1 /* $NetBSD: ip_icmp.c,v 1.120 2008/06/18 09:06:28 yamt 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 1.120 2008/06/18 09:06:28 yamt Exp $"); 98 99 #include "opt_ipsec.h" 100 101 #include <sys/param.h> 102 #include <sys/systm.h> 103 #include <sys/malloc.h> 104 #include <sys/mbuf.h> 105 #include <sys/protosw.h> 106 #include <sys/socket.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 <netinet6/ipsec.h> 128 #include <netkey/key.h> 129 #endif 130 131 #ifdef FAST_IPSEC 132 #include <netipsec/ipsec.h> 133 #include <netipsec/key.h> 134 #endif /* FAST_IPSEC*/ 135 136 #include <machine/stdarg.h> 137 138 /* 139 * ICMP routines: error generation, receive packet processing, and 140 * routines to turnaround packets back to the originator, and 141 * host table maintenance routines. 142 */ 143 144 int icmpmaskrepl = 0; 145 #ifdef ICMPPRINTFS 146 int icmpprintfs = 0; 147 #endif 148 int icmpreturndatabytes = 8; 149 150 percpu_t *icmpstat_percpu; 151 152 /* 153 * List of callbacks to notify when Path MTU changes are made. 154 */ 155 struct icmp_mtudisc_callback { 156 LIST_ENTRY(icmp_mtudisc_callback) mc_list; 157 void (*mc_func)(struct in_addr); 158 }; 159 160 LIST_HEAD(, icmp_mtudisc_callback) icmp_mtudisc_callbacks = 161 LIST_HEAD_INITIALIZER(&icmp_mtudisc_callbacks); 162 163 #if 0 164 static u_int ip_next_mtu(u_int, int); 165 #else 166 /*static*/ u_int ip_next_mtu(u_int, int); 167 #endif 168 169 extern int icmperrppslim; 170 static int icmperrpps_count = 0; 171 static struct timeval icmperrppslim_last; 172 static int icmp_rediraccept = 1; 173 static int icmp_redirtimeout = 600; 174 static struct rttimer_queue *icmp_redirect_timeout_q = NULL; 175 176 static void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *); 177 static void icmp_redirect_timeout(struct rtentry *, struct rttimer *); 178 179 static int icmp_ratelimit(const struct in_addr *, const int, const int); 180 181 182 void 183 icmp_init(void) 184 { 185 /* 186 * This is only useful if the user initializes redirtimeout to 187 * something other than zero. 188 */ 189 if (icmp_redirtimeout != 0) { 190 icmp_redirect_timeout_q = 191 rt_timer_queue_create(icmp_redirtimeout); 192 } 193 194 icmpstat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP_NSTATS); 195 } 196 197 /* 198 * Register a Path MTU Discovery callback. 199 */ 200 void 201 icmp_mtudisc_callback_register(void (*func)(struct in_addr)) 202 { 203 struct icmp_mtudisc_callback *mc; 204 205 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL; 206 mc = LIST_NEXT(mc, mc_list)) { 207 if (mc->mc_func == func) 208 return; 209 } 210 211 mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT); 212 if (mc == NULL) 213 panic("icmp_mtudisc_callback_register"); 214 215 mc->mc_func = func; 216 LIST_INSERT_HEAD(&icmp_mtudisc_callbacks, mc, mc_list); 217 } 218 219 /* 220 * Generate an error packet of type error 221 * in response to bad packet ip. 222 */ 223 void 224 icmp_error(struct mbuf *n, int type, int code, n_long dest, 225 int destmtu) 226 { 227 struct ip *oip = mtod(n, struct ip *), *nip; 228 unsigned oiplen = oip->ip_hl << 2; 229 struct icmp *icp; 230 struct mbuf *m; 231 struct m_tag *mtag; 232 unsigned icmplen, mblen; 233 234 #ifdef ICMPPRINTFS 235 if (icmpprintfs) 236 printf("icmp_error(%p, type:%d, code:%d)\n", oip, type, code); 237 #endif 238 if (type != ICMP_REDIRECT) 239 ICMP_STATINC(ICMP_STAT_ERROR); 240 /* 241 * Don't send error if the original packet was encrypted. 242 * Don't send error if not the first fragment of message. 243 * Don't error if the old packet protocol was ICMP 244 * error message, only known informational types. 245 */ 246 if (n->m_flags & M_DECRYPTED) 247 goto freeit; 248 if (oip->ip_off &~ htons(IP_MF|IP_DF)) 249 goto freeit; 250 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 251 n->m_len >= oiplen + ICMP_MINLEN && 252 !ICMP_INFOTYPE(((struct icmp *)((char *)oip + oiplen))->icmp_type)) { 253 ICMP_STATINC(ICMP_STAT_OLDICMP); 254 goto freeit; 255 } 256 /* Don't send error in response to a multicast or broadcast packet */ 257 if (n->m_flags & (M_BCAST|M_MCAST)) 258 goto freeit; 259 260 /* 261 * First, do a rate limitation check. 262 */ 263 if (icmp_ratelimit(&oip->ip_src, type, code)) { 264 /* XXX stat */ 265 goto freeit; 266 } 267 268 /* 269 * Now, formulate icmp message 270 */ 271 icmplen = oiplen + min(icmpreturndatabytes, 272 ntohs(oip->ip_len) - oiplen); 273 /* 274 * Defend against mbuf chains shorter than oip->ip_len - oiplen: 275 */ 276 mblen = 0; 277 for (m = n; m && (mblen < icmplen); m = m->m_next) 278 mblen += m->m_len; 279 icmplen = min(mblen, icmplen); 280 281 /* 282 * As we are not required to return everything we have, 283 * we return whatever we can return at ease. 284 * 285 * Note that ICMP datagrams longer than 576 octets are out of spec 286 * according to RFC1812; the limit on icmpreturndatabytes below in 287 * icmp_sysctl will keep things below that limit. 288 */ 289 290 KASSERT(ICMP_MINLEN <= MCLBYTES); 291 292 if (icmplen + ICMP_MINLEN > MCLBYTES) 293 icmplen = MCLBYTES - ICMP_MINLEN; 294 295 m = m_gethdr(M_DONTWAIT, MT_HEADER); 296 if (m && (icmplen + ICMP_MINLEN > MHLEN)) { 297 MCLGET(m, M_DONTWAIT); 298 if ((m->m_flags & M_EXT) == 0) { 299 m_freem(m); 300 m = NULL; 301 } 302 } 303 if (m == NULL) 304 goto freeit; 305 MCLAIM(m, n->m_owner); 306 m->m_len = icmplen + ICMP_MINLEN; 307 if ((m->m_flags & M_EXT) == 0) 308 MH_ALIGN(m, m->m_len); 309 icp = mtod(m, struct icmp *); 310 if ((u_int)type > ICMP_MAXTYPE) 311 panic("icmp_error"); 312 ICMP_STATINC(ICMP_STAT_OUTHIST + type); 313 icp->icmp_type = type; 314 if (type == ICMP_REDIRECT) 315 icp->icmp_gwaddr.s_addr = dest; 316 else { 317 icp->icmp_void = 0; 318 /* 319 * The following assignments assume an overlay with the 320 * zeroed icmp_void field. 321 */ 322 if (type == ICMP_PARAMPROB) { 323 icp->icmp_pptr = code; 324 code = 0; 325 } else if (type == ICMP_UNREACH && 326 code == ICMP_UNREACH_NEEDFRAG && destmtu) 327 icp->icmp_nextmtu = htons(destmtu); 328 } 329 330 icp->icmp_code = code; 331 m_copydata(n, 0, icmplen, (void *)&icp->icmp_ip); 332 333 /* 334 * Now, copy old ip header (without options) 335 * in front of icmp message. 336 */ 337 if (m->m_data - sizeof(struct ip) < m->m_pktdat) 338 panic("icmp len"); 339 m->m_data -= sizeof(struct ip); 340 m->m_len += sizeof(struct ip); 341 m->m_pkthdr.len = m->m_len; 342 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 343 nip = mtod(m, struct ip *); 344 /* ip_v set in ip_output */ 345 nip->ip_hl = sizeof(struct ip) >> 2; 346 nip->ip_tos = 0; 347 nip->ip_len = htons(m->m_len); 348 /* ip_id set in ip_output */ 349 nip->ip_off = htons(0); 350 /* ip_ttl set in icmp_reflect */ 351 nip->ip_p = IPPROTO_ICMP; 352 nip->ip_src = oip->ip_src; 353 nip->ip_dst = oip->ip_dst; 354 /* move PF m_tag to new packet, if it exists */ 355 mtag = m_tag_find(n, PACKET_TAG_PF, NULL); 356 if (mtag != NULL) { 357 m_tag_unlink(n, mtag); 358 m_tag_prepend(m, mtag); 359 } 360 icmp_reflect(m); 361 362 freeit: 363 m_freem(n); 364 } 365 366 struct sockaddr_in icmpsrc = { 367 .sin_len = sizeof (struct sockaddr_in), 368 .sin_family = AF_INET, 369 }; 370 static struct sockaddr_in icmpdst = { 371 .sin_len = sizeof (struct sockaddr_in), 372 .sin_family = AF_INET, 373 }; 374 static struct sockaddr_in icmpgw = { 375 .sin_len = sizeof (struct sockaddr_in), 376 .sin_family = AF_INET, 377 }; 378 struct sockaddr_in icmpmask = { 379 .sin_len = 8, 380 .sin_family = 0, 381 }; 382 383 /* 384 * Process a received ICMP message. 385 */ 386 void 387 icmp_input(struct mbuf *m, ...) 388 { 389 int proto; 390 struct icmp *icp; 391 struct ip *ip = mtod(m, struct ip *); 392 int icmplen; 393 int i; 394 struct in_ifaddr *ia; 395 void *(*ctlfunc)(int, const struct sockaddr *, void *); 396 int code; 397 int hlen; 398 va_list ap; 399 struct rtentry *rt; 400 401 va_start(ap, m); 402 hlen = va_arg(ap, int); 403 proto = va_arg(ap, int); 404 va_end(ap); 405 406 /* 407 * Locate icmp structure in mbuf, and check 408 * that not corrupted and of at least minimum length. 409 */ 410 icmplen = ntohs(ip->ip_len) - hlen; 411 #ifdef ICMPPRINTFS 412 if (icmpprintfs) { 413 printf("icmp_input from `%s' to ", inet_ntoa(ip->ip_src)); 414 printf("`%s', len %d\n", inet_ntoa(ip->ip_dst), icmplen); 415 } 416 #endif 417 if (icmplen < ICMP_MINLEN) { 418 ICMP_STATINC(ICMP_STAT_TOOSHORT); 419 goto freeit; 420 } 421 i = hlen + min(icmplen, ICMP_ADVLENMIN); 422 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) { 423 ICMP_STATINC(ICMP_STAT_TOOSHORT); 424 return; 425 } 426 ip = mtod(m, struct ip *); 427 m->m_len -= hlen; 428 m->m_data += hlen; 429 icp = mtod(m, struct icmp *); 430 /* Don't need to assert alignment, here. */ 431 if (in_cksum(m, icmplen)) { 432 ICMP_STATINC(ICMP_STAT_CHECKSUM); 433 goto freeit; 434 } 435 m->m_len += hlen; 436 m->m_data -= hlen; 437 438 #ifdef ICMPPRINTFS 439 /* 440 * Message type specific processing. 441 */ 442 if (icmpprintfs) 443 printf("icmp_input(type:%d, code:%d)\n", icp->icmp_type, 444 icp->icmp_code); 445 #endif 446 if (icp->icmp_type > ICMP_MAXTYPE) 447 goto raw; 448 ICMP_STATINC(ICMP_STAT_INHIST + icp->icmp_type); 449 code = icp->icmp_code; 450 switch (icp->icmp_type) { 451 452 case ICMP_UNREACH: 453 switch (code) { 454 case ICMP_UNREACH_NET: 455 code = PRC_UNREACH_NET; 456 break; 457 458 case ICMP_UNREACH_HOST: 459 code = PRC_UNREACH_HOST; 460 break; 461 462 case ICMP_UNREACH_PROTOCOL: 463 code = PRC_UNREACH_PROTOCOL; 464 break; 465 466 case ICMP_UNREACH_PORT: 467 code = PRC_UNREACH_PORT; 468 break; 469 470 case ICMP_UNREACH_SRCFAIL: 471 code = PRC_UNREACH_SRCFAIL; 472 break; 473 474 case ICMP_UNREACH_NEEDFRAG: 475 code = PRC_MSGSIZE; 476 break; 477 478 case ICMP_UNREACH_NET_UNKNOWN: 479 case ICMP_UNREACH_NET_PROHIB: 480 case ICMP_UNREACH_TOSNET: 481 code = PRC_UNREACH_NET; 482 break; 483 484 case ICMP_UNREACH_HOST_UNKNOWN: 485 case ICMP_UNREACH_ISOLATED: 486 case ICMP_UNREACH_HOST_PROHIB: 487 case ICMP_UNREACH_TOSHOST: 488 code = PRC_UNREACH_HOST; 489 break; 490 491 default: 492 goto badcode; 493 } 494 goto deliver; 495 496 case ICMP_TIMXCEED: 497 if (code > 1) 498 goto badcode; 499 code += PRC_TIMXCEED_INTRANS; 500 goto deliver; 501 502 case ICMP_PARAMPROB: 503 if (code > 1) 504 goto badcode; 505 code = PRC_PARAMPROB; 506 goto deliver; 507 508 case ICMP_SOURCEQUENCH: 509 if (code) 510 goto badcode; 511 code = PRC_QUENCH; 512 goto deliver; 513 514 deliver: 515 /* 516 * Problem with datagram; advise higher level routines. 517 */ 518 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 519 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 520 ICMP_STATINC(ICMP_STAT_BADLEN); 521 goto freeit; 522 } 523 if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr)) 524 goto badcode; 525 #ifdef ICMPPRINTFS 526 if (icmpprintfs) 527 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 528 #endif 529 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 530 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 531 if (ctlfunc) 532 (void) (*ctlfunc)(code, sintosa(&icmpsrc), 533 &icp->icmp_ip); 534 break; 535 536 badcode: 537 ICMP_STATINC(ICMP_STAT_BADCODE); 538 break; 539 540 case ICMP_ECHO: 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 icp->icmp_type = ICMP_TSTAMPREPLY; 550 icp->icmp_rtime = iptime(); 551 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 552 goto reflect; 553 554 case ICMP_MASKREQ: 555 if (icmpmaskrepl == 0) 556 break; 557 /* 558 * We are not able to respond with all ones broadcast 559 * unless we receive it over a point-to-point interface. 560 */ 561 if (icmplen < ICMP_MASKLEN) { 562 ICMP_STATINC(ICMP_STAT_BADLEN); 563 break; 564 } 565 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 566 in_nullhost(ip->ip_dst)) 567 icmpdst.sin_addr = ip->ip_src; 568 else 569 icmpdst.sin_addr = ip->ip_dst; 570 ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst), 571 m->m_pkthdr.rcvif)); 572 if (ia == 0) 573 break; 574 icp->icmp_type = ICMP_MASKREPLY; 575 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 576 if (in_nullhost(ip->ip_src)) { 577 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 578 ip->ip_src = ia->ia_broadaddr.sin_addr; 579 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 580 ip->ip_src = ia->ia_dstaddr.sin_addr; 581 } 582 reflect: 583 { 584 uint64_t *icps = percpu_getref(icmpstat_percpu); 585 icps[ICMP_STAT_REFLECT]++; 586 icps[ICMP_STAT_OUTHIST + icp->icmp_type]++; 587 percpu_putref(icmpstat_percpu); 588 } 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 ICMP_STATINC(ICMP_STAT_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 NULL, RTF_GATEWAY | RTF_HOST, sintosa(&icmpgw), &rt); 622 if (rt != NULL && icmp_redirtimeout != 0) { 623 i = rt_timer_add(rt, icmp_redirect_timeout, 624 icmp_redirect_timeout_q); 625 if (i) 626 log(LOG_ERR, "ICMP: redirect failed to " 627 "register timeout for route to %x, " 628 "code %d\n", 629 icp->icmp_ip.ip_dst.s_addr, i); 630 } 631 if (rt != NULL) 632 rtfree(rt); 633 634 pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc)); 635 #if defined(IPSEC) || defined(FAST_IPSEC) 636 key_sa_routechange((struct sockaddr *)&icmpsrc); 637 #endif 638 break; 639 640 /* 641 * No kernel processing for the following; 642 * just fall through to send to raw listener. 643 */ 644 case ICMP_ECHOREPLY: 645 case ICMP_ROUTERADVERT: 646 case ICMP_ROUTERSOLICIT: 647 case ICMP_TSTAMPREPLY: 648 case ICMP_IREQREPLY: 649 case ICMP_MASKREPLY: 650 default: 651 break; 652 } 653 654 raw: 655 rip_input(m, hlen, proto); 656 return; 657 658 freeit: 659 m_freem(m); 660 return; 661 } 662 663 /* 664 * Reflect the ip packet back to the source 665 */ 666 void 667 icmp_reflect(struct mbuf *m) 668 { 669 struct ip *ip = mtod(m, struct ip *); 670 struct in_ifaddr *ia; 671 struct ifaddr *ifa; 672 struct sockaddr_in *sin = 0; 673 struct in_addr t; 674 struct mbuf *opts = 0; 675 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 676 677 if (!in_canforward(ip->ip_src) && 678 ((ip->ip_src.s_addr & IN_CLASSA_NET) != 679 htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { 680 m_freem(m); /* Bad return address */ 681 goto done; /* ip_output() will check for broadcast */ 682 } 683 t = ip->ip_dst; 684 ip->ip_dst = ip->ip_src; 685 /* 686 * If the incoming packet was addressed directly to us, use 687 * dst as the src for the reply. Otherwise (broadcast or 688 * anonymous), use an address which corresponds to the 689 * incoming interface, with a preference for the address which 690 * corresponds to the route to the destination of the ICMP. 691 */ 692 693 /* Look for packet addressed to us */ 694 INADDR_TO_IA(t, ia); 695 696 /* look for packet sent to broadcast address */ 697 if (ia == NULL && m->m_pkthdr.rcvif && 698 (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST)) { 699 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 700 if (ifa->ifa_addr->sa_family != AF_INET) 701 continue; 702 if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) { 703 ia = ifatoia(ifa); 704 break; 705 } 706 } 707 } 708 709 if (ia) 710 sin = &ia->ia_addr; 711 712 icmpdst.sin_addr = t; 713 714 /* 715 * if the packet is addressed somewhere else, compute the 716 * source address for packets routed back to the source, and 717 * use that, if it's an address on the interface which 718 * received the packet 719 */ 720 if (sin == NULL && m->m_pkthdr.rcvif) { 721 struct sockaddr_in sin_dst; 722 struct route icmproute; 723 int errornum; 724 725 sockaddr_in_init(&sin_dst, &ip->ip_dst, 0); 726 memset(&icmproute, 0, sizeof(icmproute)); 727 errornum = 0; 728 sin = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum); 729 /* errornum is never used */ 730 rtcache_free(&icmproute); 731 /* check to make sure sin is a source address on rcvif */ 732 if (sin) { 733 t = sin->sin_addr; 734 sin = NULL; 735 INADDR_TO_IA(t, ia); 736 while (ia) { 737 if (ia->ia_ifp == m->m_pkthdr.rcvif) { 738 sin = &ia->ia_addr; 739 break; 740 } 741 NEXT_IA_WITH_SAME_ADDR(ia); 742 } 743 } 744 } 745 746 /* 747 * if it was not addressed to us, but the route doesn't go out 748 * the source interface, pick an address on the source 749 * interface. This can happen when routing is asymmetric, or 750 * when the incoming packet was encapsulated 751 */ 752 if (sin == NULL && m->m_pkthdr.rcvif) { 753 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 754 if (ifa->ifa_addr->sa_family != AF_INET) 755 continue; 756 sin = &(ifatoia(ifa)->ia_addr); 757 break; 758 } 759 } 760 761 /* 762 * The following happens if the packet was not addressed to us, 763 * and was received on an interface with no IP address: 764 * We find the first AF_INET address on the first non-loopback 765 * interface. 766 */ 767 if (sin == NULL) 768 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) { 769 if (ia->ia_ifp->if_flags & IFF_LOOPBACK) 770 continue; 771 sin = &ia->ia_addr; 772 break; 773 } 774 775 /* 776 * If we still didn't find an address, punt. We could have an 777 * interface up (and receiving packets) with no address. 778 */ 779 if (sin == NULL) { 780 m_freem(m); 781 goto done; 782 } 783 784 ip->ip_src = sin->sin_addr; 785 ip->ip_ttl = MAXTTL; 786 787 if (optlen > 0) { 788 u_char *cp; 789 int opt, cnt; 790 u_int len; 791 792 /* 793 * Retrieve any source routing from the incoming packet; 794 * add on any record-route or timestamp options. 795 */ 796 cp = (u_char *) (ip + 1); 797 if ((opts = ip_srcroute()) == 0 && 798 (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { 799 MCLAIM(opts, m->m_owner); 800 opts->m_len = sizeof(struct in_addr); 801 *mtod(opts, struct in_addr *) = zeroin_addr; 802 } 803 if (opts) { 804 #ifdef ICMPPRINTFS 805 if (icmpprintfs) 806 printf("icmp_reflect optlen %d rt %d => ", 807 optlen, opts->m_len); 808 #endif 809 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 810 opt = cp[IPOPT_OPTVAL]; 811 if (opt == IPOPT_EOL) 812 break; 813 if (opt == IPOPT_NOP) 814 len = 1; 815 else { 816 if (cnt < IPOPT_OLEN + sizeof(*cp)) 817 break; 818 len = cp[IPOPT_OLEN]; 819 if (len < IPOPT_OLEN + sizeof(*cp) || 820 len > cnt) 821 break; 822 } 823 /* 824 * Should check for overflow, but it "can't happen" 825 */ 826 if (opt == IPOPT_RR || opt == IPOPT_TS || 827 opt == IPOPT_SECURITY) { 828 memmove(mtod(opts, char *) + opts->m_len, 829 cp, len); 830 opts->m_len += len; 831 } 832 } 833 /* Terminate & pad, if necessary */ 834 if ((cnt = opts->m_len % 4) != 0) { 835 for (; cnt < 4; cnt++) { 836 *(mtod(opts, char *) + opts->m_len) = 837 IPOPT_EOL; 838 opts->m_len++; 839 } 840 } 841 #ifdef ICMPPRINTFS 842 if (icmpprintfs) 843 printf("%d\n", opts->m_len); 844 #endif 845 } 846 /* 847 * Now strip out original options by copying rest of first 848 * mbuf's data back, and adjust the IP length. 849 */ 850 ip->ip_len = htons(ntohs(ip->ip_len) - optlen); 851 ip->ip_hl = sizeof(struct ip) >> 2; 852 m->m_len -= optlen; 853 if (m->m_flags & M_PKTHDR) 854 m->m_pkthdr.len -= optlen; 855 optlen += sizeof(struct ip); 856 memmove(ip + 1, (char *)ip + optlen, 857 (unsigned)(m->m_len - sizeof(struct ip))); 858 } 859 m_tag_delete_nonpersistent(m); 860 m->m_flags &= ~(M_BCAST|M_MCAST); 861 862 /* 863 * Clear any in-bound checksum flags for this packet. 864 */ 865 if (m->m_flags & M_PKTHDR) 866 m->m_pkthdr.csum_flags = 0; 867 868 icmp_send(m, opts); 869 done: 870 if (opts) 871 (void)m_free(opts); 872 } 873 874 /* 875 * Send an icmp packet back to the ip level, 876 * after supplying a checksum. 877 */ 878 void 879 icmp_send(struct mbuf *m, struct mbuf *opts) 880 { 881 struct ip *ip = mtod(m, struct ip *); 882 int hlen; 883 struct icmp *icp; 884 885 hlen = ip->ip_hl << 2; 886 m->m_data += hlen; 887 m->m_len -= hlen; 888 icp = mtod(m, struct icmp *); 889 icp->icmp_cksum = 0; 890 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 891 m->m_data -= hlen; 892 m->m_len += hlen; 893 #ifdef ICMPPRINTFS 894 if (icmpprintfs) { 895 printf("icmp_send to destination `%s' from `%s'\n", 896 inet_ntoa(ip->ip_dst), inet_ntoa(ip->ip_src)); 897 } 898 #endif 899 (void)ip_output(m, opts, NULL, 0, NULL, NULL); 900 } 901 902 n_time 903 iptime(void) 904 { 905 struct timeval atv; 906 u_long t; 907 908 microtime(&atv); 909 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 910 return (htonl(t)); 911 } 912 913 /* 914 * sysctl helper routine for net.inet.icmp.returndatabytes. ensures 915 * that the new value is in the correct range. 916 */ 917 static int 918 sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS) 919 { 920 int error, t; 921 struct sysctlnode node; 922 923 node = *rnode; 924 node.sysctl_data = &t; 925 t = icmpreturndatabytes; 926 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 927 if (error || newp == NULL) 928 return (error); 929 930 if (t < 8 || t > 512) 931 return (EINVAL); 932 icmpreturndatabytes = t; 933 934 return (0); 935 } 936 937 /* 938 * sysctl helper routine for net.inet.icmp.redirtimeout. ensures that 939 * the given value is not less than zero and then resets the timeout 940 * queue. 941 */ 942 static int 943 sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS) 944 { 945 int error, tmp; 946 struct sysctlnode node; 947 948 node = *rnode; 949 node.sysctl_data = &tmp; 950 tmp = icmp_redirtimeout; 951 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 952 if (error || newp == NULL) 953 return (error); 954 if (tmp < 0) 955 return (EINVAL); 956 icmp_redirtimeout = tmp; 957 958 /* 959 * was it a *defined* side-effect that anyone even *reading* 960 * this value causes these things to happen? 961 */ 962 if (icmp_redirect_timeout_q != NULL) { 963 if (icmp_redirtimeout == 0) { 964 rt_timer_queue_destroy(icmp_redirect_timeout_q, 965 true); 966 icmp_redirect_timeout_q = NULL; 967 } else { 968 rt_timer_queue_change(icmp_redirect_timeout_q, 969 icmp_redirtimeout); 970 } 971 } else if (icmp_redirtimeout > 0) { 972 icmp_redirect_timeout_q = 973 rt_timer_queue_create(icmp_redirtimeout); 974 } 975 976 return (0); 977 } 978 979 static int 980 sysctl_net_inet_icmp_stats(SYSCTLFN_ARGS) 981 { 982 983 return (NETSTAT_SYSCTL(icmpstat_percpu, ICMP_NSTATS)); 984 } 985 986 SYSCTL_SETUP(sysctl_net_inet_icmp_setup, "sysctl net.inet.icmp subtree setup") 987 { 988 989 sysctl_createv(clog, 0, NULL, NULL, 990 CTLFLAG_PERMANENT, 991 CTLTYPE_NODE, "net", NULL, 992 NULL, 0, NULL, 0, 993 CTL_NET, CTL_EOL); 994 sysctl_createv(clog, 0, NULL, NULL, 995 CTLFLAG_PERMANENT, 996 CTLTYPE_NODE, "inet", NULL, 997 NULL, 0, NULL, 0, 998 CTL_NET, PF_INET, CTL_EOL); 999 sysctl_createv(clog, 0, NULL, NULL, 1000 CTLFLAG_PERMANENT, 1001 CTLTYPE_NODE, "icmp", 1002 SYSCTL_DESCR("ICMPv4 related settings"), 1003 NULL, 0, NULL, 0, 1004 CTL_NET, PF_INET, IPPROTO_ICMP, CTL_EOL); 1005 1006 sysctl_createv(clog, 0, NULL, NULL, 1007 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1008 CTLTYPE_INT, "maskrepl", 1009 SYSCTL_DESCR("Respond to ICMP_MASKREQ messages"), 1010 NULL, 0, &icmpmaskrepl, 0, 1011 CTL_NET, PF_INET, IPPROTO_ICMP, 1012 ICMPCTL_MASKREPL, CTL_EOL); 1013 sysctl_createv(clog, 0, NULL, NULL, 1014 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1015 CTLTYPE_INT, "returndatabytes", 1016 SYSCTL_DESCR("Number of bytes to return in an ICMP " 1017 "error message"), 1018 sysctl_net_inet_icmp_returndatabytes, 0, 1019 &icmpreturndatabytes, 0, 1020 CTL_NET, PF_INET, IPPROTO_ICMP, 1021 ICMPCTL_RETURNDATABYTES, CTL_EOL); 1022 sysctl_createv(clog, 0, NULL, NULL, 1023 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1024 CTLTYPE_INT, "errppslimit", 1025 SYSCTL_DESCR("Maximum number of outgoing ICMP error " 1026 "messages per second"), 1027 NULL, 0, &icmperrppslim, 0, 1028 CTL_NET, PF_INET, IPPROTO_ICMP, 1029 ICMPCTL_ERRPPSLIMIT, CTL_EOL); 1030 sysctl_createv(clog, 0, NULL, NULL, 1031 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1032 CTLTYPE_INT, "rediraccept", 1033 SYSCTL_DESCR("Accept ICMP_REDIRECT messages"), 1034 NULL, 0, &icmp_rediraccept, 0, 1035 CTL_NET, PF_INET, IPPROTO_ICMP, 1036 ICMPCTL_REDIRACCEPT, CTL_EOL); 1037 sysctl_createv(clog, 0, NULL, NULL, 1038 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1039 CTLTYPE_INT, "redirtimeout", 1040 SYSCTL_DESCR("Lifetime of ICMP_REDIRECT generated " 1041 "routes"), 1042 sysctl_net_inet_icmp_redirtimeout, 0, 1043 &icmp_redirtimeout, 0, 1044 CTL_NET, PF_INET, IPPROTO_ICMP, 1045 ICMPCTL_REDIRTIMEOUT, CTL_EOL); 1046 sysctl_createv(clog, 0, NULL, NULL, 1047 CTLFLAG_PERMANENT, 1048 CTLTYPE_STRUCT, "stats", 1049 SYSCTL_DESCR("ICMP statistics"), 1050 sysctl_net_inet_icmp_stats, 0, NULL, 0, 1051 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_STATS, 1052 CTL_EOL); 1053 } 1054 1055 void 1056 icmp_statinc(u_int stat) 1057 { 1058 1059 KASSERT(stat < ICMP_NSTATS); 1060 ICMP_STATINC(stat); 1061 } 1062 1063 /* Table of common MTUs: */ 1064 1065 static const u_int mtu_table[] = { 1066 65535, 65280, 32000, 17914, 9180, 8166, 1067 4352, 2002, 1492, 1006, 508, 296, 68, 0 1068 }; 1069 1070 void 1071 icmp_mtudisc(struct icmp *icp, struct in_addr faddr) 1072 { 1073 struct icmp_mtudisc_callback *mc; 1074 struct sockaddr *dst = sintosa(&icmpsrc); 1075 struct rtentry *rt; 1076 u_long mtu = ntohs(icp->icmp_nextmtu); /* Why a long? IPv6 */ 1077 int error; 1078 1079 rt = rtalloc1(dst, 1); 1080 if (rt == 0) 1081 return; 1082 1083 /* If we didn't get a host route, allocate one */ 1084 1085 if ((rt->rt_flags & RTF_HOST) == 0) { 1086 struct rtentry *nrt; 1087 1088 error = rtrequest((int) RTM_ADD, dst, 1089 (struct sockaddr *) rt->rt_gateway, 1090 (struct sockaddr *) 0, 1091 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 1092 if (error) { 1093 rtfree(rt); 1094 return; 1095 } 1096 nrt->rt_rmx = rt->rt_rmx; 1097 rtfree(rt); 1098 rt = nrt; 1099 } 1100 error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q); 1101 if (error) { 1102 rtfree(rt); 1103 return; 1104 } 1105 1106 if (mtu == 0) { 1107 int i = 0; 1108 1109 mtu = ntohs(icp->icmp_ip.ip_len); 1110 /* Some 4.2BSD-based routers incorrectly adjust the ip_len */ 1111 if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0) 1112 mtu -= (icp->icmp_ip.ip_hl << 2); 1113 1114 /* If we still can't guess a value, try the route */ 1115 1116 if (mtu == 0) { 1117 mtu = rt->rt_rmx.rmx_mtu; 1118 1119 /* If no route mtu, default to the interface mtu */ 1120 1121 if (mtu == 0) 1122 mtu = rt->rt_ifp->if_mtu; 1123 } 1124 1125 for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++) 1126 if (mtu > mtu_table[i]) { 1127 mtu = mtu_table[i]; 1128 break; 1129 } 1130 } 1131 1132 /* 1133 * XXX: RTV_MTU is overloaded, since the admin can set it 1134 * to turn off PMTU for a route, and the kernel can 1135 * set it to indicate a serious problem with PMTU 1136 * on a route. We should be using a separate flag 1137 * for the kernel to indicate this. 1138 */ 1139 1140 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1141 if (mtu < 296 || mtu > rt->rt_ifp->if_mtu) 1142 rt->rt_rmx.rmx_locks |= RTV_MTU; 1143 else if (rt->rt_rmx.rmx_mtu > mtu || 1144 rt->rt_rmx.rmx_mtu == 0) { 1145 ICMP_STATINC(ICMP_STAT_PMTUCHG); 1146 rt->rt_rmx.rmx_mtu = mtu; 1147 } 1148 } 1149 1150 if (rt) 1151 rtfree(rt); 1152 1153 /* 1154 * Notify protocols that the MTU for this destination 1155 * has changed. 1156 */ 1157 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL; 1158 mc = LIST_NEXT(mc, mc_list)) 1159 (*mc->mc_func)(faddr); 1160 } 1161 1162 /* 1163 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1164 * given current value MTU. If DIR is less than zero, a larger plateau 1165 * is returned; otherwise, a smaller value is returned. 1166 */ 1167 u_int 1168 ip_next_mtu(u_int mtu, int dir) /* XXX */ 1169 { 1170 int i; 1171 1172 for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) { 1173 if (mtu >= mtu_table[i]) 1174 break; 1175 } 1176 1177 if (dir < 0) { 1178 if (i == 0) { 1179 return 0; 1180 } else { 1181 return mtu_table[i - 1]; 1182 } 1183 } else { 1184 if (mtu_table[i] == 0) { 1185 return 0; 1186 } else if (mtu > mtu_table[i]) { 1187 return mtu_table[i]; 1188 } else { 1189 return mtu_table[i + 1]; 1190 } 1191 } 1192 } 1193 1194 static void 1195 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 1196 { 1197 if (rt == NULL) 1198 panic("icmp_mtudisc_timeout: bad route to timeout"); 1199 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1200 (RTF_DYNAMIC | RTF_HOST)) { 1201 rtrequest((int) RTM_DELETE, rt_getkey(rt), 1202 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 1203 } else { 1204 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1205 rt->rt_rmx.rmx_mtu = 0; 1206 } 1207 } 1208 } 1209 1210 static void 1211 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r) 1212 { 1213 if (rt == NULL) 1214 panic("icmp_redirect_timeout: bad route to timeout"); 1215 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1216 (RTF_DYNAMIC | RTF_HOST)) { 1217 rtrequest((int) RTM_DELETE, rt_getkey(rt), 1218 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 1219 } 1220 } 1221 1222 /* 1223 * Perform rate limit check. 1224 * Returns 0 if it is okay to send the icmp packet. 1225 * Returns 1 if the router SHOULD NOT send this icmp packet due to rate 1226 * limitation. 1227 * 1228 * XXX per-destination/type check necessary? 1229 */ 1230 static int 1231 icmp_ratelimit(const struct in_addr *dst, const int type, 1232 const int code) 1233 { 1234 1235 /* PPS limit */ 1236 if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count, 1237 icmperrppslim)) { 1238 /* The packet is subject to rate limit */ 1239 return 1; 1240 } 1241 1242 /* okay to send */ 1243 return 0; 1244 } 1245