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/sysctl.h> 48 #include <sys/syslog.h> 49 50 #include <net/if.h> 51 #include <net/if_var.h> 52 #include <net/if_types.h> 53 #include <net/route.h> 54 #include <net/route/route_ctl.h> 55 #include <net/route/nhop.h> 56 #include <net/vnet.h> 57 58 #include <netinet/in.h> 59 #include <netinet/in_fib.h> 60 #include <netinet/in_pcb.h> 61 #include <netinet/in_systm.h> 62 #include <netinet/in_var.h> 63 #include <netinet/ip.h> 64 #include <netinet/ip_icmp.h> 65 #include <netinet/ip_var.h> 66 #include <netinet/ip_options.h> 67 #include <netinet/sctp.h> 68 #include <netinet/tcp.h> 69 #include <netinet/tcp_var.h> 70 #include <netinet/tcpip.h> 71 #include <netinet/icmp_var.h> 72 73 #ifdef INET 74 75 #include <machine/in_cksum.h> 76 77 #include <security/mac/mac_framework.h> 78 #endif /* INET */ 79 80 extern ipproto_ctlinput_t *ip_ctlprotox[]; 81 82 /* 83 * ICMP routines: error generation, receive packet processing, and 84 * routines to turnaround packets back to the originator, and 85 * host table maintenance routines. 86 */ 87 VNET_DEFINE_STATIC(int, icmplim) = 200; 88 #define V_icmplim VNET(icmplim) 89 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW, 90 &VNET_NAME(icmplim), 0, 91 "Maximum number of ICMP responses per second"); 92 93 VNET_DEFINE_STATIC(int, icmplim_curr_jitter) = 0; 94 #define V_icmplim_curr_jitter VNET(icmplim_curr_jitter) 95 VNET_DEFINE_STATIC(int, icmplim_jitter) = 16; 96 #define V_icmplim_jitter VNET(icmplim_jitter) 97 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_jitter, CTLFLAG_VNET | CTLFLAG_RW, 98 &VNET_NAME(icmplim_jitter), 0, 99 "Random icmplim jitter adjustment limit"); 100 101 VNET_DEFINE_STATIC(int, icmplim_output) = 1; 102 #define V_icmplim_output VNET(icmplim_output) 103 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW, 104 &VNET_NAME(icmplim_output), 0, 105 "Enable logging of ICMP response rate limiting"); 106 107 #ifdef INET 108 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat); 109 VNET_PCPUSTAT_SYSINIT(icmpstat); 110 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat, 111 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)"); 112 113 #ifdef VIMAGE 114 VNET_PCPUSTAT_SYSUNINIT(icmpstat); 115 #endif /* VIMAGE */ 116 117 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0; 118 #define V_icmpmaskrepl VNET(icmpmaskrepl) 119 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW, 120 &VNET_NAME(icmpmaskrepl), 0, 121 "Reply to ICMP Address Mask Request packets"); 122 123 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0; 124 #define V_icmpmaskfake VNET(icmpmaskfake) 125 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW, 126 &VNET_NAME(icmpmaskfake), 0, 127 "Fake reply to ICMP Address Mask Request packets"); 128 129 VNET_DEFINE(int, drop_redirect) = 0; 130 #define V_drop_redirect VNET(drop_redirect) 131 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW, 132 &VNET_NAME(drop_redirect), 0, 133 "Ignore ICMP redirects"); 134 135 VNET_DEFINE_STATIC(int, log_redirect) = 0; 136 #define V_log_redirect VNET(log_redirect) 137 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW, 138 &VNET_NAME(log_redirect), 0, 139 "Log ICMP redirects to the console"); 140 141 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */ 142 #define V_redirtimeout VNET(redirtimeout) 143 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW, 144 &VNET_NAME(redirtimeout), 0, 145 "Delay in seconds before expiring redirect route"); 146 147 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]); 148 #define V_reply_src VNET(reply_src) 149 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW, 150 &VNET_NAME(reply_src), IFNAMSIZ, 151 "ICMP reply source for non-local packets"); 152 153 VNET_DEFINE_STATIC(int, icmp_rfi) = 0; 154 #define V_icmp_rfi VNET(icmp_rfi) 155 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW, 156 &VNET_NAME(icmp_rfi), 0, 157 "ICMP reply from incoming interface for non-local packets"); 158 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */ 159 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548; 160 #define V_icmp_quotelen VNET(icmp_quotelen) 161 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW, 162 &VNET_NAME(icmp_quotelen), 0, 163 "Number of bytes from original packet to quote in ICMP reply"); 164 165 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0; 166 #define V_icmpbmcastecho VNET(icmpbmcastecho) 167 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW, 168 &VNET_NAME(icmpbmcastecho), 0, 169 "Reply to multicast ICMP Echo Request and Timestamp packets"); 170 171 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1; 172 #define V_icmptstamprepl VNET(icmptstamprepl) 173 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW, 174 &VNET_NAME(icmptstamprepl), 0, 175 "Respond to ICMP Timestamp packets"); 176 177 VNET_DEFINE_STATIC(int, error_keeptags) = 0; 178 #define V_error_keeptags VNET(error_keeptags) 179 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW, 180 &VNET_NAME(error_keeptags), 0, 181 "ICMP error response keeps copy of mbuf_tags of original packet"); 182 183 #ifdef ICMPPRINTFS 184 int icmpprintfs = 0; 185 #endif 186 187 static void icmp_reflect(struct mbuf *); 188 static void icmp_send(struct mbuf *, struct mbuf *); 189 static int icmp_verify_redirect_gateway(struct sockaddr_in *, 190 struct sockaddr_in *, struct sockaddr_in *, u_int); 191 192 /* 193 * Kernel module interface for updating icmpstat. The argument is an index 194 * into icmpstat treated as an array of u_long. While this encodes the 195 * general layout of icmpstat into the caller, it doesn't encode its 196 * location, so that future changes to add, for example, per-CPU stats 197 * support won't cause binary compatibility problems for kernel modules. 198 */ 199 void 200 kmod_icmpstat_inc(int statnum) 201 { 202 203 counter_u64_add(VNET(icmpstat)[statnum], 1); 204 } 205 206 /* 207 * Generate an error packet of type error 208 * in response to bad packet ip. 209 */ 210 void 211 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu) 212 { 213 struct ip *oip, *nip; 214 struct icmp *icp; 215 struct mbuf *m; 216 unsigned icmplen, icmpelen, nlen, oiphlen; 217 218 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", 219 __func__)); 220 221 if (type != ICMP_REDIRECT) 222 ICMPSTAT_INC(icps_error); 223 /* 224 * Don't send error: 225 * if the original packet was encrypted. 226 * if not the first fragment of message. 227 * in response to a multicast or broadcast packet. 228 * if the old packet protocol was an ICMP error message. 229 */ 230 if (n->m_flags & M_DECRYPTED) 231 goto freeit; 232 if (n->m_flags & (M_BCAST|M_MCAST)) 233 goto freeit; 234 235 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */ 236 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) 237 goto freeit; 238 oip = mtod(n, struct ip *); 239 oiphlen = oip->ip_hl << 2; 240 if (n->m_len < oiphlen + ICMP_MINLEN) 241 goto freeit; 242 #ifdef ICMPPRINTFS 243 if (icmpprintfs) 244 printf("icmp_error(%p, %x, %d)\n", oip, type, code); 245 #endif 246 if (oip->ip_off & htons(~(IP_MF|IP_DF))) 247 goto freeit; 248 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 249 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + 250 oiphlen))->icmp_type)) { 251 ICMPSTAT_INC(icps_oldicmp); 252 goto freeit; 253 } 254 /* 255 * Calculate length to quote from original packet and 256 * prevent the ICMP mbuf from overflowing. 257 * Unfortunately this is non-trivial since ip_forward() 258 * sends us truncated packets. 259 */ 260 nlen = m_length(n, NULL); 261 if (oip->ip_p == IPPROTO_TCP) { 262 struct tcphdr *th; 263 int tcphlen; 264 265 if (oiphlen + sizeof(struct tcphdr) > n->m_len && 266 n->m_next == NULL) 267 goto stdreply; 268 if (n->m_len < oiphlen + sizeof(struct tcphdr) && 269 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL) 270 goto freeit; 271 oip = mtod(n, struct ip *); 272 th = mtodo(n, oiphlen); 273 tcphlen = th->th_off << 2; 274 if (tcphlen < sizeof(struct tcphdr)) 275 goto freeit; 276 if (ntohs(oip->ip_len) < oiphlen + tcphlen) 277 goto freeit; 278 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL) 279 goto stdreply; 280 if (n->m_len < oiphlen + tcphlen && 281 (n = m_pullup(n, oiphlen + tcphlen)) == NULL) 282 goto freeit; 283 oip = mtod(n, struct ip *); 284 icmpelen = max(tcphlen, min(V_icmp_quotelen, 285 ntohs(oip->ip_len) - oiphlen)); 286 } else if (oip->ip_p == IPPROTO_SCTP) { 287 struct sctphdr *sh; 288 struct sctp_chunkhdr *ch; 289 290 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr)) 291 goto stdreply; 292 if (oiphlen + sizeof(struct sctphdr) > n->m_len && 293 n->m_next == NULL) 294 goto stdreply; 295 if (n->m_len < oiphlen + sizeof(struct sctphdr) && 296 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL) 297 goto freeit; 298 oip = mtod(n, struct ip *); 299 icmpelen = max(sizeof(struct sctphdr), 300 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen)); 301 sh = mtodo(n, oiphlen); 302 if (ntohl(sh->v_tag) == 0 && 303 ntohs(oip->ip_len) >= oiphlen + 304 sizeof(struct sctphdr) + 8 && 305 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 || 306 n->m_next != NULL)) { 307 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 && 308 (n = m_pullup(n, oiphlen + 309 sizeof(struct sctphdr) + 8)) == NULL) 310 goto freeit; 311 oip = mtod(n, struct ip *); 312 sh = mtodo(n, oiphlen); 313 ch = (struct sctp_chunkhdr *)(sh + 1); 314 if (ch->chunk_type == SCTP_INITIATION) { 315 icmpelen = max(sizeof(struct sctphdr) + 8, 316 min(V_icmp_quotelen, ntohs(oip->ip_len) - 317 oiphlen)); 318 } 319 } 320 } else 321 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - 322 oiphlen)); 323 324 icmplen = min(oiphlen + icmpelen, nlen); 325 if (icmplen < sizeof(struct ip)) 326 goto freeit; 327 328 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen) 329 m = m_gethdr(M_NOWAIT, MT_DATA); 330 else 331 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 332 if (m == NULL) 333 goto freeit; 334 #ifdef MAC 335 mac_netinet_icmp_reply(n, m); 336 #endif 337 icmplen = min(icmplen, M_TRAILINGSPACE(m) - 338 sizeof(struct ip) - ICMP_MINLEN); 339 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen); 340 m->m_data += sizeof(struct ip); 341 m->m_len = ICMP_MINLEN + icmplen; 342 343 /* XXX MRT make the outgoing packet use the same FIB 344 * that was associated with the incoming packet 345 */ 346 M_SETFIB(m, M_GETFIB(n)); 347 icp = mtod(m, struct icmp *); 348 ICMPSTAT_INC(icps_outhist[type]); 349 icp->icmp_type = type; 350 if (type == ICMP_REDIRECT) 351 icp->icmp_gwaddr.s_addr = dest; 352 else { 353 icp->icmp_void = 0; 354 /* 355 * The following assignments assume an overlay with the 356 * just zeroed icmp_void field. 357 */ 358 if (type == ICMP_PARAMPROB) { 359 icp->icmp_pptr = code; 360 code = 0; 361 } else if (type == ICMP_UNREACH && 362 code == ICMP_UNREACH_NEEDFRAG && mtu) { 363 icp->icmp_nextmtu = htons(mtu); 364 } 365 } 366 icp->icmp_code = code; 367 368 /* 369 * Copy the quotation into ICMP message and 370 * convert quoted IP header back to network representation. 371 */ 372 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 373 nip = &icp->icmp_ip; 374 375 /* 376 * Set up ICMP message mbuf and copy old IP header (without options 377 * in front of ICMP message. 378 * If the original mbuf was meant to bypass the firewall, the error 379 * reply should bypass as well. 380 */ 381 m->m_flags |= n->m_flags & M_SKIP_FIREWALL; 382 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip), 383 ("insufficient space for ip header")); 384 m->m_data -= sizeof(struct ip); 385 m->m_len += sizeof(struct ip); 386 m->m_pkthdr.len = m->m_len; 387 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 388 nip = mtod(m, struct ip *); 389 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); 390 nip->ip_len = htons(m->m_len); 391 nip->ip_v = IPVERSION; 392 nip->ip_hl = 5; 393 nip->ip_p = IPPROTO_ICMP; 394 nip->ip_tos = 0; 395 nip->ip_off = 0; 396 397 if (V_error_keeptags) 398 m_tag_copy_chain(m, n, M_NOWAIT); 399 400 icmp_reflect(m); 401 402 freeit: 403 m_freem(n); 404 } 405 406 int 407 icmp_errmap(const struct icmp *icp) 408 { 409 410 switch (icp->icmp_type) { 411 case ICMP_UNREACH: 412 switch (icp->icmp_code) { 413 case ICMP_UNREACH_NET: 414 case ICMP_UNREACH_HOST: 415 case ICMP_UNREACH_SRCFAIL: 416 case ICMP_UNREACH_NET_UNKNOWN: 417 case ICMP_UNREACH_HOST_UNKNOWN: 418 case ICMP_UNREACH_ISOLATED: 419 case ICMP_UNREACH_TOSNET: 420 case ICMP_UNREACH_TOSHOST: 421 case ICMP_UNREACH_HOST_PRECEDENCE: 422 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 423 return (EHOSTUNREACH); 424 case ICMP_UNREACH_NEEDFRAG: 425 return (EMSGSIZE); 426 case ICMP_UNREACH_PROTOCOL: 427 case ICMP_UNREACH_PORT: 428 case ICMP_UNREACH_NET_PROHIB: 429 case ICMP_UNREACH_HOST_PROHIB: 430 case ICMP_UNREACH_FILTER_PROHIB: 431 return (ECONNREFUSED); 432 default: 433 return (0); 434 } 435 case ICMP_TIMXCEED: 436 switch (icp->icmp_code) { 437 case ICMP_TIMXCEED_INTRANS: 438 return (EHOSTUNREACH); 439 default: 440 return (0); 441 } 442 case ICMP_PARAMPROB: 443 switch (icp->icmp_code) { 444 case ICMP_PARAMPROB_ERRATPTR: 445 case ICMP_PARAMPROB_OPTABSENT: 446 return (ENOPROTOOPT); 447 default: 448 return (0); 449 } 450 default: 451 return (0); 452 } 453 } 454 455 /* 456 * Process a received ICMP message. 457 */ 458 int 459 icmp_input(struct mbuf **mp, int *offp, int proto) 460 { 461 struct icmp *icp; 462 struct in_ifaddr *ia; 463 struct mbuf *m = *mp; 464 struct ip *ip = mtod(m, struct ip *); 465 struct sockaddr_in icmpsrc, icmpdst, icmpgw; 466 int hlen = *offp; 467 int icmplen = ntohs(ip->ip_len) - *offp; 468 int i, code; 469 int fibnum; 470 471 NET_EPOCH_ASSERT(); 472 473 *mp = NULL; 474 475 /* 476 * Locate icmp structure in mbuf, and check 477 * that not corrupted and of at least minimum length. 478 */ 479 #ifdef ICMPPRINTFS 480 if (icmpprintfs) { 481 char srcbuf[INET_ADDRSTRLEN]; 482 char dstbuf[INET_ADDRSTRLEN]; 483 484 printf("icmp_input from %s to %s, len %d\n", 485 inet_ntoa_r(ip->ip_src, srcbuf), 486 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen); 487 } 488 #endif 489 if (icmplen < ICMP_MINLEN) { 490 ICMPSTAT_INC(icps_tooshort); 491 goto freeit; 492 } 493 i = hlen + min(icmplen, ICMP_ADVLENMIN); 494 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 495 ICMPSTAT_INC(icps_tooshort); 496 return (IPPROTO_DONE); 497 } 498 ip = mtod(m, struct ip *); 499 m->m_len -= hlen; 500 m->m_data += hlen; 501 icp = mtod(m, struct icmp *); 502 if (in_cksum(m, icmplen)) { 503 ICMPSTAT_INC(icps_checksum); 504 goto freeit; 505 } 506 m->m_len += hlen; 507 m->m_data -= hlen; 508 509 #ifdef ICMPPRINTFS 510 if (icmpprintfs) 511 printf("icmp_input, type %d code %d\n", icp->icmp_type, 512 icp->icmp_code); 513 #endif 514 515 /* 516 * Message type specific processing. 517 */ 518 if (icp->icmp_type > ICMP_MAXTYPE) 519 goto raw; 520 521 /* Initialize */ 522 bzero(&icmpsrc, sizeof(icmpsrc)); 523 icmpsrc.sin_len = sizeof(struct sockaddr_in); 524 icmpsrc.sin_family = AF_INET; 525 bzero(&icmpdst, sizeof(icmpdst)); 526 icmpdst.sin_len = sizeof(struct sockaddr_in); 527 icmpdst.sin_family = AF_INET; 528 bzero(&icmpgw, sizeof(icmpgw)); 529 icmpgw.sin_len = sizeof(struct sockaddr_in); 530 icmpgw.sin_family = AF_INET; 531 532 ICMPSTAT_INC(icps_inhist[icp->icmp_type]); 533 code = icp->icmp_code; 534 switch (icp->icmp_type) { 535 case ICMP_UNREACH: 536 if (code > ICMP_UNREACH_PRECEDENCE_CUTOFF) 537 goto badcode; 538 else 539 goto deliver; 540 541 case ICMP_TIMXCEED: 542 if (code > ICMP_TIMXCEED_REASS) 543 goto badcode; 544 else 545 goto deliver; 546 547 case ICMP_PARAMPROB: 548 if (code > ICMP_PARAMPROB_LENGTH) 549 goto badcode; 550 551 deliver: 552 /* 553 * Problem with datagram; advise higher level routines. 554 */ 555 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 556 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 557 ICMPSTAT_INC(icps_badlen); 558 goto freeit; 559 } 560 /* Discard ICMP's in response to multicast packets */ 561 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 562 goto badcode; 563 /* Filter out responses to INADDR_ANY, protocols ignore it. */ 564 if (icp->icmp_ip.ip_dst.s_addr == INADDR_ANY) 565 goto freeit; 566 #ifdef ICMPPRINTFS 567 if (icmpprintfs) 568 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 569 #endif 570 /* 571 * XXX if the packet contains [IPv4 AH TCP], we can't make a 572 * notification to TCP layer. 573 */ 574 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp)); 575 ip_stripoptions(m); 576 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 577 /* This should actually not happen */ 578 ICMPSTAT_INC(icps_tooshort); 579 return (IPPROTO_DONE); 580 } 581 ip = mtod(m, struct ip *); 582 icp = (struct icmp *)(ip + 1); 583 /* 584 * The upper layer handler can rely on: 585 * - The outer IP header has no options. 586 * - The outer IP header, the ICMP header, the inner IP header, 587 * and the first n bytes of the inner payload are contiguous. 588 * n is at least 8, but might be larger based on 589 * ICMP_ADVLENPREF. See its definition in ip_icmp.h. 590 */ 591 if (ip_ctlprotox[icp->icmp_ip.ip_p] != NULL) 592 ip_ctlprotox[icp->icmp_ip.ip_p](icp); 593 break; 594 595 badcode: 596 ICMPSTAT_INC(icps_badcode); 597 break; 598 599 case ICMP_ECHO: 600 if (!V_icmpbmcastecho 601 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 602 ICMPSTAT_INC(icps_bmcastecho); 603 break; 604 } 605 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 606 goto freeit; 607 icp->icmp_type = ICMP_ECHOREPLY; 608 goto reflect; 609 610 case ICMP_TSTAMP: 611 if (V_icmptstamprepl == 0) 612 break; 613 if (!V_icmpbmcastecho 614 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 615 ICMPSTAT_INC(icps_bmcasttstamp); 616 break; 617 } 618 if (icmplen < ICMP_TSLEN) { 619 ICMPSTAT_INC(icps_badlen); 620 break; 621 } 622 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 623 goto freeit; 624 icp->icmp_type = ICMP_TSTAMPREPLY; 625 icp->icmp_rtime = iptime(); 626 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 627 goto reflect; 628 629 case ICMP_MASKREQ: 630 if (V_icmpmaskrepl == 0) 631 break; 632 /* 633 * We are not able to respond with all ones broadcast 634 * unless we receive it over a point-to-point interface. 635 */ 636 if (icmplen < ICMP_MASKLEN) 637 break; 638 switch (ip->ip_dst.s_addr) { 639 case INADDR_BROADCAST: 640 case INADDR_ANY: 641 icmpdst.sin_addr = ip->ip_src; 642 break; 643 644 default: 645 icmpdst.sin_addr = ip->ip_dst; 646 } 647 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 648 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 649 if (ia == NULL) 650 break; 651 if (ia->ia_ifp == NULL) 652 break; 653 icp->icmp_type = ICMP_MASKREPLY; 654 if (V_icmpmaskfake == 0) 655 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 656 else 657 icp->icmp_mask = V_icmpmaskfake; 658 if (ip->ip_src.s_addr == 0) { 659 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 660 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 661 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 662 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 663 } 664 reflect: 665 ICMPSTAT_INC(icps_reflect); 666 ICMPSTAT_INC(icps_outhist[icp->icmp_type]); 667 icmp_reflect(m); 668 return (IPPROTO_DONE); 669 670 case ICMP_REDIRECT: 671 if (V_log_redirect) { 672 u_long src, dst, gw; 673 674 src = ntohl(ip->ip_src.s_addr); 675 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 676 gw = ntohl(icp->icmp_gwaddr.s_addr); 677 printf("icmp redirect from %d.%d.%d.%d: " 678 "%d.%d.%d.%d => %d.%d.%d.%d\n", 679 (int)(src >> 24), (int)((src >> 16) & 0xff), 680 (int)((src >> 8) & 0xff), (int)(src & 0xff), 681 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 682 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 683 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 684 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 685 } 686 /* 687 * RFC1812 says we must ignore ICMP redirects if we 688 * are acting as router. 689 */ 690 if (V_drop_redirect || V_ipforwarding) 691 break; 692 if (code > 3) 693 goto badcode; 694 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 695 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 696 ICMPSTAT_INC(icps_badlen); 697 break; 698 } 699 /* 700 * Short circuit routing redirects to force 701 * immediate change in the kernel's routing 702 * tables. The message is also handed to anyone 703 * listening on a raw socket (e.g. the routing 704 * daemon for use in updating its tables). 705 */ 706 icmpgw.sin_addr = ip->ip_src; 707 icmpdst.sin_addr = icp->icmp_gwaddr; 708 #ifdef ICMPPRINTFS 709 if (icmpprintfs) { 710 char dstbuf[INET_ADDRSTRLEN]; 711 char gwbuf[INET_ADDRSTRLEN]; 712 713 printf("redirect dst %s to %s\n", 714 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf), 715 inet_ntoa_r(icp->icmp_gwaddr, gwbuf)); 716 } 717 #endif 718 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 719 720 /* 721 * RFC 1122 says network (code 0,2) redirects SHOULD 722 * be treated identically to the host redirects. 723 * Given that, ignore network masks. 724 */ 725 726 /* 727 * Variable values: 728 * icmpsrc: route destination 729 * icmpdst: route gateway 730 * icmpgw: message source 731 */ 732 733 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst, 734 M_GETFIB(m)) != 0) { 735 /* TODO: increment bad redirects here */ 736 break; 737 } 738 739 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 740 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc, 741 (struct sockaddr *)&icmpdst, 742 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif, 743 RTF_GATEWAY, V_redirtimeout); 744 } 745 break; 746 747 /* 748 * No kernel processing for the following; 749 * just fall through to send to raw listener. 750 */ 751 case ICMP_ECHOREPLY: 752 case ICMP_ROUTERADVERT: 753 case ICMP_ROUTERSOLICIT: 754 case ICMP_TSTAMPREPLY: 755 case ICMP_IREQREPLY: 756 case ICMP_MASKREPLY: 757 case ICMP_SOURCEQUENCH: 758 default: 759 break; 760 } 761 762 raw: 763 *mp = m; 764 rip_input(mp, offp, proto); 765 return (IPPROTO_DONE); 766 767 freeit: 768 m_freem(m); 769 return (IPPROTO_DONE); 770 } 771 772 /* 773 * Reflect the ip packet back to the source 774 */ 775 static void 776 icmp_reflect(struct mbuf *m) 777 { 778 struct ip *ip = mtod(m, struct ip *); 779 struct ifaddr *ifa; 780 struct ifnet *ifp; 781 struct in_ifaddr *ia; 782 struct in_addr t; 783 struct nhop_object *nh; 784 struct mbuf *opts = NULL; 785 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 786 787 NET_EPOCH_ASSERT(); 788 789 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 790 (IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net240) || 791 (IN_ZERONET(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net0) ) { 792 m_freem(m); /* Bad return address */ 793 ICMPSTAT_INC(icps_badaddr); 794 goto done; /* Ip_output() will check for broadcast */ 795 } 796 797 t = ip->ip_dst; 798 ip->ip_dst = ip->ip_src; 799 800 /* 801 * Source selection for ICMP replies: 802 * 803 * If the incoming packet was addressed directly to one of our 804 * own addresses, use dst as the src for the reply. 805 */ 806 CK_LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { 807 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { 808 t = IA_SIN(ia)->sin_addr; 809 goto match; 810 } 811 } 812 813 /* 814 * If the incoming packet was addressed to one of our broadcast 815 * addresses, use the first non-broadcast address which corresponds 816 * to the incoming interface. 817 */ 818 ifp = m->m_pkthdr.rcvif; 819 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 820 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 821 if (ifa->ifa_addr->sa_family != AF_INET) 822 continue; 823 ia = ifatoia(ifa); 824 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 825 t.s_addr) { 826 t = IA_SIN(ia)->sin_addr; 827 goto match; 828 } 829 } 830 } 831 /* 832 * If the packet was transiting through us, use the address of 833 * the interface the packet came through in. If that interface 834 * doesn't have a suitable IP address, the normal selection 835 * criteria apply. 836 */ 837 if (V_icmp_rfi && ifp != NULL) { 838 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 839 if (ifa->ifa_addr->sa_family != AF_INET) 840 continue; 841 ia = ifatoia(ifa); 842 t = IA_SIN(ia)->sin_addr; 843 goto match; 844 } 845 } 846 /* 847 * If the incoming packet was not addressed directly to us, use 848 * designated interface for icmp replies specified by sysctl 849 * net.inet.icmp.reply_src (default not set). Otherwise continue 850 * with normal source selection. 851 */ 852 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) { 853 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 854 if (ifa->ifa_addr->sa_family != AF_INET) 855 continue; 856 ia = ifatoia(ifa); 857 t = IA_SIN(ia)->sin_addr; 858 goto match; 859 } 860 } 861 /* 862 * If the packet was transiting through us, use the address of 863 * the interface that is the closest to the packet source. 864 * When we don't have a route back to the packet source, stop here 865 * and drop the packet. 866 */ 867 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0); 868 if (nh == NULL) { 869 m_freem(m); 870 ICMPSTAT_INC(icps_noroute); 871 goto done; 872 } 873 t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr; 874 match: 875 #ifdef MAC 876 mac_netinet_icmp_replyinplace(m); 877 #endif 878 ip->ip_src = t; 879 ip->ip_ttl = V_ip_defttl; 880 881 if (optlen > 0) { 882 u_char *cp; 883 int opt, cnt; 884 u_int len; 885 886 /* 887 * Retrieve any source routing from the incoming packet; 888 * add on any record-route or timestamp options. 889 */ 890 cp = (u_char *) (ip + 1); 891 if ((opts = ip_srcroute(m)) == NULL && 892 (opts = m_gethdr(M_NOWAIT, MT_DATA))) { 893 opts->m_len = sizeof(struct in_addr); 894 mtod(opts, struct in_addr *)->s_addr = 0; 895 } 896 if (opts) { 897 #ifdef ICMPPRINTFS 898 if (icmpprintfs) 899 printf("icmp_reflect optlen %d rt %d => ", 900 optlen, opts->m_len); 901 #endif 902 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 903 opt = cp[IPOPT_OPTVAL]; 904 if (opt == IPOPT_EOL) 905 break; 906 if (opt == IPOPT_NOP) 907 len = 1; 908 else { 909 if (cnt < IPOPT_OLEN + sizeof(*cp)) 910 break; 911 len = cp[IPOPT_OLEN]; 912 if (len < IPOPT_OLEN + sizeof(*cp) || 913 len > cnt) 914 break; 915 } 916 /* 917 * Should check for overflow, but it "can't happen" 918 */ 919 if (opt == IPOPT_RR || opt == IPOPT_TS || 920 opt == IPOPT_SECURITY) { 921 bcopy((caddr_t)cp, 922 mtod(opts, caddr_t) + opts->m_len, len); 923 opts->m_len += len; 924 } 925 } 926 /* Terminate & pad, if necessary */ 927 cnt = opts->m_len % 4; 928 if (cnt) { 929 for (; cnt < 4; cnt++) { 930 *(mtod(opts, caddr_t) + opts->m_len) = 931 IPOPT_EOL; 932 opts->m_len++; 933 } 934 } 935 #ifdef ICMPPRINTFS 936 if (icmpprintfs) 937 printf("%d\n", opts->m_len); 938 #endif 939 } 940 ip_stripoptions(m); 941 } 942 m_tag_delete_nonpersistent(m); 943 m->m_flags &= ~(M_BCAST|M_MCAST); 944 icmp_send(m, opts); 945 done: 946 if (opts) 947 (void)m_free(opts); 948 } 949 950 /* 951 * Verifies if redirect message is valid, according to RFC 1122 952 * 953 * @src: sockaddr with address of redirect originator 954 * @dst: sockaddr with destination in question 955 * @gateway: new proposed gateway 956 * 957 * Returns 0 on success. 958 */ 959 static int 960 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst, 961 struct sockaddr_in *gateway, u_int fibnum) 962 { 963 struct nhop_object *nh; 964 struct ifaddr *ifa; 965 966 NET_EPOCH_ASSERT(); 967 968 /* Verify the gateway is directly reachable. */ 969 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL) 970 return (ENETUNREACH); 971 972 /* TODO: fib-aware. */ 973 if (ifa_ifwithaddr_check((struct sockaddr *)gateway)) 974 return (EHOSTUNREACH); 975 976 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0); 977 if (nh == NULL) 978 return (EINVAL); 979 980 /* 981 * If the redirect isn't from our current router for this dst, 982 * it's either old or wrong. If it redirects us to ourselves, 983 * we have a routing loop, perhaps as a result of an interface 984 * going down recently. 985 */ 986 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa)) 987 return (EINVAL); 988 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) 989 return (EINVAL); 990 991 /* If host route already exists, ignore redirect. */ 992 if (nh->nh_flags & NHF_HOST) 993 return (EEXIST); 994 995 /* If the prefix is directly reachable, ignore redirect. */ 996 if (!(nh->nh_flags & NHF_GATEWAY)) 997 return (EEXIST); 998 999 return (0); 1000 } 1001 1002 /* 1003 * Send an icmp packet back to the ip level, 1004 * after supplying a checksum. 1005 */ 1006 static void 1007 icmp_send(struct mbuf *m, struct mbuf *opts) 1008 { 1009 struct ip *ip = mtod(m, struct ip *); 1010 int hlen; 1011 struct icmp *icp; 1012 1013 hlen = ip->ip_hl << 2; 1014 m->m_data += hlen; 1015 m->m_len -= hlen; 1016 icp = mtod(m, struct icmp *); 1017 icp->icmp_cksum = 0; 1018 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 1019 m->m_data -= hlen; 1020 m->m_len += hlen; 1021 m->m_pkthdr.rcvif = (struct ifnet *)0; 1022 #ifdef ICMPPRINTFS 1023 if (icmpprintfs) { 1024 char dstbuf[INET_ADDRSTRLEN]; 1025 char srcbuf[INET_ADDRSTRLEN]; 1026 1027 printf("icmp_send dst %s src %s\n", 1028 inet_ntoa_r(ip->ip_dst, dstbuf), 1029 inet_ntoa_r(ip->ip_src, srcbuf)); 1030 } 1031 #endif 1032 (void) ip_output(m, opts, NULL, 0, NULL, NULL); 1033 } 1034 1035 /* 1036 * Return milliseconds since 00:00 UTC in network format. 1037 */ 1038 uint32_t 1039 iptime(void) 1040 { 1041 struct timeval atv; 1042 u_long t; 1043 1044 getmicrotime(&atv); 1045 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1046 return (htonl(t)); 1047 } 1048 1049 /* 1050 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1051 * given current value MTU. If DIR is less than zero, a larger plateau 1052 * is returned; otherwise, a smaller value is returned. 1053 */ 1054 int 1055 ip_next_mtu(int mtu, int dir) 1056 { 1057 static int mtutab[] = { 1058 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 1059 296, 68, 0 1060 }; 1061 int i, size; 1062 1063 size = (sizeof mtutab) / (sizeof mtutab[0]); 1064 if (dir >= 0) { 1065 for (i = 0; i < size; i++) 1066 if (mtu > mtutab[i]) 1067 return mtutab[i]; 1068 } else { 1069 for (i = size - 1; i >= 0; i--) 1070 if (mtu < mtutab[i]) 1071 return mtutab[i]; 1072 if (mtu == mtutab[0]) 1073 return mtutab[0]; 1074 } 1075 return 0; 1076 } 1077 #endif /* INET */ 1078 1079 /* 1080 * badport_bandlim() - check for ICMP bandwidth limit 1081 * 1082 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1083 * hit our bandwidth limit and it is not ok. 1084 * 1085 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1086 * 1087 * For now we separate the TCP and UDP subsystems w/ different 'which' 1088 * values. We may eventually remove this separation (and simplify the 1089 * code further). 1090 * 1091 * Note that the printing of the error message is delayed so we can 1092 * properly print the icmp error rate that the system was trying to do 1093 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1094 * the 'final' error, but it doesn't make sense to solve the printing 1095 * delay with more complex code. 1096 */ 1097 struct icmp_rate { 1098 const char *descr; 1099 struct counter_rate cr; 1100 }; 1101 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = { 1102 { "icmp unreach response" }, 1103 { "icmp ping response" }, 1104 { "icmp tstamp response" }, 1105 { "closed port RST response" }, 1106 { "open port RST response" }, 1107 { "icmp6 unreach response" }, 1108 { "sctp ootb response" } 1109 }; 1110 #define V_icmp_rates VNET(icmp_rates) 1111 1112 static void 1113 icmp_bandlimit_init(void) 1114 { 1115 1116 for (int i = 0; i < BANDLIM_MAX; i++) { 1117 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK); 1118 V_icmp_rates[i].cr.cr_ticks = ticks; 1119 } 1120 } 1121 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, 1122 icmp_bandlimit_init, NULL); 1123 1124 static void 1125 icmp_bandlimit_uninit(void) 1126 { 1127 1128 for (int i = 0; i < BANDLIM_MAX; i++) 1129 counter_u64_free(V_icmp_rates[i].cr.cr_rate); 1130 } 1131 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1132 icmp_bandlimit_uninit, NULL); 1133 1134 int 1135 badport_bandlim(int which) 1136 { 1137 int64_t pps; 1138 1139 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED) 1140 return (0); 1141 1142 KASSERT(which >= 0 && which < BANDLIM_MAX, 1143 ("%s: which %d", __func__, which)); 1144 1145 if ((V_icmplim + V_icmplim_curr_jitter) <= 0) 1146 V_icmplim_curr_jitter = -V_icmplim + 1; 1147 1148 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim + 1149 V_icmplim_curr_jitter); 1150 if (pps > 0) { 1151 /* 1152 * Adjust limit +/- to jitter the measurement to deny a 1153 * side-channel port scan as in CVE-2020-25705 1154 */ 1155 if (V_icmplim_jitter > 0) { 1156 int32_t inc = 1157 arc4random_uniform(V_icmplim_jitter * 2 +1) 1158 - V_icmplim_jitter; 1159 1160 V_icmplim_curr_jitter = inc; 1161 } 1162 } 1163 if (pps == -1) 1164 return (-1); 1165 if (pps > 0 && V_icmplim_output) 1166 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n", 1167 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim + 1168 V_icmplim_curr_jitter); 1169 return (0); 1170 } 1171