1 /* $OpenBSD: ip_icmp.c,v 1.151 2015/12/09 09:27:40 mpi Exp $ */ 2 /* $NetBSD: ip_icmp.c,v 1.19 1996/02/13 23:42:22 christos Exp $ */ 3 4 /* 5 * Copyright (c) 1982, 1986, 1988, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)COPYRIGHT 1.1 (NRL) 17 January 1995 33 * 34 * NRL grants permission for redistribution and use in source and binary 35 * forms, with or without modification, of the software and documentation 36 * created at NRL provided that the following conditions are met: 37 * 38 * 1. Redistributions of source code must retain the above copyright 39 * notice, this list of conditions and the following disclaimer. 40 * 2. Redistributions in binary form must reproduce the above copyright 41 * notice, this list of conditions and the following disclaimer in the 42 * documentation and/or other materials provided with the distribution. 43 * 3. All advertising materials mentioning features or use of this software 44 * must display the following acknowledgements: 45 * This product includes software developed by the University of 46 * California, Berkeley and its contributors. 47 * This product includes software developed at the Information 48 * Technology Division, US Naval Research Laboratory. 49 * 4. Neither the name of the NRL nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS 54 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 55 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A 56 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR 57 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 58 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 59 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 60 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 61 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 62 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 63 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 64 * 65 * The views and conclusions contained in the software and documentation 66 * are those of the authors and should not be interpreted as representing 67 * official policies, either expressed or implied, of the US Naval 68 * Research Laboratory (NRL). 69 */ 70 71 #include "carp.h" 72 #include "pf.h" 73 74 #include <sys/param.h> 75 #include <sys/systm.h> 76 #include <sys/mbuf.h> 77 #include <sys/protosw.h> 78 #include <sys/socket.h> 79 #include <sys/sysctl.h> 80 81 #include <net/if.h> 82 #include <net/if_var.h> 83 #include <net/route.h> 84 85 #include <netinet/in.h> 86 #include <netinet/in_systm.h> 87 #include <netinet/in_var.h> 88 #include <netinet/ip.h> 89 #include <netinet/ip_icmp.h> 90 #include <netinet/ip_var.h> 91 #include <netinet/icmp_var.h> 92 93 #if NCARP > 0 94 #include <net/if_types.h> 95 #include <netinet/ip_carp.h> 96 #endif 97 98 #if NPF > 0 99 #include <net/pfvar.h> 100 #endif 101 102 /* 103 * ICMP routines: error generation, receive packet processing, and 104 * routines to turnaround packets back to the originator, and 105 * host table maintenance routines. 106 */ 107 108 #ifdef ICMPPRINTFS 109 int icmpprintfs = 0; /* Settable from ddb */ 110 #endif 111 112 /* values controllable via sysctl */ 113 int icmpmaskrepl = 0; 114 int icmpbmcastecho = 0; 115 int icmptstamprepl = 1; 116 int icmperrppslim = 100; 117 int icmp_rediraccept = 0; 118 int icmp_redirtimeout = 10 * 60; 119 120 static int icmperrpps_count = 0; 121 static struct timeval icmperrppslim_last; 122 123 static struct rttimer_queue *icmp_redirect_timeout_q = NULL; 124 struct icmpstat icmpstat; 125 126 int *icmpctl_vars[ICMPCTL_MAXID] = ICMPCTL_VARS; 127 128 void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *); 129 int icmp_ratelimit(const struct in_addr *, const int, const int); 130 void icmp_redirect_timeout(struct rtentry *, struct rttimer *); 131 void icmp_input_if(struct ifnet *, struct mbuf *, int); 132 133 void 134 icmp_init(void) 135 { 136 /* 137 * This is only useful if the user initializes redirtimeout to 138 * something other than zero. 139 */ 140 if (icmp_redirtimeout != 0) { 141 icmp_redirect_timeout_q = 142 rt_timer_queue_create(icmp_redirtimeout); 143 } 144 } 145 146 struct mbuf * 147 icmp_do_error(struct mbuf *n, int type, int code, u_int32_t dest, int destmtu) 148 { 149 struct ip *oip = mtod(n, struct ip *), *nip; 150 unsigned oiplen = oip->ip_hl << 2; 151 struct icmp *icp; 152 struct mbuf *m; 153 unsigned icmplen, mblen; 154 155 #ifdef ICMPPRINTFS 156 if (icmpprintfs) 157 printf("icmp_error(%x, %d, %d)\n", oip, type, code); 158 #endif 159 if (type != ICMP_REDIRECT) 160 icmpstat.icps_error++; 161 /* 162 * Don't send error if not the first fragment of message. 163 * Don't error if the old packet protocol was ICMP 164 * error message, only known informational types. 165 */ 166 if (oip->ip_off & htons(IP_OFFMASK)) 167 goto freeit; 168 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 169 n->m_len >= oiplen + ICMP_MINLEN && 170 !ICMP_INFOTYPE(((struct icmp *) 171 ((caddr_t)oip + oiplen))->icmp_type)) { 172 icmpstat.icps_oldicmp++; 173 goto freeit; 174 } 175 /* Don't send error in response to a multicast or broadcast packet */ 176 if (n->m_flags & (M_BCAST|M_MCAST)) 177 goto freeit; 178 179 /* 180 * First, do a rate limitation check. 181 */ 182 if (icmp_ratelimit(&oip->ip_src, type, code)) { 183 icmpstat.icps_toofreq++; 184 goto freeit; 185 } 186 187 /* 188 * Now, formulate icmp message 189 */ 190 icmplen = oiplen + min(8, ntohs(oip->ip_len)); 191 /* 192 * Defend against mbuf chains shorter than oip->ip_len: 193 */ 194 mblen = 0; 195 for (m = n; m && (mblen < icmplen); m = m->m_next) 196 mblen += m->m_len; 197 icmplen = min(mblen, icmplen); 198 199 /* 200 * As we are not required to return everything we have, 201 * we return whatever we can return at ease. 202 * 203 * Note that ICMP datagrams longer than 576 octets are out of spec 204 * according to RFC1812; 205 */ 206 207 KASSERT(ICMP_MINLEN <= MCLBYTES); 208 209 if (icmplen + ICMP_MINLEN > MCLBYTES) 210 icmplen = MCLBYTES - ICMP_MINLEN - sizeof (struct ip); 211 212 m = m_gethdr(M_DONTWAIT, MT_HEADER); 213 if (m && (sizeof (struct ip) + icmplen + ICMP_MINLEN > MHLEN)) { 214 MCLGET(m, M_DONTWAIT); 215 if ((m->m_flags & M_EXT) == 0) { 216 m_freem(m); 217 m = NULL; 218 } 219 } 220 if (m == NULL) 221 goto freeit; 222 /* keep in same rtable */ 223 m->m_pkthdr.ph_rtableid = n->m_pkthdr.ph_rtableid; 224 m->m_len = icmplen + ICMP_MINLEN; 225 if ((m->m_flags & M_EXT) == 0) 226 MH_ALIGN(m, m->m_len); 227 icp = mtod(m, struct icmp *); 228 if ((u_int)type > ICMP_MAXTYPE) 229 panic("icmp_error"); 230 icmpstat.icps_outhist[type]++; 231 icp->icmp_type = type; 232 if (type == ICMP_REDIRECT) 233 icp->icmp_gwaddr.s_addr = dest; 234 else { 235 icp->icmp_void = 0; 236 /* 237 * The following assignments assume an overlay with the 238 * zeroed icmp_void field. 239 */ 240 if (type == ICMP_PARAMPROB) { 241 icp->icmp_pptr = code; 242 code = 0; 243 } else if (type == ICMP_UNREACH && 244 code == ICMP_UNREACH_NEEDFRAG && destmtu) 245 icp->icmp_nextmtu = htons(destmtu); 246 } 247 248 icp->icmp_code = code; 249 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 250 251 /* 252 * Now, copy old ip header (without options) 253 * in front of icmp message. 254 */ 255 if ((m->m_flags & M_EXT) == 0 && 256 m->m_data - sizeof(struct ip) < m->m_pktdat) 257 panic("icmp len"); 258 m->m_data -= sizeof(struct ip); 259 m->m_len += sizeof(struct ip); 260 m->m_pkthdr.len = m->m_len; 261 m->m_pkthdr.ph_ifidx = n->m_pkthdr.ph_ifidx; 262 nip = mtod(m, struct ip *); 263 /* ip_v set in ip_output */ 264 nip->ip_hl = sizeof(struct ip) >> 2; 265 nip->ip_tos = 0; 266 nip->ip_len = htons(m->m_len); 267 /* ip_id set in ip_output */ 268 nip->ip_off = 0; 269 /* ip_ttl set in icmp_reflect */ 270 nip->ip_p = IPPROTO_ICMP; 271 nip->ip_src = oip->ip_src; 272 nip->ip_dst = oip->ip_dst; 273 274 /* move PF_GENERATED to new packet, if existent XXX preserve more? */ 275 if (n->m_pkthdr.pf.flags & PF_TAG_GENERATED) 276 m->m_pkthdr.pf.flags |= PF_TAG_GENERATED; 277 278 m_freem(n); 279 return (m); 280 281 freeit: 282 m_freem(n); 283 return (NULL); 284 } 285 286 /* 287 * Generate an error packet of type error 288 * in response to bad packet ip. 289 * 290 * The ip packet inside has ip_off and ip_len in host byte order. 291 */ 292 void 293 icmp_error(struct mbuf *n, int type, int code, u_int32_t dest, int destmtu) 294 { 295 struct mbuf *m; 296 297 m = icmp_do_error(n, type, code, dest, destmtu); 298 if (m != NULL) 299 if (!icmp_reflect(m, NULL, NULL)) 300 icmp_send(m, NULL); 301 } 302 303 /* 304 * Process a received ICMP message. 305 */ 306 void 307 icmp_input(struct mbuf *m, ...) 308 { 309 struct ifnet *ifp; 310 int hlen; 311 va_list ap; 312 313 va_start(ap, m); 314 hlen = va_arg(ap, int); 315 va_end(ap); 316 317 ifp = if_get(m->m_pkthdr.ph_ifidx); 318 if (ifp == NULL) { 319 m_freem(m); 320 return; 321 } 322 323 icmp_input_if(ifp, m, hlen); 324 if_put(ifp); 325 } 326 327 void 328 icmp_input_if(struct ifnet *ifp, struct mbuf *m, int hlen) 329 { 330 struct icmp *icp; 331 struct ip *ip = mtod(m, struct ip *); 332 struct sockaddr_in sin; 333 int icmplen, i, code; 334 struct in_ifaddr *ia; 335 void *(*ctlfunc)(int, struct sockaddr *, u_int, void *); 336 struct mbuf *opts; 337 338 /* 339 * Locate icmp structure in mbuf, and check 340 * that not corrupted and of at least minimum length. 341 */ 342 icmplen = ntohs(ip->ip_len) - hlen; 343 #ifdef ICMPPRINTFS 344 if (icmpprintfs) { 345 char dst[INET_ADDRSTRLEN], src[INET_ADDRSTRLEN]; 346 347 inet_ntop(AF_INET, &ip->ip_dst, dst, sizeof(dst)); 348 inet_ntop(AF_INET, &ip->ip_src, src, sizeof(src)); 349 350 printf("icmp_input from %s to %s, len %d\n", src, dst, icmplen); 351 } 352 #endif 353 if (icmplen < ICMP_MINLEN) { 354 icmpstat.icps_tooshort++; 355 goto freeit; 356 } 357 i = hlen + min(icmplen, ICMP_ADVLENMIN); 358 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 359 icmpstat.icps_tooshort++; 360 return; 361 } 362 ip = mtod(m, struct ip *); 363 if (in4_cksum(m, 0, hlen, icmplen)) { 364 icmpstat.icps_checksum++; 365 goto freeit; 366 } 367 368 icp = (struct icmp *)(mtod(m, caddr_t) + hlen); 369 #ifdef ICMPPRINTFS 370 /* 371 * Message type specific processing. 372 */ 373 if (icmpprintfs) 374 printf("icmp_input, type %d code %d\n", icp->icmp_type, 375 icp->icmp_code); 376 #endif 377 if (icp->icmp_type > ICMP_MAXTYPE) 378 goto raw; 379 #if NPF > 0 380 if (m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) { 381 switch (icp->icmp_type) { 382 /* 383 * These ICMP types map to other connections. They must be 384 * delivered to pr_ctlinput() also for diverted connections. 385 */ 386 case ICMP_UNREACH: 387 case ICMP_TIMXCEED: 388 case ICMP_PARAMPROB: 389 case ICMP_SOURCEQUENCH: 390 break; 391 /* 392 * Although pf_icmp_mapping() considers redirects belonging 393 * to a diverted connection, we must process it here anyway. 394 */ 395 case ICMP_REDIRECT: 396 break; 397 default: 398 goto raw; 399 } 400 } 401 #endif /* NPF */ 402 icmpstat.icps_inhist[icp->icmp_type]++; 403 code = icp->icmp_code; 404 switch (icp->icmp_type) { 405 406 case ICMP_UNREACH: 407 switch (code) { 408 case ICMP_UNREACH_NET: 409 case ICMP_UNREACH_HOST: 410 case ICMP_UNREACH_PROTOCOL: 411 case ICMP_UNREACH_PORT: 412 case ICMP_UNREACH_SRCFAIL: 413 code += PRC_UNREACH_NET; 414 break; 415 416 case ICMP_UNREACH_NEEDFRAG: 417 code = PRC_MSGSIZE; 418 break; 419 420 case ICMP_UNREACH_NET_UNKNOWN: 421 case ICMP_UNREACH_NET_PROHIB: 422 case ICMP_UNREACH_TOSNET: 423 code = PRC_UNREACH_NET; 424 break; 425 426 case ICMP_UNREACH_HOST_UNKNOWN: 427 case ICMP_UNREACH_ISOLATED: 428 case ICMP_UNREACH_HOST_PROHIB: 429 case ICMP_UNREACH_TOSHOST: 430 case ICMP_UNREACH_FILTER_PROHIB: 431 case ICMP_UNREACH_HOST_PRECEDENCE: 432 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 433 code = PRC_UNREACH_HOST; 434 break; 435 436 default: 437 goto badcode; 438 } 439 goto deliver; 440 441 case ICMP_TIMXCEED: 442 if (code > 1) 443 goto badcode; 444 code += PRC_TIMXCEED_INTRANS; 445 goto deliver; 446 447 case ICMP_PARAMPROB: 448 if (code > 1) 449 goto badcode; 450 code = PRC_PARAMPROB; 451 goto deliver; 452 453 case ICMP_SOURCEQUENCH: 454 if (code) 455 goto badcode; 456 code = PRC_QUENCH; 457 deliver: 458 /* Free packet atttributes */ 459 if (m->m_flags & M_PKTHDR) 460 m_tag_delete_chain(m); 461 462 /* 463 * Problem with datagram; advise higher level routines. 464 */ 465 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 466 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 467 icmpstat.icps_badlen++; 468 goto freeit; 469 } 470 if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr)) 471 goto badcode; 472 #ifdef INET6 473 /* Get more contiguous data for a v6 in v4 ICMP message. */ 474 if (icp->icmp_ip.ip_p == IPPROTO_IPV6) { 475 if (icmplen < ICMP_V6ADVLENMIN || 476 icmplen < ICMP_V6ADVLEN(icp)) { 477 icmpstat.icps_badlen++; 478 goto freeit; 479 } else { 480 if ((m = m_pullup(m, (ip->ip_hl << 2) + 481 ICMP_V6ADVLEN(icp))) == NULL) { 482 icmpstat.icps_tooshort++; 483 return; 484 } 485 ip = mtod(m, struct ip *); 486 icp = (struct icmp *) 487 (m->m_data + (ip->ip_hl << 2)); 488 } 489 } 490 #endif /* INET6 */ 491 #ifdef ICMPPRINTFS 492 if (icmpprintfs) 493 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 494 #endif 495 memset(&sin, 0, sizeof(sin)); 496 sin.sin_family = AF_INET; 497 sin.sin_len = sizeof(struct sockaddr_in); 498 sin.sin_addr = icp->icmp_ip.ip_dst; 499 #if NCARP > 0 500 if (ifp->if_type == IFT_CARP && 501 carp_lsdrop(m, AF_INET, &sin.sin_addr.s_addr, 502 &ip->ip_dst.s_addr)) 503 goto freeit; 504 #endif 505 /* 506 * XXX if the packet contains [IPv4 AH TCP], we can't make a 507 * notification to TCP layer. 508 */ 509 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 510 if (ctlfunc) 511 (*ctlfunc)(code, sintosa(&sin), m->m_pkthdr.ph_rtableid, 512 &icp->icmp_ip); 513 break; 514 515 badcode: 516 icmpstat.icps_badcode++; 517 break; 518 519 case ICMP_ECHO: 520 if (!icmpbmcastecho && 521 (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 522 icmpstat.icps_bmcastecho++; 523 break; 524 } 525 icp->icmp_type = ICMP_ECHOREPLY; 526 goto reflect; 527 528 case ICMP_TSTAMP: 529 if (icmptstamprepl == 0) 530 break; 531 532 if (!icmpbmcastecho && 533 (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 534 icmpstat.icps_bmcastecho++; 535 break; 536 } 537 if (icmplen < ICMP_TSLEN) { 538 icmpstat.icps_badlen++; 539 break; 540 } 541 icp->icmp_type = ICMP_TSTAMPREPLY; 542 icp->icmp_rtime = iptime(); 543 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 544 goto reflect; 545 546 case ICMP_MASKREQ: 547 if (icmpmaskrepl == 0) 548 break; 549 if (icmplen < ICMP_MASKLEN) { 550 icmpstat.icps_badlen++; 551 break; 552 } 553 /* 554 * We are not able to respond with all ones broadcast 555 * unless we receive it over a point-to-point interface. 556 */ 557 memset(&sin, 0, sizeof(sin)); 558 sin.sin_family = AF_INET; 559 sin.sin_len = sizeof(struct sockaddr_in); 560 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 561 ip->ip_dst.s_addr == INADDR_ANY) 562 sin.sin_addr = ip->ip_src; 563 else 564 sin.sin_addr = ip->ip_dst; 565 if (ifp == NULL) 566 break; 567 ia = ifatoia(ifaof_ifpforaddr(sintosa(&sin), ifp)); 568 if (ia == NULL) 569 break; 570 icp->icmp_type = ICMP_MASKREPLY; 571 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 572 if (ip->ip_src.s_addr == 0) { 573 if (ifp->if_flags & IFF_BROADCAST) { 574 if (ia->ia_broadaddr.sin_addr.s_addr) 575 ip->ip_src = ia->ia_broadaddr.sin_addr; 576 else 577 ip->ip_src.s_addr = INADDR_BROADCAST; 578 } 579 else if (ifp->if_flags & IFF_POINTOPOINT) 580 ip->ip_src = ia->ia_dstaddr.sin_addr; 581 } 582 reflect: 583 #if NCARP > 0 584 if (ifp->if_type == IFT_CARP && 585 carp_lsdrop(m, AF_INET, &ip->ip_src.s_addr, 586 &ip->ip_dst.s_addr)) 587 goto freeit; 588 #endif 589 /* Free packet atttributes */ 590 if (m->m_flags & M_PKTHDR) 591 m_tag_delete_chain(m); 592 593 icmpstat.icps_reflect++; 594 icmpstat.icps_outhist[icp->icmp_type]++; 595 if (!icmp_reflect(m, &opts, NULL)) 596 icmp_send(m, opts); 597 return; 598 599 case ICMP_REDIRECT: 600 { 601 struct sockaddr_in sdst; 602 struct sockaddr_in sgw; 603 struct sockaddr_in ssrc; 604 struct rtentry *newrt = NULL; 605 606 /* Free packet atttributes */ 607 if (m->m_flags & M_PKTHDR) 608 m_tag_delete_chain(m); 609 if (icmp_rediraccept == 0 || ipforwarding == 1) 610 goto freeit; 611 if (code > 3) 612 goto badcode; 613 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 614 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 615 icmpstat.icps_badlen++; 616 break; 617 } 618 /* 619 * Short circuit routing redirects to force 620 * immediate change in the kernel's routing 621 * tables. The message is also handed to anyone 622 * listening on a raw socket (e.g. the routing 623 * daemon for use in updating its tables). 624 */ 625 memset(&sdst, 0, sizeof(sdst)); 626 memset(&sgw, 0, sizeof(sgw)); 627 memset(&ssrc, 0, sizeof(ssrc)); 628 sdst.sin_family = sgw.sin_family = ssrc.sin_family = AF_INET; 629 sdst.sin_len = sgw.sin_len = ssrc.sin_len = sizeof(sdst); 630 memcpy(&sdst.sin_addr, &icp->icmp_ip.ip_dst, 631 sizeof(sdst.sin_addr)); 632 memcpy(&sgw.sin_addr, &icp->icmp_gwaddr, 633 sizeof(sgw.sin_addr)); 634 memcpy(&ssrc.sin_addr, &ip->ip_src, 635 sizeof(ssrc.sin_addr)); 636 637 #ifdef ICMPPRINTFS 638 if (icmpprintfs) { 639 char gw[INET_ADDRSTRLEN], dst[INET_ADDRSTRLEN]; 640 641 inet_ntop(AF_INET, &icp->icmp_gwaddr, gw, sizeof(gw)); 642 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, 643 dst, sizeof(dst)); 644 645 printf("redirect dst %s to %s\n", dst, gw); 646 } 647 #endif 648 649 #if NCARP > 0 650 if (ifp->if_type == IFT_CARP && 651 carp_lsdrop(m, AF_INET, &sdst.sin_addr.s_addr, 652 &ip->ip_dst.s_addr)) 653 goto freeit; 654 #endif 655 rtredirect(sintosa(&sdst), sintosa(&sgw), 656 sintosa(&ssrc), &newrt, m->m_pkthdr.ph_rtableid); 657 if (newrt != NULL && icmp_redirtimeout != 0) { 658 (void)rt_timer_add(newrt, icmp_redirect_timeout, 659 icmp_redirect_timeout_q, m->m_pkthdr.ph_rtableid); 660 } 661 if (newrt != NULL) 662 rtfree(newrt); 663 pfctlinput(PRC_REDIRECT_HOST, sintosa(&sdst)); 664 break; 665 } 666 /* 667 * No kernel processing for the following; 668 * just fall through to send to raw listener. 669 */ 670 case ICMP_ECHOREPLY: 671 case ICMP_ROUTERADVERT: 672 case ICMP_ROUTERSOLICIT: 673 case ICMP_TSTAMPREPLY: 674 case ICMP_IREQREPLY: 675 case ICMP_MASKREPLY: 676 case ICMP_TRACEROUTE: 677 case ICMP_DATACONVERR: 678 case ICMP_MOBILE_REDIRECT: 679 case ICMP_IPV6_WHEREAREYOU: 680 case ICMP_IPV6_IAMHERE: 681 case ICMP_MOBILE_REGREQUEST: 682 case ICMP_MOBILE_REGREPLY: 683 case ICMP_PHOTURIS: 684 default: 685 break; 686 } 687 688 raw: 689 rip_input(m); 690 return; 691 692 freeit: 693 m_freem(m); 694 } 695 696 /* 697 * Reflect the ip packet back to the source 698 */ 699 int 700 icmp_reflect(struct mbuf *m, struct mbuf **op, struct in_ifaddr *ia) 701 { 702 struct ip *ip = mtod(m, struct ip *); 703 struct mbuf *opts = NULL; 704 struct sockaddr_in sin; 705 struct rtentry *rt; 706 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 707 u_int rtableid; 708 709 if (!in_canforward(ip->ip_src) && 710 ((ip->ip_src.s_addr & IN_CLASSA_NET) != 711 htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { 712 m_freem(m); /* Bad return address */ 713 return (EHOSTUNREACH); 714 } 715 716 #if NPF > 0 717 pf_pkt_addr_changed(m); 718 #endif 719 rtableid = m->m_pkthdr.ph_rtableid; 720 721 /* 722 * If the incoming packet was addressed directly to us, 723 * use dst as the src for the reply. For broadcast, use 724 * the address which corresponds to the incoming interface. 725 */ 726 if (ia == NULL) { 727 memset(&sin, 0, sizeof(sin)); 728 sin.sin_len = sizeof(sin); 729 sin.sin_family = AF_INET; 730 sin.sin_addr = ip->ip_dst; 731 732 rt = rtalloc(sintosa(&sin), 0, rtableid); 733 if (rtisvalid(rt) && 734 ISSET(rt->rt_flags, RTF_LOCAL|RTF_BROADCAST)) 735 ia = ifatoia(rt->rt_ifa); 736 rtfree(rt); 737 } 738 739 /* 740 * The following happens if the packet was not addressed to us. 741 * Use the new source address and do a route lookup. If it fails 742 * drop the packet as there is no path to the host. 743 */ 744 if (ia == NULL) { 745 memset(&sin, 0, sizeof(sin)); 746 sin.sin_len = sizeof(sin); 747 sin.sin_family = AF_INET; 748 sin.sin_addr = ip->ip_src; 749 750 /* keep packet in the original virtual instance */ 751 rt = rtalloc(sintosa(&sin), RT_RESOLVE, rtableid); 752 if (rt == NULL) { 753 ipstat.ips_noroute++; 754 m_freem(m); 755 return (EHOSTUNREACH); 756 } 757 758 ia = ifatoia(rt->rt_ifa); 759 rtfree(rt); 760 } 761 762 ip->ip_dst = ip->ip_src; 763 ip->ip_src = ia->ia_addr.sin_addr; 764 ip->ip_ttl = MAXTTL; 765 766 if (optlen > 0) { 767 u_char *cp; 768 int opt, cnt; 769 u_int len; 770 771 /* 772 * Retrieve any source routing from the incoming packet; 773 * add on any record-route or timestamp options. 774 */ 775 cp = (u_char *) (ip + 1); 776 if (op && (opts = ip_srcroute(m)) == NULL && 777 (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { 778 opts->m_len = sizeof(struct in_addr); 779 mtod(opts, struct in_addr *)->s_addr = 0; 780 } 781 if (op && opts) { 782 #ifdef ICMPPRINTFS 783 if (icmpprintfs) 784 printf("icmp_reflect optlen %d rt %d => ", 785 optlen, opts->m_len); 786 #endif 787 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 788 opt = cp[IPOPT_OPTVAL]; 789 if (opt == IPOPT_EOL) 790 break; 791 if (opt == IPOPT_NOP) 792 len = 1; 793 else { 794 if (cnt < IPOPT_OLEN + sizeof(*cp)) 795 break; 796 len = cp[IPOPT_OLEN]; 797 if (len < IPOPT_OLEN + sizeof(*cp) || 798 len > cnt) 799 break; 800 } 801 /* 802 * Should check for overflow, but it 803 * "can't happen" 804 */ 805 if (opt == IPOPT_RR || opt == IPOPT_TS || 806 opt == IPOPT_SECURITY) { 807 memcpy(mtod(opts, caddr_t) + 808 opts->m_len, cp, len); 809 opts->m_len += len; 810 } 811 } 812 /* Terminate & pad, if necessary */ 813 if ((cnt = opts->m_len % 4) != 0) 814 for (; cnt < 4; cnt++) { 815 *(mtod(opts, caddr_t) + opts->m_len) = 816 IPOPT_EOL; 817 opts->m_len++; 818 } 819 #ifdef ICMPPRINTFS 820 if (icmpprintfs) 821 printf("%d\n", opts->m_len); 822 #endif 823 } 824 ip_stripoptions(m); 825 } 826 m->m_flags &= ~(M_BCAST|M_MCAST); 827 if (op) 828 *op = opts; 829 830 return (0); 831 } 832 833 /* 834 * Send an icmp packet back to the ip level 835 */ 836 void 837 icmp_send(struct mbuf *m, struct mbuf *opts) 838 { 839 struct ip *ip = mtod(m, struct ip *); 840 int hlen; 841 struct icmp *icp; 842 843 hlen = ip->ip_hl << 2; 844 icp = (struct icmp *)(mtod(m, caddr_t) + hlen); 845 icp->icmp_cksum = 0; 846 m->m_pkthdr.csum_flags = M_ICMP_CSUM_OUT; 847 #ifdef ICMPPRINTFS 848 if (icmpprintfs) { 849 char dst[INET_ADDRSTRLEN], src[INET_ADDRSTRLEN]; 850 851 inet_ntop(AF_INET, &ip->ip_dst, dst, sizeof(dst)); 852 inet_ntop(AF_INET, &ip->ip_src, src, sizeof(src)); 853 854 printf("icmp_send dst %s src %s\n", dst, src); 855 } 856 #endif 857 if (opts != NULL) 858 m = ip_insertoptions(m, opts, &hlen); 859 860 ip_send(m); 861 } 862 863 u_int32_t 864 iptime(void) 865 { 866 struct timeval atv; 867 u_long t; 868 869 microtime(&atv); 870 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 871 return (htonl(t)); 872 } 873 874 int 875 icmp_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 876 size_t newlen) 877 { 878 int s, error; 879 880 /* All sysctl names at this level are terminal. */ 881 if (namelen != 1) 882 return (ENOTDIR); 883 884 s = splsoftnet(); 885 switch (name[0]) { 886 case ICMPCTL_REDIRTIMEOUT: 887 888 error = sysctl_int(oldp, oldlenp, newp, newlen, 889 &icmp_redirtimeout); 890 if (icmp_redirect_timeout_q != NULL) { 891 if (icmp_redirtimeout == 0) { 892 rt_timer_queue_destroy(icmp_redirect_timeout_q); 893 icmp_redirect_timeout_q = NULL; 894 } else 895 rt_timer_queue_change(icmp_redirect_timeout_q, 896 icmp_redirtimeout); 897 } else if (icmp_redirtimeout > 0) { 898 icmp_redirect_timeout_q = 899 rt_timer_queue_create(icmp_redirtimeout); 900 } 901 break; 902 903 case ICMPCTL_STATS: 904 if (newp != NULL) { 905 error = EPERM; 906 break; 907 } 908 error = sysctl_struct(oldp, oldlenp, newp, newlen, 909 &icmpstat, sizeof(icmpstat)); 910 break; 911 912 default: 913 if (name[0] < ICMPCTL_MAXID) { 914 error = sysctl_int_arr(icmpctl_vars, name, namelen, 915 oldp, oldlenp, newp, newlen); 916 break; 917 } 918 error = ENOPROTOOPT; 919 break; 920 } 921 splx(s); 922 923 return (error); 924 } 925 926 927 struct rtentry * 928 icmp_mtudisc_clone(struct in_addr dst, u_int rtableid) 929 { 930 struct sockaddr_in sin; 931 struct rtentry *rt; 932 int error; 933 934 memset(&sin, 0, sizeof(sin)); 935 sin.sin_family = AF_INET; 936 sin.sin_len = sizeof(sin); 937 sin.sin_addr = dst; 938 939 rt = rtalloc(sintosa(&sin), RT_RESOLVE, rtableid); 940 941 /* Check if the route is actually usable */ 942 if (!rtisvalid(rt) || (rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE))) { 943 rtfree(rt); 944 return (NULL); 945 } 946 947 /* If we didn't get a host route, allocate one */ 948 if ((rt->rt_flags & RTF_HOST) == 0) { 949 struct rtentry *nrt; 950 struct rt_addrinfo info; 951 952 memset(&info, 0, sizeof(info)); 953 info.rti_info[RTAX_DST] = sintosa(&sin); 954 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 955 info.rti_flags = RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC; 956 957 error = rtrequest(RTM_ADD, &info, RTP_DEFAULT, &nrt, rtableid); 958 if (error) { 959 rtfree(rt); 960 return (NULL); 961 } 962 nrt->rt_rmx = rt->rt_rmx; 963 rtfree(rt); 964 rt = nrt; 965 } 966 error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q, 967 rtableid); 968 if (error) { 969 rtfree(rt); 970 return (NULL); 971 } 972 973 return (rt); 974 } 975 976 /* Table of common MTUs: */ 977 static const u_short mtu_table[] = { 978 65535, 65280, 32000, 17914, 9180, 8166, 979 4352, 2002, 1492, 1006, 508, 296, 68, 0 980 }; 981 982 void 983 icmp_mtudisc(struct icmp *icp, u_int rtableid) 984 { 985 struct rtentry *rt; 986 struct ifnet *ifp; 987 u_long mtu = ntohs(icp->icmp_nextmtu); /* Why a long? IPv6 */ 988 989 rt = icmp_mtudisc_clone(icp->icmp_ip.ip_dst, rtableid); 990 if (rt == NULL) 991 return; 992 993 ifp = if_get(rt->rt_ifidx); 994 if (ifp == NULL) { 995 rtfree(rt); 996 return; 997 } 998 999 if (mtu == 0) { 1000 int i = 0; 1001 1002 mtu = ntohs(icp->icmp_ip.ip_len); 1003 /* Some 4.2BSD-based routers incorrectly adjust the ip_len */ 1004 if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0) 1005 mtu -= (icp->icmp_ip.ip_hl << 2); 1006 1007 /* If we still can't guess a value, try the route */ 1008 if (mtu == 0) { 1009 mtu = rt->rt_rmx.rmx_mtu; 1010 1011 /* If no route mtu, default to the interface mtu */ 1012 1013 if (mtu == 0) 1014 mtu = ifp->if_mtu; 1015 } 1016 1017 for (i = 0; i < nitems(mtu_table); i++) 1018 if (mtu > mtu_table[i]) { 1019 mtu = mtu_table[i]; 1020 break; 1021 } 1022 } 1023 1024 /* 1025 * XXX: RTV_MTU is overloaded, since the admin can set it 1026 * to turn off PMTU for a route, and the kernel can 1027 * set it to indicate a serious problem with PMTU 1028 * on a route. We should be using a separate flag 1029 * for the kernel to indicate this. 1030 */ 1031 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1032 if (mtu < 296 || mtu > ifp->if_mtu) 1033 rt->rt_rmx.rmx_locks |= RTV_MTU; 1034 else if (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0) 1035 rt->rt_rmx.rmx_mtu = mtu; 1036 } 1037 1038 if_put(ifp); 1039 rtfree(rt); 1040 } 1041 1042 void 1043 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 1044 { 1045 struct ifnet *ifp; 1046 int s; 1047 1048 ifp = if_get(rt->rt_ifidx); 1049 if (ifp == NULL) 1050 return; 1051 1052 if ((rt->rt_flags & (RTF_DYNAMIC|RTF_HOST)) == (RTF_DYNAMIC|RTF_HOST)) { 1053 void *(*ctlfunc)(int, struct sockaddr *, u_int, void *); 1054 struct sockaddr_in sin; 1055 1056 sin = *satosin(rt_key(rt)); 1057 1058 s = splsoftnet(); 1059 rtdeletemsg(rt, ifp, r->rtt_tableid); 1060 1061 /* Notify TCP layer of increased Path MTU estimate */ 1062 ctlfunc = inetsw[ip_protox[IPPROTO_TCP]].pr_ctlinput; 1063 if (ctlfunc) 1064 (*ctlfunc)(PRC_MTUINC, sintosa(&sin), 1065 r->rtt_tableid, NULL); 1066 splx(s); 1067 } else { 1068 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 1069 rt->rt_rmx.rmx_mtu = 0; 1070 } 1071 1072 if_put(ifp); 1073 } 1074 1075 /* 1076 * Perform rate limit check. 1077 * Returns 0 if it is okay to send the icmp packet. 1078 * Returns 1 if the router SHOULD NOT send this icmp packet due to rate 1079 * limitation. 1080 * 1081 * XXX per-destination/type check necessary? 1082 */ 1083 int 1084 icmp_ratelimit(const struct in_addr *dst, const int type, const int code) 1085 { 1086 /* PPS limit */ 1087 if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count, 1088 icmperrppslim)) 1089 return 1; /* The packet is subject to rate limit */ 1090 return 0; /* okay to send */ 1091 } 1092 1093 void 1094 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r) 1095 { 1096 struct ifnet *ifp; 1097 int s; 1098 1099 ifp = if_get(rt->rt_ifidx); 1100 if (ifp == NULL) 1101 return; 1102 1103 if ((rt->rt_flags & (RTF_DYNAMIC|RTF_HOST)) == (RTF_DYNAMIC|RTF_HOST)) { 1104 s = splsoftnet(); 1105 rtdeletemsg(rt, ifp, r->rtt_tableid); 1106 splx(s); 1107 } 1108 1109 if_put(ifp); 1110 } 1111 1112 int 1113 icmp_do_exthdr(struct mbuf *m, u_int16_t class, u_int8_t ctype, void *buf, 1114 size_t len) 1115 { 1116 struct ip *ip = mtod(m, struct ip *); 1117 int hlen, off; 1118 struct mbuf *n; 1119 struct icmp *icp; 1120 struct icmp_ext_hdr *ieh; 1121 struct { 1122 struct icmp_ext_hdr ieh; 1123 struct icmp_ext_obj_hdr ieo; 1124 } hdr; 1125 1126 hlen = ip->ip_hl << 2; 1127 icp = (struct icmp *)(mtod(m, caddr_t) + hlen); 1128 if (icp->icmp_type != ICMP_TIMXCEED && icp->icmp_type != ICMP_UNREACH && 1129 icp->icmp_type != ICMP_PARAMPROB) 1130 /* exthdr not supported */ 1131 return (0); 1132 1133 if (icp->icmp_length != 0) 1134 /* exthdr already present, giving up */ 1135 return (0); 1136 1137 /* the actual offset starts after the common ICMP header */ 1138 hlen += ICMP_MINLEN; 1139 /* exthdr must start on a word boundary */ 1140 off = roundup(ntohs(ip->ip_len) - hlen, sizeof(u_int32_t)); 1141 /* ... and at an offset of ICMP_EXT_OFFSET or bigger */ 1142 off = max(off, ICMP_EXT_OFFSET); 1143 icp->icmp_length = off / sizeof(u_int32_t); 1144 1145 memset(&hdr, 0, sizeof(hdr)); 1146 hdr.ieh.ieh_version = ICMP_EXT_HDR_VERSION; 1147 hdr.ieo.ieo_length = htons(sizeof(struct icmp_ext_obj_hdr) + len); 1148 hdr.ieo.ieo_cnum = class; 1149 hdr.ieo.ieo_ctype = ctype; 1150 1151 if (m_copyback(m, hlen + off, sizeof(hdr), &hdr, M_NOWAIT) || 1152 m_copyback(m, hlen + off + sizeof(hdr), len, buf, M_NOWAIT)) { 1153 m_freem(m); 1154 return (ENOBUFS); 1155 } 1156 1157 /* calculate checksum */ 1158 n = m_getptr(m, hlen + off, &off); 1159 if (n == NULL) 1160 panic("icmp_do_exthdr: m_getptr failure"); 1161 ieh = (struct icmp_ext_hdr *)(mtod(n, caddr_t) + off); 1162 ieh->ieh_cksum = in4_cksum(n, 0, off, sizeof(hdr) + len); 1163 1164 ip->ip_len = htons(m->m_pkthdr.len); 1165 1166 return (0); 1167 } 1168