xref: /netbsd-src/sys/netinet/ip_icmp.c (revision fdecd6a253f999ae92b139670d9e15cc9df4497c)
1 /*	$NetBSD: ip_icmp.c,v 1.23 1997/06/24 01:26:19 thorpej 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_icmp.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/protosw.h>
43 #include <sys/socket.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/proc.h>
47 
48 #include <vm/vm.h>
49 #include <sys/sysctl.h>
50 
51 #include <net/if.h>
52 #include <net/route.h>
53 
54 #include <netinet/in.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/in_var.h>
57 #include <netinet/ip.h>
58 #include <netinet/ip_icmp.h>
59 #include <netinet/ip_var.h>
60 #include <netinet/icmp_var.h>
61 
62 #include <machine/stdarg.h>
63 
64 /*
65  * ICMP routines: error generation, receive packet processing, and
66  * routines to turnaround packets back to the originator, and
67  * host table maintenance routines.
68  */
69 
70 int	icmpmaskrepl = 0;
71 #ifdef ICMPPRINTFS
72 int	icmpprintfs = 0;
73 #endif
74 
75 extern	struct protosw inetsw[];
76 
77 /*
78  * Generate an error packet of type error
79  * in response to bad packet ip.
80  */
81 void
82 icmp_error(n, type, code, dest, destifp)
83 	struct mbuf *n;
84 	int type, code;
85 	n_long dest;
86 	struct ifnet *destifp;
87 {
88 	register struct ip *oip = mtod(n, struct ip *), *nip;
89 	register unsigned oiplen = oip->ip_hl << 2;
90 	register struct icmp *icp;
91 	register struct mbuf *m;
92 	unsigned icmplen;
93 
94 #ifdef ICMPPRINTFS
95 	if (icmpprintfs)
96 		printf("icmp_error(%x, %d, %d)\n", oip, type, code);
97 #endif
98 	if (type != ICMP_REDIRECT)
99 		icmpstat.icps_error++;
100 	/*
101 	 * Don't send error if not the first fragment of message.
102 	 * Don't error if the old packet protocol was ICMP
103 	 * error message, only known informational types.
104 	 */
105 	if (oip->ip_off &~ (IP_MF|IP_DF))
106 		goto freeit;
107 	if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
108 	  n->m_len >= oiplen + ICMP_MINLEN &&
109 	  !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
110 		icmpstat.icps_oldicmp++;
111 		goto freeit;
112 	}
113 	/* Don't send error in response to a multicast or broadcast packet */
114 	if (n->m_flags & (M_BCAST|M_MCAST))
115 		goto freeit;
116 	/*
117 	 * First, formulate icmp message
118 	 */
119 	m = m_gethdr(M_DONTWAIT, MT_HEADER);
120 	if (m == NULL)
121 		goto freeit;
122 	icmplen = oiplen + min(8, oip->ip_len);
123 	m->m_len = icmplen + ICMP_MINLEN;
124 	MH_ALIGN(m, m->m_len);
125 	icp = mtod(m, struct icmp *);
126 	if ((u_int)type > ICMP_MAXTYPE)
127 		panic("icmp_error");
128 	icmpstat.icps_outhist[type]++;
129 	icp->icmp_type = type;
130 	if (type == ICMP_REDIRECT)
131 		icp->icmp_gwaddr.s_addr = dest;
132 	else {
133 		icp->icmp_void = 0;
134 		/*
135 		 * The following assignments assume an overlay with the
136 		 * zeroed icmp_void field.
137 		 */
138 		if (type == ICMP_PARAMPROB) {
139 			icp->icmp_pptr = code;
140 			code = 0;
141 		} else if (type == ICMP_UNREACH &&
142 		    code == ICMP_UNREACH_NEEDFRAG && destifp)
143 			icp->icmp_nextmtu = htons(destifp->if_mtu);
144 	}
145 
146 	icp->icmp_code = code;
147 	bcopy((caddr_t)oip, (caddr_t)&icp->icmp_ip, icmplen);
148 	nip = &icp->icmp_ip;
149 	nip->ip_len = htons((u_int16_t)(nip->ip_len + oiplen));
150 
151 	/*
152 	 * Now, copy old ip header (without options)
153 	 * in front of icmp message.
154 	 */
155 	if (m->m_data - sizeof(struct ip) < m->m_pktdat)
156 		panic("icmp len");
157 	m->m_data -= sizeof(struct ip);
158 	m->m_len += sizeof(struct ip);
159 	m->m_pkthdr.len = m->m_len;
160 	m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
161 	nip = mtod(m, struct ip *);
162 	bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
163 	nip->ip_len = m->m_len;
164 	nip->ip_hl = sizeof(struct ip) >> 2;
165 	nip->ip_p = IPPROTO_ICMP;
166 	nip->ip_tos = 0;
167 	icmp_reflect(m);
168 
169 freeit:
170 	m_freem(n);
171 }
172 
173 static struct sockaddr_in icmpsrc = { sizeof (struct sockaddr_in), AF_INET };
174 static struct sockaddr_in icmpdst = { sizeof (struct sockaddr_in), AF_INET };
175 static struct sockaddr_in icmpgw = { sizeof (struct sockaddr_in), AF_INET };
176 struct sockaddr_in icmpmask = { 8, 0 };
177 
178 /*
179  * Process a received ICMP message.
180  */
181 void
182 #if __STDC__
183 icmp_input(struct mbuf *m, ...)
184 #else
185 icmp_input(m, va_alist)
186 	struct mbuf *m;
187 	va_dcl
188 #endif
189 {
190 	register struct icmp *icp;
191 	register struct ip *ip = mtod(m, struct ip *);
192 	int icmplen = ip->ip_len;
193 	register int i;
194 	struct in_ifaddr *ia;
195 	void *(*ctlfunc) __P((int, struct sockaddr *, void *));
196 	int code;
197 	extern u_char ip_protox[];
198 	int hlen;
199 	va_list ap;
200 
201 	va_start(ap, m);
202 	hlen = va_arg(ap, int);
203 	va_end(ap);
204 
205 	/*
206 	 * Locate icmp structure in mbuf, and check
207 	 * that not corrupted and of at least minimum length.
208 	 */
209 #ifdef ICMPPRINTFS
210 	if (icmpprintfs)
211 		printf("icmp_input from %x to %x, len %d\n",
212 		    ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr),
213 		    icmplen);
214 #endif
215 	if (icmplen < ICMP_MINLEN) {
216 		icmpstat.icps_tooshort++;
217 		goto freeit;
218 	}
219 	i = hlen + min(icmplen, ICMP_ADVLENMIN);
220 	if (m->m_len < i && (m = m_pullup(m, i)) == 0) {
221 		icmpstat.icps_tooshort++;
222 		return;
223 	}
224 	ip = mtod(m, struct ip *);
225 	m->m_len -= hlen;
226 	m->m_data += hlen;
227 	icp = mtod(m, struct icmp *);
228 	if (in_cksum(m, icmplen)) {
229 		icmpstat.icps_checksum++;
230 		goto freeit;
231 	}
232 	m->m_len += hlen;
233 	m->m_data -= hlen;
234 
235 #ifdef ICMPPRINTFS
236 	/*
237 	 * Message type specific processing.
238 	 */
239 	if (icmpprintfs)
240 		printf("icmp_input, type %d code %d\n", icp->icmp_type,
241 		    icp->icmp_code);
242 #endif
243 	if (icp->icmp_type > ICMP_MAXTYPE)
244 		goto raw;
245 	icmpstat.icps_inhist[icp->icmp_type]++;
246 	code = icp->icmp_code;
247 	switch (icp->icmp_type) {
248 
249 	case ICMP_UNREACH:
250 		switch (code) {
251 			case ICMP_UNREACH_NET:
252 			case ICMP_UNREACH_HOST:
253 			case ICMP_UNREACH_PROTOCOL:
254 			case ICMP_UNREACH_PORT:
255 			case ICMP_UNREACH_SRCFAIL:
256 				code += PRC_UNREACH_NET;
257 				break;
258 
259 			case ICMP_UNREACH_NEEDFRAG:
260 				code = PRC_MSGSIZE;
261 				break;
262 
263 			case ICMP_UNREACH_NET_UNKNOWN:
264 			case ICMP_UNREACH_NET_PROHIB:
265 			case ICMP_UNREACH_TOSNET:
266 				code = PRC_UNREACH_NET;
267 				break;
268 
269 			case ICMP_UNREACH_HOST_UNKNOWN:
270 			case ICMP_UNREACH_ISOLATED:
271 			case ICMP_UNREACH_HOST_PROHIB:
272 			case ICMP_UNREACH_TOSHOST:
273 				code = PRC_UNREACH_HOST;
274 				break;
275 
276 			default:
277 				goto badcode;
278 		}
279 		goto deliver;
280 
281 	case ICMP_TIMXCEED:
282 		if (code > 1)
283 			goto badcode;
284 		code += PRC_TIMXCEED_INTRANS;
285 		goto deliver;
286 
287 	case ICMP_PARAMPROB:
288 		if (code > 1)
289 			goto badcode;
290 		code = PRC_PARAMPROB;
291 		goto deliver;
292 
293 	case ICMP_SOURCEQUENCH:
294 		if (code)
295 			goto badcode;
296 		code = PRC_QUENCH;
297 		goto deliver;
298 
299 	deliver:
300 		/*
301 		 * Problem with datagram; advise higher level routines.
302 		 */
303 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
304 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
305 			icmpstat.icps_badlen++;
306 			goto freeit;
307 		}
308 		if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
309 			goto badcode;
310 		NTOHS(icp->icmp_ip.ip_len);
311 #ifdef ICMPPRINTFS
312 		if (icmpprintfs)
313 			printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
314 #endif
315 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
316 		ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
317 		if (ctlfunc)
318 			(*ctlfunc)(code, sintosa(&icmpsrc), &icp->icmp_ip);
319 		break;
320 
321 	badcode:
322 		icmpstat.icps_badcode++;
323 		break;
324 
325 	case ICMP_ECHO:
326 		icp->icmp_type = ICMP_ECHOREPLY;
327 		goto reflect;
328 
329 	case ICMP_TSTAMP:
330 		if (icmplen < ICMP_TSLEN) {
331 			icmpstat.icps_badlen++;
332 			break;
333 		}
334 		icp->icmp_type = ICMP_TSTAMPREPLY;
335 		icp->icmp_rtime = iptime();
336 		icp->icmp_ttime = icp->icmp_rtime;	/* bogus, do later! */
337 		goto reflect;
338 
339 	case ICMP_MASKREQ:
340 		if (icmpmaskrepl == 0)
341 			break;
342 		/*
343 		 * We are not able to respond with all ones broadcast
344 		 * unless we receive it over a point-to-point interface.
345 		 */
346 		if (icmplen < ICMP_MASKLEN) {
347 			icmpstat.icps_badlen++;
348 			break;
349 		}
350 		if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
351 		    in_nullhost(ip->ip_dst))
352 			icmpdst.sin_addr = ip->ip_src;
353 		else
354 			icmpdst.sin_addr = ip->ip_dst;
355 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
356 		    m->m_pkthdr.rcvif));
357 		if (ia == 0)
358 			break;
359 		icp->icmp_type = ICMP_MASKREPLY;
360 		icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
361 		if (in_nullhost(ip->ip_src)) {
362 			if (ia->ia_ifp->if_flags & IFF_BROADCAST)
363 				ip->ip_src = ia->ia_broadaddr.sin_addr;
364 			else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
365 				ip->ip_src = ia->ia_dstaddr.sin_addr;
366 		}
367 reflect:
368 		ip->ip_len += hlen;	/* since ip_input deducts this */
369 		icmpstat.icps_reflect++;
370 		icmpstat.icps_outhist[icp->icmp_type]++;
371 		icmp_reflect(m);
372 		return;
373 
374 	case ICMP_REDIRECT:
375 		if (code > 3)
376 			goto badcode;
377 		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
378 		    icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
379 			icmpstat.icps_badlen++;
380 			break;
381 		}
382 		/*
383 		 * Short circuit routing redirects to force
384 		 * immediate change in the kernel's routing
385 		 * tables.  The message is also handed to anyone
386 		 * listening on a raw socket (e.g. the routing
387 		 * daemon for use in updating its tables).
388 		 */
389 		icmpgw.sin_addr = ip->ip_src;
390 		icmpdst.sin_addr = icp->icmp_gwaddr;
391 #ifdef	ICMPPRINTFS
392 		if (icmpprintfs)
393 			printf("redirect dst %x to %x\n", icp->icmp_ip.ip_dst,
394 			    icp->icmp_gwaddr);
395 #endif
396 		icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
397 		rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst),
398 		    (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
399 		    sintosa(&icmpgw), (struct rtentry **)0);
400 		pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc));
401 		break;
402 
403 	/*
404 	 * No kernel processing for the following;
405 	 * just fall through to send to raw listener.
406 	 */
407 	case ICMP_ECHOREPLY:
408 	case ICMP_ROUTERADVERT:
409 	case ICMP_ROUTERSOLICIT:
410 	case ICMP_TSTAMPREPLY:
411 	case ICMP_IREQREPLY:
412 	case ICMP_MASKREPLY:
413 	default:
414 		break;
415 	}
416 
417 raw:
418 	rip_input(m);
419 	return;
420 
421 freeit:
422 	m_freem(m);
423 }
424 
425 /*
426  * Reflect the ip packet back to the source
427  */
428 void
429 icmp_reflect(m)
430 	struct mbuf *m;
431 {
432 	register struct ip *ip = mtod(m, struct ip *);
433 	register struct in_ifaddr *ia;
434 	struct in_addr t;
435 	struct mbuf *opts = 0;
436 	int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
437 
438 	if (!in_canforward(ip->ip_src) &&
439 	    ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
440 	     htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
441 		m_freem(m);	/* Bad return address */
442 		goto done;	/* ip_output() will check for broadcast */
443 	}
444 	t = ip->ip_dst;
445 	ip->ip_dst = ip->ip_src;
446 	/*
447 	 * If the incoming packet was addressed directly to us,
448 	 * use dst as the src for the reply.  Otherwise (broadcast
449 	 * or anonymous), use the address which corresponds
450 	 * to the incoming interface.
451 	 */
452 	for (ia = in_ifaddr.tqh_first; ia; ia = ia->ia_list.tqe_next) {
453 		if (in_hosteq(t, ia->ia_addr.sin_addr))
454 			break;
455 		if ((ia->ia_ifp->if_flags & IFF_BROADCAST) &&
456 		    in_hosteq(t, ia->ia_broadaddr.sin_addr))
457 			break;
458 	}
459 	icmpdst.sin_addr = t;
460 	if (ia == (struct in_ifaddr *)0)
461 		ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
462 		    m->m_pkthdr.rcvif));
463 	/*
464 	 * The following happens if the packet was not addressed to us,
465 	 * and was received on an interface with no IP address.
466 	 */
467 	if (ia == (struct in_ifaddr *)0)
468 		ia = in_ifaddr.tqh_first;
469 	t = ia->ia_addr.sin_addr;
470 	ip->ip_src = t;
471 	ip->ip_ttl = MAXTTL;
472 
473 	if (optlen > 0) {
474 		register u_char *cp;
475 		int opt, cnt;
476 		u_int len;
477 
478 		/*
479 		 * Retrieve any source routing from the incoming packet;
480 		 * add on any record-route or timestamp options.
481 		 */
482 		cp = (u_char *) (ip + 1);
483 		if ((opts = ip_srcroute()) == 0 &&
484 		    (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
485 			opts->m_len = sizeof(struct in_addr);
486 			*mtod(opts, struct in_addr *) = zeroin_addr;
487 		}
488 		if (opts) {
489 #ifdef ICMPPRINTFS
490 		    if (icmpprintfs)
491 			    printf("icmp_reflect optlen %d rt %d => ",
492 				optlen, opts->m_len);
493 #endif
494 		    for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
495 			    opt = cp[IPOPT_OPTVAL];
496 			    if (opt == IPOPT_EOL)
497 				    break;
498 			    if (opt == IPOPT_NOP)
499 				    len = 1;
500 			    else {
501 				    len = cp[IPOPT_OLEN];
502 				    if (len <= 0 || len > cnt)
503 					    break;
504 			    }
505 			    /*
506 			     * Should check for overflow, but it "can't happen"
507 			     */
508 			    if (opt == IPOPT_RR || opt == IPOPT_TS ||
509 				opt == IPOPT_SECURITY) {
510 				    bcopy((caddr_t)cp,
511 					mtod(opts, caddr_t) + opts->m_len, len);
512 				    opts->m_len += len;
513 			    }
514 		    }
515 		    /* Terminate & pad, if necessary */
516 		    if ((cnt = opts->m_len % 4) != 0) {
517 			    for (; cnt < 4; cnt++) {
518 				    *(mtod(opts, caddr_t) + opts->m_len) =
519 					IPOPT_EOL;
520 				    opts->m_len++;
521 			    }
522 		    }
523 #ifdef ICMPPRINTFS
524 		    if (icmpprintfs)
525 			    printf("%d\n", opts->m_len);
526 #endif
527 		}
528 		/*
529 		 * Now strip out original options by copying rest of first
530 		 * mbuf's data back, and adjust the IP length.
531 		 */
532 		ip->ip_len -= optlen;
533 		ip->ip_hl = sizeof(struct ip) >> 2;
534 		m->m_len -= optlen;
535 		if (m->m_flags & M_PKTHDR)
536 			m->m_pkthdr.len -= optlen;
537 		optlen += sizeof(struct ip);
538 		bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
539 			 (unsigned)(m->m_len - sizeof(struct ip)));
540 	}
541 	m->m_flags &= ~(M_BCAST|M_MCAST);
542 	icmp_send(m, opts);
543 done:
544 	if (opts)
545 		(void)m_free(opts);
546 }
547 
548 /*
549  * Send an icmp packet back to the ip level,
550  * after supplying a checksum.
551  */
552 void
553 icmp_send(m, opts)
554 	register struct mbuf *m;
555 	struct mbuf *opts;
556 {
557 	register struct ip *ip = mtod(m, struct ip *);
558 	register int hlen;
559 	register struct icmp *icp;
560 
561 	hlen = ip->ip_hl << 2;
562 	m->m_data += hlen;
563 	m->m_len -= hlen;
564 	icp = mtod(m, struct icmp *);
565 	icp->icmp_cksum = 0;
566 	icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
567 	m->m_data -= hlen;
568 	m->m_len += hlen;
569 #ifdef ICMPPRINTFS
570 	if (icmpprintfs)
571 		printf("icmp_send dst %x src %x\n", ip->ip_dst, ip->ip_src);
572 #endif
573 	(void) ip_output(m, opts, NULL, 0, NULL);
574 }
575 
576 n_time
577 iptime()
578 {
579 	struct timeval atv;
580 	u_long t;
581 
582 	microtime(&atv);
583 	t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
584 	return (htonl(t));
585 }
586 
587 int
588 icmp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
589 	int *name;
590 	u_int namelen;
591 	void *oldp;
592 	size_t *oldlenp;
593 	void *newp;
594 	size_t newlen;
595 {
596 
597 	/* All sysctl names at this level are terminal. */
598 	if (namelen != 1)
599 		return (ENOTDIR);
600 
601 	switch (name[0]) {
602 	case ICMPCTL_MASKREPL:
603 		return (sysctl_int(oldp, oldlenp, newp, newlen, &icmpmaskrepl));
604 	default:
605 		return (ENOPROTOOPT);
606 	}
607 	/* NOTREACHED */
608 }
609