ip_icmp.c revision 1.132.2.1 1 /* $NetBSD: ip_icmp.c,v 1.132.2.1 2014/08/10 06:56:25 tls Exp $ */
2
3 /*
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * 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. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32 /*-
33 * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc.
34 * All rights reserved.
35 *
36 * This code is derived from software contributed to The NetBSD Foundation
37 * by Public Access Networks Corporation ("Panix"). It was developed under
38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon.
39 *
40 * This code is derived from software contributed to The NetBSD Foundation
41 * by Jason R. Thorpe of Zembu Labs, Inc.
42 *
43 * Redistribution and use in source and binary forms, with or without
44 * modification, are permitted provided that the following conditions
45 * are met:
46 * 1. Redistributions of source code must retain the above copyright
47 * notice, this list of conditions and the following disclaimer.
48 * 2. Redistributions in binary form must reproduce the above copyright
49 * notice, this list of conditions and the following disclaimer in the
50 * documentation and/or other materials provided with the distribution.
51 *
52 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
53 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
54 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
55 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
56 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
57 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
58 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
59 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
60 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
61 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
62 * POSSIBILITY OF SUCH DAMAGE.
63 */
64
65 /*
66 * Copyright (c) 1982, 1986, 1988, 1993
67 * The Regents of the University of California. All rights reserved.
68 *
69 * Redistribution and use in source and binary forms, with or without
70 * modification, are permitted provided that the following conditions
71 * are met:
72 * 1. Redistributions of source code must retain the above copyright
73 * notice, this list of conditions and the following disclaimer.
74 * 2. Redistributions in binary form must reproduce the above copyright
75 * notice, this list of conditions and the following disclaimer in the
76 * documentation and/or other materials provided with the distribution.
77 * 3. Neither the name of the University nor the names of its contributors
78 * may be used to endorse or promote products derived from this software
79 * without specific prior written permission.
80 *
81 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
82 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
83 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
84 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
85 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
86 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
87 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
88 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
89 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
90 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
91 * SUCH DAMAGE.
92 *
93 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94
94 */
95
96 #include <sys/cdefs.h>
97 __KERNEL_RCSID(0, "$NetBSD: ip_icmp.c,v 1.132.2.1 2014/08/10 06:56:25 tls Exp $");
98
99 #include "opt_ipsec.h"
100
101 #include <sys/param.h>
102 #include <sys/systm.h>
103 #include <sys/mbuf.h>
104 #include <sys/protosw.h>
105 #include <sys/socket.h>
106 #include <sys/kmem.h>
107 #include <sys/time.h>
108 #include <sys/kernel.h>
109 #include <sys/syslog.h>
110 #include <sys/sysctl.h>
111
112 #include <net/if.h>
113 #include <net/route.h>
114
115 #include <netinet/in.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/in_var.h>
118 #include <netinet/ip.h>
119 #include <netinet/ip_icmp.h>
120 #include <netinet/ip_var.h>
121 #include <netinet/in_pcb.h>
122 #include <netinet/in_proto.h>
123 #include <netinet/icmp_var.h>
124 #include <netinet/icmp_private.h>
125
126 #ifdef IPSEC
127 #include <netipsec/ipsec.h>
128 #include <netipsec/key.h>
129 #endif /* IPSEC*/
130
131 /*
132 * ICMP routines: error generation, receive packet processing, and
133 * routines to turnaround packets back to the originator, and
134 * host table maintenance routines.
135 */
136
137 int icmpmaskrepl = 0;
138 int icmpbmcastecho = 0;
139 #ifdef ICMPPRINTFS
140 int icmpprintfs = 0;
141 #endif
142 int icmpreturndatabytes = 8;
143
144 percpu_t *icmpstat_percpu;
145
146 /*
147 * List of callbacks to notify when Path MTU changes are made.
148 */
149 struct icmp_mtudisc_callback {
150 LIST_ENTRY(icmp_mtudisc_callback) mc_list;
151 void (*mc_func)(struct in_addr);
152 };
153
154 LIST_HEAD(, icmp_mtudisc_callback) icmp_mtudisc_callbacks =
155 LIST_HEAD_INITIALIZER(&icmp_mtudisc_callbacks);
156
157 #if 0
158 static u_int ip_next_mtu(u_int, int);
159 #else
160 /*static*/ u_int ip_next_mtu(u_int, int);
161 #endif
162
163 extern int icmperrppslim;
164 static int icmperrpps_count = 0;
165 static struct timeval icmperrppslim_last;
166 static int icmp_rediraccept = 1;
167 static int icmp_redirtimeout = 600;
168 static struct rttimer_queue *icmp_redirect_timeout_q = NULL;
169
170 static void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *);
171 static void icmp_redirect_timeout(struct rtentry *, struct rttimer *);
172
173 static void sysctl_netinet_icmp_setup(struct sysctllog **);
174
175 void
176 icmp_init(void)
177 {
178
179 sysctl_netinet_icmp_setup(NULL);
180
181 /*
182 * This is only useful if the user initializes redirtimeout to
183 * something other than zero.
184 */
185 if (icmp_redirtimeout != 0) {
186 icmp_redirect_timeout_q =
187 rt_timer_queue_create(icmp_redirtimeout);
188 }
189
190 icmpstat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP_NSTATS);
191 }
192
193 /*
194 * Register a Path MTU Discovery callback.
195 */
196 void
197 icmp_mtudisc_callback_register(void (*func)(struct in_addr))
198 {
199 struct icmp_mtudisc_callback *mc;
200
201 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
202 mc = LIST_NEXT(mc, mc_list)) {
203 if (mc->mc_func == func)
204 return;
205 }
206
207 mc = kmem_alloc(sizeof(*mc), KM_SLEEP);
208 mc->mc_func = func;
209 LIST_INSERT_HEAD(&icmp_mtudisc_callbacks, mc, mc_list);
210 }
211
212 /*
213 * Generate an error packet of type error
214 * in response to bad packet ip.
215 */
216 void
217 icmp_error(struct mbuf *n, int type, int code, n_long dest,
218 int destmtu)
219 {
220 struct ip *oip = mtod(n, struct ip *), *nip;
221 unsigned oiplen = oip->ip_hl << 2;
222 struct icmp *icp;
223 struct mbuf *m;
224 struct m_tag *mtag;
225 unsigned icmplen, mblen;
226
227 #ifdef ICMPPRINTFS
228 if (icmpprintfs)
229 printf("icmp_error(%p, type:%d, code:%d)\n", oip, type, code);
230 #endif
231 if (type != ICMP_REDIRECT)
232 ICMP_STATINC(ICMP_STAT_ERROR);
233 /*
234 * Don't send error if the original packet was encrypted.
235 * Don't send error if not the first fragment of message.
236 * Don't error if the old packet protocol was ICMP
237 * error message, only known informational types.
238 */
239 if (n->m_flags & M_DECRYPTED)
240 goto freeit;
241 if (oip->ip_off &~ htons(IP_MF|IP_DF))
242 goto freeit;
243 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
244 n->m_len >= oiplen + ICMP_MINLEN &&
245 !ICMP_INFOTYPE(((struct icmp *)((char *)oip + oiplen))->icmp_type)) {
246 ICMP_STATINC(ICMP_STAT_OLDICMP);
247 goto freeit;
248 }
249 /* Don't send error in response to a multicast or broadcast packet */
250 if (n->m_flags & (M_BCAST|M_MCAST))
251 goto freeit;
252
253 /*
254 * First, do a rate limitation check.
255 */
256 if (icmp_ratelimit(&oip->ip_src, type, code)) {
257 /* XXX stat */
258 goto freeit;
259 }
260
261 /*
262 * Now, formulate icmp message
263 */
264 icmplen = oiplen + min(icmpreturndatabytes,
265 ntohs(oip->ip_len) - oiplen);
266 /*
267 * Defend against mbuf chains shorter than oip->ip_len - oiplen:
268 */
269 mblen = 0;
270 for (m = n; m && (mblen < icmplen); m = m->m_next)
271 mblen += m->m_len;
272 icmplen = min(mblen, icmplen);
273
274 /*
275 * As we are not required to return everything we have,
276 * we return whatever we can return at ease.
277 *
278 * Note that ICMP datagrams longer than 576 octets are out of spec
279 * according to RFC1812; the limit on icmpreturndatabytes below in
280 * icmp_sysctl will keep things below that limit.
281 */
282
283 KASSERT(ICMP_MINLEN <= MCLBYTES);
284
285 if (icmplen + ICMP_MINLEN > MCLBYTES)
286 icmplen = MCLBYTES - ICMP_MINLEN;
287
288 m = m_gethdr(M_DONTWAIT, MT_HEADER);
289 if (m && (icmplen + ICMP_MINLEN > MHLEN)) {
290 MCLGET(m, M_DONTWAIT);
291 if ((m->m_flags & M_EXT) == 0) {
292 m_freem(m);
293 m = NULL;
294 }
295 }
296 if (m == NULL)
297 goto freeit;
298 MCLAIM(m, n->m_owner);
299 m->m_len = icmplen + ICMP_MINLEN;
300 if ((m->m_flags & M_EXT) == 0)
301 MH_ALIGN(m, m->m_len);
302 else {
303 m->m_data += sizeof(struct ip);
304 m->m_len -= sizeof(struct ip);
305 }
306 icp = mtod(m, struct icmp *);
307 if ((u_int)type > ICMP_MAXTYPE)
308 panic("icmp_error");
309 ICMP_STATINC(ICMP_STAT_OUTHIST + type);
310 icp->icmp_type = type;
311 if (type == ICMP_REDIRECT)
312 icp->icmp_gwaddr.s_addr = dest;
313 else {
314 icp->icmp_void = 0;
315 /*
316 * The following assignments assume an overlay with the
317 * zeroed icmp_void field.
318 */
319 if (type == ICMP_PARAMPROB) {
320 icp->icmp_pptr = code;
321 code = 0;
322 } else if (type == ICMP_UNREACH &&
323 code == ICMP_UNREACH_NEEDFRAG && destmtu)
324 icp->icmp_nextmtu = htons(destmtu);
325 }
326
327 icp->icmp_code = code;
328 m_copydata(n, 0, icmplen, (void *)&icp->icmp_ip);
329
330 /*
331 * Now, copy old ip header (without options)
332 * in front of icmp message.
333 */
334 if ((m->m_flags & M_EXT) == 0 &&
335 m->m_data - sizeof(struct ip) < m->m_pktdat)
336 panic("icmp len");
337 m->m_data -= sizeof(struct ip);
338 m->m_len += sizeof(struct ip);
339 m->m_pkthdr.len = m->m_len;
340 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
341 nip = mtod(m, struct ip *);
342 /* ip_v set in ip_output */
343 nip->ip_hl = sizeof(struct ip) >> 2;
344 nip->ip_tos = 0;
345 nip->ip_len = htons(m->m_len);
346 /* ip_id set in ip_output */
347 nip->ip_off = htons(0);
348 /* ip_ttl set in icmp_reflect */
349 nip->ip_p = IPPROTO_ICMP;
350 nip->ip_src = oip->ip_src;
351 nip->ip_dst = oip->ip_dst;
352 /* move PF m_tag to new packet, if it exists */
353 mtag = m_tag_find(n, PACKET_TAG_PF, NULL);
354 if (mtag != NULL) {
355 m_tag_unlink(n, mtag);
356 m_tag_prepend(m, mtag);
357 }
358 icmp_reflect(m);
359
360 freeit:
361 m_freem(n);
362 }
363
364 struct sockaddr_in icmpsrc = {
365 .sin_len = sizeof (struct sockaddr_in),
366 .sin_family = AF_INET,
367 };
368 static struct sockaddr_in icmpdst = {
369 .sin_len = sizeof (struct sockaddr_in),
370 .sin_family = AF_INET,
371 };
372 static struct sockaddr_in icmpgw = {
373 .sin_len = sizeof (struct sockaddr_in),
374 .sin_family = AF_INET,
375 };
376 struct sockaddr_in icmpmask = {
377 .sin_len = 8,
378 .sin_family = 0,
379 };
380
381 /*
382 * Process a received ICMP message.
383 */
384 void
385 icmp_input(struct mbuf *m, ...)
386 {
387 int proto;
388 struct icmp *icp;
389 struct ip *ip = mtod(m, struct ip *);
390 int icmplen;
391 int i;
392 struct in_ifaddr *ia;
393 void *(*ctlfunc)(int, const struct sockaddr *, void *);
394 int code;
395 int hlen;
396 va_list ap;
397 struct rtentry *rt;
398
399 va_start(ap, m);
400 hlen = va_arg(ap, int);
401 proto = va_arg(ap, int);
402 va_end(ap);
403
404 /*
405 * Locate icmp structure in mbuf, and check
406 * that not corrupted and of at least minimum length.
407 */
408 icmplen = ntohs(ip->ip_len) - hlen;
409 #ifdef ICMPPRINTFS
410 if (icmpprintfs) {
411 printf("icmp_input from `%s' to ", inet_ntoa(ip->ip_src));
412 printf("`%s', len %d\n", inet_ntoa(ip->ip_dst), icmplen);
413 }
414 #endif
415 if (icmplen < ICMP_MINLEN) {
416 ICMP_STATINC(ICMP_STAT_TOOSHORT);
417 goto freeit;
418 }
419 i = hlen + min(icmplen, ICMP_ADVLENMIN);
420 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == NULL) {
421 ICMP_STATINC(ICMP_STAT_TOOSHORT);
422 return;
423 }
424 ip = mtod(m, struct ip *);
425 m->m_len -= hlen;
426 m->m_data += hlen;
427 icp = mtod(m, struct icmp *);
428 /* Don't need to assert alignment, here. */
429 if (in_cksum(m, icmplen)) {
430 ICMP_STATINC(ICMP_STAT_CHECKSUM);
431 goto freeit;
432 }
433 m->m_len += hlen;
434 m->m_data -= hlen;
435
436 #ifdef ICMPPRINTFS
437 /*
438 * Message type specific processing.
439 */
440 if (icmpprintfs)
441 printf("icmp_input(type:%d, code:%d)\n", icp->icmp_type,
442 icp->icmp_code);
443 #endif
444 if (icp->icmp_type > ICMP_MAXTYPE)
445 goto raw;
446 ICMP_STATINC(ICMP_STAT_INHIST + icp->icmp_type);
447 code = icp->icmp_code;
448 switch (icp->icmp_type) {
449
450 case ICMP_UNREACH:
451 switch (code) {
452 case ICMP_UNREACH_PROTOCOL:
453 code = PRC_UNREACH_PROTOCOL;
454 break;
455
456 case ICMP_UNREACH_PORT:
457 code = PRC_UNREACH_PORT;
458 break;
459
460 case ICMP_UNREACH_SRCFAIL:
461 code = PRC_UNREACH_SRCFAIL;
462 break;
463
464 case ICMP_UNREACH_NEEDFRAG:
465 code = PRC_MSGSIZE;
466 break;
467
468 case ICMP_UNREACH_NET:
469 case ICMP_UNREACH_NET_UNKNOWN:
470 case ICMP_UNREACH_NET_PROHIB:
471 case ICMP_UNREACH_TOSNET:
472 code = PRC_UNREACH_NET;
473 break;
474
475 case ICMP_UNREACH_HOST:
476 case ICMP_UNREACH_HOST_UNKNOWN:
477 case ICMP_UNREACH_ISOLATED:
478 case ICMP_UNREACH_HOST_PROHIB:
479 case ICMP_UNREACH_TOSHOST:
480 case ICMP_UNREACH_ADMIN_PROHIBIT:
481 case ICMP_UNREACH_HOST_PREC:
482 case ICMP_UNREACH_PREC_CUTOFF:
483 code = PRC_UNREACH_HOST;
484 break;
485
486 default:
487 goto badcode;
488 }
489 goto deliver;
490
491 case ICMP_TIMXCEED:
492 if (code > 1)
493 goto badcode;
494 code += PRC_TIMXCEED_INTRANS;
495 goto deliver;
496
497 case ICMP_PARAMPROB:
498 if (code > 1)
499 goto badcode;
500 code = PRC_PARAMPROB;
501 goto deliver;
502
503 case ICMP_SOURCEQUENCH:
504 if (code)
505 goto badcode;
506 code = PRC_QUENCH;
507 goto deliver;
508
509 deliver:
510 /*
511 * Problem with datagram; advise higher level routines.
512 */
513 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
514 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
515 ICMP_STATINC(ICMP_STAT_BADLEN);
516 goto freeit;
517 }
518 if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr))
519 goto badcode;
520 #ifdef ICMPPRINTFS
521 if (icmpprintfs)
522 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
523 #endif
524 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
525 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
526 if (ctlfunc)
527 (void) (*ctlfunc)(code, sintosa(&icmpsrc),
528 &icp->icmp_ip);
529 break;
530
531 badcode:
532 ICMP_STATINC(ICMP_STAT_BADCODE);
533 break;
534
535 case ICMP_ECHO:
536 if (!icmpbmcastecho &&
537 (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
538 ICMP_STATINC(ICMP_STAT_BMCASTECHO);
539 break;
540 }
541 icp->icmp_type = ICMP_ECHOREPLY;
542 goto reflect;
543
544 case ICMP_TSTAMP:
545 if (icmplen < ICMP_TSLEN) {
546 ICMP_STATINC(ICMP_STAT_BADLEN);
547 break;
548 }
549 if (!icmpbmcastecho &&
550 (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
551 ICMP_STATINC(ICMP_STAT_BMCASTTSTAMP);
552 break;
553 }
554 icp->icmp_type = ICMP_TSTAMPREPLY;
555 icp->icmp_rtime = iptime();
556 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */
557 goto reflect;
558
559 case ICMP_MASKREQ:
560 if (icmpmaskrepl == 0)
561 break;
562 /*
563 * We are not able to respond with all ones broadcast
564 * unless we receive it over a point-to-point interface.
565 */
566 if (icmplen < ICMP_MASKLEN) {
567 ICMP_STATINC(ICMP_STAT_BADLEN);
568 break;
569 }
570 if (ip->ip_dst.s_addr == INADDR_BROADCAST ||
571 in_nullhost(ip->ip_dst))
572 icmpdst.sin_addr = ip->ip_src;
573 else
574 icmpdst.sin_addr = ip->ip_dst;
575 ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst),
576 m->m_pkthdr.rcvif));
577 if (ia == 0)
578 break;
579 icp->icmp_type = ICMP_MASKREPLY;
580 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
581 if (in_nullhost(ip->ip_src)) {
582 if (ia->ia_ifp->if_flags & IFF_BROADCAST)
583 ip->ip_src = ia->ia_broadaddr.sin_addr;
584 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
585 ip->ip_src = ia->ia_dstaddr.sin_addr;
586 }
587 reflect:
588 {
589 uint64_t *icps = percpu_getref(icmpstat_percpu);
590 icps[ICMP_STAT_REFLECT]++;
591 icps[ICMP_STAT_OUTHIST + icp->icmp_type]++;
592 percpu_putref(icmpstat_percpu);
593 }
594 icmp_reflect(m);
595 return;
596
597 case ICMP_REDIRECT:
598 if (code > 3)
599 goto badcode;
600 if (icmp_rediraccept == 0)
601 goto freeit;
602 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
603 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
604 ICMP_STATINC(ICMP_STAT_BADLEN);
605 break;
606 }
607 /*
608 * Short circuit routing redirects to force
609 * immediate change in the kernel's routing
610 * tables. The message is also handed to anyone
611 * listening on a raw socket (e.g. the routing
612 * daemon for use in updating its tables).
613 */
614 icmpgw.sin_addr = ip->ip_src;
615 icmpdst.sin_addr = icp->icmp_gwaddr;
616 #ifdef ICMPPRINTFS
617 if (icmpprintfs) {
618 printf("redirect dst `%s' to `%s'\n",
619 inet_ntoa(icp->icmp_ip.ip_dst),
620 inet_ntoa(icp->icmp_gwaddr));
621 }
622 #endif
623 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
624 rt = NULL;
625 rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst),
626 NULL, RTF_GATEWAY | RTF_HOST, sintosa(&icmpgw), &rt);
627 if (rt != NULL && icmp_redirtimeout != 0) {
628 i = rt_timer_add(rt, icmp_redirect_timeout,
629 icmp_redirect_timeout_q);
630 if (i)
631 log(LOG_ERR, "ICMP: redirect failed to "
632 "register timeout for route to %x, "
633 "code %d\n",
634 icp->icmp_ip.ip_dst.s_addr, i);
635 }
636 if (rt != NULL)
637 rtfree(rt);
638
639 pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc));
640 #if defined(IPSEC)
641 if (ipsec_used)
642 key_sa_routechange((struct sockaddr *)&icmpsrc);
643 #endif
644 break;
645
646 /*
647 * No kernel processing for the following;
648 * just fall through to send to raw listener.
649 */
650 case ICMP_ECHOREPLY:
651 case ICMP_ROUTERADVERT:
652 case ICMP_ROUTERSOLICIT:
653 case ICMP_TSTAMPREPLY:
654 case ICMP_IREQREPLY:
655 case ICMP_MASKREPLY:
656 default:
657 break;
658 }
659
660 raw:
661 rip_input(m, hlen, proto);
662 return;
663
664 freeit:
665 m_freem(m);
666 return;
667 }
668
669 /*
670 * Reflect the ip packet back to the source
671 */
672 void
673 icmp_reflect(struct mbuf *m)
674 {
675 struct ip *ip = mtod(m, struct ip *);
676 struct in_ifaddr *ia;
677 struct ifaddr *ifa;
678 struct sockaddr_in *sin = 0;
679 struct in_addr t;
680 struct mbuf *opts = 0;
681 int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
682
683 if (!in_canforward(ip->ip_src) &&
684 ((ip->ip_src.s_addr & IN_CLASSA_NET) !=
685 htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
686 m_freem(m); /* Bad return address */
687 goto done; /* ip_output() will check for broadcast */
688 }
689 t = ip->ip_dst;
690 ip->ip_dst = ip->ip_src;
691 /*
692 * If the incoming packet was addressed directly to us, use
693 * dst as the src for the reply. Otherwise (broadcast or
694 * anonymous), use an address which corresponds to the
695 * incoming interface, with a preference for the address which
696 * corresponds to the route to the destination of the ICMP.
697 */
698
699 /* Look for packet addressed to us */
700 INADDR_TO_IA(t, ia);
701
702 /* look for packet sent to broadcast address */
703 if (ia == NULL && m->m_pkthdr.rcvif &&
704 (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST)) {
705 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) {
706 if (ifa->ifa_addr->sa_family != AF_INET)
707 continue;
708 if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) {
709 ia = ifatoia(ifa);
710 break;
711 }
712 }
713 }
714
715 if (ia)
716 sin = &ia->ia_addr;
717
718 icmpdst.sin_addr = t;
719
720 /*
721 * if the packet is addressed somewhere else, compute the
722 * source address for packets routed back to the source, and
723 * use that, if it's an address on the interface which
724 * received the packet
725 */
726 if (sin == NULL && m->m_pkthdr.rcvif) {
727 struct sockaddr_in sin_dst;
728 struct route icmproute;
729 int errornum;
730
731 sockaddr_in_init(&sin_dst, &ip->ip_dst, 0);
732 memset(&icmproute, 0, sizeof(icmproute));
733 errornum = 0;
734 sin = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum);
735 /* errornum is never used */
736 rtcache_free(&icmproute);
737 /* check to make sure sin is a source address on rcvif */
738 if (sin) {
739 t = sin->sin_addr;
740 sin = NULL;
741 INADDR_TO_IA(t, ia);
742 while (ia) {
743 if (ia->ia_ifp == m->m_pkthdr.rcvif) {
744 sin = &ia->ia_addr;
745 break;
746 }
747 NEXT_IA_WITH_SAME_ADDR(ia);
748 }
749 }
750 }
751
752 /*
753 * if it was not addressed to us, but the route doesn't go out
754 * the source interface, pick an address on the source
755 * interface. This can happen when routing is asymmetric, or
756 * when the incoming packet was encapsulated
757 */
758 if (sin == NULL && m->m_pkthdr.rcvif) {
759 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) {
760 if (ifa->ifa_addr->sa_family != AF_INET)
761 continue;
762 sin = &(ifatoia(ifa)->ia_addr);
763 break;
764 }
765 }
766
767 /*
768 * The following happens if the packet was not addressed to us,
769 * and was received on an interface with no IP address:
770 * We find the first AF_INET address on the first non-loopback
771 * interface.
772 */
773 if (sin == NULL)
774 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) {
775 if (ia->ia_ifp->if_flags & IFF_LOOPBACK)
776 continue;
777 sin = &ia->ia_addr;
778 break;
779 }
780
781 /*
782 * If we still didn't find an address, punt. We could have an
783 * interface up (and receiving packets) with no address.
784 */
785 if (sin == NULL) {
786 m_freem(m);
787 goto done;
788 }
789
790 ip->ip_src = sin->sin_addr;
791 ip->ip_ttl = MAXTTL;
792
793 if (optlen > 0) {
794 u_char *cp;
795 int opt, cnt;
796 u_int len;
797
798 /*
799 * Retrieve any source routing from the incoming packet;
800 * add on any record-route or timestamp options.
801 */
802 cp = (u_char *) (ip + 1);
803 if ((opts = ip_srcroute()) == 0 &&
804 (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) {
805 MCLAIM(opts, m->m_owner);
806 opts->m_len = sizeof(struct in_addr);
807 *mtod(opts, struct in_addr *) = zeroin_addr;
808 }
809 if (opts) {
810 #ifdef ICMPPRINTFS
811 if (icmpprintfs)
812 printf("icmp_reflect optlen %d rt %d => ",
813 optlen, opts->m_len);
814 #endif
815 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
816 opt = cp[IPOPT_OPTVAL];
817 if (opt == IPOPT_EOL)
818 break;
819 if (opt == IPOPT_NOP)
820 len = 1;
821 else {
822 if (cnt < IPOPT_OLEN + sizeof(*cp))
823 break;
824 len = cp[IPOPT_OLEN];
825 if (len < IPOPT_OLEN + sizeof(*cp) ||
826 len > cnt)
827 break;
828 }
829 /*
830 * Should check for overflow, but it "can't happen"
831 */
832 if (opt == IPOPT_RR || opt == IPOPT_TS ||
833 opt == IPOPT_SECURITY) {
834 memmove(mtod(opts, char *) + opts->m_len,
835 cp, len);
836 opts->m_len += len;
837 }
838 }
839 /* Terminate & pad, if necessary */
840 if ((cnt = opts->m_len % 4) != 0) {
841 for (; cnt < 4; cnt++) {
842 *(mtod(opts, char *) + opts->m_len) =
843 IPOPT_EOL;
844 opts->m_len++;
845 }
846 }
847 #ifdef ICMPPRINTFS
848 if (icmpprintfs)
849 printf("%d\n", opts->m_len);
850 #endif
851 }
852 /*
853 * Now strip out original options by copying rest of first
854 * mbuf's data back, and adjust the IP length.
855 */
856 ip->ip_len = htons(ntohs(ip->ip_len) - optlen);
857 ip->ip_hl = sizeof(struct ip) >> 2;
858 m->m_len -= optlen;
859 if (m->m_flags & M_PKTHDR)
860 m->m_pkthdr.len -= optlen;
861 optlen += sizeof(struct ip);
862 memmove(ip + 1, (char *)ip + optlen,
863 (unsigned)(m->m_len - sizeof(struct ip)));
864 }
865 m_tag_delete_nonpersistent(m);
866 m->m_flags &= ~(M_BCAST|M_MCAST);
867
868 /*
869 * Clear any in-bound checksum flags for this packet.
870 */
871 if (m->m_flags & M_PKTHDR)
872 m->m_pkthdr.csum_flags = 0;
873
874 icmp_send(m, opts);
875 done:
876 if (opts)
877 (void)m_free(opts);
878 }
879
880 /*
881 * Send an icmp packet back to the ip level,
882 * after supplying a checksum.
883 */
884 void
885 icmp_send(struct mbuf *m, struct mbuf *opts)
886 {
887 struct ip *ip = mtod(m, struct ip *);
888 int hlen;
889 struct icmp *icp;
890
891 hlen = ip->ip_hl << 2;
892 m->m_data += hlen;
893 m->m_len -= hlen;
894 icp = mtod(m, struct icmp *);
895 icp->icmp_cksum = 0;
896 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
897 m->m_data -= hlen;
898 m->m_len += hlen;
899 #ifdef ICMPPRINTFS
900 if (icmpprintfs) {
901 printf("icmp_send to destination `%s' from `%s'\n",
902 inet_ntoa(ip->ip_dst), inet_ntoa(ip->ip_src));
903 }
904 #endif
905 (void)ip_output(m, opts, NULL, 0, NULL, NULL);
906 }
907
908 n_time
909 iptime(void)
910 {
911 struct timeval atv;
912 u_long t;
913
914 microtime(&atv);
915 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
916 return (htonl(t));
917 }
918
919 /*
920 * sysctl helper routine for net.inet.icmp.returndatabytes. ensures
921 * that the new value is in the correct range.
922 */
923 static int
924 sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS)
925 {
926 int error, t;
927 struct sysctlnode node;
928
929 node = *rnode;
930 node.sysctl_data = &t;
931 t = icmpreturndatabytes;
932 error = sysctl_lookup(SYSCTLFN_CALL(&node));
933 if (error || newp == NULL)
934 return (error);
935
936 if (t < 8 || t > 512)
937 return (EINVAL);
938 icmpreturndatabytes = t;
939
940 return (0);
941 }
942
943 /*
944 * sysctl helper routine for net.inet.icmp.redirtimeout. ensures that
945 * the given value is not less than zero and then resets the timeout
946 * queue.
947 */
948 static int
949 sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS)
950 {
951 int error, tmp;
952 struct sysctlnode node;
953
954 node = *rnode;
955 node.sysctl_data = &tmp;
956 tmp = icmp_redirtimeout;
957 error = sysctl_lookup(SYSCTLFN_CALL(&node));
958 if (error || newp == NULL)
959 return (error);
960 if (tmp < 0)
961 return (EINVAL);
962 icmp_redirtimeout = tmp;
963
964 /*
965 * was it a *defined* side-effect that anyone even *reading*
966 * this value causes these things to happen?
967 */
968 if (icmp_redirect_timeout_q != NULL) {
969 if (icmp_redirtimeout == 0) {
970 rt_timer_queue_destroy(icmp_redirect_timeout_q,
971 true);
972 icmp_redirect_timeout_q = NULL;
973 } else {
974 rt_timer_queue_change(icmp_redirect_timeout_q,
975 icmp_redirtimeout);
976 }
977 } else if (icmp_redirtimeout > 0) {
978 icmp_redirect_timeout_q =
979 rt_timer_queue_create(icmp_redirtimeout);
980 }
981
982 return (0);
983 }
984
985 static int
986 sysctl_net_inet_icmp_stats(SYSCTLFN_ARGS)
987 {
988
989 return (NETSTAT_SYSCTL(icmpstat_percpu, ICMP_NSTATS));
990 }
991
992 static void
993 sysctl_netinet_icmp_setup(struct sysctllog **clog)
994 {
995
996 sysctl_createv(clog, 0, NULL, NULL,
997 CTLFLAG_PERMANENT,
998 CTLTYPE_NODE, "inet", NULL,
999 NULL, 0, NULL, 0,
1000 CTL_NET, PF_INET, CTL_EOL);
1001 sysctl_createv(clog, 0, NULL, NULL,
1002 CTLFLAG_PERMANENT,
1003 CTLTYPE_NODE, "icmp",
1004 SYSCTL_DESCR("ICMPv4 related settings"),
1005 NULL, 0, NULL, 0,
1006 CTL_NET, PF_INET, IPPROTO_ICMP, CTL_EOL);
1007
1008 sysctl_createv(clog, 0, NULL, NULL,
1009 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1010 CTLTYPE_INT, "maskrepl",
1011 SYSCTL_DESCR("Respond to ICMP_MASKREQ messages"),
1012 NULL, 0, &icmpmaskrepl, 0,
1013 CTL_NET, PF_INET, IPPROTO_ICMP,
1014 ICMPCTL_MASKREPL, CTL_EOL);
1015 sysctl_createv(clog, 0, NULL, NULL,
1016 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1017 CTLTYPE_INT, "returndatabytes",
1018 SYSCTL_DESCR("Number of bytes to return in an ICMP "
1019 "error message"),
1020 sysctl_net_inet_icmp_returndatabytes, 0,
1021 &icmpreturndatabytes, 0,
1022 CTL_NET, PF_INET, IPPROTO_ICMP,
1023 ICMPCTL_RETURNDATABYTES, CTL_EOL);
1024 sysctl_createv(clog, 0, NULL, NULL,
1025 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1026 CTLTYPE_INT, "errppslimit",
1027 SYSCTL_DESCR("Maximum number of outgoing ICMP error "
1028 "messages per second"),
1029 NULL, 0, &icmperrppslim, 0,
1030 CTL_NET, PF_INET, IPPROTO_ICMP,
1031 ICMPCTL_ERRPPSLIMIT, CTL_EOL);
1032 sysctl_createv(clog, 0, NULL, NULL,
1033 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1034 CTLTYPE_INT, "rediraccept",
1035 SYSCTL_DESCR("Accept ICMP_REDIRECT messages"),
1036 NULL, 0, &icmp_rediraccept, 0,
1037 CTL_NET, PF_INET, IPPROTO_ICMP,
1038 ICMPCTL_REDIRACCEPT, CTL_EOL);
1039 sysctl_createv(clog, 0, NULL, NULL,
1040 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1041 CTLTYPE_INT, "redirtimeout",
1042 SYSCTL_DESCR("Lifetime of ICMP_REDIRECT generated "
1043 "routes"),
1044 sysctl_net_inet_icmp_redirtimeout, 0,
1045 &icmp_redirtimeout, 0,
1046 CTL_NET, PF_INET, IPPROTO_ICMP,
1047 ICMPCTL_REDIRTIMEOUT, CTL_EOL);
1048 sysctl_createv(clog, 0, NULL, NULL,
1049 CTLFLAG_PERMANENT,
1050 CTLTYPE_STRUCT, "stats",
1051 SYSCTL_DESCR("ICMP statistics"),
1052 sysctl_net_inet_icmp_stats, 0, NULL, 0,
1053 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_STATS,
1054 CTL_EOL);
1055 sysctl_createv(clog, 0, NULL, NULL,
1056 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1057 CTLTYPE_INT, "bmcastecho",
1058 SYSCTL_DESCR("Respond to ICMP_ECHO or ICMP_TIMESTAMP "
1059 "message to the broadcast or multicast"),
1060 NULL, 0, &icmpbmcastecho, 0,
1061 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_BMCASTECHO,
1062 CTL_EOL);
1063 }
1064
1065 void
1066 icmp_statinc(u_int stat)
1067 {
1068
1069 KASSERT(stat < ICMP_NSTATS);
1070 ICMP_STATINC(stat);
1071 }
1072
1073 /* Table of common MTUs: */
1074
1075 static const u_int mtu_table[] = {
1076 65535, 65280, 32000, 17914, 9180, 8166,
1077 4352, 2002, 1492, 1006, 508, 296, 68, 0
1078 };
1079
1080 void
1081 icmp_mtudisc(struct icmp *icp, struct in_addr faddr)
1082 {
1083 struct icmp_mtudisc_callback *mc;
1084 struct sockaddr *dst = sintosa(&icmpsrc);
1085 struct rtentry *rt;
1086 u_long mtu = ntohs(icp->icmp_nextmtu); /* Why a long? IPv6 */
1087 int error;
1088
1089 rt = rtalloc1(dst, 1);
1090 if (rt == 0)
1091 return;
1092
1093 /* If we didn't get a host route, allocate one */
1094
1095 if ((rt->rt_flags & RTF_HOST) == 0) {
1096 struct rtentry *nrt;
1097
1098 error = rtrequest((int) RTM_ADD, dst,
1099 (struct sockaddr *) rt->rt_gateway, NULL,
1100 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
1101 if (error) {
1102 rtfree(rt);
1103 return;
1104 }
1105 nrt->rt_rmx = rt->rt_rmx;
1106 rtfree(rt);
1107 rt = nrt;
1108 }
1109 error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q);
1110 if (error) {
1111 rtfree(rt);
1112 return;
1113 }
1114
1115 if (mtu == 0) {
1116 int i = 0;
1117
1118 mtu = ntohs(icp->icmp_ip.ip_len);
1119 /* Some 4.2BSD-based routers incorrectly adjust the ip_len */
1120 if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0)
1121 mtu -= (icp->icmp_ip.ip_hl << 2);
1122
1123 /* If we still can't guess a value, try the route */
1124
1125 if (mtu == 0) {
1126 mtu = rt->rt_rmx.rmx_mtu;
1127
1128 /* If no route mtu, default to the interface mtu */
1129
1130 if (mtu == 0)
1131 mtu = rt->rt_ifp->if_mtu;
1132 }
1133
1134 for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++)
1135 if (mtu > mtu_table[i]) {
1136 mtu = mtu_table[i];
1137 break;
1138 }
1139 }
1140
1141 /*
1142 * XXX: RTV_MTU is overloaded, since the admin can set it
1143 * to turn off PMTU for a route, and the kernel can
1144 * set it to indicate a serious problem with PMTU
1145 * on a route. We should be using a separate flag
1146 * for the kernel to indicate this.
1147 */
1148
1149 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1150 if (mtu < 296 || mtu > rt->rt_ifp->if_mtu)
1151 rt->rt_rmx.rmx_locks |= RTV_MTU;
1152 else if (rt->rt_rmx.rmx_mtu > mtu ||
1153 rt->rt_rmx.rmx_mtu == 0) {
1154 ICMP_STATINC(ICMP_STAT_PMTUCHG);
1155 rt->rt_rmx.rmx_mtu = mtu;
1156 }
1157 }
1158
1159 if (rt)
1160 rtfree(rt);
1161
1162 /*
1163 * Notify protocols that the MTU for this destination
1164 * has changed.
1165 */
1166 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL;
1167 mc = LIST_NEXT(mc, mc_list))
1168 (*mc->mc_func)(faddr);
1169 }
1170
1171 /*
1172 * Return the next larger or smaller MTU plateau (table from RFC 1191)
1173 * given current value MTU. If DIR is less than zero, a larger plateau
1174 * is returned; otherwise, a smaller value is returned.
1175 */
1176 u_int
1177 ip_next_mtu(u_int mtu, int dir) /* XXX */
1178 {
1179 int i;
1180
1181 for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) {
1182 if (mtu >= mtu_table[i])
1183 break;
1184 }
1185
1186 if (dir < 0) {
1187 if (i == 0) {
1188 return 0;
1189 } else {
1190 return mtu_table[i - 1];
1191 }
1192 } else {
1193 if (mtu_table[i] == 0) {
1194 return 0;
1195 } else if (mtu > mtu_table[i]) {
1196 return mtu_table[i];
1197 } else {
1198 return mtu_table[i + 1];
1199 }
1200 }
1201 }
1202
1203 static void
1204 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
1205 {
1206 if (rt == NULL)
1207 panic("icmp_mtudisc_timeout: bad route to timeout");
1208 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1209 (RTF_DYNAMIC | RTF_HOST)) {
1210 rtrequest((int) RTM_DELETE, rt_getkey(rt),
1211 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1212 } else {
1213 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
1214 rt->rt_rmx.rmx_mtu = 0;
1215 }
1216 }
1217 }
1218
1219 static void
1220 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r)
1221 {
1222 if (rt == NULL)
1223 panic("icmp_redirect_timeout: bad route to timeout");
1224 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
1225 (RTF_DYNAMIC | RTF_HOST)) {
1226 rtrequest((int) RTM_DELETE, rt_getkey(rt),
1227 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
1228 }
1229 }
1230
1231 /*
1232 * Perform rate limit check.
1233 * Returns 0 if it is okay to send the icmp packet.
1234 * Returns 1 if the router SHOULD NOT send this icmp packet due to rate
1235 * limitation.
1236 *
1237 * XXX per-destination/type check necessary?
1238 */
1239 int
1240 icmp_ratelimit(const struct in_addr *dst, const int type,
1241 const int code)
1242 {
1243
1244 /* PPS limit */
1245 if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count,
1246 icmperrppslim)) {
1247 /* The packet is subject to rate limit */
1248 return 1;
1249 }
1250
1251 /* okay to send */
1252 return 0;
1253 }
1254