ip_icmp.c revision 1.120.6.1
1 /* $NetBSD: ip_icmp.c,v 1.120.6.1 2013/06/09 16:39:43 msaitoh 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 1.120.6.1 2013/06/09 16:39:43 msaitoh 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_PROTOCOL: 455 code = PRC_UNREACH_PROTOCOL; 456 break; 457 458 case ICMP_UNREACH_PORT: 459 code = PRC_UNREACH_PORT; 460 break; 461 462 case ICMP_UNREACH_SRCFAIL: 463 code = PRC_UNREACH_SRCFAIL; 464 break; 465 466 case ICMP_UNREACH_NEEDFRAG: 467 code = PRC_MSGSIZE; 468 break; 469 470 case ICMP_UNREACH_NET: 471 case ICMP_UNREACH_NET_UNKNOWN: 472 case ICMP_UNREACH_NET_PROHIB: 473 case ICMP_UNREACH_TOSNET: 474 code = PRC_UNREACH_NET; 475 break; 476 477 case ICMP_UNREACH_HOST: 478 case ICMP_UNREACH_HOST_UNKNOWN: 479 case ICMP_UNREACH_ISOLATED: 480 case ICMP_UNREACH_HOST_PROHIB: 481 case ICMP_UNREACH_TOSHOST: 482 case ICMP_UNREACH_ADMIN_PROHIBIT: 483 case ICMP_UNREACH_HOST_PREC: 484 case ICMP_UNREACH_PREC_CUTOFF: 485 code = PRC_UNREACH_HOST; 486 break; 487 488 default: 489 goto badcode; 490 } 491 goto deliver; 492 493 case ICMP_TIMXCEED: 494 if (code > 1) 495 goto badcode; 496 code += PRC_TIMXCEED_INTRANS; 497 goto deliver; 498 499 case ICMP_PARAMPROB: 500 if (code > 1) 501 goto badcode; 502 code = PRC_PARAMPROB; 503 goto deliver; 504 505 case ICMP_SOURCEQUENCH: 506 if (code) 507 goto badcode; 508 code = PRC_QUENCH; 509 goto deliver; 510 511 deliver: 512 /* 513 * Problem with datagram; advise higher level routines. 514 */ 515 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 516 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 517 ICMP_STATINC(ICMP_STAT_BADLEN); 518 goto freeit; 519 } 520 if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr)) 521 goto badcode; 522 #ifdef ICMPPRINTFS 523 if (icmpprintfs) 524 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 525 #endif 526 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 527 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 528 if (ctlfunc) 529 (void) (*ctlfunc)(code, sintosa(&icmpsrc), 530 &icp->icmp_ip); 531 break; 532 533 badcode: 534 ICMP_STATINC(ICMP_STAT_BADCODE); 535 break; 536 537 case ICMP_ECHO: 538 icp->icmp_type = ICMP_ECHOREPLY; 539 goto reflect; 540 541 case ICMP_TSTAMP: 542 if (icmplen < ICMP_TSLEN) { 543 ICMP_STATINC(ICMP_STAT_BADLEN); 544 break; 545 } 546 icp->icmp_type = ICMP_TSTAMPREPLY; 547 icp->icmp_rtime = iptime(); 548 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 549 goto reflect; 550 551 case ICMP_MASKREQ: 552 if (icmpmaskrepl == 0) 553 break; 554 /* 555 * We are not able to respond with all ones broadcast 556 * unless we receive it over a point-to-point interface. 557 */ 558 if (icmplen < ICMP_MASKLEN) { 559 ICMP_STATINC(ICMP_STAT_BADLEN); 560 break; 561 } 562 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 563 in_nullhost(ip->ip_dst)) 564 icmpdst.sin_addr = ip->ip_src; 565 else 566 icmpdst.sin_addr = ip->ip_dst; 567 ia = ifatoia(ifaof_ifpforaddr(sintosa(&icmpdst), 568 m->m_pkthdr.rcvif)); 569 if (ia == 0) 570 break; 571 icp->icmp_type = ICMP_MASKREPLY; 572 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 573 if (in_nullhost(ip->ip_src)) { 574 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 575 ip->ip_src = ia->ia_broadaddr.sin_addr; 576 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 577 ip->ip_src = ia->ia_dstaddr.sin_addr; 578 } 579 reflect: 580 { 581 uint64_t *icps = percpu_getref(icmpstat_percpu); 582 icps[ICMP_STAT_REFLECT]++; 583 icps[ICMP_STAT_OUTHIST + icp->icmp_type]++; 584 percpu_putref(icmpstat_percpu); 585 } 586 icmp_reflect(m); 587 return; 588 589 case ICMP_REDIRECT: 590 if (code > 3) 591 goto badcode; 592 if (icmp_rediraccept == 0) 593 goto freeit; 594 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 595 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 596 ICMP_STATINC(ICMP_STAT_BADLEN); 597 break; 598 } 599 /* 600 * Short circuit routing redirects to force 601 * immediate change in the kernel's routing 602 * tables. The message is also handed to anyone 603 * listening on a raw socket (e.g. the routing 604 * daemon for use in updating its tables). 605 */ 606 icmpgw.sin_addr = ip->ip_src; 607 icmpdst.sin_addr = icp->icmp_gwaddr; 608 #ifdef ICMPPRINTFS 609 if (icmpprintfs) { 610 printf("redirect dst `%s' to `%s'\n", 611 inet_ntoa(icp->icmp_ip.ip_dst), 612 inet_ntoa(icp->icmp_gwaddr)); 613 } 614 #endif 615 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 616 rt = NULL; 617 rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst), 618 NULL, RTF_GATEWAY | RTF_HOST, sintosa(&icmpgw), &rt); 619 if (rt != NULL && icmp_redirtimeout != 0) { 620 i = rt_timer_add(rt, icmp_redirect_timeout, 621 icmp_redirect_timeout_q); 622 if (i) 623 log(LOG_ERR, "ICMP: redirect failed to " 624 "register timeout for route to %x, " 625 "code %d\n", 626 icp->icmp_ip.ip_dst.s_addr, i); 627 } 628 if (rt != NULL) 629 rtfree(rt); 630 631 pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc)); 632 #if defined(IPSEC) || defined(FAST_IPSEC) 633 key_sa_routechange((struct sockaddr *)&icmpsrc); 634 #endif 635 break; 636 637 /* 638 * No kernel processing for the following; 639 * just fall through to send to raw listener. 640 */ 641 case ICMP_ECHOREPLY: 642 case ICMP_ROUTERADVERT: 643 case ICMP_ROUTERSOLICIT: 644 case ICMP_TSTAMPREPLY: 645 case ICMP_IREQREPLY: 646 case ICMP_MASKREPLY: 647 default: 648 break; 649 } 650 651 raw: 652 rip_input(m, hlen, proto); 653 return; 654 655 freeit: 656 m_freem(m); 657 return; 658 } 659 660 /* 661 * Reflect the ip packet back to the source 662 */ 663 void 664 icmp_reflect(struct mbuf *m) 665 { 666 struct ip *ip = mtod(m, struct ip *); 667 struct in_ifaddr *ia; 668 struct ifaddr *ifa; 669 struct sockaddr_in *sin = 0; 670 struct in_addr t; 671 struct mbuf *opts = 0; 672 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 673 674 if (!in_canforward(ip->ip_src) && 675 ((ip->ip_src.s_addr & IN_CLASSA_NET) != 676 htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { 677 m_freem(m); /* Bad return address */ 678 goto done; /* ip_output() will check for broadcast */ 679 } 680 t = ip->ip_dst; 681 ip->ip_dst = ip->ip_src; 682 /* 683 * If the incoming packet was addressed directly to us, use 684 * dst as the src for the reply. Otherwise (broadcast or 685 * anonymous), use an address which corresponds to the 686 * incoming interface, with a preference for the address which 687 * corresponds to the route to the destination of the ICMP. 688 */ 689 690 /* Look for packet addressed to us */ 691 INADDR_TO_IA(t, ia); 692 693 /* look for packet sent to broadcast address */ 694 if (ia == NULL && m->m_pkthdr.rcvif && 695 (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST)) { 696 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 697 if (ifa->ifa_addr->sa_family != AF_INET) 698 continue; 699 if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) { 700 ia = ifatoia(ifa); 701 break; 702 } 703 } 704 } 705 706 if (ia) 707 sin = &ia->ia_addr; 708 709 icmpdst.sin_addr = t; 710 711 /* 712 * if the packet is addressed somewhere else, compute the 713 * source address for packets routed back to the source, and 714 * use that, if it's an address on the interface which 715 * received the packet 716 */ 717 if (sin == NULL && m->m_pkthdr.rcvif) { 718 struct sockaddr_in sin_dst; 719 struct route icmproute; 720 int errornum; 721 722 sockaddr_in_init(&sin_dst, &ip->ip_dst, 0); 723 memset(&icmproute, 0, sizeof(icmproute)); 724 errornum = 0; 725 sin = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum); 726 /* errornum is never used */ 727 rtcache_free(&icmproute); 728 /* check to make sure sin is a source address on rcvif */ 729 if (sin) { 730 t = sin->sin_addr; 731 sin = NULL; 732 INADDR_TO_IA(t, ia); 733 while (ia) { 734 if (ia->ia_ifp == m->m_pkthdr.rcvif) { 735 sin = &ia->ia_addr; 736 break; 737 } 738 NEXT_IA_WITH_SAME_ADDR(ia); 739 } 740 } 741 } 742 743 /* 744 * if it was not addressed to us, but the route doesn't go out 745 * the source interface, pick an address on the source 746 * interface. This can happen when routing is asymmetric, or 747 * when the incoming packet was encapsulated 748 */ 749 if (sin == NULL && m->m_pkthdr.rcvif) { 750 IFADDR_FOREACH(ifa, m->m_pkthdr.rcvif) { 751 if (ifa->ifa_addr->sa_family != AF_INET) 752 continue; 753 sin = &(ifatoia(ifa)->ia_addr); 754 break; 755 } 756 } 757 758 /* 759 * The following happens if the packet was not addressed to us, 760 * and was received on an interface with no IP address: 761 * We find the first AF_INET address on the first non-loopback 762 * interface. 763 */ 764 if (sin == NULL) 765 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_list) { 766 if (ia->ia_ifp->if_flags & IFF_LOOPBACK) 767 continue; 768 sin = &ia->ia_addr; 769 break; 770 } 771 772 /* 773 * If we still didn't find an address, punt. We could have an 774 * interface up (and receiving packets) with no address. 775 */ 776 if (sin == NULL) { 777 m_freem(m); 778 goto done; 779 } 780 781 ip->ip_src = sin->sin_addr; 782 ip->ip_ttl = MAXTTL; 783 784 if (optlen > 0) { 785 u_char *cp; 786 int opt, cnt; 787 u_int len; 788 789 /* 790 * Retrieve any source routing from the incoming packet; 791 * add on any record-route or timestamp options. 792 */ 793 cp = (u_char *) (ip + 1); 794 if ((opts = ip_srcroute()) == 0 && 795 (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { 796 MCLAIM(opts, m->m_owner); 797 opts->m_len = sizeof(struct in_addr); 798 *mtod(opts, struct in_addr *) = zeroin_addr; 799 } 800 if (opts) { 801 #ifdef ICMPPRINTFS 802 if (icmpprintfs) 803 printf("icmp_reflect optlen %d rt %d => ", 804 optlen, opts->m_len); 805 #endif 806 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 807 opt = cp[IPOPT_OPTVAL]; 808 if (opt == IPOPT_EOL) 809 break; 810 if (opt == IPOPT_NOP) 811 len = 1; 812 else { 813 if (cnt < IPOPT_OLEN + sizeof(*cp)) 814 break; 815 len = cp[IPOPT_OLEN]; 816 if (len < IPOPT_OLEN + sizeof(*cp) || 817 len > cnt) 818 break; 819 } 820 /* 821 * Should check for overflow, but it "can't happen" 822 */ 823 if (opt == IPOPT_RR || opt == IPOPT_TS || 824 opt == IPOPT_SECURITY) { 825 memmove(mtod(opts, char *) + opts->m_len, 826 cp, len); 827 opts->m_len += len; 828 } 829 } 830 /* Terminate & pad, if necessary */ 831 if ((cnt = opts->m_len % 4) != 0) { 832 for (; cnt < 4; cnt++) { 833 *(mtod(opts, char *) + opts->m_len) = 834 IPOPT_EOL; 835 opts->m_len++; 836 } 837 } 838 #ifdef ICMPPRINTFS 839 if (icmpprintfs) 840 printf("%d\n", opts->m_len); 841 #endif 842 } 843 /* 844 * Now strip out original options by copying rest of first 845 * mbuf's data back, and adjust the IP length. 846 */ 847 ip->ip_len = htons(ntohs(ip->ip_len) - optlen); 848 ip->ip_hl = sizeof(struct ip) >> 2; 849 m->m_len -= optlen; 850 if (m->m_flags & M_PKTHDR) 851 m->m_pkthdr.len -= optlen; 852 optlen += sizeof(struct ip); 853 memmove(ip + 1, (char *)ip + optlen, 854 (unsigned)(m->m_len - sizeof(struct ip))); 855 } 856 m_tag_delete_nonpersistent(m); 857 m->m_flags &= ~(M_BCAST|M_MCAST); 858 859 /* 860 * Clear any in-bound checksum flags for this packet. 861 */ 862 if (m->m_flags & M_PKTHDR) 863 m->m_pkthdr.csum_flags = 0; 864 865 icmp_send(m, opts); 866 done: 867 if (opts) 868 (void)m_free(opts); 869 } 870 871 /* 872 * Send an icmp packet back to the ip level, 873 * after supplying a checksum. 874 */ 875 void 876 icmp_send(struct mbuf *m, struct mbuf *opts) 877 { 878 struct ip *ip = mtod(m, struct ip *); 879 int hlen; 880 struct icmp *icp; 881 882 hlen = ip->ip_hl << 2; 883 m->m_data += hlen; 884 m->m_len -= hlen; 885 icp = mtod(m, struct icmp *); 886 icp->icmp_cksum = 0; 887 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 888 m->m_data -= hlen; 889 m->m_len += hlen; 890 #ifdef ICMPPRINTFS 891 if (icmpprintfs) { 892 printf("icmp_send to destination `%s' from `%s'\n", 893 inet_ntoa(ip->ip_dst), inet_ntoa(ip->ip_src)); 894 } 895 #endif 896 (void)ip_output(m, opts, NULL, 0, NULL, NULL); 897 } 898 899 n_time 900 iptime(void) 901 { 902 struct timeval atv; 903 u_long t; 904 905 microtime(&atv); 906 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 907 return (htonl(t)); 908 } 909 910 /* 911 * sysctl helper routine for net.inet.icmp.returndatabytes. ensures 912 * that the new value is in the correct range. 913 */ 914 static int 915 sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS) 916 { 917 int error, t; 918 struct sysctlnode node; 919 920 node = *rnode; 921 node.sysctl_data = &t; 922 t = icmpreturndatabytes; 923 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 924 if (error || newp == NULL) 925 return (error); 926 927 if (t < 8 || t > 512) 928 return (EINVAL); 929 icmpreturndatabytes = t; 930 931 return (0); 932 } 933 934 /* 935 * sysctl helper routine for net.inet.icmp.redirtimeout. ensures that 936 * the given value is not less than zero and then resets the timeout 937 * queue. 938 */ 939 static int 940 sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS) 941 { 942 int error, tmp; 943 struct sysctlnode node; 944 945 node = *rnode; 946 node.sysctl_data = &tmp; 947 tmp = icmp_redirtimeout; 948 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 949 if (error || newp == NULL) 950 return (error); 951 if (tmp < 0) 952 return (EINVAL); 953 icmp_redirtimeout = tmp; 954 955 /* 956 * was it a *defined* side-effect that anyone even *reading* 957 * this value causes these things to happen? 958 */ 959 if (icmp_redirect_timeout_q != NULL) { 960 if (icmp_redirtimeout == 0) { 961 rt_timer_queue_destroy(icmp_redirect_timeout_q, 962 true); 963 icmp_redirect_timeout_q = NULL; 964 } else { 965 rt_timer_queue_change(icmp_redirect_timeout_q, 966 icmp_redirtimeout); 967 } 968 } else if (icmp_redirtimeout > 0) { 969 icmp_redirect_timeout_q = 970 rt_timer_queue_create(icmp_redirtimeout); 971 } 972 973 return (0); 974 } 975 976 static int 977 sysctl_net_inet_icmp_stats(SYSCTLFN_ARGS) 978 { 979 980 return (NETSTAT_SYSCTL(icmpstat_percpu, ICMP_NSTATS)); 981 } 982 983 SYSCTL_SETUP(sysctl_net_inet_icmp_setup, "sysctl net.inet.icmp subtree setup") 984 { 985 986 sysctl_createv(clog, 0, NULL, NULL, 987 CTLFLAG_PERMANENT, 988 CTLTYPE_NODE, "net", NULL, 989 NULL, 0, NULL, 0, 990 CTL_NET, CTL_EOL); 991 sysctl_createv(clog, 0, NULL, NULL, 992 CTLFLAG_PERMANENT, 993 CTLTYPE_NODE, "inet", NULL, 994 NULL, 0, NULL, 0, 995 CTL_NET, PF_INET, CTL_EOL); 996 sysctl_createv(clog, 0, NULL, NULL, 997 CTLFLAG_PERMANENT, 998 CTLTYPE_NODE, "icmp", 999 SYSCTL_DESCR("ICMPv4 related settings"), 1000 NULL, 0, NULL, 0, 1001 CTL_NET, PF_INET, IPPROTO_ICMP, CTL_EOL); 1002 1003 sysctl_createv(clog, 0, NULL, NULL, 1004 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1005 CTLTYPE_INT, "maskrepl", 1006 SYSCTL_DESCR("Respond to ICMP_MASKREQ messages"), 1007 NULL, 0, &icmpmaskrepl, 0, 1008 CTL_NET, PF_INET, IPPROTO_ICMP, 1009 ICMPCTL_MASKREPL, CTL_EOL); 1010 sysctl_createv(clog, 0, NULL, NULL, 1011 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1012 CTLTYPE_INT, "returndatabytes", 1013 SYSCTL_DESCR("Number of bytes to return in an ICMP " 1014 "error message"), 1015 sysctl_net_inet_icmp_returndatabytes, 0, 1016 &icmpreturndatabytes, 0, 1017 CTL_NET, PF_INET, IPPROTO_ICMP, 1018 ICMPCTL_RETURNDATABYTES, CTL_EOL); 1019 sysctl_createv(clog, 0, NULL, NULL, 1020 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1021 CTLTYPE_INT, "errppslimit", 1022 SYSCTL_DESCR("Maximum number of outgoing ICMP error " 1023 "messages per second"), 1024 NULL, 0, &icmperrppslim, 0, 1025 CTL_NET, PF_INET, IPPROTO_ICMP, 1026 ICMPCTL_ERRPPSLIMIT, CTL_EOL); 1027 sysctl_createv(clog, 0, NULL, NULL, 1028 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1029 CTLTYPE_INT, "rediraccept", 1030 SYSCTL_DESCR("Accept ICMP_REDIRECT messages"), 1031 NULL, 0, &icmp_rediraccept, 0, 1032 CTL_NET, PF_INET, IPPROTO_ICMP, 1033 ICMPCTL_REDIRACCEPT, CTL_EOL); 1034 sysctl_createv(clog, 0, NULL, NULL, 1035 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1036 CTLTYPE_INT, "redirtimeout", 1037 SYSCTL_DESCR("Lifetime of ICMP_REDIRECT generated " 1038 "routes"), 1039 sysctl_net_inet_icmp_redirtimeout, 0, 1040 &icmp_redirtimeout, 0, 1041 CTL_NET, PF_INET, IPPROTO_ICMP, 1042 ICMPCTL_REDIRTIMEOUT, CTL_EOL); 1043 sysctl_createv(clog, 0, NULL, NULL, 1044 CTLFLAG_PERMANENT, 1045 CTLTYPE_STRUCT, "stats", 1046 SYSCTL_DESCR("ICMP statistics"), 1047 sysctl_net_inet_icmp_stats, 0, NULL, 0, 1048 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_STATS, 1049 CTL_EOL); 1050 } 1051 1052 void 1053 icmp_statinc(u_int stat) 1054 { 1055 1056 KASSERT(stat < ICMP_NSTATS); 1057 ICMP_STATINC(stat); 1058 } 1059 1060 /* Table of common MTUs: */ 1061 1062 static const u_int mtu_table[] = { 1063 65535, 65280, 32000, 17914, 9180, 8166, 1064 4352, 2002, 1492, 1006, 508, 296, 68, 0 1065 }; 1066 1067 void 1068 icmp_mtudisc(struct icmp *icp, struct in_addr faddr) 1069 { 1070 struct icmp_mtudisc_callback *mc; 1071 struct sockaddr *dst = sintosa(&icmpsrc); 1072 struct rtentry *rt; 1073 u_long mtu = ntohs(icp->icmp_nextmtu); /* Why a long? IPv6 */ 1074 int error; 1075 1076 rt = rtalloc1(dst, 1); 1077 if (rt == 0) 1078 return; 1079 1080 /* If we didn't get a host route, allocate one */ 1081 1082 if ((rt->rt_flags & RTF_HOST) == 0) { 1083 struct rtentry *nrt; 1084 1085 error = rtrequest((int) RTM_ADD, dst, 1086 (struct sockaddr *) rt->rt_gateway, 1087 (struct sockaddr *) 0, 1088 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 1089 if (error) { 1090 rtfree(rt); 1091 return; 1092 } 1093 nrt->rt_rmx = rt->rt_rmx; 1094 rtfree(rt); 1095 rt = nrt; 1096 } 1097 error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q); 1098 if (error) { 1099 rtfree(rt); 1100 return; 1101 } 1102 1103 if (mtu == 0) { 1104 int i = 0; 1105 1106 mtu = ntohs(icp->icmp_ip.ip_len); 1107 /* Some 4.2BSD-based routers incorrectly adjust the ip_len */ 1108 if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0) 1109 mtu -= (icp->icmp_ip.ip_hl << 2); 1110 1111 /* If we still can't guess a value, try the route */ 1112 1113 if (mtu == 0) { 1114 mtu = rt->rt_rmx.rmx_mtu; 1115 1116 /* If no route mtu, default to the interface mtu */ 1117 1118 if (mtu == 0) 1119 mtu = rt->rt_ifp->if_mtu; 1120 } 1121 1122 for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++) 1123 if (mtu > mtu_table[i]) { 1124 mtu = mtu_table[i]; 1125 break; 1126 } 1127 } 1128 1129 /* 1130 * XXX: RTV_MTU is overloaded, since the admin can set it 1131 * to turn off PMTU for a route, and the kernel can 1132 * set it to indicate a serious problem with PMTU 1133 * on a route. We should be using a separate flag 1134 * for the kernel to indicate this. 1135 */ 1136 1137 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1138 if (mtu < 296 || mtu > rt->rt_ifp->if_mtu) 1139 rt->rt_rmx.rmx_locks |= RTV_MTU; 1140 else if (rt->rt_rmx.rmx_mtu > mtu || 1141 rt->rt_rmx.rmx_mtu == 0) { 1142 ICMP_STATINC(ICMP_STAT_PMTUCHG); 1143 rt->rt_rmx.rmx_mtu = mtu; 1144 } 1145 } 1146 1147 if (rt) 1148 rtfree(rt); 1149 1150 /* 1151 * Notify protocols that the MTU for this destination 1152 * has changed. 1153 */ 1154 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL; 1155 mc = LIST_NEXT(mc, mc_list)) 1156 (*mc->mc_func)(faddr); 1157 } 1158 1159 /* 1160 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1161 * given current value MTU. If DIR is less than zero, a larger plateau 1162 * is returned; otherwise, a smaller value is returned. 1163 */ 1164 u_int 1165 ip_next_mtu(u_int mtu, int dir) /* XXX */ 1166 { 1167 int i; 1168 1169 for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) { 1170 if (mtu >= mtu_table[i]) 1171 break; 1172 } 1173 1174 if (dir < 0) { 1175 if (i == 0) { 1176 return 0; 1177 } else { 1178 return mtu_table[i - 1]; 1179 } 1180 } else { 1181 if (mtu_table[i] == 0) { 1182 return 0; 1183 } else if (mtu > mtu_table[i]) { 1184 return mtu_table[i]; 1185 } else { 1186 return mtu_table[i + 1]; 1187 } 1188 } 1189 } 1190 1191 static void 1192 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 1193 { 1194 if (rt == NULL) 1195 panic("icmp_mtudisc_timeout: bad route to timeout"); 1196 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1197 (RTF_DYNAMIC | RTF_HOST)) { 1198 rtrequest((int) RTM_DELETE, rt_getkey(rt), 1199 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 1200 } else { 1201 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1202 rt->rt_rmx.rmx_mtu = 0; 1203 } 1204 } 1205 } 1206 1207 static void 1208 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r) 1209 { 1210 if (rt == NULL) 1211 panic("icmp_redirect_timeout: bad route to timeout"); 1212 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1213 (RTF_DYNAMIC | RTF_HOST)) { 1214 rtrequest((int) RTM_DELETE, rt_getkey(rt), 1215 rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); 1216 } 1217 } 1218 1219 /* 1220 * Perform rate limit check. 1221 * Returns 0 if it is okay to send the icmp packet. 1222 * Returns 1 if the router SHOULD NOT send this icmp packet due to rate 1223 * limitation. 1224 * 1225 * XXX per-destination/type check necessary? 1226 */ 1227 static int 1228 icmp_ratelimit(const struct in_addr *dst, const int type, 1229 const int code) 1230 { 1231 1232 /* PPS limit */ 1233 if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count, 1234 icmperrppslim)) { 1235 /* The packet is subject to rate limit */ 1236 return 1; 1237 } 1238 1239 /* okay to send */ 1240 return 0; 1241 } 1242
Indexes created Fri Oct 03 02:09:57 GMT 2025