Home | History | Annotate | Line # | Download | only in netinet
ip_input.c revision 1.25
      1 /*	$NetBSD: ip_input.c,v 1.25 1995/11/21 01:07:34 cgd Exp $	*/
      2 
      3 /*
      4  * Copyright (c) 1982, 1986, 1988, 1993
      5  *	The Regents of the University of California.  All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. All advertising materials mentioning features or use of this software
     16  *    must display the following acknowledgement:
     17  *	This product includes software developed by the University of
     18  *	California, Berkeley and its contributors.
     19  * 4. Neither the name of the University nor the names of its contributors
     20  *    may be used to endorse or promote products derived from this software
     21  *    without specific prior written permission.
     22  *
     23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     33  * SUCH DAMAGE.
     34  *
     35  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
     36  */
     37 
     38 #include <sys/param.h>
     39 #include <sys/systm.h>
     40 #include <sys/malloc.h>
     41 #include <sys/mbuf.h>
     42 #include <sys/domain.h>
     43 #include <sys/protosw.h>
     44 #include <sys/socket.h>
     45 #include <sys/errno.h>
     46 #include <sys/time.h>
     47 #include <sys/kernel.h>
     48 
     49 #include <net/if.h>
     50 #include <net/route.h>
     51 
     52 #include <netinet/in.h>
     53 #include <netinet/in_systm.h>
     54 #include <netinet/ip.h>
     55 #include <netinet/in_pcb.h>
     56 #include <netinet/in_var.h>
     57 #include <netinet/ip_var.h>
     58 #include <netinet/ip_icmp.h>
     59 
     60 #ifndef	IPFORWARDING
     61 #ifdef GATEWAY
     62 #define	IPFORWARDING	1	/* forward IP packets not for us */
     63 #else /* GATEWAY */
     64 #define	IPFORWARDING	0	/* don't forward IP packets not for us */
     65 #endif /* GATEWAY */
     66 #endif /* IPFORWARDING */
     67 #ifndef	IPSENDREDIRECTS
     68 #define	IPSENDREDIRECTS	1
     69 #endif
     70 int	ipforwarding = IPFORWARDING;
     71 int	ipsendredirects = IPSENDREDIRECTS;
     72 int	ip_defttl = IPDEFTTL;
     73 #ifdef DIAGNOSTIC
     74 int	ipprintfs = 0;
     75 #endif
     76 
     77 extern	struct domain inetdomain;
     78 extern	struct protosw inetsw[];
     79 u_char	ip_protox[IPPROTO_MAX];
     80 int	ipqmaxlen = IFQ_MAXLEN;
     81 struct	in_ifaddrhead in_ifaddr;
     82 struct	ifqueue ipintrq;
     83 
     84 /*
     85  * We need to save the IP options in case a protocol wants to respond
     86  * to an incoming packet over the same route if the packet got here
     87  * using IP source routing.  This allows connection establishment and
     88  * maintenance when the remote end is on a network that is not known
     89  * to us.
     90  */
     91 int	ip_nhops = 0;
     92 static	struct ip_srcrt {
     93 	struct	in_addr dst;			/* final destination */
     94 	char	nop;				/* one NOP to align */
     95 	char	srcopt[IPOPT_OFFSET + 1];	/* OPTVAL, OLEN and OFFSET */
     96 	struct	in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
     97 } ip_srcrt;
     98 
     99 static void save_rte __P((u_char *, struct in_addr));
    100 /*
    101  * IP initialization: fill in IP protocol switch table.
    102  * All protocols not implemented in kernel go to raw IP protocol handler.
    103  */
    104 void
    105 ip_init()
    106 {
    107 	register struct protosw *pr;
    108 	register int i;
    109 
    110 	pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
    111 	if (pr == 0)
    112 		panic("ip_init");
    113 	for (i = 0; i < IPPROTO_MAX; i++)
    114 		ip_protox[i] = pr - inetsw;
    115 	for (pr = inetdomain.dom_protosw;
    116 	    pr < inetdomain.dom_protoswNPROTOSW; pr++)
    117 		if (pr->pr_domain->dom_family == PF_INET &&
    118 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
    119 			ip_protox[pr->pr_protocol] = pr - inetsw;
    120 	LIST_INIT(&ipq);
    121 	ip_id = time.tv_sec & 0xffff;
    122 	ipintrq.ifq_maxlen = ipqmaxlen;
    123 	TAILQ_INIT(&in_ifaddr);
    124 }
    125 
    126 struct	sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
    127 struct	route ipforward_rt;
    128 
    129 /*
    130  * Ip input routine.  Checksum and byte swap header.  If fragmented
    131  * try to reassemble.  Process options.  Pass to next level.
    132  */
    133 void
    134 ipintr()
    135 {
    136 	register struct ip *ip;
    137 	register struct mbuf *m;
    138 	register struct ipq *fp;
    139 	register struct in_ifaddr *ia;
    140 	struct ipqent *ipqe;
    141 	int hlen, mff, s;
    142 
    143 next:
    144 	/*
    145 	 * Get next datagram off input queue and get IP header
    146 	 * in first mbuf.
    147 	 */
    148 	s = splimp();
    149 	IF_DEQUEUE(&ipintrq, m);
    150 	splx(s);
    151 	if (m == 0)
    152 		return;
    153 #ifdef	DIAGNOSTIC
    154 	if ((m->m_flags & M_PKTHDR) == 0)
    155 		panic("ipintr no HDR");
    156 #endif
    157 	/*
    158 	 * If no IP addresses have been set yet but the interfaces
    159 	 * are receiving, can't do anything with incoming packets yet.
    160 	 */
    161 	if (in_ifaddr.tqh_first == 0)
    162 		goto bad;
    163 	ipstat.ips_total++;
    164 	if (m->m_len < sizeof (struct ip) &&
    165 	    (m = m_pullup(m, sizeof (struct ip))) == 0) {
    166 		ipstat.ips_toosmall++;
    167 		goto next;
    168 	}
    169 	ip = mtod(m, struct ip *);
    170 	if (ip->ip_v != IPVERSION) {
    171 		ipstat.ips_badvers++;
    172 		goto bad;
    173 	}
    174 	hlen = ip->ip_hl << 2;
    175 	if (hlen < sizeof(struct ip)) {	/* minimum header length */
    176 		ipstat.ips_badhlen++;
    177 		goto bad;
    178 	}
    179 	if (hlen > m->m_len) {
    180 		if ((m = m_pullup(m, hlen)) == 0) {
    181 			ipstat.ips_badhlen++;
    182 			goto next;
    183 		}
    184 		ip = mtod(m, struct ip *);
    185 	}
    186 	if (ip->ip_sum = in_cksum(m, hlen)) {
    187 		ipstat.ips_badsum++;
    188 		goto bad;
    189 	}
    190 
    191 	/*
    192 	 * Convert fields to host representation.
    193 	 */
    194 	NTOHS(ip->ip_len);
    195 	if (ip->ip_len < hlen) {
    196 		ipstat.ips_badlen++;
    197 		goto bad;
    198 	}
    199 	NTOHS(ip->ip_id);
    200 	NTOHS(ip->ip_off);
    201 
    202 	/*
    203 	 * Check that the amount of data in the buffers
    204 	 * is as at least much as the IP header would have us expect.
    205 	 * Trim mbufs if longer than we expect.
    206 	 * Drop packet if shorter than we expect.
    207 	 */
    208 	if (m->m_pkthdr.len < ip->ip_len) {
    209 		ipstat.ips_tooshort++;
    210 		goto bad;
    211 	}
    212 	if (m->m_pkthdr.len > ip->ip_len) {
    213 		if (m->m_len == m->m_pkthdr.len) {
    214 			m->m_len = ip->ip_len;
    215 			m->m_pkthdr.len = ip->ip_len;
    216 		} else
    217 			m_adj(m, ip->ip_len - m->m_pkthdr.len);
    218 	}
    219 
    220 	/*
    221 	 * Process options and, if not destined for us,
    222 	 * ship it on.  ip_dooptions returns 1 when an
    223 	 * error was detected (causing an icmp message
    224 	 * to be sent and the original packet to be freed).
    225 	 */
    226 	ip_nhops = 0;		/* for source routed packets */
    227 	if (hlen > sizeof (struct ip) && ip_dooptions(m))
    228 		goto next;
    229 
    230 	/*
    231 	 * Check our list of addresses, to see if the packet is for us.
    232 	 */
    233 	for (ia = in_ifaddr.tqh_first; ia; ia = ia->ia_list.tqe_next) {
    234 		if (ip->ip_dst.s_addr == ia->ia_addr.sin_addr.s_addr)
    235 			goto ours;
    236 		if (
    237 #ifdef	DIRECTED_BROADCAST
    238 		    ia->ia_ifp == m->m_pkthdr.rcvif &&
    239 #endif
    240 		    (ia->ia_ifp->if_flags & IFF_BROADCAST)) {
    241 			if (ip->ip_dst.s_addr == ia->ia_broadaddr.sin_addr.s_addr ||
    242 			    ip->ip_dst.s_addr == ia->ia_netbroadcast.s_addr ||
    243 			    /*
    244 			     * Look for all-0's host part (old broadcast addr),
    245 			     * either for subnet or net.
    246 			     */
    247 			    ip->ip_dst.s_addr == ia->ia_subnet ||
    248 			    ip->ip_dst.s_addr == ia->ia_net)
    249 				goto ours;
    250 		}
    251 	}
    252 	if (IN_MULTICAST(ip->ip_dst.s_addr)) {
    253 		struct in_multi *inm;
    254 #ifdef MROUTING
    255 		extern struct socket *ip_mrouter;
    256 
    257 		if (m->m_flags & M_EXT) {
    258 			if ((m = m_pullup(m, hlen)) == 0) {
    259 				ipstat.ips_toosmall++;
    260 				goto next;
    261 			}
    262 			ip = mtod(m, struct ip *);
    263 		}
    264 
    265 		if (ip_mrouter) {
    266 			/*
    267 			 * If we are acting as a multicast router, all
    268 			 * incoming multicast packets are passed to the
    269 			 * kernel-level multicast forwarding function.
    270 			 * The packet is returned (relatively) intact; if
    271 			 * ip_mforward() returns a non-zero value, the packet
    272 			 * must be discarded, else it may be accepted below.
    273 			 *
    274 			 * (The IP ident field is put in the same byte order
    275 			 * as expected when ip_mforward() is called from
    276 			 * ip_output().)
    277 			 */
    278 			ip->ip_id = htons(ip->ip_id);
    279 			if (ip_mforward(m, m->m_pkthdr.rcvif) != 0) {
    280 				ipstat.ips_cantforward++;
    281 				m_freem(m);
    282 				goto next;
    283 			}
    284 			ip->ip_id = ntohs(ip->ip_id);
    285 
    286 			/*
    287 			 * The process-level routing demon needs to receive
    288 			 * all multicast IGMP packets, whether or not this
    289 			 * host belongs to their destination groups.
    290 			 */
    291 			if (ip->ip_p == IPPROTO_IGMP)
    292 				goto ours;
    293 			ipstat.ips_forward++;
    294 		}
    295 #endif
    296 		/*
    297 		 * See if we belong to the destination multicast group on the
    298 		 * arrival interface.
    299 		 */
    300 		IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
    301 		if (inm == NULL) {
    302 			ipstat.ips_cantforward++;
    303 			m_freem(m);
    304 			goto next;
    305 		}
    306 		goto ours;
    307 	}
    308 	if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
    309 	    ip->ip_dst.s_addr == INADDR_ANY)
    310 		goto ours;
    311 
    312 	/*
    313 	 * Not for us; forward if possible and desirable.
    314 	 */
    315 	if (ipforwarding == 0) {
    316 		ipstat.ips_cantforward++;
    317 		m_freem(m);
    318 	} else
    319 		ip_forward(m, 0);
    320 	goto next;
    321 
    322 ours:
    323 	/*
    324 	 * If offset or IP_MF are set, must reassemble.
    325 	 * Otherwise, nothing need be done.
    326 	 * (We could look in the reassembly queue to see
    327 	 * if the packet was previously fragmented,
    328 	 * but it's not worth the time; just let them time out.)
    329 	 */
    330 	if (ip->ip_off &~ IP_DF) {
    331 		if (m->m_flags & M_EXT) {		/* XXX */
    332 			if ((m = m_pullup(m, sizeof (struct ip))) == 0) {
    333 				ipstat.ips_toosmall++;
    334 				goto next;
    335 			}
    336 			ip = mtod(m, struct ip *);
    337 		}
    338 		/*
    339 		 * Look for queue of fragments
    340 		 * of this datagram.
    341 		 */
    342 		for (fp = ipq.lh_first; fp != NULL; fp = fp->ipq_q.le_next)
    343 			if (ip->ip_id == fp->ipq_id &&
    344 			    ip->ip_src.s_addr == fp->ipq_src.s_addr &&
    345 			    ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
    346 			    ip->ip_p == fp->ipq_p)
    347 				goto found;
    348 		fp = 0;
    349 found:
    350 
    351 		/*
    352 		 * Adjust ip_len to not reflect header,
    353 		 * set ipqe_mff if more fragments are expected,
    354 		 * convert offset of this to bytes.
    355 		 */
    356 		ip->ip_len -= hlen;
    357 		mff = (ip->ip_off & IP_MF) != 0;
    358 		if (mff) {
    359 		        /*
    360 		         * Make sure that fragments have a data length
    361 			 * that's a non-zero multiple of 8 bytes.
    362 		         */
    363 			if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
    364 				ipstat.ips_badfrags++;
    365 				goto bad;
    366 			}
    367 		}
    368 		ip->ip_off <<= 3;
    369 
    370 		/*
    371 		 * If datagram marked as having more fragments
    372 		 * or if this is not the first fragment,
    373 		 * attempt reassembly; if it succeeds, proceed.
    374 		 */
    375 		if (mff || ip->ip_off) {
    376 			ipstat.ips_fragments++;
    377 			MALLOC(ipqe, struct ipqent *, sizeof (struct ipqent),
    378 			    M_IPQ, M_NOWAIT);
    379 			if (ipqe == NULL) {
    380 				ipstat.ips_rcvmemdrop++;
    381 				goto bad;
    382 			}
    383 			ipqe->ipqe_mff = mff;
    384 			ipqe->ipqe_ip = ip;
    385 			ip = ip_reass(ipqe, fp);
    386 			if (ip == 0)
    387 				goto next;
    388 			ipstat.ips_reassembled++;
    389 			m = dtom(ip);
    390 		} else
    391 			if (fp)
    392 				ip_freef(fp);
    393 	} else
    394 		ip->ip_len -= hlen;
    395 
    396 	/*
    397 	 * Switch out to protocol's input routine.
    398 	 */
    399 	ipstat.ips_delivered++;
    400 	(*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
    401 	goto next;
    402 bad:
    403 	m_freem(m);
    404 	goto next;
    405 }
    406 
    407 /*
    408  * Take incoming datagram fragment and try to
    409  * reassemble it into whole datagram.  If a chain for
    410  * reassembly of this datagram already exists, then it
    411  * is given as fp; otherwise have to make a chain.
    412  */
    413 struct ip *
    414 ip_reass(ipqe, fp)
    415 	register struct ipqent *ipqe;
    416 	register struct ipq *fp;
    417 {
    418 	register struct mbuf *m = dtom(ipqe->ipqe_ip);
    419 	register struct ipqent *nq, *p, *q;
    420 	struct ip *ip;
    421 	struct mbuf *t;
    422 	int hlen = ipqe->ipqe_ip->ip_hl << 2;
    423 	int i, next;
    424 
    425 	/*
    426 	 * Presence of header sizes in mbufs
    427 	 * would confuse code below.
    428 	 */
    429 	m->m_data += hlen;
    430 	m->m_len -= hlen;
    431 
    432 	/*
    433 	 * If first fragment to arrive, create a reassembly queue.
    434 	 */
    435 	if (fp == 0) {
    436 		if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
    437 			goto dropfrag;
    438 		fp = mtod(t, struct ipq *);
    439 		LIST_INSERT_HEAD(&ipq, fp, ipq_q);
    440 		fp->ipq_ttl = IPFRAGTTL;
    441 		fp->ipq_p = ipqe->ipqe_ip->ip_p;
    442 		fp->ipq_id = ipqe->ipqe_ip->ip_id;
    443 		LIST_INIT(&fp->ipq_fragq);
    444 		fp->ipq_src = ipqe->ipqe_ip->ip_src;
    445 		fp->ipq_dst = ipqe->ipqe_ip->ip_dst;
    446 		p = NULL;
    447 		goto insert;
    448 	}
    449 
    450 	/*
    451 	 * Find a segment which begins after this one does.
    452 	 */
    453 	for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL;
    454 	    p = q, q = q->ipqe_q.le_next)
    455 		if (q->ipqe_ip->ip_off > ipqe->ipqe_ip->ip_off)
    456 			break;
    457 
    458 	/*
    459 	 * If there is a preceding segment, it may provide some of
    460 	 * our data already.  If so, drop the data from the incoming
    461 	 * segment.  If it provides all of our data, drop us.
    462 	 */
    463 	if (p != NULL) {
    464 		i = p->ipqe_ip->ip_off + p->ipqe_ip->ip_len -
    465 		    ipqe->ipqe_ip->ip_off;
    466 		if (i > 0) {
    467 			if (i >= ipqe->ipqe_ip->ip_len)
    468 				goto dropfrag;
    469 			m_adj(dtom(ipqe->ipqe_ip), i);
    470 			ipqe->ipqe_ip->ip_off += i;
    471 			ipqe->ipqe_ip->ip_len -= i;
    472 		}
    473 	}
    474 
    475 	/*
    476 	 * While we overlap succeeding segments trim them or,
    477 	 * if they are completely covered, dequeue them.
    478 	 */
    479 	for (; q != NULL && ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len >
    480 	    q->ipqe_ip->ip_off; q = nq) {
    481 		i = (ipqe->ipqe_ip->ip_off + ipqe->ipqe_ip->ip_len) -
    482 		    q->ipqe_ip->ip_off;
    483 		if (i < q->ipqe_ip->ip_len) {
    484 			q->ipqe_ip->ip_len -= i;
    485 			q->ipqe_ip->ip_off += i;
    486 			m_adj(dtom(q->ipqe_ip), i);
    487 			break;
    488 		}
    489 		nq = q->ipqe_q.le_next;
    490 		m_freem(dtom(q->ipqe_ip));
    491 		LIST_REMOVE(q, ipqe_q);
    492 		FREE(q, M_IPQ);
    493 	}
    494 
    495 insert:
    496 	/*
    497 	 * Stick new segment in its place;
    498 	 * check for complete reassembly.
    499 	 */
    500 	if (p == NULL) {
    501 		LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q);
    502 	} else {
    503 		LIST_INSERT_AFTER(p, ipqe, ipqe_q);
    504 	}
    505 	next = 0;
    506 	for (p = NULL, q = fp->ipq_fragq.lh_first; q != NULL;
    507 	    p = q, q = q->ipqe_q.le_next) {
    508 		if (q->ipqe_ip->ip_off != next)
    509 			return (0);
    510 		next += q->ipqe_ip->ip_len;
    511 	}
    512 	if (p->ipqe_mff)
    513 		return (0);
    514 
    515 	/*
    516 	 * Reassembly is complete; concatenate fragments.
    517 	 */
    518 	q = fp->ipq_fragq.lh_first;
    519 	ip = q->ipqe_ip;
    520 	m = dtom(q->ipqe_ip);
    521 	t = m->m_next;
    522 	m->m_next = 0;
    523 	m_cat(m, t);
    524 	nq = q->ipqe_q.le_next;
    525 	FREE(q, M_IPQ);
    526 	for (q = nq; q != NULL; q = nq) {
    527 		t = dtom(q->ipqe_ip);
    528 		nq = q->ipqe_q.le_next;
    529 		FREE(q, M_IPQ);
    530 		m_cat(m, t);
    531 	}
    532 
    533 	/*
    534 	 * Create header for new ip packet by
    535 	 * modifying header of first packet;
    536 	 * dequeue and discard fragment reassembly header.
    537 	 * Make header visible.
    538 	 */
    539 	ip->ip_len = next;
    540 	ip->ip_src = fp->ipq_src;
    541 	ip->ip_dst = fp->ipq_dst;
    542 	LIST_REMOVE(fp, ipq_q);
    543 	(void) m_free(dtom(fp));
    544 	m->m_len += (ip->ip_hl << 2);
    545 	m->m_data -= (ip->ip_hl << 2);
    546 	/* some debugging cruft by sklower, below, will go away soon */
    547 	if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
    548 		register int plen = 0;
    549 		for (t = m; m; m = m->m_next)
    550 			plen += m->m_len;
    551 		t->m_pkthdr.len = plen;
    552 	}
    553 	return (ip);
    554 
    555 dropfrag:
    556 	ipstat.ips_fragdropped++;
    557 	m_freem(m);
    558 	FREE(ipqe, M_IPQ);
    559 	return (0);
    560 }
    561 
    562 /*
    563  * Free a fragment reassembly header and all
    564  * associated datagrams.
    565  */
    566 void
    567 ip_freef(fp)
    568 	struct ipq *fp;
    569 {
    570 	register struct ipqent *q, *p;
    571 
    572 	for (q = fp->ipq_fragq.lh_first; q != NULL; q = p) {
    573 		p = q->ipqe_q.le_next;
    574 		m_freem(dtom(q->ipqe_ip));
    575 		LIST_REMOVE(q, ipqe_q);
    576 		FREE(q, M_IPQ);
    577 	}
    578 	LIST_REMOVE(fp, ipq_q);
    579 	(void) m_free(dtom(fp));
    580 }
    581 
    582 /*
    583  * IP timer processing;
    584  * if a timer expires on a reassembly
    585  * queue, discard it.
    586  */
    587 void
    588 ip_slowtimo()
    589 {
    590 	register struct ipq *fp, *nfp;
    591 	int s = splsoftnet();
    592 
    593 	for (fp = ipq.lh_first; fp != NULL; fp = nfp) {
    594 		nfp = fp->ipq_q.le_next;
    595 		if (--fp->ipq_ttl == 0) {
    596 			ipstat.ips_fragtimeout++;
    597 			ip_freef(fp);
    598 		}
    599 	}
    600 	splx(s);
    601 }
    602 
    603 /*
    604  * Drain off all datagram fragments.
    605  */
    606 void
    607 ip_drain()
    608 {
    609 
    610 	while (ipq.lh_first != NULL) {
    611 		ipstat.ips_fragdropped++;
    612 		ip_freef(ipq.lh_first);
    613 	}
    614 }
    615 
    616 /*
    617  * Do option processing on a datagram,
    618  * possibly discarding it if bad options are encountered,
    619  * or forwarding it if source-routed.
    620  * Returns 1 if packet has been forwarded/freed,
    621  * 0 if the packet should be processed further.
    622  */
    623 int
    624 ip_dooptions(m)
    625 	struct mbuf *m;
    626 {
    627 	register struct ip *ip = mtod(m, struct ip *);
    628 	register u_char *cp;
    629 	register struct ip_timestamp *ipt;
    630 	register struct in_ifaddr *ia;
    631 	int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
    632 	struct in_addr *sin, dst;
    633 	n_time ntime;
    634 
    635 	dst = ip->ip_dst;
    636 	cp = (u_char *)(ip + 1);
    637 	cnt = (ip->ip_hl << 2) - sizeof (struct ip);
    638 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
    639 		opt = cp[IPOPT_OPTVAL];
    640 		if (opt == IPOPT_EOL)
    641 			break;
    642 		if (opt == IPOPT_NOP)
    643 			optlen = 1;
    644 		else {
    645 			optlen = cp[IPOPT_OLEN];
    646 			if (optlen <= 0 || optlen > cnt) {
    647 				code = &cp[IPOPT_OLEN] - (u_char *)ip;
    648 				goto bad;
    649 			}
    650 		}
    651 		switch (opt) {
    652 
    653 		default:
    654 			break;
    655 
    656 		/*
    657 		 * Source routing with record.
    658 		 * Find interface with current destination address.
    659 		 * If none on this machine then drop if strictly routed,
    660 		 * or do nothing if loosely routed.
    661 		 * Record interface address and bring up next address
    662 		 * component.  If strictly routed make sure next
    663 		 * address is on directly accessible net.
    664 		 */
    665 		case IPOPT_LSRR:
    666 		case IPOPT_SSRR:
    667 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
    668 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
    669 				goto bad;
    670 			}
    671 			ipaddr.sin_addr = ip->ip_dst;
    672 			ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr)));
    673 			if (ia == 0) {
    674 				if (opt == IPOPT_SSRR) {
    675 					type = ICMP_UNREACH;
    676 					code = ICMP_UNREACH_SRCFAIL;
    677 					goto bad;
    678 				}
    679 				/*
    680 				 * Loose routing, and not at next destination
    681 				 * yet; nothing to do except forward.
    682 				 */
    683 				break;
    684 			}
    685 			off--;			/* 0 origin */
    686 			if (off > optlen - sizeof(struct in_addr)) {
    687 				/*
    688 				 * End of source route.  Should be for us.
    689 				 */
    690 				save_rte(cp, ip->ip_src);
    691 				break;
    692 			}
    693 			/*
    694 			 * locate outgoing interface
    695 			 */
    696 			bcopy((caddr_t)(cp + off), (caddr_t)&ipaddr.sin_addr,
    697 			    sizeof(ipaddr.sin_addr));
    698 			if (opt == IPOPT_SSRR) {
    699 #define	INA	struct in_ifaddr *
    700 #define	SA	struct sockaddr *
    701 			    if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
    702 				ia = (INA)ifa_ifwithnet((SA)&ipaddr);
    703 			} else
    704 				ia = ip_rtaddr(ipaddr.sin_addr);
    705 			if (ia == 0) {
    706 				type = ICMP_UNREACH;
    707 				code = ICMP_UNREACH_SRCFAIL;
    708 				goto bad;
    709 			}
    710 			ip->ip_dst = ipaddr.sin_addr;
    711 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
    712 			    (caddr_t)(cp + off), sizeof(struct in_addr));
    713 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
    714 			/*
    715 			 * Let ip_intr's mcast routing check handle mcast pkts
    716 			 */
    717 			forward = !IN_MULTICAST(ip->ip_dst.s_addr);
    718 			break;
    719 
    720 		case IPOPT_RR:
    721 			if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
    722 				code = &cp[IPOPT_OFFSET] - (u_char *)ip;
    723 				goto bad;
    724 			}
    725 			/*
    726 			 * If no space remains, ignore.
    727 			 */
    728 			off--;			/* 0 origin */
    729 			if (off > optlen - sizeof(struct in_addr))
    730 				break;
    731 			bcopy((caddr_t)(&ip->ip_dst), (caddr_t)&ipaddr.sin_addr,
    732 			    sizeof(ipaddr.sin_addr));
    733 			/*
    734 			 * locate outgoing interface; if we're the destination,
    735 			 * use the incoming interface (should be same).
    736 			 */
    737 			if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
    738 			    (ia = ip_rtaddr(ipaddr.sin_addr)) == 0) {
    739 				type = ICMP_UNREACH;
    740 				code = ICMP_UNREACH_HOST;
    741 				goto bad;
    742 			}
    743 			bcopy((caddr_t)&ia->ia_addr.sin_addr,
    744 			    (caddr_t)(cp + off), sizeof(struct in_addr));
    745 			cp[IPOPT_OFFSET] += sizeof(struct in_addr);
    746 			break;
    747 
    748 		case IPOPT_TS:
    749 			code = cp - (u_char *)ip;
    750 			ipt = (struct ip_timestamp *)cp;
    751 			if (ipt->ipt_len < 5)
    752 				goto bad;
    753 			if (ipt->ipt_ptr > ipt->ipt_len - sizeof (int32_t)) {
    754 				if (++ipt->ipt_oflw == 0)
    755 					goto bad;
    756 				break;
    757 			}
    758 			sin = (struct in_addr *)(cp + ipt->ipt_ptr - 1);
    759 			switch (ipt->ipt_flg) {
    760 
    761 			case IPOPT_TS_TSONLY:
    762 				break;
    763 
    764 			case IPOPT_TS_TSANDADDR:
    765 				if (ipt->ipt_ptr + sizeof(n_time) +
    766 				    sizeof(struct in_addr) > ipt->ipt_len)
    767 					goto bad;
    768 				ipaddr.sin_addr = dst;
    769 				ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
    770 							    m->m_pkthdr.rcvif);
    771 				if (ia == 0)
    772 					continue;
    773 				bcopy((caddr_t)&ia->ia_addr.sin_addr,
    774 				    (caddr_t)sin, sizeof(struct in_addr));
    775 				ipt->ipt_ptr += sizeof(struct in_addr);
    776 				break;
    777 
    778 			case IPOPT_TS_PRESPEC:
    779 				if (ipt->ipt_ptr + sizeof(n_time) +
    780 				    sizeof(struct in_addr) > ipt->ipt_len)
    781 					goto bad;
    782 				bcopy((caddr_t)sin, (caddr_t)&ipaddr.sin_addr,
    783 				    sizeof(struct in_addr));
    784 				if (ifa_ifwithaddr((SA)&ipaddr) == 0)
    785 					continue;
    786 				ipt->ipt_ptr += sizeof(struct in_addr);
    787 				break;
    788 
    789 			default:
    790 				goto bad;
    791 			}
    792 			ntime = iptime();
    793 			bcopy((caddr_t)&ntime, (caddr_t)cp + ipt->ipt_ptr - 1,
    794 			    sizeof(n_time));
    795 			ipt->ipt_ptr += sizeof(n_time);
    796 		}
    797 	}
    798 	if (forward) {
    799 		ip_forward(m, 1);
    800 		return (1);
    801 	}
    802 	return (0);
    803 bad:
    804 	ip->ip_len -= ip->ip_hl << 2;   /* XXX icmp_error adds in hdr length */
    805 	icmp_error(m, type, code, 0, 0);
    806 	ipstat.ips_badoptions++;
    807 	return (1);
    808 }
    809 
    810 /*
    811  * Given address of next destination (final or next hop),
    812  * return internet address info of interface to be used to get there.
    813  */
    814 struct in_ifaddr *
    815 ip_rtaddr(dst)
    816 	 struct in_addr dst;
    817 {
    818 	register struct sockaddr_in *sin;
    819 
    820 	sin = satosin(&ipforward_rt.ro_dst);
    821 
    822 	if (ipforward_rt.ro_rt == 0 || dst.s_addr != sin->sin_addr.s_addr) {
    823 		if (ipforward_rt.ro_rt) {
    824 			RTFREE(ipforward_rt.ro_rt);
    825 			ipforward_rt.ro_rt = 0;
    826 		}
    827 		sin->sin_family = AF_INET;
    828 		sin->sin_len = sizeof(*sin);
    829 		sin->sin_addr = dst;
    830 
    831 		rtalloc(&ipforward_rt);
    832 	}
    833 	if (ipforward_rt.ro_rt == 0)
    834 		return ((struct in_ifaddr *)0);
    835 	return (ifatoia(ipforward_rt.ro_rt->rt_ifa));
    836 }
    837 
    838 /*
    839  * Save incoming source route for use in replies,
    840  * to be picked up later by ip_srcroute if the receiver is interested.
    841  */
    842 void
    843 save_rte(option, dst)
    844 	u_char *option;
    845 	struct in_addr dst;
    846 {
    847 	unsigned olen;
    848 
    849 	olen = option[IPOPT_OLEN];
    850 #ifdef DIAGNOSTIC
    851 	if (ipprintfs)
    852 		printf("save_rte: olen %d\n", olen);
    853 #endif
    854 	if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
    855 		return;
    856 	bcopy((caddr_t)option, (caddr_t)ip_srcrt.srcopt, olen);
    857 	ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
    858 	ip_srcrt.dst = dst;
    859 }
    860 
    861 /*
    862  * Retrieve incoming source route for use in replies,
    863  * in the same form used by setsockopt.
    864  * The first hop is placed before the options, will be removed later.
    865  */
    866 struct mbuf *
    867 ip_srcroute()
    868 {
    869 	register struct in_addr *p, *q;
    870 	register struct mbuf *m;
    871 
    872 	if (ip_nhops == 0)
    873 		return ((struct mbuf *)0);
    874 	m = m_get(M_DONTWAIT, MT_SOOPTS);
    875 	if (m == 0)
    876 		return ((struct mbuf *)0);
    877 
    878 #define OPTSIZ	(sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
    879 
    880 	/* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
    881 	m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
    882 	    OPTSIZ;
    883 #ifdef DIAGNOSTIC
    884 	if (ipprintfs)
    885 		printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
    886 #endif
    887 
    888 	/*
    889 	 * First save first hop for return route
    890 	 */
    891 	p = &ip_srcrt.route[ip_nhops - 1];
    892 	*(mtod(m, struct in_addr *)) = *p--;
    893 #ifdef DIAGNOSTIC
    894 	if (ipprintfs)
    895 		printf(" hops %lx", ntohl(mtod(m, struct in_addr *)->s_addr));
    896 #endif
    897 
    898 	/*
    899 	 * Copy option fields and padding (nop) to mbuf.
    900 	 */
    901 	ip_srcrt.nop = IPOPT_NOP;
    902 	ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
    903 	bcopy((caddr_t)&ip_srcrt.nop,
    904 	    mtod(m, caddr_t) + sizeof(struct in_addr), OPTSIZ);
    905 	q = (struct in_addr *)(mtod(m, caddr_t) +
    906 	    sizeof(struct in_addr) + OPTSIZ);
    907 #undef OPTSIZ
    908 	/*
    909 	 * Record return path as an IP source route,
    910 	 * reversing the path (pointers are now aligned).
    911 	 */
    912 	while (p >= ip_srcrt.route) {
    913 #ifdef DIAGNOSTIC
    914 		if (ipprintfs)
    915 			printf(" %lx", ntohl(q->s_addr));
    916 #endif
    917 		*q++ = *p--;
    918 	}
    919 	/*
    920 	 * Last hop goes to final destination.
    921 	 */
    922 	*q = ip_srcrt.dst;
    923 #ifdef DIAGNOSTIC
    924 	if (ipprintfs)
    925 		printf(" %lx\n", ntohl(q->s_addr));
    926 #endif
    927 	return (m);
    928 }
    929 
    930 /*
    931  * Strip out IP options, at higher
    932  * level protocol in the kernel.
    933  * Second argument is buffer to which options
    934  * will be moved, and return value is their length.
    935  * XXX should be deleted; last arg currently ignored.
    936  */
    937 void
    938 ip_stripoptions(m, mopt)
    939 	register struct mbuf *m;
    940 	struct mbuf *mopt;
    941 {
    942 	register int i;
    943 	struct ip *ip = mtod(m, struct ip *);
    944 	register caddr_t opts;
    945 	int olen;
    946 
    947 	olen = (ip->ip_hl<<2) - sizeof (struct ip);
    948 	opts = (caddr_t)(ip + 1);
    949 	i = m->m_len - (sizeof (struct ip) + olen);
    950 	bcopy(opts  + olen, opts, (unsigned)i);
    951 	m->m_len -= olen;
    952 	if (m->m_flags & M_PKTHDR)
    953 		m->m_pkthdr.len -= olen;
    954 	ip->ip_hl = sizeof(struct ip) >> 2;
    955 }
    956 
    957 int inetctlerrmap[PRC_NCMDS] = {
    958 	0,		0,		0,		0,
    959 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
    960 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
    961 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
    962 	0,		0,		0,		0,
    963 	ENOPROTOOPT
    964 };
    965 
    966 /*
    967  * Forward a packet.  If some error occurs return the sender
    968  * an icmp packet.  Note we can't always generate a meaningful
    969  * icmp message because icmp doesn't have a large enough repertoire
    970  * of codes and types.
    971  *
    972  * If not forwarding, just drop the packet.  This could be confusing
    973  * if ipforwarding was zero but some routing protocol was advancing
    974  * us as a gateway to somewhere.  However, we must let the routing
    975  * protocol deal with that.
    976  *
    977  * The srcrt parameter indicates whether the packet is being forwarded
    978  * via a source route.
    979  */
    980 void
    981 ip_forward(m, srcrt)
    982 	struct mbuf *m;
    983 	int srcrt;
    984 {
    985 	register struct ip *ip = mtod(m, struct ip *);
    986 	register struct sockaddr_in *sin;
    987 	register struct rtentry *rt;
    988 	int error, type = 0, code;
    989 	struct mbuf *mcopy;
    990 	n_long dest;
    991 	struct ifnet *destifp;
    992 
    993 	dest = 0;
    994 #ifdef DIAGNOSTIC
    995 	if (ipprintfs)
    996 		printf("forward: src %x dst %x ttl %x\n", ip->ip_src,
    997 			ip->ip_dst, ip->ip_ttl);
    998 #endif
    999 	if (m->m_flags & M_BCAST || in_canforward(ip->ip_dst) == 0) {
   1000 		ipstat.ips_cantforward++;
   1001 		m_freem(m);
   1002 		return;
   1003 	}
   1004 	HTONS(ip->ip_id);
   1005 	if (ip->ip_ttl <= IPTTLDEC) {
   1006 		icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
   1007 		return;
   1008 	}
   1009 	ip->ip_ttl -= IPTTLDEC;
   1010 
   1011 	sin = satosin(&ipforward_rt.ro_dst);
   1012 	if ((rt = ipforward_rt.ro_rt) == 0 ||
   1013 	    ip->ip_dst.s_addr != sin->sin_addr.s_addr) {
   1014 		if (ipforward_rt.ro_rt) {
   1015 			RTFREE(ipforward_rt.ro_rt);
   1016 			ipforward_rt.ro_rt = 0;
   1017 		}
   1018 		sin->sin_family = AF_INET;
   1019 		sin->sin_len = sizeof(*sin);
   1020 		sin->sin_addr = ip->ip_dst;
   1021 
   1022 		rtalloc(&ipforward_rt);
   1023 		if (ipforward_rt.ro_rt == 0) {
   1024 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
   1025 			return;
   1026 		}
   1027 		rt = ipforward_rt.ro_rt;
   1028 	}
   1029 
   1030 	/*
   1031 	 * Save at most 64 bytes of the packet in case
   1032 	 * we need to generate an ICMP message to the src.
   1033 	 */
   1034 	mcopy = m_copy(m, 0, imin((int)ip->ip_len, 64));
   1035 
   1036 	/*
   1037 	 * If forwarding packet using same interface that it came in on,
   1038 	 * perhaps should send a redirect to sender to shortcut a hop.
   1039 	 * Only send redirect if source is sending directly to us,
   1040 	 * and if packet was not source routed (or has any options).
   1041 	 * Also, don't send redirect if forwarding using a default route
   1042 	 * or a route modified by a redirect.
   1043 	 */
   1044 	if (rt->rt_ifp == m->m_pkthdr.rcvif &&
   1045 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
   1046 	    satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
   1047 	    ipsendredirects && !srcrt) {
   1048 		if (rt->rt_ifa &&
   1049 		    (ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_subnetmask) ==
   1050 		    ifatoia(rt->rt_ifa)->ia_subnet) {
   1051 		    if (rt->rt_flags & RTF_GATEWAY)
   1052 			dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
   1053 		    else
   1054 			dest = ip->ip_dst.s_addr;
   1055 		    /* Router requirements says to only send host redirects */
   1056 		    type = ICMP_REDIRECT;
   1057 		    code = ICMP_REDIRECT_HOST;
   1058 #ifdef DIAGNOSTIC
   1059 		    if (ipprintfs)
   1060 		        printf("redirect (%d) to %lx\n", code, (u_int32_t)dest);
   1061 #endif
   1062 		}
   1063 	}
   1064 
   1065 	error = ip_output(m, (struct mbuf *)0, &ipforward_rt, IP_FORWARDING
   1066 #ifdef DIRECTED_BROADCAST
   1067 			    | IP_ALLOWBROADCAST
   1068 #endif
   1069 						, 0);
   1070 	if (error)
   1071 		ipstat.ips_cantforward++;
   1072 	else {
   1073 		ipstat.ips_forward++;
   1074 		if (type)
   1075 			ipstat.ips_redirectsent++;
   1076 		else {
   1077 			if (mcopy)
   1078 				m_freem(mcopy);
   1079 			return;
   1080 		}
   1081 	}
   1082 	if (mcopy == NULL)
   1083 		return;
   1084 	destifp = NULL;
   1085 
   1086 	switch (error) {
   1087 
   1088 	case 0:				/* forwarded, but need redirect */
   1089 		/* type, code set above */
   1090 		break;
   1091 
   1092 	case ENETUNREACH:		/* shouldn't happen, checked above */
   1093 	case EHOSTUNREACH:
   1094 	case ENETDOWN:
   1095 	case EHOSTDOWN:
   1096 	default:
   1097 		type = ICMP_UNREACH;
   1098 		code = ICMP_UNREACH_HOST;
   1099 		break;
   1100 
   1101 	case EMSGSIZE:
   1102 		type = ICMP_UNREACH;
   1103 		code = ICMP_UNREACH_NEEDFRAG;
   1104 		if (ipforward_rt.ro_rt)
   1105 			destifp = ipforward_rt.ro_rt->rt_ifp;
   1106 		ipstat.ips_cantfrag++;
   1107 		break;
   1108 
   1109 	case ENOBUFS:
   1110 		type = ICMP_SOURCEQUENCH;
   1111 		code = 0;
   1112 		break;
   1113 	}
   1114 	icmp_error(mcopy, type, code, dest, destifp);
   1115 }
   1116 
   1117 int
   1118 ip_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
   1119 	int *name;
   1120 	u_int namelen;
   1121 	void *oldp;
   1122 	size_t *oldlenp;
   1123 	void *newp;
   1124 	size_t newlen;
   1125 {
   1126 	/* All sysctl names at this level are terminal. */
   1127 	if (namelen != 1)
   1128 		return (ENOTDIR);
   1129 
   1130 	switch (name[0]) {
   1131 	case IPCTL_FORWARDING:
   1132 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ipforwarding));
   1133 	case IPCTL_SENDREDIRECTS:
   1134 		return (sysctl_int(oldp, oldlenp, newp, newlen,
   1135 			&ipsendredirects));
   1136 	case IPCTL_DEFTTL:
   1137 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_defttl));
   1138 #ifdef notyet
   1139 	case IPCTL_DEFMTU:
   1140 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu));
   1141 #endif
   1142 	default:
   1143 		return (EOPNOTSUPP);
   1144 	}
   1145 	/* NOTREACHED */
   1146 }
   1147