1 /* $NetBSD: ip_icmp.c,v 1.104 2006/10/12 01:32:38 christos 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 * 3. All advertising materials mentioning features or use of this software 52 * must display the following acknowledgement: 53 * This product includes software developed by the NetBSD 54 * Foundation, Inc. and its contributors. 55 * 4. Neither the name of The NetBSD Foundation nor the names of its 56 * contributors may be used to endorse or promote products derived 57 * from this software without specific prior written permission. 58 * 59 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 60 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 61 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 62 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 63 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 64 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 65 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 66 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 67 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 68 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 69 * POSSIBILITY OF SUCH DAMAGE. 70 */ 71 72 /* 73 * Copyright (c) 1982, 1986, 1988, 1993 74 * The Regents of the University of California. All rights reserved. 75 * 76 * Redistribution and use in source and binary forms, with or without 77 * modification, are permitted provided that the following conditions 78 * are met: 79 * 1. Redistributions of source code must retain the above copyright 80 * notice, this list of conditions and the following disclaimer. 81 * 2. Redistributions in binary form must reproduce the above copyright 82 * notice, this list of conditions and the following disclaimer in the 83 * documentation and/or other materials provided with the distribution. 84 * 3. Neither the name of the University nor the names of its contributors 85 * may be used to endorse or promote products derived from this software 86 * without specific prior written permission. 87 * 88 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 89 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 90 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 91 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 92 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 93 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 94 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 95 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 96 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 97 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 98 * SUCH DAMAGE. 99 * 100 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 101 */ 102 103 #include <sys/cdefs.h> 104 __KERNEL_RCSID(0, "$NetBSD: ip_icmp.c,v 1.104 2006/10/12 01:32:38 christos Exp $"); 105 106 #include "opt_ipsec.h" 107 108 #include <sys/param.h> 109 #include <sys/systm.h> 110 #include <sys/malloc.h> 111 #include <sys/mbuf.h> 112 #include <sys/protosw.h> 113 #include <sys/socket.h> 114 #include <sys/time.h> 115 #include <sys/kernel.h> 116 #include <sys/syslog.h> 117 #include <sys/sysctl.h> 118 119 #include <net/if.h> 120 #include <net/route.h> 121 122 #include <netinet/in.h> 123 #include <netinet/in_systm.h> 124 #include <netinet/in_var.h> 125 #include <netinet/ip.h> 126 #include <netinet/ip_icmp.h> 127 #include <netinet/ip_var.h> 128 #include <netinet/in_pcb.h> 129 #include <netinet/in_proto.h> 130 #include <netinet/icmp_var.h> 131 132 #ifdef IPSEC 133 #include <netinet6/ipsec.h> 134 #include <netkey/key.h> 135 #endif 136 137 #ifdef FAST_IPSEC 138 #include <netipsec/ipsec.h> 139 #include <netipsec/key.h> 140 #endif /* FAST_IPSEC*/ 141 142 #include <machine/stdarg.h> 143 144 /* 145 * ICMP routines: error generation, receive packet processing, and 146 * routines to turnaround packets back to the originator, and 147 * host table maintenance routines. 148 */ 149 150 int icmpmaskrepl = 0; 151 #ifdef ICMPPRINTFS 152 int icmpprintfs = 0; 153 #endif 154 int icmpreturndatabytes = 8; 155 156 struct icmpstat icmpstat; 157 158 /* 159 * List of callbacks to notify when Path MTU changes are made. 160 */ 161 struct icmp_mtudisc_callback { 162 LIST_ENTRY(icmp_mtudisc_callback) mc_list; 163 void (*mc_func)(struct in_addr); 164 }; 165 166 LIST_HEAD(, icmp_mtudisc_callback) icmp_mtudisc_callbacks = 167 LIST_HEAD_INITIALIZER(&icmp_mtudisc_callbacks); 168 169 #if 0 170 static u_int ip_next_mtu(u_int, int); 171 #else 172 /*static*/ u_int ip_next_mtu(u_int, int); 173 #endif 174 175 extern int icmperrppslim; 176 static int icmperrpps_count = 0; 177 static struct timeval icmperrppslim_last; 178 static int icmp_rediraccept = 1; 179 static int icmp_redirtimeout = 600; 180 static struct rttimer_queue *icmp_redirect_timeout_q = NULL; 181 182 static void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *); 183 static void icmp_redirect_timeout(struct rtentry *, struct rttimer *); 184 185 static int icmp_ratelimit(const struct in_addr *, const int, const int); 186 187 188 void 189 icmp_init(void) 190 { 191 /* 192 * This is only useful if the user initializes redirtimeout to 193 * something other than zero. 194 */ 195 if (icmp_redirtimeout != 0) { 196 icmp_redirect_timeout_q = 197 rt_timer_queue_create(icmp_redirtimeout); 198 } 199 } 200 201 /* 202 * Register a Path MTU Discovery callback. 203 */ 204 void 205 icmp_mtudisc_callback_register(void (*func)(struct in_addr)) 206 { 207 struct icmp_mtudisc_callback *mc; 208 209 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL; 210 mc = LIST_NEXT(mc, mc_list)) { 211 if (mc->mc_func == func) 212 return; 213 } 214 215 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 216 if (mc == NULL) 217 panic("icmp_mtudisc_callback_register"); 218 219 mc->mc_func = func; 220 LIST_INSERT_HEAD(&icmp_mtudisc_callbacks, mc, mc_list); 221 } 222 223 /* 224 * Generate an error packet of type error 225 * in response to bad packet ip. 226 */ 227 void 228 icmp_error(struct mbuf *n, int type, int code, n_long dest, 229 int destmtu) 230 { 231 struct ip *oip = mtod(n, struct ip *), *nip; 232 unsigned oiplen = oip->ip_hl << 2; 233 struct icmp *icp; 234 struct mbuf *m; 235 struct m_tag *mtag; 236 unsigned icmplen, mblen; 237 238 #ifdef ICMPPRINTFS 239 if (icmpprintfs) 240 printf("icmp_error(%p, type:%d, code:%d)\n", oip, type, code); 241 #endif 242 if (type != ICMP_REDIRECT) 243 icmpstat.icps_error++; 244 /* 245 * Don't send error if the original packet was encrypted. 246 * Don't send error if not the first fragment of message. 247 * Don't error if the old packet protocol was ICMP 248 * error message, only known informational types. 249 */ 250 if (n->m_flags & M_DECRYPTED) 251 goto freeit; 252 if (oip->ip_off &~ htons(IP_MF|IP_DF)) 253 goto freeit; 254 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 255 n->m_len >= oiplen + ICMP_MINLEN && 256 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) { 257 icmpstat.icps_oldicmp++; 258 goto freeit; 259 } 260 /* Don't send error in response to a multicast or broadcast packet */ 261 if (n->m_flags & (M_BCAST|M_MCAST)) 262 goto freeit; 263 264 /* 265 * First, do a rate limitation check. 266 */ 267 if (icmp_ratelimit(&oip->ip_src, type, code)) { 268 /* XXX stat */ 269 goto freeit; 270 } 271 272 /* 273 * Now, formulate icmp message 274 */ 275 icmplen = oiplen + min(icmpreturndatabytes, 276 ntohs(oip->ip_len) - oiplen); 277 /* 278 * Defend against mbuf chains shorter than oip->ip_len - oiplen: 279 */ 280 mblen = 0; 281 for (m = n; m && (mblen < icmplen); m = m->m_next) 282 mblen += m->m_len; 283 icmplen = min(mblen, icmplen); 284 285 /* 286 * As we are not required to return everything we have, 287 * we return whatever we can return at ease. 288 * 289 * Note that ICMP datagrams longer than 576 octets are out of spec 290 * according to RFC1812; the limit on icmpreturndatabytes below in 291 * icmp_sysctl will keep things below that limit. 292 */ 293 294 KASSERT(ICMP_MINLEN <= MCLBYTES); 295 296 if (icmplen + ICMP_MINLEN > MCLBYTES) 297 icmplen = MCLBYTES - ICMP_MINLEN; 298 299 m = m_gethdr(M_DONTWAIT, MT_HEADER); 300 if (m && (icmplen + ICMP_MINLEN > MHLEN)) { 301 MCLGET(m, M_DONTWAIT); 302 if ((m->m_flags & M_EXT) == 0) { 303 m_freem(m); 304 m = NULL; 305 } 306 } 307 if (m == NULL) 308 goto freeit; 309 MCLAIM(m, n->m_owner); 310 m->m_len = icmplen + ICMP_MINLEN; 311 if ((m->m_flags & M_EXT) == 0) 312 MH_ALIGN(m, m->m_len); 313 icp = mtod(m, struct icmp *); 314 if ((u_int)type > ICMP_MAXTYPE) 315 panic("icmp_error"); 316 icmpstat.icps_outhist[type]++; 317 icp->icmp_type = type; 318 if (type == ICMP_REDIRECT) 319 icp->icmp_gwaddr.s_addr = dest; 320 else { 321 icp->icmp_void = 0; 322 /* 323 * The following assignments assume an overlay with the 324 * zeroed icmp_void field. 325 */ 326 if (type == ICMP_PARAMPROB) { 327 icp->icmp_pptr = code; 328 code = 0; 329 } else if (type == ICMP_UNREACH && 330 code == ICMP_UNREACH_NEEDFRAG && destmtu) 331 icp->icmp_nextmtu = htons(destmtu); 332 } 333 334 icp->icmp_code = code; 335 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 336 337 /* 338 * Now, copy old ip header (without options) 339 * in front of icmp message. 340 */ 341 if (m->m_data - sizeof(struct ip) < m->m_pktdat) 342 panic("icmp len"); 343 m->m_data -= sizeof(struct ip); 344 m->m_len += sizeof(struct ip); 345 m->m_pkthdr.len = m->m_len; 346 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 347 nip = mtod(m, struct ip *); 348 /* ip_v set in ip_output */ 349 nip->ip_hl = sizeof(struct ip) >> 2; 350 nip->ip_tos = 0; 351 nip->ip_len = htons(m->m_len); 352 /* ip_id set in ip_output */ 353 nip->ip_off = htons(0); 354 /* ip_ttl set in icmp_reflect */ 355 nip->ip_p = IPPROTO_ICMP; 356 nip->ip_src = oip->ip_src; 357 nip->ip_dst = oip->ip_dst; 358 /* move PF_GENERATED m_tag to new packet, if it exists */ 359 mtag = m_tag_find(n, PACKET_TAG_PF_GENERATED, NULL); 360 if (mtag != NULL) { 361 m_tag_unlink(n, mtag); 362 m_tag_prepend(m, mtag); 363 } 364 icmp_reflect(m); 365 366 freeit: 367 m_freem(n); 368 } 369 370 struct sockaddr_in icmpsrc = { 371 .sin_len = sizeof (struct sockaddr_in), 372 .sin_family = AF_INET, 373 }; 374 static struct sockaddr_in icmpdst = { 375 .sin_len = sizeof (struct sockaddr_in), 376 .sin_family = AF_INET, 377 }; 378 static struct sockaddr_in icmpgw = { 379 .sin_len = sizeof (struct sockaddr_in), 380 .sin_family = AF_INET, 381 }; 382 struct sockaddr_in icmpmask = { 383 .sin_len = 8, 384 .sin_family = 0, 385 }; 386 387 /* 388 * Process a received ICMP message. 389 */ 390 void 391 icmp_input(struct mbuf *m, ...) 392 { 393 int proto; 394 struct icmp *icp; 395 struct ip *ip = mtod(m, struct ip *); 396 int icmplen; 397 int i; 398 struct in_ifaddr *ia; 399 void *(*ctlfunc)(int, struct sockaddr *, void *); 400 int code; 401 int hlen; 402 va_list ap; 403 struct rtentry *rt; 404 405 va_start(ap, m); 406 hlen = va_arg(ap, int); 407 proto = va_arg(ap, int); 408 va_end(ap); 409 410 /* 411 * Locate icmp structure in mbuf, and check 412 * that not corrupted and of at least minimum length. 413 */ 414 icmplen = ntohs(ip->ip_len) - hlen; 415 #ifdef ICMPPRINTFS 416 if (icmpprintfs) { 417 printf("icmp_input from `%s' to ", inet_ntoa(ip->ip_src)); 418 printf("`%s', len %d\n", inet_ntoa(ip->ip_dst), icmplen); 419 } 420 #endif 421 if (icmplen < ICMP_MINLEN) { 422 icmpstat.icps_tooshort++; 423 goto freeit; 424 } 425 i = hlen + min(icmplen, ICMP_ADVLENMIN); 426 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) { 427 icmpstat.icps_tooshort++; 428 return; 429 } 430 ip = mtod(m, struct ip *); 431 m->m_len -= hlen; 432 m->m_data += hlen; 433 icp = mtod(m, struct icmp *); 434 /* Don't need to assert alignment, here. */ 435 if (in_cksum(m, icmplen)) { 436 icmpstat.icps_checksum++; 437 goto freeit; 438 } 439 m->m_len += hlen; 440 m->m_data -= hlen; 441 442 #ifdef ICMPPRINTFS 443 /* 444 * Message type specific processing. 445 */ 446 if (icmpprintfs) 447 printf("icmp_input(type:%d, code:%d)\n", icp->icmp_type, 448 icp->icmp_code); 449 #endif 450 if (icp->icmp_type > ICMP_MAXTYPE) 451 goto raw; 452 icmpstat.icps_inhist[icp->icmp_type]++; 453 code = icp->icmp_code; 454 switch (icp->icmp_type) { 455 456 case ICMP_UNREACH: 457 switch (code) { 458 case ICMP_UNREACH_NET: 459 code = PRC_UNREACH_NET; 460 break; 461 462 case ICMP_UNREACH_HOST: 463 code = PRC_UNREACH_HOST; 464 break; 465 466 case ICMP_UNREACH_PROTOCOL: 467 code = PRC_UNREACH_PROTOCOL; 468 break; 469 470 case ICMP_UNREACH_PORT: 471 code = PRC_UNREACH_PORT; 472 break; 473 474 case ICMP_UNREACH_SRCFAIL: 475 code = PRC_UNREACH_SRCFAIL; 476 break; 477 478 case ICMP_UNREACH_NEEDFRAG: 479 code = PRC_MSGSIZE; 480 break; 481 482 case ICMP_UNREACH_NET_UNKNOWN: 483 case ICMP_UNREACH_NET_PROHIB: 484 case ICMP_UNREACH_TOSNET: 485 code = PRC_UNREACH_NET; 486 break; 487 488 case ICMP_UNREACH_HOST_UNKNOWN: 489 case ICMP_UNREACH_ISOLATED: 490 case ICMP_UNREACH_HOST_PROHIB: 491 case ICMP_UNREACH_TOSHOST: 492 code = PRC_UNREACH_HOST; 493 break; 494 495 default: 496 goto badcode; 497 } 498 goto deliver; 499 500 case ICMP_TIMXCEED: 501 if (code > 1) 502 goto badcode; 503 code += PRC_TIMXCEED_INTRANS; 504 goto deliver; 505 506 case ICMP_PARAMPROB: 507 if (code > 1) 508 goto badcode; 509 code = PRC_PARAMPROB; 510 goto deliver; 511 512 case ICMP_SOURCEQUENCH: 513 if (code) 514 goto badcode; 515 code = PRC_QUENCH; 516 goto deliver; 517 518 deliver: 519 /* 520 * Problem with datagram; advise higher level routines. 521 */ 522 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 523 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 524 icmpstat.icps_badlen++; 525 goto freeit; 526 } 527 if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr)) 528 goto badcode; 529 #ifdef ICMPPRINTFS 530 if (icmpprintfs) 531 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 532 #endif 533 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 534 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 535 if (ctlfunc) 536 (void) (*ctlfunc)(code, sintosa(&icmpsrc), 537 &icp->icmp_ip); 538 break; 539 540 badcode: 541 icmpstat.icps_badcode++; 542 break; 543 544 case ICMP_ECHO: 545 icp->icmp_type = ICMP_ECHOREPLY; 546 goto reflect; 547 548 case ICMP_TSTAMP: 549 if (icmplen < ICMP_TSLEN) { 550 icmpstat.icps_badlen++; 551 break; 552 } 553 icp->icmp_type = ICMP_TSTAMPREPLY; 554 icp->icmp_rtime = iptime(); 555 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 556 goto reflect; 557 558 case ICMP_MASKREQ: 559 if (icmpmaskrepl == 0) 560 break; 561 /* 562 * We are not able to respond with all ones broadcast 563 * unless we receive it over a point-to-point interface. 564 */ 565 if (icmplen < ICMP_MASKLEN) { 566 icmpstat.icps_badlen++; 567 break; 568 } 569 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 570 in_nullhost(ip->ip_dst)) 571 icmpdst.sin_addr = ip->ip_src; 572 else 573 icmpdst.sin_addr = ip->ip_dst; 574 ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst), 575 m->m_pkthdr.rcvif)); 576 if (ia == 0) 577 break; 578 icp->icmp_type = ICMP_MASKREPLY; 579 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 580 if (in_nullhost(ip->ip_src)) { 581 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 582 ip->ip_src = ia->ia_broadaddr.sin_addr; 583 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 584 ip->ip_src = ia->ia_dstaddr.sin_addr; 585 } 586 reflect: 587 icmpstat.icps_reflect++; 588 icmpstat.icps_outhist[icp->icmp_type]++; 589 icmp_reflect(m); 590 return; 591 592 case ICMP_REDIRECT: 593 if (code > 3) 594 goto badcode; 595 if (icmp_rediraccept == 0) 596 goto freeit; 597 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 598 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 599 icmpstat.icps_badlen++; 600 break; 601 } 602 /* 603 * Short circuit routing redirects to force 604 * immediate change in the kernel's routing 605 * tables. The message is also handed to anyone 606 * listening on a raw socket (e.g. the routing 607 * daemon for use in updating its tables). 608 */ 609 icmpgw.sin_addr = ip->ip_src; 610 icmpdst.sin_addr = icp->icmp_gwaddr; 611 #ifdef ICMPPRINTFS 612 if (icmpprintfs) { 613 printf("redirect dst `%s' to `%s'\n", 614 inet_ntoa(icp->icmp_ip.ip_dst), 615 inet_ntoa(icp->icmp_gwaddr)); 616 } 617 #endif 618 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 619 rt = NULL; 620 rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst), 621 (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST, 622 sintosa(&icmpgw), (struct rtentry **)&rt); 623 if (rt != NULL && icmp_redirtimeout != 0) { 624 i = rt_timer_add(rt, icmp_redirect_timeout, 625 icmp_redirect_timeout_q); 626 if (i) 627 log(LOG_ERR, "ICMP: redirect failed to " 628 "register timeout for route to %x, " 629 "code %d\n", 630 icp->icmp_ip.ip_dst.s_addr, i); 631 } 632 if (rt != NULL) 633 rtfree(rt); 634 635 pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc)); 636 #if defined(IPSEC) || defined(FAST_IPSEC) 637 key_sa_routechange((struct sockaddr *)&icmpsrc); 638 #endif 639 break; 640 641 /* 642 * No kernel processing for the following; 643 * just fall through to send to raw listener. 644 */ 645 case ICMP_ECHOREPLY: 646 case ICMP_ROUTERADVERT: 647 case ICMP_ROUTERSOLICIT: 648 case ICMP_TSTAMPREPLY: 649 case ICMP_IREQREPLY: 650 case ICMP_MASKREPLY: 651 default: 652 break; 653 } 654 655 raw: 656 rip_input(m, hlen, proto); 657 return; 658 659 freeit: 660 m_freem(m); 661 return; 662 } 663 664 /* 665 * Reflect the ip packet back to the source 666 */ 667 void 668 icmp_reflect(struct mbuf *m) 669 { 670 struct ip *ip = mtod(m, struct ip *); 671 struct in_ifaddr *ia; 672 struct ifaddr *ifa; 673 struct sockaddr_in *sin = 0; 674 struct in_addr t; 675 struct mbuf *opts = 0; 676 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 677 678 if (!in_canforward(ip->ip_src) && 679 ((ip->ip_src.s_addr & IN_CLASSA_NET) != 680 htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { 681 m_freem(m); /* Bad return address */ 682 goto done; /* ip_output() will check for broadcast */ 683 } 684 t = ip->ip_dst; 685 ip->ip_dst = ip->ip_src; 686 /* 687 * If the incoming packet was addressed directly to us, use 688 * dst as the src for the reply. Otherwise (broadcast or 689 * anonymous), use an address which corresponds to the 690 * incoming interface, with a preference for the address which 691 * corresponds to the route to the destination of the ICMP. 692 */ 693 694 /* Look for packet addressed to us */ 695 INADDR_TO_IA(t, ia); 696 697 /* look for packet sent to broadcast address */ 698 if (ia == NULL && m->m_pkthdr.rcvif && 699 (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST)) { 700 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 701 if (ifa->ifa_addr->sa_family != AF_INET) 702 continue; 703 if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) { 704 ia = ifatoia(ifa); 705 break; 706 } 707 } 708 } 709 710 if (ia) 711 sin = &ia->ia_addr; 712 713 icmpdst.sin_addr = t; 714 715 /* 716 * if the packet is addressed somewhere else, compute the 717 * source address for packets routed back to the source, and 718 * use that, if it's an address on the interface which 719 * received the packet 720 */ 721 if (sin == (struct sockaddr_in *)0 && m->m_pkthdr.rcvif) { 722 struct sockaddr_in sin_dst; 723 struct route icmproute; 724 int errornum; 725 726 sin_dst.sin_family = AF_INET; 727 sin_dst.sin_len = sizeof(struct sockaddr_in); 728 sin_dst.sin_addr = ip->ip_dst; 729 bzero(&icmproute, sizeof(icmproute)); 730 errornum = 0; 731 sin = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum); 732 /* errornum is never used */ 733 if (icmproute.ro_rt) 734 RTFREE(icmproute.ro_rt); 735 /* check to make sure sin is a source address on rcvif */ 736 if (sin) { 737 t = sin->sin_addr; 738 sin = (struct sockaddr_in *)0; 739 INADDR_TO_IA(t, ia); 740 while (ia) { 741 if (ia->ia_ifp == m->m_pkthdr.rcvif) { 742 sin = &ia->ia_addr; 743 break; 744 } 745 NEXT_IA_WITH_SAME_ADDR(ia); 746 } 747 } 748 } 749 750 /* 751 * if it was not addressed to us, but the route doesn't go out 752 * the source interface, pick an address on the source 753 * interface. This can happen when routing is asymmetric, or 754 * when the incoming packet was encapsulated 755 */ 756 if (sin == (struct sockaddr_in *)0 && m->m_pkthdr.rcvif) { 757 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 758 if (ifa->ifa_addr->sa_family != AF_INET) 759 continue; 760 sin = &(ifatoia(ifa)->ia_addr); 761 break; 762 } 763 } 764 765 /* 766 * The following happens if the packet was not addressed to us, 767 * and was received on an interface with no IP address: 768 * We find the first AF_INET address on the first non-loopback 769 * interface. 770 */ 771 if (sin == (struct sockaddr_in *)0) 772 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) { 773 if (ia->ia_ifp->if_flags & IFF_LOOPBACK) 774 continue; 775 sin = &ia->ia_addr; 776 break; 777 } 778 779 /* 780 * If we still didn't find an address, punt. We could have an 781 * interface up (and receiving packets) with no address. 782 */ 783 if (sin == (struct sockaddr_in *)0) { 784 m_freem(m); 785 goto done; 786 } 787 788 ip->ip_src = sin->sin_addr; 789 ip->ip_ttl = MAXTTL; 790 791 if (optlen > 0) { 792 u_char *cp; 793 int opt, cnt; 794 u_int len; 795 796 /* 797 * Retrieve any source routing from the incoming packet; 798 * add on any record-route or timestamp options. 799 */ 800 cp = (u_char *) (ip + 1); 801 if ((opts = ip_srcroute()) == 0 && 802 (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { 803 MCLAIM(opts, m->m_owner); 804 opts->m_len = sizeof(struct in_addr); 805 *mtod(opts, struct in_addr *) = zeroin_addr; 806 } 807 if (opts) { 808 #ifdef ICMPPRINTFS 809 if (icmpprintfs) 810 printf("icmp_reflect optlen %d rt %d => ", 811 optlen, opts->m_len); 812 #endif 813 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 814 opt = cp[IPOPT_OPTVAL]; 815 if (opt == IPOPT_EOL) 816 break; 817 if (opt == IPOPT_NOP) 818 len = 1; 819 else { 820 if (cnt < IPOPT_OLEN + sizeof(*cp)) 821 break; 822 len = cp[IPOPT_OLEN]; 823 if (len < IPOPT_OLEN + sizeof(*cp) || 824 len > cnt) 825 break; 826 } 827 /* 828 * Should check for overflow, but it "can't happen" 829 */ 830 if (opt == IPOPT_RR || opt == IPOPT_TS || 831 opt == IPOPT_SECURITY) { 832 bcopy((caddr_t)cp, 833 mtod(opts, caddr_t) + opts->m_len, len); 834 opts->m_len += len; 835 } 836 } 837 /* Terminate & pad, if necessary */ 838 if ((cnt = opts->m_len % 4) != 0) { 839 for (; cnt < 4; cnt++) { 840 *(mtod(opts, caddr_t) + opts->m_len) = 841 IPOPT_EOL; 842 opts->m_len++; 843 } 844 } 845 #ifdef ICMPPRINTFS 846 if (icmpprintfs) 847 printf("%d\n", opts->m_len); 848 #endif 849 } 850 /* 851 * Now strip out original options by copying rest of first 852 * mbuf's data back, and adjust the IP length. 853 */ 854 ip->ip_len = htons(ntohs(ip->ip_len) - optlen); 855 ip->ip_hl = sizeof(struct ip) >> 2; 856 m->m_len -= optlen; 857 if (m->m_flags & M_PKTHDR) 858 m->m_pkthdr.len -= optlen; 859 optlen += sizeof(struct ip); 860 bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1), 861 (unsigned)(m->m_len - sizeof(struct ip))); 862 } 863 m_tag_delete_nonpersistent(m); 864 m->m_flags &= ~(M_BCAST|M_MCAST); 865 866 /* 867 * Clear any in-bound checksum flags for this packet. 868 */ 869 if (m->m_flags & M_PKTHDR) 870 m->m_pkthdr.csum_flags = 0; 871 872 icmp_send(m, opts); 873 done: 874 if (opts) 875 (void)m_free(opts); 876 } 877 878 /* 879 * Send an icmp packet back to the ip level, 880 * after supplying a checksum. 881 */ 882 void 883 icmp_send(struct mbuf *m, struct mbuf *opts) 884 { 885 struct ip *ip = mtod(m, struct ip *); 886 int hlen; 887 struct icmp *icp; 888 889 hlen = ip->ip_hl << 2; 890 m->m_data += hlen; 891 m->m_len -= hlen; 892 icp = mtod(m, struct icmp *); 893 icp->icmp_cksum = 0; 894 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 895 m->m_data -= hlen; 896 m->m_len += hlen; 897 #ifdef ICMPPRINTFS 898 if (icmpprintfs) { 899 printf("icmp_send to destination `%s' from `%s'\n", 900 inet_ntoa(ip->ip_dst), inet_ntoa(ip->ip_src)); 901 } 902 #endif 903 (void) ip_output(m, opts, NULL, 0, 904 (struct ip_moptions *)NULL, (struct socket *)NULL); 905 } 906 907 n_time 908 iptime(void) 909 { 910 struct timeval atv; 911 u_long t; 912 913 microtime(&atv); 914 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 915 return (htonl(t)); 916 } 917 918 /* 919 * sysctl helper routine for net.inet.icmp.returndatabytes. ensures 920 * that the new value is in the correct range. 921 */ 922 static int 923 sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS) 924 { 925 int error, t; 926 struct sysctlnode node; 927 928 node = *rnode; 929 node.sysctl_data = &t; 930 t = icmpreturndatabytes; 931 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 932 if (error || newp == NULL) 933 return (error); 934 935 if (t < 8 || t > 512) 936 return (EINVAL); 937 icmpreturndatabytes = t; 938 939 return (0); 940 } 941 942 /* 943 * sysctl helper routine for net.inet.icmp.redirtimeout. ensures that 944 * the given value is not less than zero and then resets the timeout 945 * queue. 946 */ 947 static int 948 sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS) 949 { 950 int error, tmp; 951 struct sysctlnode node; 952 953 node = *rnode; 954 node.sysctl_data = &tmp; 955 tmp = icmp_redirtimeout; 956 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 957 if (error || newp == NULL) 958 return (error); 959 if (tmp < 0) 960 return (EINVAL); 961 icmp_redirtimeout = tmp; 962 963 /* 964 * was it a *defined* side-effect that anyone even *reading* 965 * this value causes these things to happen? 966 */ 967 if (icmp_redirect_timeout_q != NULL) { 968 if (icmp_redirtimeout == 0) { 969 rt_timer_queue_destroy(icmp_redirect_timeout_q, 970 TRUE); 971 icmp_redirect_timeout_q = NULL; 972 } else { 973 rt_timer_queue_change(icmp_redirect_timeout_q, 974 icmp_redirtimeout); 975 } 976 } else if (icmp_redirtimeout > 0) { 977 icmp_redirect_timeout_q = 978 rt_timer_queue_create(icmp_redirtimeout); 979 } 980 981 return (0); 982 } 983 984 SYSCTL_SETUP(sysctl_net_inet_icmp_setup, "sysctl net.inet.icmp subtree setup") 985 { 986 987 sysctl_createv(clog, 0, NULL, NULL, 988 CTLFLAG_PERMANENT, 989 CTLTYPE_NODE, "net", NULL, 990 NULL, 0, NULL, 0, 991 CTL_NET, CTL_EOL); 992 sysctl_createv(clog, 0, NULL, NULL, 993 CTLFLAG_PERMANENT, 994 CTLTYPE_NODE, "inet", NULL, 995 NULL, 0, NULL, 0, 996 CTL_NET, PF_INET, CTL_EOL); 997 sysctl_createv(clog, 0, NULL, NULL, 998 CTLFLAG_PERMANENT, 999 CTLTYPE_NODE, "icmp", 1000 SYSCTL_DESCR("ICMPv4 related settings"), 1001 NULL, 0, NULL, 0, 1002 CTL_NET, PF_INET, IPPROTO_ICMP, CTL_EOL); 1003 1004 sysctl_createv(clog, 0, NULL, NULL, 1005 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1006 CTLTYPE_INT, "maskrepl", 1007 SYSCTL_DESCR("Respond to ICMP_MASKREQ messages"), 1008 NULL, 0, &icmpmaskrepl, 0, 1009 CTL_NET, PF_INET, IPPROTO_ICMP, 1010 ICMPCTL_MASKREPL, CTL_EOL); 1011 sysctl_createv(clog, 0, NULL, NULL, 1012 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1013 CTLTYPE_INT, "returndatabytes", 1014 SYSCTL_DESCR("Number of bytes to return in an ICMP " 1015 "error message"), 1016 sysctl_net_inet_icmp_returndatabytes, 0, 1017 &icmpreturndatabytes, 0, 1018 CTL_NET, PF_INET, IPPROTO_ICMP, 1019 ICMPCTL_RETURNDATABYTES, CTL_EOL); 1020 sysctl_createv(clog, 0, NULL, NULL, 1021 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1022 CTLTYPE_INT, "errppslimit", 1023 SYSCTL_DESCR("Maximum number of outgoing ICMP error " 1024 "messages per second"), 1025 NULL, 0, &icmperrppslim, 0, 1026 CTL_NET, PF_INET, IPPROTO_ICMP, 1027 ICMPCTL_ERRPPSLIMIT, CTL_EOL); 1028 sysctl_createv(clog, 0, NULL, NULL, 1029 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1030 CTLTYPE_INT, "rediraccept", 1031 SYSCTL_DESCR("Accept ICMP_REDIRECT messages"), 1032 NULL, 0, &icmp_rediraccept, 0, 1033 CTL_NET, PF_INET, IPPROTO_ICMP, 1034 ICMPCTL_REDIRACCEPT, CTL_EOL); 1035 sysctl_createv(clog, 0, NULL, NULL, 1036 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1037 CTLTYPE_INT, "redirtimeout", 1038 SYSCTL_DESCR("Lifetime of ICMP_REDIRECT generated " 1039 "routes"), 1040 sysctl_net_inet_icmp_redirtimeout, 0, 1041 &icmp_redirtimeout, 0, 1042 CTL_NET, PF_INET, IPPROTO_ICMP, 1043 ICMPCTL_REDIRTIMEOUT, CTL_EOL); 1044 sysctl_createv(clog, 0, NULL, NULL, 1045 CTLFLAG_PERMANENT, 1046 CTLTYPE_STRUCT, "stats", 1047 SYSCTL_DESCR("ICMP statistics"), 1048 NULL, 0, &icmpstat, sizeof(icmpstat), 1049 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_STATS, 1050 CTL_EOL); 1051 } 1052 1053 /* Table of common MTUs: */ 1054 1055 static const u_int mtu_table[] = { 1056 65535, 65280, 32000, 17914, 9180, 8166, 1057 4352, 2002, 1492, 1006, 508, 296, 68, 0 1058 }; 1059 1060 void 1061 icmp_mtudisc(struct icmp *icp, struct in_addr faddr) 1062 { 1063 struct icmp_mtudisc_callback *mc; 1064 struct sockaddr *dst = sintosa(&icmpsrc); 1065 struct rtentry *rt; 1066 u_long mtu = ntohs(icp->icmp_nextmtu); /* Why a long? IPv6 */ 1067 int error; 1068 1069 rt = rtalloc1(dst, 1); 1070 if (rt == 0) 1071 return; 1072 1073 /* If we didn't get a host route, allocate one */ 1074 1075 if ((rt->rt_flags & RTF_HOST) == 0) { 1076 struct rtentry *nrt; 1077 1078 error = rtrequest((int) RTM_ADD, dst, 1079 (struct sockaddr *) rt->rt_gateway, 1080 (struct sockaddr *) 0, 1081 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 1082 if (error) { 1083 rtfree(rt); 1084 return; 1085 } 1086 nrt->rt_rmx = rt->rt_rmx; 1087 rtfree(rt); 1088 rt = nrt; 1089 } 1090 error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q); 1091 if (error) { 1092 rtfree(rt); 1093 return; 1094 } 1095 1096 if (mtu == 0) { 1097 int i = 0; 1098 1099 mtu = ntohs(icp->icmp_ip.ip_len); 1100 /* Some 4.2BSD-based routers incorrectly adjust the ip_len */ 1101 if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0) 1102 mtu -= (icp->icmp_ip.ip_hl << 2); 1103 1104 /* If we still can't guess a value, try the route */ 1105 1106 if (mtu == 0) { 1107 mtu = rt->rt_rmx.rmx_mtu; 1108 1109 /* If no route mtu, default to the interface mtu */ 1110 1111 if (mtu == 0) 1112 mtu = rt->rt_ifp->if_mtu; 1113 } 1114 1115 for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++) 1116 if (mtu > mtu_table[i]) { 1117 mtu = mtu_table[i]; 1118 break; 1119 } 1120 } 1121 1122 /* 1123 * XXX: RTV_MTU is overloaded, since the admin can set it 1124 * to turn off PMTU for a route, and the kernel can 1125 * set it to indicate a serious problem with PMTU 1126 * on a route. We should be using a separate flag 1127 * for the kernel to indicate this. 1128 */ 1129 1130 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1131 if (mtu < 296 || mtu > rt->rt_ifp->if_mtu) 1132 rt->rt_rmx.rmx_locks |= RTV_MTU; 1133 else if (rt->rt_rmx.rmx_mtu > mtu || 1134 rt->rt_rmx.rmx_mtu == 0) { 1135 icmpstat.icps_pmtuchg++; 1136 rt->rt_rmx.rmx_mtu = mtu; 1137 } 1138 } 1139 1140 if (rt) 1141 rtfree(rt); 1142 1143 /* 1144 * Notify protocols that the MTU for this destination 1145 * has changed. 1146 */ 1147 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL; 1148 mc = LIST_NEXT(mc, mc_list)) 1149 (*mc->mc_func)(faddr); 1150 } 1151 1152 /* 1153 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1154 * given current value MTU. If DIR is less than zero, a larger plateau 1155 * is returned; otherwise, a smaller value is returned. 1156 */ 1157 u_int 1158 ip_next_mtu(u_int mtu, int dir) /* XXX */ 1159 { 1160 int i; 1161 1162 for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) { 1163 if (mtu >= mtu_table[i]) 1164 break; 1165 } 1166 1167 if (dir < 0) { 1168 if (i == 0) { 1169 return 0; 1170 } else { 1171 return mtu_table[i - 1]; 1172 } 1173 } else { 1174 if (mtu_table[i] == 0) { 1175 return 0; 1176 } else if (mtu > mtu_table[i]) { 1177 return mtu_table[i]; 1178 } else { 1179 return mtu_table[i + 1]; 1180 } 1181 } 1182 } 1183 1184 static void 1185 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r __unused) 1186 { 1187 if (rt == NULL) 1188 panic("icmp_mtudisc_timeout: bad route to timeout"); 1189 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1190 (RTF_DYNAMIC | RTF_HOST)) { 1191 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 1192 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 1193 } else { 1194 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1195 rt->rt_rmx.rmx_mtu = 0; 1196 } 1197 } 1198 } 1199 1200 static void 1201 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r __unused) 1202 { 1203 if (rt == NULL) 1204 panic("icmp_redirect_timeout: bad route to timeout"); 1205 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1206 (RTF_DYNAMIC | RTF_HOST)) { 1207 rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt), 1208 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 1209 } 1210 } 1211 1212 /* 1213 * Perform rate limit check. 1214 * Returns 0 if it is okay to send the icmp packet. 1215 * Returns 1 if the router SHOULD NOT send this icmp packet due to rate 1216 * limitation. 1217 * 1218 * XXX per-destination/type check necessary? 1219 */ 1220 static int 1221 icmp_ratelimit(const struct in_addr *dst __unused, const int type __unused, 1222 const int code __unused) 1223 { 1224 1225 /* PPS limit */ 1226 if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count, 1227 icmperrppslim)) { 1228 /* The packet is subject to rate limit */ 1229 return 1; 1230 } 1231 1232 /* okay to send */ 1233 return 0; 1234 } 1235