprintfs)
    353 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
    354 #endif
    355 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
    356 		if (code == PRC_MSGSIZE && ip_mtudisc)
    357 			icmp_mtudisc(icp);
    358 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
    359 		if (ctlfunc)
    360 			(*ctlfunc)(code, sintosa(&icmpsrc), &icp->icmp_ip);
    361 		break;
    362 
    363 	badcode:
    364 		icmpstat.icps_badcode++;
    365 		break;
    366 
    367 	case ICMP_ECHO:
    368 		icp->icmp_type = ICMP_ECHOREPLY;
    369 		goto reflect;
    370 
    371 	case ICMP_TSTAMP:
    372 		if (icmplen < ICMP_TSLEN) {
    373 			icmpstat.icps_badlen++;
    374 			break;
    375 		}
    376 		icp->icmp_type = ICMP_TSTAMPREPLY;
    377 		icp->icmp_rtime = iptime();
    378 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
    379 		goto reflect;
    380 
    381 	case ICMP_MASKREQ:
    382 		if (icmpmaskrepl == 0)
    383 			break;
    384 		/*
    385 		 * We are not able to respond with all ones broadcast
    386 		 * unless we receive it over a point-to-point interface.
    387 		 */
    388 		if (icmplen < ICMP_MASKLEN) {
    389 			icmpstat.icps_badlen++;
    390 			break;
    391 		}
    392 		if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
    393 		    in_nullhost(ip->ip_dst))
    394 			icmpdst.sin_addr = ip->ip_src;
    395 		else
    396 			icmpdst.sin_addr = ip->ip_dst;
    397 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
    398 		    m->m_pkthdr.rcvif));
    399 		if (ia == 0)
    400 			break;
    401 		icp->icmp_type = ICMP_MASKREPLY;
    402 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
    403 		if (in_nullhost(ip->ip_src)) {
    404 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
    405 				ip->ip_src = ia->ia_broadaddr.sin_addr;
    406 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
    407 				ip->ip_src = ia->ia_dstaddr.sin_addr;
    408 		}
    409 reflect:
    410 		ip->ip_len += hlen;	/* since ip_input deducts this */
    411 		icmpstat.icps_reflect++;
    412 		icmpstat.icps_outhist[icp->icmp_type]++;
    413 		icmp_reflect(m);
    414 		return;
    415 
    416 	case ICMP_REDIRECT:
    417 		if (code > 3)
    418 			goto badcode;
    419 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
    420 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
    421 			icmpstat.icps_badlen++;
    422 			break;
    423 		}
    424 		/*
    425 		 * Short circuit routing redirects to force
    426 		 * immediate change in the kernel's routing
    427 		 * tables.  The message is also handed to anyone
    428 		 * listening on a raw socket (e.g. the routing
    429 		 * daemon for use in updating its tables).
    430 		 */
    431 		icmpgw.sin_addr = ip->ip_src;
    432 		icmpdst.sin_addr = icp->icmp_gwaddr;
    433 #ifdef	ICMPPRINTFS
    434 		if (icmpprintfs)
    435 			printf("redirect dst %x to %x\n", icp->icmp_ip.ip_dst,
    436 			    icp->icmp_gwaddr);
    437 #endif
    438 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
    439 		rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst),
    440 		    (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
    441 		    sintosa(&icmpgw), (struct rtentry **)0);
    442 		pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc));
    443 		break;
    444 
    445 	/*
    446 	 * No kernel processing for the following;
    447 	 * just fall through to send to raw listener.
    448 	 */
    449 	case ICMP_ECHOREPLY:
    450 	case ICMP_ROUTERADVERT:
    451 	case ICMP_ROUTERSOLICIT:
    452 	case ICMP_TSTAMPREPLY:
    453 	case ICMP_IREQREPLY:
    454 	case ICMP_MASKREPLY:
    455 	default:
    456 		break;
    457 	}
    458 
    459 raw:
    460 	rip_input(m);
    461 	return;
    462 
    463 freeit:
    464 	m_freem(m);
    465 }
    466 
    467 /*
    468  * Reflect the ip packet back to the source
    469  */
    470 void
    471 icmp_reflect(m)
    472 	struct mbuf *m;
    473 {
    474 	register struct ip *ip = mtod(m, struct ip *);
    475 	register struct in_ifaddr *ia;
    476 	register struct ifaddr *ifa;
    477 	struct in_addr t;
    478 	struct mbuf *opts = 0;
    479 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
    480 
    481 	if (!in_canforward(ip->ip_src) &&
    482 	    ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
    483 	     htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
    484 		m_freem(m);	/* Bad return address */
    485 		goto done;	/* ip_output() will check for broadcast */
    486 	}
    487 	t = ip->ip_dst;
    488 	ip->ip_dst = ip->ip_src;
    489 	/*
    490 	 * If the incoming packet was addressed directly to us,
    491 	 * use dst as the src for the reply.  Otherwise (broadcast
    492 	 * or anonymous), use the address which corresponds
    493 	 * to the incoming interface.
    494 	 */
    495 	INADDR_TO_IA(t, ia);
    496 	if (ia == NULL && (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST)) {
    497 		for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
    498 		    ifa != NULL; ifa = ifa->ifa_list.tqe_next) {
    499 			if (ifa->ifa_addr->sa_family != AF_INET)
    500 				continue;
    501 			ia = ifatoia(ifa);
    502 			if (in_hosteq(t, ia->ia_broadaddr.sin_addr))
    503 				break;
    504 		}
    505 	}
    506 
    507 	icmpdst.sin_addr = t;
    508 	if (ia == (struct in_ifaddr *)0)
    509 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
    510 		    m->m_pkthdr.rcvif));
    511 	/*
    512 	 * The following happens if the packet was not addressed to us,
    513 	 * and was received on an interface with no IP address:
    514 	 * We find the first AF_INET address on the first non-loopback
    515 	 * interface.
    516 	 */
    517 	if (ia == (struct in_ifaddr *)0)
    518 		for (ia = in_ifaddr.tqh_first; ia != NULL;
    519 		    ia = ia->ia_list.tqe_next) {
    520 			if (ia->ia_ifp->if_flags & IFF_LOOPBACK)
    521 				continue;
    522 			break;
    523 		}
    524 	t = ia->ia_addr.sin_addr;
    525 	ip->ip_src = t;
    526 	ip->ip_ttl = MAXTTL;
    527 
    528 	if (optlen > 0) {
    529 		register u_char *cp;
    530 		int opt, cnt;
    531 		u_int len;
    532 
    533 		/*
    534 		 * Retrieve any source routing from the incoming packet;
    535 		 * add on any record-route or timestamp options.
    536 		 */
    537 		cp = (u_char *) (ip + 1);
    538 		if ((opts = ip_srcroute()) == 0 &&
    539 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
    540 			opts->m_len = sizeof(struct in_addr);
    541 			*mtod(opts, struct in_addr *) = zeroin_addr;
    542 		}
    543 		if (opts) {
    544 #ifdef ICMPPRINTFS
    545 		    if (icmpprintfs)
    546 			    printf("icmp_reflect optlen %d rt %d => ",
    547 				optlen, opts->m_len);
    548 #endif
    549 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
    550 			    opt = cp[IPOPT_OPTVAL];
    551 			    if (opt == IPOPT_EOL)
    552 				    break;
    553 			    if (opt == IPOPT_NOP)
    554 				    len = 1;
    555 			    else {
    556 				    len = cp[IPOPT_OLEN];
    557 				    if (len <= 0 || len > cnt)
    558 					    break;
    559 			    }
    560 			    /*
    561 			     * Should check for overflow, but it "can't happen"
    562 			     */
    563 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
    564 				opt == IPOPT_SECURITY) {
    565 				    bcopy((caddr_t)cp,
    566 					mtod(opts, caddr_t) + opts->m_len, len);
    567 				    opts->m_len += len;
    568 			    }
    569 		    }
    570 		    /* Terminate & pad, if necessary */
    571 		    if ((cnt = opts->m_len % 4) != 0) {
    572 			    for (; cnt < 4; cnt++) {
    573 				    *(mtod(opts, caddr_t) + opts->m_len) =
    574 					IPOPT_EOL;
    575 				    opts->m_len++;
    576 			    }
    577 		    }
    578 #ifdef ICMPPRINTFS
    579 		    if (icmpprintfs)
    580 			    printf("%d\n", opts->m_len);
    581 #endif
    582 		}
    583 		/*
    584 		 * Now strip out original options by copying rest of first
    585 		 * mbuf's data back, and adjust the IP length.
    586 		 */
    587 		ip->ip_len -= optlen;
    588 		ip->ip_hl = sizeof(struct ip) >> 2;
    589 		m->m_len -= optlen;
    590 		if (m->m_flags & M_PKTHDR)
    591 			m->m_pkthdr.len -= optlen;
    592 		optlen += sizeof(struct ip);
    593 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
    594 			 (unsigned)(m->m_len - sizeof(struct ip)));
    595 	}
    596 	m->m_flags &= ~(M_BCAST|M_MCAST);
    597 	icmp_send(m, opts);
    598 done:
    599 	if (opts)
    600 		(void)m_free(opts);
    601 }
    602 
    603 /*
    604  * Send an icmp packet back to the ip level,
    605  * after supplying a checksum.
    606  */
    607 void
    608 icmp_send(m, opts)
    609 	register struct mbuf *m;
    610 	struct mbuf *opts;
    611 {
    612 	register struct ip *ip = mtod(m, struct ip *);
    613 	register int hlen;
    614 	register struct icmp *icp;
    615 
    616 	hlen = ip->ip_hl << 2;
    617 	m->m_data += hlen;
    618 	m->m_len -= hlen;
    619 	icp = mtod(m, struct icmp *);
    620 	icp->icmp_cksum = 0;
    621 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
    622 	m->m_data -= hlen;
    623 	m->m_len += hlen;
    624 #ifdef ICMPPRINTFS
    625 	if (icmpprintfs)
    626 		printf("icmp_send dst %x src %x\n", ip->ip_dst, ip->ip_src);
    627 #endif
    628 	(void) ip_output(m, opts, NULL, 0, NULL);
    629 }
    630 
    631 n_time
    632 iptime()
    633 {
    634 	struct timeval atv;
    635 	u_long t;
    636 
    637 	microtime(&atv);
    638 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
    639 	return (htonl(t));
    640 }
    641 
    642 int
    643 icmp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
    644 	int *name;
    645 	u_int namelen;
    646 	void *oldp;
    647 	size_t *oldlenp;
    648 	void *newp;
    649 	size_t newlen;
    650 {
    651 
    652 	/* All sysctl names at this level are terminal. */
    653 	if (namelen != 1)
    654 		return (ENOTDIR);
    655 
    656 	switch (name[0]) {
    657 	case ICMPCTL_MASKREPL:
    658 		return (sysctl_int(oldp, oldlenp, newp, newlen, &icmpmaskrepl));
    659 	default:
    660 		return (ENOPROTOOPT);
    661 	}
    662 	/* NOTREACHED */
    663 }
    664 
    665 static void
    666 icmp_mtudisc(icp)
    667 	struct icmp *icp;
    668 {
    669 	struct rtentry *rt;
    670 	struct sockaddr *dst = sintosa(&icmpsrc);
    671 	u_long mtu = ntohs(icp->icmp_nextmtu);  /* Why a long?  IPv6 */
    672 	int    error;
    673 
    674 	/* Table of common MTUs: */
    675 
    676 	static u_long mtu_table[] = {65535, 65280, 32000, 17914, 9180, 8166,
    677 				     4352, 2002, 1492, 1006, 508, 296, 68, 0};
    678 
    679 	rt = rtalloc1(dst, 1);
    680 	if (rt == 0)
    681 		return;
    682 
    683 	/* If we didn't get a host route, allocate one */
    684 
    685 	if ((rt->rt_flags & RTF_HOST) == 0) {
    686 		struct rtentry *nrt;
    687 
    688 		error = rtrequest((int) RTM_ADD, dst,
    689 		    (struct sockaddr *) rt->rt_gateway,
    690 		    (struct sockaddr *) 0,
    691 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
    692 		if (error) {
    693 			rtfree(rt);
    694 			rtfree(nrt);
    695 			return;
    696 		}
    697 		nrt->rt_rmx = rt->rt_rmx;
    698 		rtfree(rt);
    699 		rt = nrt;
    700 	}
    701 	error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q);
    702 	if (error) {
    703 		rtfree(rt);
    704 		return;
    705 	}
    706 
    707 	if (mtu == 0) {
    708 		int i = 0;
    709 
    710 		mtu = icp->icmp_ip.ip_len; /* NTOHS happened in deliver: */
    711 		/* Some 4.2BSD-based routers incorrectly adjust the ip_len */
    712 		if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0)
    713 			mtu -= (icp->icmp_ip.ip_hl << 2);
    714 
    715 		/* If we still can't guess a value, try the route */
    716 
    717 		if (mtu == 0) {
    718 			mtu = rt->rt_rmx.rmx_mtu;
    719 
    720 			/* If no route mtu, default to the interface mtu */
    721 
    722 			if (mtu == 0)
    723 				mtu = rt->rt_ifp->if_mtu;
    724 		}
    725 
    726 		for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++)
    727 			if (mtu > mtu_table[i]) {
    728 				mtu = mtu_table[i];
    729 				break;
    730 			}
    731 	}
    732 
    733 	/*
    734 	 * XXX:   RTV_MTU is overloaded, since the admin can set it
    735 	 *	  to turn off PMTU for a route, and the kernel can
    736 	 *	  set it to indicate a serious problem with PMTU
    737 	 *	  on a route.  We should be using a separate flag
    738 	 *	  for the kernel to indicate this.
    739 	 */
    740 
    741 	if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
    742 		if (mtu < 296 || mtu > rt->rt_ifp->if_mtu)
    743 			rt->rt_rmx.rmx_locks |= RTV_MTU;
    744 		else if (rt->rt_rmx.rmx_mtu > mtu ||
    745 			 rt->rt_rmx.rmx_mtu == 0)
    746 			rt->rt_rmx.rmx_mtu = mtu;
    747 	}
    748 
    749 	if (rt)
    750 		rtfree(rt);
    751 }
    752 
    753 
    754 static void
    755 icmp_mtudisc_timeout(rt, r)
    756 	struct rtentry *rt;
    757 	struct rttimer *r;
    758 {
    759 	if (rt == NULL)
    760 		panic("icmp_mtudisc_timeout:  bad route to timeout");
    761 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
    762 	    (RTF_DYNAMIC | RTF_HOST)) {
    763 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
    764 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
    765 	} else {
    766 		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
    767 			rt->rt_rmx.rmx_mtu = 0;
    768 		}
    769 	}
    770 }
    771