1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1993 5 * The Regents of the University of California. 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 University 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 REGENTS 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 REGENTS 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 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_inet.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/mbuf.h> 42 #include <sys/protosw.h> 43 #include <sys/socket.h> 44 #include <sys/time.h> 45 #include <sys/kernel.h> 46 #include <sys/lock.h> 47 #include <sys/rmlock.h> 48 #include <sys/sysctl.h> 49 #include <sys/syslog.h> 50 51 #include <net/if.h> 52 #include <net/if_var.h> 53 #include <net/if_types.h> 54 #include <net/route.h> 55 #include <net/vnet.h> 56 57 #include <netinet/in.h> 58 #include <netinet/in_fib.h> 59 #include <netinet/in_pcb.h> 60 #include <netinet/in_systm.h> 61 #include <netinet/in_var.h> 62 #include <netinet/ip.h> 63 #include <netinet/ip_icmp.h> 64 #include <netinet/ip_var.h> 65 #include <netinet/ip_options.h> 66 #include <netinet/sctp.h> 67 #include <netinet/tcp.h> 68 #include <netinet/tcp_var.h> 69 #include <netinet/tcpip.h> 70 #include <netinet/icmp_var.h> 71 72 73 #ifdef INET 74 75 #include <machine/in_cksum.h> 76 77 #include <security/mac/mac_framework.h> 78 #endif /* INET */ 79 80 /* 81 * ICMP routines: error generation, receive packet processing, and 82 * routines to turnaround packets back to the originator, and 83 * host table maintenance routines. 84 */ 85 VNET_DEFINE_STATIC(int, icmplim) = 200; 86 #define V_icmplim VNET(icmplim) 87 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW, 88 &VNET_NAME(icmplim), 0, 89 "Maximum number of ICMP responses per second"); 90 91 VNET_DEFINE_STATIC(int, icmplim_output) = 1; 92 #define V_icmplim_output VNET(icmplim_output) 93 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW, 94 &VNET_NAME(icmplim_output), 0, 95 "Enable logging of ICMP response rate limiting"); 96 97 #ifdef INET 98 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat); 99 VNET_PCPUSTAT_SYSINIT(icmpstat); 100 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat, 101 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)"); 102 103 #ifdef VIMAGE 104 VNET_PCPUSTAT_SYSUNINIT(icmpstat); 105 #endif /* VIMAGE */ 106 107 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0; 108 #define V_icmpmaskrepl VNET(icmpmaskrepl) 109 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW, 110 &VNET_NAME(icmpmaskrepl), 0, 111 "Reply to ICMP Address Mask Request packets"); 112 113 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0; 114 #define V_icmpmaskfake VNET(icmpmaskfake) 115 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW, 116 &VNET_NAME(icmpmaskfake), 0, 117 "Fake reply to ICMP Address Mask Request packets"); 118 119 VNET_DEFINE(int, drop_redirect) = 0; 120 #define V_drop_redirect VNET(drop_redirect) 121 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW, 122 &VNET_NAME(drop_redirect), 0, 123 "Ignore ICMP redirects"); 124 125 VNET_DEFINE_STATIC(int, log_redirect) = 0; 126 #define V_log_redirect VNET(log_redirect) 127 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW, 128 &VNET_NAME(log_redirect), 0, 129 "Log ICMP redirects to the console"); 130 131 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */ 132 #define V_redirtimeout VNET(redirtimeout) 133 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW, 134 &VNET_NAME(redirtimeout), 0, 135 "Delay in seconds before expiring redirect route"); 136 137 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]); 138 #define V_reply_src VNET(reply_src) 139 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW, 140 &VNET_NAME(reply_src), IFNAMSIZ, 141 "ICMP reply source for non-local packets"); 142 143 VNET_DEFINE_STATIC(int, icmp_rfi) = 0; 144 #define V_icmp_rfi VNET(icmp_rfi) 145 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW, 146 &VNET_NAME(icmp_rfi), 0, 147 "ICMP reply from incoming interface for non-local packets"); 148 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */ 149 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548; 150 #define V_icmp_quotelen VNET(icmp_quotelen) 151 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW, 152 &VNET_NAME(icmp_quotelen), 0, 153 "Number of bytes from original packet to quote in ICMP reply"); 154 155 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0; 156 #define V_icmpbmcastecho VNET(icmpbmcastecho) 157 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW, 158 &VNET_NAME(icmpbmcastecho), 0, 159 "Reply to multicast ICMP Echo Request and Timestamp packets"); 160 161 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1; 162 #define V_icmptstamprepl VNET(icmptstamprepl) 163 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW, 164 &VNET_NAME(icmptstamprepl), 0, 165 "Respond to ICMP Timestamp packets"); 166 167 VNET_DEFINE_STATIC(int, error_keeptags) = 0; 168 #define V_error_keeptags VNET(error_keeptags) 169 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW, 170 &VNET_NAME(error_keeptags), 0, 171 "ICMP error response keeps copy of mbuf_tags of original packet"); 172 173 #ifdef ICMPPRINTFS 174 int icmpprintfs = 0; 175 #endif 176 177 static void icmp_reflect(struct mbuf *); 178 static void icmp_send(struct mbuf *, struct mbuf *); 179 static int icmp_verify_redirect_gateway(struct sockaddr_in *, 180 struct sockaddr_in *, struct sockaddr_in *, u_int); 181 182 extern struct protosw inetsw[]; 183 184 /* 185 * Kernel module interface for updating icmpstat. The argument is an index 186 * into icmpstat treated as an array of u_long. While this encodes the 187 * general layout of icmpstat into the caller, it doesn't encode its 188 * location, so that future changes to add, for example, per-CPU stats 189 * support won't cause binary compatibility problems for kernel modules. 190 */ 191 void 192 kmod_icmpstat_inc(int statnum) 193 { 194 195 counter_u64_add(VNET(icmpstat)[statnum], 1); 196 } 197 198 /* 199 * Generate an error packet of type error 200 * in response to bad packet ip. 201 */ 202 void 203 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu) 204 { 205 struct ip *oip, *nip; 206 struct icmp *icp; 207 struct mbuf *m; 208 unsigned icmplen, icmpelen, nlen, oiphlen; 209 210 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", 211 __func__)); 212 213 if (type != ICMP_REDIRECT) 214 ICMPSTAT_INC(icps_error); 215 /* 216 * Don't send error: 217 * if the original packet was encrypted. 218 * if not the first fragment of message. 219 * in response to a multicast or broadcast packet. 220 * if the old packet protocol was an ICMP error message. 221 */ 222 if (n->m_flags & M_DECRYPTED) 223 goto freeit; 224 if (n->m_flags & (M_BCAST|M_MCAST)) 225 goto freeit; 226 227 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */ 228 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) 229 goto freeit; 230 oip = mtod(n, struct ip *); 231 oiphlen = oip->ip_hl << 2; 232 if (n->m_len < oiphlen + ICMP_MINLEN) 233 goto freeit; 234 #ifdef ICMPPRINTFS 235 if (icmpprintfs) 236 printf("icmp_error(%p, %x, %d)\n", oip, type, code); 237 #endif 238 if (oip->ip_off & htons(~(IP_MF|IP_DF))) 239 goto freeit; 240 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 241 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + 242 oiphlen))->icmp_type)) { 243 ICMPSTAT_INC(icps_oldicmp); 244 goto freeit; 245 } 246 /* 247 * Calculate length to quote from original packet and 248 * prevent the ICMP mbuf from overflowing. 249 * Unfortunately this is non-trivial since ip_forward() 250 * sends us truncated packets. 251 */ 252 nlen = m_length(n, NULL); 253 if (oip->ip_p == IPPROTO_TCP) { 254 struct tcphdr *th; 255 int tcphlen; 256 257 if (oiphlen + sizeof(struct tcphdr) > n->m_len && 258 n->m_next == NULL) 259 goto stdreply; 260 if (n->m_len < oiphlen + sizeof(struct tcphdr) && 261 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL) 262 goto freeit; 263 oip = mtod(n, struct ip *); 264 th = mtodo(n, oiphlen); 265 tcphlen = th->th_off << 2; 266 if (tcphlen < sizeof(struct tcphdr)) 267 goto freeit; 268 if (ntohs(oip->ip_len) < oiphlen + tcphlen) 269 goto freeit; 270 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL) 271 goto stdreply; 272 if (n->m_len < oiphlen + tcphlen && 273 (n = m_pullup(n, oiphlen + tcphlen)) == NULL) 274 goto freeit; 275 oip = mtod(n, struct ip *); 276 icmpelen = max(tcphlen, min(V_icmp_quotelen, 277 ntohs(oip->ip_len) - oiphlen)); 278 } else if (oip->ip_p == IPPROTO_SCTP) { 279 struct sctphdr *sh; 280 struct sctp_chunkhdr *ch; 281 282 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr)) 283 goto stdreply; 284 if (oiphlen + sizeof(struct sctphdr) > n->m_len && 285 n->m_next == NULL) 286 goto stdreply; 287 if (n->m_len < oiphlen + sizeof(struct sctphdr) && 288 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL) 289 goto freeit; 290 oip = mtod(n, struct ip *); 291 icmpelen = max(sizeof(struct sctphdr), 292 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen)); 293 sh = mtodo(n, oiphlen); 294 if (ntohl(sh->v_tag) == 0 && 295 ntohs(oip->ip_len) >= oiphlen + 296 sizeof(struct sctphdr) + 8 && 297 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 || 298 n->m_next != NULL)) { 299 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 && 300 (n = m_pullup(n, oiphlen + 301 sizeof(struct sctphdr) + 8)) == NULL) 302 goto freeit; 303 oip = mtod(n, struct ip *); 304 sh = mtodo(n, oiphlen); 305 ch = (struct sctp_chunkhdr *)(sh + 1); 306 if (ch->chunk_type == SCTP_INITIATION) { 307 icmpelen = max(sizeof(struct sctphdr) + 8, 308 min(V_icmp_quotelen, ntohs(oip->ip_len) - 309 oiphlen)); 310 } 311 } 312 } else 313 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - 314 oiphlen)); 315 316 icmplen = min(oiphlen + icmpelen, nlen); 317 if (icmplen < sizeof(struct ip)) 318 goto freeit; 319 320 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen) 321 m = m_gethdr(M_NOWAIT, MT_DATA); 322 else 323 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 324 if (m == NULL) 325 goto freeit; 326 #ifdef MAC 327 mac_netinet_icmp_reply(n, m); 328 #endif 329 icmplen = min(icmplen, M_TRAILINGSPACE(m) - 330 sizeof(struct ip) - ICMP_MINLEN); 331 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen); 332 m->m_data += sizeof(struct ip); 333 m->m_len = ICMP_MINLEN + icmplen; 334 335 /* XXX MRT make the outgoing packet use the same FIB 336 * that was associated with the incoming packet 337 */ 338 M_SETFIB(m, M_GETFIB(n)); 339 icp = mtod(m, struct icmp *); 340 ICMPSTAT_INC(icps_outhist[type]); 341 icp->icmp_type = type; 342 if (type == ICMP_REDIRECT) 343 icp->icmp_gwaddr.s_addr = dest; 344 else { 345 icp->icmp_void = 0; 346 /* 347 * The following assignments assume an overlay with the 348 * just zeroed icmp_void field. 349 */ 350 if (type == ICMP_PARAMPROB) { 351 icp->icmp_pptr = code; 352 code = 0; 353 } else if (type == ICMP_UNREACH && 354 code == ICMP_UNREACH_NEEDFRAG && mtu) { 355 icp->icmp_nextmtu = htons(mtu); 356 } 357 } 358 icp->icmp_code = code; 359 360 /* 361 * Copy the quotation into ICMP message and 362 * convert quoted IP header back to network representation. 363 */ 364 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 365 nip = &icp->icmp_ip; 366 367 /* 368 * Set up ICMP message mbuf and copy old IP header (without options 369 * in front of ICMP message. 370 * If the original mbuf was meant to bypass the firewall, the error 371 * reply should bypass as well. 372 */ 373 m->m_flags |= n->m_flags & M_SKIP_FIREWALL; 374 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip), 375 ("insufficient space for ip header")); 376 m->m_data -= sizeof(struct ip); 377 m->m_len += sizeof(struct ip); 378 m->m_pkthdr.len = m->m_len; 379 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 380 nip = mtod(m, struct ip *); 381 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); 382 nip->ip_len = htons(m->m_len); 383 nip->ip_v = IPVERSION; 384 nip->ip_hl = 5; 385 nip->ip_p = IPPROTO_ICMP; 386 nip->ip_tos = 0; 387 nip->ip_off = 0; 388 389 if (V_error_keeptags) 390 m_tag_copy_chain(m, n, M_NOWAIT); 391 392 icmp_reflect(m); 393 394 freeit: 395 m_freem(n); 396 } 397 398 /* 399 * Process a received ICMP message. 400 */ 401 int 402 icmp_input(struct mbuf **mp, int *offp, int proto) 403 { 404 struct icmp *icp; 405 struct in_ifaddr *ia; 406 struct mbuf *m = *mp; 407 struct ip *ip = mtod(m, struct ip *); 408 struct sockaddr_in icmpsrc, icmpdst, icmpgw; 409 int hlen = *offp; 410 int icmplen = ntohs(ip->ip_len) - *offp; 411 int i, code; 412 void (*ctlfunc)(int, struct sockaddr *, void *); 413 int fibnum; 414 415 NET_EPOCH_ASSERT(); 416 417 *mp = NULL; 418 419 /* 420 * Locate icmp structure in mbuf, and check 421 * that not corrupted and of at least minimum length. 422 */ 423 #ifdef ICMPPRINTFS 424 if (icmpprintfs) { 425 char srcbuf[INET_ADDRSTRLEN]; 426 char dstbuf[INET_ADDRSTRLEN]; 427 428 printf("icmp_input from %s to %s, len %d\n", 429 inet_ntoa_r(ip->ip_src, srcbuf), 430 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen); 431 } 432 #endif 433 if (icmplen < ICMP_MINLEN) { 434 ICMPSTAT_INC(icps_tooshort); 435 goto freeit; 436 } 437 i = hlen + min(icmplen, ICMP_ADVLENMIN); 438 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 439 ICMPSTAT_INC(icps_tooshort); 440 return (IPPROTO_DONE); 441 } 442 ip = mtod(m, struct ip *); 443 m->m_len -= hlen; 444 m->m_data += hlen; 445 icp = mtod(m, struct icmp *); 446 if (in_cksum(m, icmplen)) { 447 ICMPSTAT_INC(icps_checksum); 448 goto freeit; 449 } 450 m->m_len += hlen; 451 m->m_data -= hlen; 452 453 #ifdef ICMPPRINTFS 454 if (icmpprintfs) 455 printf("icmp_input, type %d code %d\n", icp->icmp_type, 456 icp->icmp_code); 457 #endif 458 459 /* 460 * Message type specific processing. 461 */ 462 if (icp->icmp_type > ICMP_MAXTYPE) 463 goto raw; 464 465 /* Initialize */ 466 bzero(&icmpsrc, sizeof(icmpsrc)); 467 icmpsrc.sin_len = sizeof(struct sockaddr_in); 468 icmpsrc.sin_family = AF_INET; 469 bzero(&icmpdst, sizeof(icmpdst)); 470 icmpdst.sin_len = sizeof(struct sockaddr_in); 471 icmpdst.sin_family = AF_INET; 472 bzero(&icmpgw, sizeof(icmpgw)); 473 icmpgw.sin_len = sizeof(struct sockaddr_in); 474 icmpgw.sin_family = AF_INET; 475 476 ICMPSTAT_INC(icps_inhist[icp->icmp_type]); 477 code = icp->icmp_code; 478 switch (icp->icmp_type) { 479 480 case ICMP_UNREACH: 481 switch (code) { 482 case ICMP_UNREACH_NET: 483 case ICMP_UNREACH_HOST: 484 case ICMP_UNREACH_SRCFAIL: 485 case ICMP_UNREACH_NET_UNKNOWN: 486 case ICMP_UNREACH_HOST_UNKNOWN: 487 case ICMP_UNREACH_ISOLATED: 488 case ICMP_UNREACH_TOSNET: 489 case ICMP_UNREACH_TOSHOST: 490 case ICMP_UNREACH_HOST_PRECEDENCE: 491 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 492 code = PRC_UNREACH_NET; 493 break; 494 495 case ICMP_UNREACH_NEEDFRAG: 496 code = PRC_MSGSIZE; 497 break; 498 499 /* 500 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. 501 * Treat subcodes 2,3 as immediate RST 502 */ 503 case ICMP_UNREACH_PROTOCOL: 504 code = PRC_UNREACH_PROTOCOL; 505 break; 506 case ICMP_UNREACH_PORT: 507 code = PRC_UNREACH_PORT; 508 break; 509 510 case ICMP_UNREACH_NET_PROHIB: 511 case ICMP_UNREACH_HOST_PROHIB: 512 case ICMP_UNREACH_FILTER_PROHIB: 513 code = PRC_UNREACH_ADMIN_PROHIB; 514 break; 515 516 default: 517 goto badcode; 518 } 519 goto deliver; 520 521 case ICMP_TIMXCEED: 522 if (code > 1) 523 goto badcode; 524 code += PRC_TIMXCEED_INTRANS; 525 goto deliver; 526 527 case ICMP_PARAMPROB: 528 if (code > 1) 529 goto badcode; 530 code = PRC_PARAMPROB; 531 deliver: 532 /* 533 * Problem with datagram; advise higher level routines. 534 */ 535 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 536 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 537 ICMPSTAT_INC(icps_badlen); 538 goto freeit; 539 } 540 /* Discard ICMP's in response to multicast packets */ 541 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 542 goto badcode; 543 #ifdef ICMPPRINTFS 544 if (icmpprintfs) 545 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 546 #endif 547 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 548 /* 549 * XXX if the packet contains [IPv4 AH TCP], we can't make a 550 * notification to TCP layer. 551 */ 552 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp)); 553 ip_stripoptions(m); 554 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 555 /* This should actually not happen */ 556 ICMPSTAT_INC(icps_tooshort); 557 return (IPPROTO_DONE); 558 } 559 ip = mtod(m, struct ip *); 560 icp = (struct icmp *)(ip + 1); 561 /* 562 * The upper layer handler can rely on: 563 * - The outer IP header has no options. 564 * - The outer IP header, the ICMP header, the inner IP header, 565 * and the first n bytes of the inner payload are contiguous. 566 * n is at least 8, but might be larger based on 567 * ICMP_ADVLENPREF. See its definition in ip_icmp.h. 568 */ 569 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 570 if (ctlfunc) 571 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc, 572 (void *)&icp->icmp_ip); 573 break; 574 575 badcode: 576 ICMPSTAT_INC(icps_badcode); 577 break; 578 579 case ICMP_ECHO: 580 if (!V_icmpbmcastecho 581 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 582 ICMPSTAT_INC(icps_bmcastecho); 583 break; 584 } 585 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 586 goto freeit; 587 icp->icmp_type = ICMP_ECHOREPLY; 588 goto reflect; 589 590 case ICMP_TSTAMP: 591 if (V_icmptstamprepl == 0) 592 break; 593 if (!V_icmpbmcastecho 594 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 595 ICMPSTAT_INC(icps_bmcasttstamp); 596 break; 597 } 598 if (icmplen < ICMP_TSLEN) { 599 ICMPSTAT_INC(icps_badlen); 600 break; 601 } 602 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 603 goto freeit; 604 icp->icmp_type = ICMP_TSTAMPREPLY; 605 icp->icmp_rtime = iptime(); 606 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 607 goto reflect; 608 609 case ICMP_MASKREQ: 610 if (V_icmpmaskrepl == 0) 611 break; 612 /* 613 * We are not able to respond with all ones broadcast 614 * unless we receive it over a point-to-point interface. 615 */ 616 if (icmplen < ICMP_MASKLEN) 617 break; 618 switch (ip->ip_dst.s_addr) { 619 620 case INADDR_BROADCAST: 621 case INADDR_ANY: 622 icmpdst.sin_addr = ip->ip_src; 623 break; 624 625 default: 626 icmpdst.sin_addr = ip->ip_dst; 627 } 628 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 629 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 630 if (ia == NULL) 631 break; 632 if (ia->ia_ifp == NULL) 633 break; 634 icp->icmp_type = ICMP_MASKREPLY; 635 if (V_icmpmaskfake == 0) 636 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 637 else 638 icp->icmp_mask = V_icmpmaskfake; 639 if (ip->ip_src.s_addr == 0) { 640 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 641 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 642 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 643 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 644 } 645 reflect: 646 ICMPSTAT_INC(icps_reflect); 647 ICMPSTAT_INC(icps_outhist[icp->icmp_type]); 648 icmp_reflect(m); 649 return (IPPROTO_DONE); 650 651 case ICMP_REDIRECT: 652 if (V_log_redirect) { 653 u_long src, dst, gw; 654 655 src = ntohl(ip->ip_src.s_addr); 656 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 657 gw = ntohl(icp->icmp_gwaddr.s_addr); 658 printf("icmp redirect from %d.%d.%d.%d: " 659 "%d.%d.%d.%d => %d.%d.%d.%d\n", 660 (int)(src >> 24), (int)((src >> 16) & 0xff), 661 (int)((src >> 8) & 0xff), (int)(src & 0xff), 662 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 663 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 664 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 665 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 666 } 667 /* 668 * RFC1812 says we must ignore ICMP redirects if we 669 * are acting as router. 670 */ 671 if (V_drop_redirect || V_ipforwarding) 672 break; 673 if (code > 3) 674 goto badcode; 675 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 676 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 677 ICMPSTAT_INC(icps_badlen); 678 break; 679 } 680 /* 681 * Short circuit routing redirects to force 682 * immediate change in the kernel's routing 683 * tables. The message is also handed to anyone 684 * listening on a raw socket (e.g. the routing 685 * daemon for use in updating its tables). 686 */ 687 icmpgw.sin_addr = ip->ip_src; 688 icmpdst.sin_addr = icp->icmp_gwaddr; 689 #ifdef ICMPPRINTFS 690 if (icmpprintfs) { 691 char dstbuf[INET_ADDRSTRLEN]; 692 char gwbuf[INET_ADDRSTRLEN]; 693 694 printf("redirect dst %s to %s\n", 695 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf), 696 inet_ntoa_r(icp->icmp_gwaddr, gwbuf)); 697 } 698 #endif 699 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 700 701 /* 702 * RFC 1122 says network (code 0,2) redirects SHOULD 703 * be treated identically to the host redirects. 704 * Given that, ignore network masks. 705 */ 706 707 /* 708 * Variable values: 709 * icmpsrc: route destination 710 * icmpdst: route gateway 711 * icmpgw: message source 712 */ 713 714 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst, 715 M_GETFIB(m)) != 0) { 716 /* TODO: increment bad redirects here */ 717 break; 718 } 719 720 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 721 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc, 722 (struct sockaddr *)&icmpdst, 723 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif, 724 RTF_GATEWAY, V_redirtimeout); 725 } 726 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); 727 break; 728 729 /* 730 * No kernel processing for the following; 731 * just fall through to send to raw listener. 732 */ 733 case ICMP_ECHOREPLY: 734 case ICMP_ROUTERADVERT: 735 case ICMP_ROUTERSOLICIT: 736 case ICMP_TSTAMPREPLY: 737 case ICMP_IREQREPLY: 738 case ICMP_MASKREPLY: 739 case ICMP_SOURCEQUENCH: 740 default: 741 break; 742 } 743 744 raw: 745 *mp = m; 746 rip_input(mp, offp, proto); 747 return (IPPROTO_DONE); 748 749 freeit: 750 m_freem(m); 751 return (IPPROTO_DONE); 752 } 753 754 /* 755 * Reflect the ip packet back to the source 756 */ 757 static void 758 icmp_reflect(struct mbuf *m) 759 { 760 struct rm_priotracker in_ifa_tracker; 761 struct ip *ip = mtod(m, struct ip *); 762 struct ifaddr *ifa; 763 struct ifnet *ifp; 764 struct in_ifaddr *ia; 765 struct in_addr t; 766 struct nhop4_extended nh_ext; 767 struct mbuf *opts = NULL; 768 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 769 770 NET_EPOCH_ASSERT(); 771 772 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 773 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) || 774 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) { 775 m_freem(m); /* Bad return address */ 776 ICMPSTAT_INC(icps_badaddr); 777 goto done; /* Ip_output() will check for broadcast */ 778 } 779 780 t = ip->ip_dst; 781 ip->ip_dst = ip->ip_src; 782 783 /* 784 * Source selection for ICMP replies: 785 * 786 * If the incoming packet was addressed directly to one of our 787 * own addresses, use dst as the src for the reply. 788 */ 789 IN_IFADDR_RLOCK(&in_ifa_tracker); 790 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { 791 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { 792 t = IA_SIN(ia)->sin_addr; 793 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 794 goto match; 795 } 796 } 797 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 798 799 /* 800 * If the incoming packet was addressed to one of our broadcast 801 * addresses, use the first non-broadcast address which corresponds 802 * to the incoming interface. 803 */ 804 ifp = m->m_pkthdr.rcvif; 805 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 806 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 807 if (ifa->ifa_addr->sa_family != AF_INET) 808 continue; 809 ia = ifatoia(ifa); 810 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 811 t.s_addr) { 812 t = IA_SIN(ia)->sin_addr; 813 goto match; 814 } 815 } 816 } 817 /* 818 * If the packet was transiting through us, use the address of 819 * the interface the packet came through in. If that interface 820 * doesn't have a suitable IP address, the normal selection 821 * criteria apply. 822 */ 823 if (V_icmp_rfi && ifp != NULL) { 824 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 825 if (ifa->ifa_addr->sa_family != AF_INET) 826 continue; 827 ia = ifatoia(ifa); 828 t = IA_SIN(ia)->sin_addr; 829 goto match; 830 } 831 } 832 /* 833 * If the incoming packet was not addressed directly to us, use 834 * designated interface for icmp replies specified by sysctl 835 * net.inet.icmp.reply_src (default not set). Otherwise continue 836 * with normal source selection. 837 */ 838 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) { 839 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 840 if (ifa->ifa_addr->sa_family != AF_INET) 841 continue; 842 ia = ifatoia(ifa); 843 t = IA_SIN(ia)->sin_addr; 844 goto match; 845 } 846 } 847 /* 848 * If the packet was transiting through us, use the address of 849 * the interface that is the closest to the packet source. 850 * When we don't have a route back to the packet source, stop here 851 * and drop the packet. 852 */ 853 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) { 854 m_freem(m); 855 ICMPSTAT_INC(icps_noroute); 856 goto done; 857 } 858 t = nh_ext.nh_src; 859 match: 860 #ifdef MAC 861 mac_netinet_icmp_replyinplace(m); 862 #endif 863 ip->ip_src = t; 864 ip->ip_ttl = V_ip_defttl; 865 866 if (optlen > 0) { 867 u_char *cp; 868 int opt, cnt; 869 u_int len; 870 871 /* 872 * Retrieve any source routing from the incoming packet; 873 * add on any record-route or timestamp options. 874 */ 875 cp = (u_char *) (ip + 1); 876 if ((opts = ip_srcroute(m)) == NULL && 877 (opts = m_gethdr(M_NOWAIT, MT_DATA))) { 878 opts->m_len = sizeof(struct in_addr); 879 mtod(opts, struct in_addr *)->s_addr = 0; 880 } 881 if (opts) { 882 #ifdef ICMPPRINTFS 883 if (icmpprintfs) 884 printf("icmp_reflect optlen %d rt %d => ", 885 optlen, opts->m_len); 886 #endif 887 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 888 opt = cp[IPOPT_OPTVAL]; 889 if (opt == IPOPT_EOL) 890 break; 891 if (opt == IPOPT_NOP) 892 len = 1; 893 else { 894 if (cnt < IPOPT_OLEN + sizeof(*cp)) 895 break; 896 len = cp[IPOPT_OLEN]; 897 if (len < IPOPT_OLEN + sizeof(*cp) || 898 len > cnt) 899 break; 900 } 901 /* 902 * Should check for overflow, but it "can't happen" 903 */ 904 if (opt == IPOPT_RR || opt == IPOPT_TS || 905 opt == IPOPT_SECURITY) { 906 bcopy((caddr_t)cp, 907 mtod(opts, caddr_t) + opts->m_len, len); 908 opts->m_len += len; 909 } 910 } 911 /* Terminate & pad, if necessary */ 912 cnt = opts->m_len % 4; 913 if (cnt) { 914 for (; cnt < 4; cnt++) { 915 *(mtod(opts, caddr_t) + opts->m_len) = 916 IPOPT_EOL; 917 opts->m_len++; 918 } 919 } 920 #ifdef ICMPPRINTFS 921 if (icmpprintfs) 922 printf("%d\n", opts->m_len); 923 #endif 924 } 925 ip_stripoptions(m); 926 } 927 m_tag_delete_nonpersistent(m); 928 m->m_flags &= ~(M_BCAST|M_MCAST); 929 icmp_send(m, opts); 930 done: 931 if (opts) 932 (void)m_free(opts); 933 } 934 935 /* 936 * Verifies if redirect message is valid, according to RFC 1122 937 * 938 * @src: sockaddr with address of redirect originator 939 * @dst: sockaddr with destination in question 940 * @gateway: new proposed gateway 941 * 942 * Returns 0 on success. 943 */ 944 static int 945 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst, 946 struct sockaddr_in *gateway, u_int fibnum) 947 { 948 struct rtentry *rt; 949 struct ifaddr *ifa; 950 951 NET_EPOCH_ASSERT(); 952 953 /* Verify the gateway is directly reachable. */ 954 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL) 955 return (ENETUNREACH); 956 957 /* TODO: fib-aware. */ 958 if (ifa_ifwithaddr_check((struct sockaddr *)gateway)) 959 return (EHOSTUNREACH); 960 961 rt = rtalloc1_fib((struct sockaddr *)dst, 0, 0UL, fibnum); /* NB: rt is locked */ 962 if (rt == NULL) 963 return (EINVAL); 964 965 /* 966 * If the redirect isn't from our current router for this dst, 967 * it's either old or wrong. If it redirects us to ourselves, 968 * we have a routing loop, perhaps as a result of an interface 969 * going down recently. 970 */ 971 if (!sa_equal((struct sockaddr *)src, rt->rt_gateway)) { 972 RTFREE_LOCKED(rt); 973 return (EINVAL); 974 } 975 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) { 976 RTFREE_LOCKED(rt); 977 return (EINVAL); 978 } 979 980 /* If host route already exists, ignore redirect. */ 981 if (rt->rt_flags & RTF_HOST) { 982 RTFREE_LOCKED(rt); 983 return (EEXIST); 984 } 985 986 /* If the prefix is directly reachable, ignore redirect. */ 987 if (!(rt->rt_flags & RTF_GATEWAY)) { 988 RTFREE_LOCKED(rt); 989 return (EEXIST); 990 } 991 992 RTFREE_LOCKED(rt); 993 return (0); 994 } 995 996 997 /* 998 * Send an icmp packet back to the ip level, 999 * after supplying a checksum. 1000 */ 1001 static void 1002 icmp_send(struct mbuf *m, struct mbuf *opts) 1003 { 1004 struct ip *ip = mtod(m, struct ip *); 1005 int hlen; 1006 struct icmp *icp; 1007 1008 hlen = ip->ip_hl << 2; 1009 m->m_data += hlen; 1010 m->m_len -= hlen; 1011 icp = mtod(m, struct icmp *); 1012 icp->icmp_cksum = 0; 1013 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 1014 m->m_data -= hlen; 1015 m->m_len += hlen; 1016 m->m_pkthdr.rcvif = (struct ifnet *)0; 1017 #ifdef ICMPPRINTFS 1018 if (icmpprintfs) { 1019 char dstbuf[INET_ADDRSTRLEN]; 1020 char srcbuf[INET_ADDRSTRLEN]; 1021 1022 printf("icmp_send dst %s src %s\n", 1023 inet_ntoa_r(ip->ip_dst, dstbuf), 1024 inet_ntoa_r(ip->ip_src, srcbuf)); 1025 } 1026 #endif 1027 (void) ip_output(m, opts, NULL, 0, NULL, NULL); 1028 } 1029 1030 /* 1031 * Return milliseconds since 00:00 UTC in network format. 1032 */ 1033 uint32_t 1034 iptime(void) 1035 { 1036 struct timeval atv; 1037 u_long t; 1038 1039 getmicrotime(&atv); 1040 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1041 return (htonl(t)); 1042 } 1043 1044 /* 1045 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1046 * given current value MTU. If DIR is less than zero, a larger plateau 1047 * is returned; otherwise, a smaller value is returned. 1048 */ 1049 int 1050 ip_next_mtu(int mtu, int dir) 1051 { 1052 static int mtutab[] = { 1053 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 1054 296, 68, 0 1055 }; 1056 int i, size; 1057 1058 size = (sizeof mtutab) / (sizeof mtutab[0]); 1059 if (dir >= 0) { 1060 for (i = 0; i < size; i++) 1061 if (mtu > mtutab[i]) 1062 return mtutab[i]; 1063 } else { 1064 for (i = size - 1; i >= 0; i--) 1065 if (mtu < mtutab[i]) 1066 return mtutab[i]; 1067 if (mtu == mtutab[0]) 1068 return mtutab[0]; 1069 } 1070 return 0; 1071 } 1072 #endif /* INET */ 1073 1074 1075 /* 1076 * badport_bandlim() - check for ICMP bandwidth limit 1077 * 1078 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1079 * hit our bandwidth limit and it is not ok. 1080 * 1081 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1082 * 1083 * For now we separate the TCP and UDP subsystems w/ different 'which' 1084 * values. We may eventually remove this separation (and simplify the 1085 * code further). 1086 * 1087 * Note that the printing of the error message is delayed so we can 1088 * properly print the icmp error rate that the system was trying to do 1089 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1090 * the 'final' error, but it doesn't make sense to solve the printing 1091 * delay with more complex code. 1092 */ 1093 struct icmp_rate { 1094 const char *descr; 1095 struct counter_rate cr; 1096 }; 1097 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = { 1098 { "icmp unreach response" }, 1099 { "icmp ping response" }, 1100 { "icmp tstamp response" }, 1101 { "closed port RST response" }, 1102 { "open port RST response" }, 1103 { "icmp6 unreach response" }, 1104 { "sctp ootb response" } 1105 }; 1106 #define V_icmp_rates VNET(icmp_rates) 1107 1108 static void 1109 icmp_bandlimit_init(void) 1110 { 1111 1112 for (int i = 0; i < BANDLIM_MAX; i++) { 1113 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK); 1114 V_icmp_rates[i].cr.cr_ticks = ticks; 1115 } 1116 } 1117 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, 1118 icmp_bandlimit_init, NULL); 1119 1120 static void 1121 icmp_bandlimit_uninit(void) 1122 { 1123 1124 for (int i = 0; i < BANDLIM_MAX; i++) 1125 counter_u64_free(V_icmp_rates[i].cr.cr_rate); 1126 } 1127 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1128 icmp_bandlimit_uninit, NULL); 1129 1130 int 1131 badport_bandlim(int which) 1132 { 1133 int64_t pps; 1134 1135 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED) 1136 return (0); 1137 1138 KASSERT(which >= 0 && which < BANDLIM_MAX, 1139 ("%s: which %d", __func__, which)); 1140 1141 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim); 1142 if (pps == -1) 1143 return (-1); 1144 if (pps > 0 && V_icmplim_output) 1145 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n", 1146 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim); 1147 return (0); 1148 } 1149