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 #include "opt_inet.h" 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/mbuf.h> 40 #include <sys/protosw.h> 41 #include <sys/socket.h> 42 #include <sys/time.h> 43 #include <sys/kernel.h> 44 #include <sys/lock.h> 45 #include <sys/sysctl.h> 46 #include <sys/syslog.h> 47 48 #include <net/if.h> 49 #include <net/if_var.h> 50 #include <net/if_private.h> 51 #include <net/if_types.h> 52 #include <net/route.h> 53 #include <net/route/route_ctl.h> 54 #include <net/route/nhop.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/tcpip.h> 69 #include <netinet/icmp_var.h> 70 71 #ifdef INET 72 73 #include <machine/in_cksum.h> 74 75 #include <security/mac/mac_framework.h> 76 #endif /* INET */ 77 78 extern ipproto_ctlinput_t *ip_ctlprotox[]; 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_curr_jitter) = 0; 92 #define V_icmplim_curr_jitter VNET(icmplim_curr_jitter) 93 VNET_DEFINE_STATIC(int, icmplim_jitter) = 16; 94 #define V_icmplim_jitter VNET(icmplim_jitter) 95 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_jitter, CTLFLAG_VNET | CTLFLAG_RW, 96 &VNET_NAME(icmplim_jitter), 0, 97 "Random icmplim jitter adjustment limit"); 98 99 VNET_DEFINE_STATIC(int, icmplim_output) = 1; 100 #define V_icmplim_output VNET(icmplim_output) 101 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW, 102 &VNET_NAME(icmplim_output), 0, 103 "Enable logging of ICMP response rate limiting"); 104 105 #ifdef INET 106 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat); 107 VNET_PCPUSTAT_SYSINIT(icmpstat); 108 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat, 109 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)"); 110 111 #ifdef VIMAGE 112 VNET_PCPUSTAT_SYSUNINIT(icmpstat); 113 #endif /* VIMAGE */ 114 115 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0; 116 #define V_icmpmaskrepl VNET(icmpmaskrepl) 117 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW, 118 &VNET_NAME(icmpmaskrepl), 0, 119 "Reply to ICMP Address Mask Request packets"); 120 121 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0; 122 #define V_icmpmaskfake VNET(icmpmaskfake) 123 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW, 124 &VNET_NAME(icmpmaskfake), 0, 125 "Fake reply to ICMP Address Mask Request packets"); 126 127 VNET_DEFINE(int, drop_redirect) = 0; 128 #define V_drop_redirect VNET(drop_redirect) 129 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW, 130 &VNET_NAME(drop_redirect), 0, 131 "Ignore ICMP redirects"); 132 133 VNET_DEFINE_STATIC(int, log_redirect) = 0; 134 #define V_log_redirect VNET(log_redirect) 135 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW, 136 &VNET_NAME(log_redirect), 0, 137 "Log ICMP redirects to the console"); 138 139 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */ 140 #define V_redirtimeout VNET(redirtimeout) 141 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW, 142 &VNET_NAME(redirtimeout), 0, 143 "Delay in seconds before expiring redirect route"); 144 145 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]); 146 #define V_reply_src VNET(reply_src) 147 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW, 148 &VNET_NAME(reply_src), IFNAMSIZ, 149 "ICMP reply source for non-local packets"); 150 151 VNET_DEFINE_STATIC(int, icmp_rfi) = 0; 152 #define V_icmp_rfi VNET(icmp_rfi) 153 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW, 154 &VNET_NAME(icmp_rfi), 0, 155 "ICMP reply from incoming interface for non-local packets"); 156 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */ 157 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548; 158 #define V_icmp_quotelen VNET(icmp_quotelen) 159 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW, 160 &VNET_NAME(icmp_quotelen), 0, 161 "Number of bytes from original packet to quote in ICMP reply"); 162 163 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0; 164 #define V_icmpbmcastecho VNET(icmpbmcastecho) 165 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW, 166 &VNET_NAME(icmpbmcastecho), 0, 167 "Reply to multicast ICMP Echo Request and Timestamp packets"); 168 169 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1; 170 #define V_icmptstamprepl VNET(icmptstamprepl) 171 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW, 172 &VNET_NAME(icmptstamprepl), 0, 173 "Respond to ICMP Timestamp packets"); 174 175 VNET_DEFINE_STATIC(int, error_keeptags) = 0; 176 #define V_error_keeptags VNET(error_keeptags) 177 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW, 178 &VNET_NAME(error_keeptags), 0, 179 "ICMP error response keeps copy of mbuf_tags of original packet"); 180 181 #ifdef ICMPPRINTFS 182 int icmpprintfs = 0; 183 #endif 184 185 static void icmp_reflect(struct mbuf *); 186 static void icmp_send(struct mbuf *, struct mbuf *); 187 static int icmp_verify_redirect_gateway(struct sockaddr_in *, 188 struct sockaddr_in *, struct sockaddr_in *, u_int); 189 190 /* 191 * Kernel module interface for updating icmpstat. The argument is an index 192 * into icmpstat treated as an array of u_long. While this encodes the 193 * general layout of icmpstat into the caller, it doesn't encode its 194 * location, so that future changes to add, for example, per-CPU stats 195 * support won't cause binary compatibility problems for kernel modules. 196 */ 197 void 198 kmod_icmpstat_inc(int statnum) 199 { 200 201 counter_u64_add(VNET(icmpstat)[statnum], 1); 202 } 203 204 /* 205 * Generate an error packet of type error 206 * in response to bad packet ip. 207 */ 208 void 209 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu) 210 { 211 struct ip *oip, *nip; 212 struct icmp *icp; 213 struct mbuf *m; 214 unsigned icmplen, icmpelen, nlen, oiphlen; 215 216 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", 217 __func__)); 218 219 if (type != ICMP_REDIRECT) 220 ICMPSTAT_INC(icps_error); 221 /* 222 * Don't send error: 223 * if the original packet was encrypted. 224 * if not the first fragment of message. 225 * in response to a multicast or broadcast packet. 226 * if the old packet protocol was an ICMP error message. 227 */ 228 if (n->m_flags & M_DECRYPTED) 229 goto freeit; 230 if (n->m_flags & (M_BCAST|M_MCAST)) 231 goto freeit; 232 233 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */ 234 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) 235 goto freeit; 236 oip = mtod(n, struct ip *); 237 oiphlen = oip->ip_hl << 2; 238 if (n->m_len < oiphlen + ICMP_MINLEN) 239 goto freeit; 240 #ifdef ICMPPRINTFS 241 if (icmpprintfs) 242 printf("icmp_error(%p, %x, %d)\n", oip, type, code); 243 #endif 244 if (oip->ip_off & htons(~(IP_MF|IP_DF))) 245 goto freeit; 246 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 247 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + 248 oiphlen))->icmp_type)) { 249 ICMPSTAT_INC(icps_oldicmp); 250 goto freeit; 251 } 252 /* 253 * Calculate length to quote from original packet and 254 * prevent the ICMP mbuf from overflowing. 255 * Unfortunately this is non-trivial since ip_forward() 256 * sends us truncated packets. 257 */ 258 nlen = m_length(n, NULL); 259 if (oip->ip_p == IPPROTO_TCP) { 260 struct tcphdr *th; 261 int tcphlen; 262 263 if (oiphlen + sizeof(struct tcphdr) > n->m_len && 264 n->m_next == NULL) 265 goto stdreply; 266 if (n->m_len < oiphlen + sizeof(struct tcphdr) && 267 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL) 268 goto freeit; 269 oip = mtod(n, struct ip *); 270 th = mtodo(n, oiphlen); 271 tcphlen = th->th_off << 2; 272 if (tcphlen < sizeof(struct tcphdr)) 273 goto freeit; 274 if (ntohs(oip->ip_len) < oiphlen + tcphlen) 275 goto freeit; 276 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL) 277 goto stdreply; 278 if (n->m_len < oiphlen + tcphlen && 279 (n = m_pullup(n, oiphlen + tcphlen)) == NULL) 280 goto freeit; 281 oip = mtod(n, struct ip *); 282 icmpelen = max(tcphlen, min(V_icmp_quotelen, 283 ntohs(oip->ip_len) - oiphlen)); 284 } else if (oip->ip_p == IPPROTO_SCTP) { 285 struct sctphdr *sh; 286 struct sctp_chunkhdr *ch; 287 288 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr)) 289 goto stdreply; 290 if (oiphlen + sizeof(struct sctphdr) > n->m_len && 291 n->m_next == NULL) 292 goto stdreply; 293 if (n->m_len < oiphlen + sizeof(struct sctphdr) && 294 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL) 295 goto freeit; 296 oip = mtod(n, struct ip *); 297 icmpelen = max(sizeof(struct sctphdr), 298 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen)); 299 sh = mtodo(n, oiphlen); 300 if (ntohl(sh->v_tag) == 0 && 301 ntohs(oip->ip_len) >= oiphlen + 302 sizeof(struct sctphdr) + 8 && 303 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 || 304 n->m_next != NULL)) { 305 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 && 306 (n = m_pullup(n, oiphlen + 307 sizeof(struct sctphdr) + 8)) == NULL) 308 goto freeit; 309 oip = mtod(n, struct ip *); 310 sh = mtodo(n, oiphlen); 311 ch = (struct sctp_chunkhdr *)(sh + 1); 312 if (ch->chunk_type == SCTP_INITIATION) { 313 icmpelen = max(sizeof(struct sctphdr) + 8, 314 min(V_icmp_quotelen, ntohs(oip->ip_len) - 315 oiphlen)); 316 } 317 } 318 } else 319 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - 320 oiphlen)); 321 322 icmplen = min(oiphlen + icmpelen, nlen); 323 if (icmplen < sizeof(struct ip)) 324 goto freeit; 325 326 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen) 327 m = m_gethdr(M_NOWAIT, MT_DATA); 328 else 329 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 330 if (m == NULL) 331 goto freeit; 332 #ifdef MAC 333 mac_netinet_icmp_reply(n, m); 334 #endif 335 icmplen = min(icmplen, M_TRAILINGSPACE(m) - 336 sizeof(struct ip) - ICMP_MINLEN); 337 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen); 338 m->m_data += sizeof(struct ip); 339 m->m_len = ICMP_MINLEN + icmplen; 340 341 /* XXX MRT make the outgoing packet use the same FIB 342 * that was associated with the incoming packet 343 */ 344 M_SETFIB(m, M_GETFIB(n)); 345 icp = mtod(m, struct icmp *); 346 ICMPSTAT_INC(icps_outhist[type]); 347 icp->icmp_type = type; 348 if (type == ICMP_REDIRECT) 349 icp->icmp_gwaddr.s_addr = dest; 350 else { 351 icp->icmp_void = 0; 352 /* 353 * The following assignments assume an overlay with the 354 * just zeroed icmp_void field. 355 */ 356 if (type == ICMP_PARAMPROB) { 357 icp->icmp_pptr = code; 358 code = 0; 359 } else if (type == ICMP_UNREACH && 360 code == ICMP_UNREACH_NEEDFRAG && mtu) { 361 icp->icmp_nextmtu = htons(mtu); 362 } 363 } 364 icp->icmp_code = code; 365 366 /* 367 * Copy the quotation into ICMP message and 368 * convert quoted IP header back to network representation. 369 */ 370 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 371 nip = &icp->icmp_ip; 372 373 /* 374 * Set up ICMP message mbuf and copy old IP header (without options 375 * in front of ICMP message. 376 * If the original mbuf was meant to bypass the firewall, the error 377 * reply should bypass as well. 378 */ 379 m->m_flags |= n->m_flags & M_SKIP_FIREWALL; 380 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip), 381 ("insufficient space for ip header")); 382 m->m_data -= sizeof(struct ip); 383 m->m_len += sizeof(struct ip); 384 m->m_pkthdr.len = m->m_len; 385 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 386 nip = mtod(m, struct ip *); 387 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); 388 nip->ip_len = htons(m->m_len); 389 nip->ip_v = IPVERSION; 390 nip->ip_hl = 5; 391 nip->ip_p = IPPROTO_ICMP; 392 nip->ip_tos = 0; 393 nip->ip_off = 0; 394 395 if (V_error_keeptags) 396 m_tag_copy_chain(m, n, M_NOWAIT); 397 398 icmp_reflect(m); 399 400 freeit: 401 m_freem(n); 402 } 403 404 int 405 icmp_errmap(const struct icmp *icp) 406 { 407 408 switch (icp->icmp_type) { 409 case ICMP_UNREACH: 410 switch (icp->icmp_code) { 411 case ICMP_UNREACH_NET: 412 case ICMP_UNREACH_HOST: 413 case ICMP_UNREACH_SRCFAIL: 414 case ICMP_UNREACH_NET_UNKNOWN: 415 case ICMP_UNREACH_HOST_UNKNOWN: 416 case ICMP_UNREACH_ISOLATED: 417 case ICMP_UNREACH_TOSNET: 418 case ICMP_UNREACH_TOSHOST: 419 case ICMP_UNREACH_HOST_PRECEDENCE: 420 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 421 return (EHOSTUNREACH); 422 case ICMP_UNREACH_NEEDFRAG: 423 return (EMSGSIZE); 424 case ICMP_UNREACH_PROTOCOL: 425 case ICMP_UNREACH_PORT: 426 case ICMP_UNREACH_NET_PROHIB: 427 case ICMP_UNREACH_HOST_PROHIB: 428 case ICMP_UNREACH_FILTER_PROHIB: 429 return (ECONNREFUSED); 430 default: 431 return (0); 432 } 433 case ICMP_TIMXCEED: 434 switch (icp->icmp_code) { 435 case ICMP_TIMXCEED_INTRANS: 436 return (EHOSTUNREACH); 437 default: 438 return (0); 439 } 440 case ICMP_PARAMPROB: 441 switch (icp->icmp_code) { 442 case ICMP_PARAMPROB_ERRATPTR: 443 case ICMP_PARAMPROB_OPTABSENT: 444 return (ENOPROTOOPT); 445 default: 446 return (0); 447 } 448 default: 449 return (0); 450 } 451 } 452 453 /* 454 * Process a received ICMP message. 455 */ 456 int 457 icmp_input(struct mbuf **mp, int *offp, int proto) 458 { 459 struct icmp *icp; 460 struct in_ifaddr *ia; 461 struct mbuf *m = *mp; 462 struct ip *ip = mtod(m, struct ip *); 463 struct sockaddr_in icmpsrc, icmpdst, icmpgw; 464 int hlen = *offp; 465 int icmplen = ntohs(ip->ip_len) - *offp; 466 int i, code; 467 int fibnum; 468 469 NET_EPOCH_ASSERT(); 470 471 *mp = NULL; 472 473 /* 474 * Locate icmp structure in mbuf, and check 475 * that not corrupted and of at least minimum length. 476 */ 477 #ifdef ICMPPRINTFS 478 if (icmpprintfs) { 479 char srcbuf[INET_ADDRSTRLEN]; 480 char dstbuf[INET_ADDRSTRLEN]; 481 482 printf("icmp_input from %s to %s, len %d\n", 483 inet_ntoa_r(ip->ip_src, srcbuf), 484 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen); 485 } 486 #endif 487 if (icmplen < ICMP_MINLEN) { 488 ICMPSTAT_INC(icps_tooshort); 489 goto freeit; 490 } 491 i = hlen + min(icmplen, ICMP_ADVLENMIN); 492 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 493 ICMPSTAT_INC(icps_tooshort); 494 return (IPPROTO_DONE); 495 } 496 ip = mtod(m, struct ip *); 497 m->m_len -= hlen; 498 m->m_data += hlen; 499 icp = mtod(m, struct icmp *); 500 if (in_cksum(m, icmplen)) { 501 ICMPSTAT_INC(icps_checksum); 502 goto freeit; 503 } 504 m->m_len += hlen; 505 m->m_data -= hlen; 506 507 #ifdef ICMPPRINTFS 508 if (icmpprintfs) 509 printf("icmp_input, type %d code %d\n", icp->icmp_type, 510 icp->icmp_code); 511 #endif 512 513 /* 514 * Message type specific processing. 515 */ 516 if (icp->icmp_type > ICMP_MAXTYPE) 517 goto raw; 518 519 /* Initialize */ 520 bzero(&icmpsrc, sizeof(icmpsrc)); 521 icmpsrc.sin_len = sizeof(struct sockaddr_in); 522 icmpsrc.sin_family = AF_INET; 523 bzero(&icmpdst, sizeof(icmpdst)); 524 icmpdst.sin_len = sizeof(struct sockaddr_in); 525 icmpdst.sin_family = AF_INET; 526 bzero(&icmpgw, sizeof(icmpgw)); 527 icmpgw.sin_len = sizeof(struct sockaddr_in); 528 icmpgw.sin_family = AF_INET; 529 530 ICMPSTAT_INC(icps_inhist[icp->icmp_type]); 531 code = icp->icmp_code; 532 switch (icp->icmp_type) { 533 case ICMP_UNREACH: 534 if (code > ICMP_UNREACH_PRECEDENCE_CUTOFF) 535 goto badcode; 536 else 537 goto deliver; 538 539 case ICMP_TIMXCEED: 540 if (code > ICMP_TIMXCEED_REASS) 541 goto badcode; 542 else 543 goto deliver; 544 545 case ICMP_PARAMPROB: 546 if (code > ICMP_PARAMPROB_LENGTH) 547 goto badcode; 548 549 deliver: 550 /* 551 * Problem with datagram; advise higher level routines. 552 */ 553 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 554 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 555 ICMPSTAT_INC(icps_badlen); 556 goto freeit; 557 } 558 /* Discard ICMP's in response to multicast packets */ 559 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 560 goto badcode; 561 /* Filter out responses to INADDR_ANY, protocols ignore it. */ 562 if (icp->icmp_ip.ip_dst.s_addr == INADDR_ANY || 563 icp->icmp_ip.ip_src.s_addr == INADDR_ANY) 564 goto freeit; 565 #ifdef ICMPPRINTFS 566 if (icmpprintfs) 567 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 568 #endif 569 /* 570 * XXX if the packet contains [IPv4 AH TCP], we can't make a 571 * notification to TCP layer. 572 */ 573 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp)); 574 ip_stripoptions(m); 575 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 576 /* This should actually not happen */ 577 ICMPSTAT_INC(icps_tooshort); 578 return (IPPROTO_DONE); 579 } 580 ip = mtod(m, struct ip *); 581 icp = (struct icmp *)(ip + 1); 582 /* 583 * The upper layer handler can rely on: 584 * - The outer IP header has no options. 585 * - The outer IP header, the ICMP header, the inner IP header, 586 * and the first n bytes of the inner payload are contiguous. 587 * n is at least 8, but might be larger based on 588 * ICMP_ADVLENPREF. See its definition in ip_icmp.h. 589 */ 590 if (ip_ctlprotox[icp->icmp_ip.ip_p] != NULL) 591 ip_ctlprotox[icp->icmp_ip.ip_p](icp); 592 break; 593 594 badcode: 595 ICMPSTAT_INC(icps_badcode); 596 break; 597 598 case ICMP_ECHO: 599 if (!V_icmpbmcastecho 600 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 601 ICMPSTAT_INC(icps_bmcastecho); 602 break; 603 } 604 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 605 goto freeit; 606 icp->icmp_type = ICMP_ECHOREPLY; 607 goto reflect; 608 609 case ICMP_TSTAMP: 610 if (V_icmptstamprepl == 0) 611 break; 612 if (!V_icmpbmcastecho 613 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 614 ICMPSTAT_INC(icps_bmcasttstamp); 615 break; 616 } 617 if (icmplen < ICMP_TSLEN) { 618 ICMPSTAT_INC(icps_badlen); 619 break; 620 } 621 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 622 goto freeit; 623 icp->icmp_type = ICMP_TSTAMPREPLY; 624 icp->icmp_rtime = iptime(); 625 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 626 goto reflect; 627 628 case ICMP_MASKREQ: 629 if (V_icmpmaskrepl == 0) 630 break; 631 /* 632 * We are not able to respond with all ones broadcast 633 * unless we receive it over a point-to-point interface. 634 */ 635 if (icmplen < ICMP_MASKLEN) 636 break; 637 switch (ip->ip_dst.s_addr) { 638 case INADDR_BROADCAST: 639 case INADDR_ANY: 640 icmpdst.sin_addr = ip->ip_src; 641 break; 642 643 default: 644 icmpdst.sin_addr = ip->ip_dst; 645 } 646 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 647 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 648 if (ia == NULL) 649 break; 650 if (ia->ia_ifp == NULL) 651 break; 652 icp->icmp_type = ICMP_MASKREPLY; 653 if (V_icmpmaskfake == 0) 654 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 655 else 656 icp->icmp_mask = V_icmpmaskfake; 657 if (ip->ip_src.s_addr == 0) { 658 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 659 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 660 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 661 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 662 } 663 reflect: 664 ICMPSTAT_INC(icps_reflect); 665 ICMPSTAT_INC(icps_outhist[icp->icmp_type]); 666 icmp_reflect(m); 667 return (IPPROTO_DONE); 668 669 case ICMP_REDIRECT: 670 if (V_log_redirect) { 671 u_long src, dst, gw; 672 673 src = ntohl(ip->ip_src.s_addr); 674 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 675 gw = ntohl(icp->icmp_gwaddr.s_addr); 676 printf("icmp redirect from %d.%d.%d.%d: " 677 "%d.%d.%d.%d => %d.%d.%d.%d\n", 678 (int)(src >> 24), (int)((src >> 16) & 0xff), 679 (int)((src >> 8) & 0xff), (int)(src & 0xff), 680 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 681 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 682 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 683 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 684 } 685 /* 686 * RFC1812 says we must ignore ICMP redirects if we 687 * are acting as router. 688 */ 689 if (V_drop_redirect || V_ipforwarding) 690 break; 691 if (code > 3) 692 goto badcode; 693 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 694 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 695 ICMPSTAT_INC(icps_badlen); 696 break; 697 } 698 /* 699 * Short circuit routing redirects to force 700 * immediate change in the kernel's routing 701 * tables. The message is also handed to anyone 702 * listening on a raw socket (e.g. the routing 703 * daemon for use in updating its tables). 704 */ 705 icmpgw.sin_addr = ip->ip_src; 706 icmpdst.sin_addr = icp->icmp_gwaddr; 707 #ifdef ICMPPRINTFS 708 if (icmpprintfs) { 709 char dstbuf[INET_ADDRSTRLEN]; 710 char gwbuf[INET_ADDRSTRLEN]; 711 712 printf("redirect dst %s to %s\n", 713 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf), 714 inet_ntoa_r(icp->icmp_gwaddr, gwbuf)); 715 } 716 #endif 717 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 718 719 /* 720 * RFC 1122 says network (code 0,2) redirects SHOULD 721 * be treated identically to the host redirects. 722 * Given that, ignore network masks. 723 */ 724 725 /* 726 * Variable values: 727 * icmpsrc: route destination 728 * icmpdst: route gateway 729 * icmpgw: message source 730 */ 731 732 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst, 733 M_GETFIB(m)) != 0) { 734 /* TODO: increment bad redirects here */ 735 break; 736 } 737 738 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 739 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc, 740 (struct sockaddr *)&icmpdst, 741 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif, 742 RTF_GATEWAY, V_redirtimeout); 743 } 744 break; 745 746 /* 747 * No kernel processing for the following; 748 * just fall through to send to raw listener. 749 */ 750 case ICMP_ECHOREPLY: 751 case ICMP_ROUTERADVERT: 752 case ICMP_ROUTERSOLICIT: 753 case ICMP_TSTAMPREPLY: 754 case ICMP_IREQREPLY: 755 case ICMP_MASKREPLY: 756 case ICMP_SOURCEQUENCH: 757 default: 758 break; 759 } 760 761 raw: 762 *mp = m; 763 rip_input(mp, offp, proto); 764 return (IPPROTO_DONE); 765 766 freeit: 767 m_freem(m); 768 return (IPPROTO_DONE); 769 } 770 771 /* 772 * Reflect the ip packet back to the source 773 */ 774 static void 775 icmp_reflect(struct mbuf *m) 776 { 777 struct ip *ip = mtod(m, struct ip *); 778 struct ifaddr *ifa; 779 struct ifnet *ifp; 780 struct in_ifaddr *ia; 781 struct in_addr t; 782 struct nhop_object *nh; 783 struct mbuf *opts = NULL; 784 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 785 786 NET_EPOCH_ASSERT(); 787 788 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 789 (IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net240) || 790 (IN_ZERONET(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net0) ) { 791 m_freem(m); /* Bad return address */ 792 ICMPSTAT_INC(icps_badaddr); 793 goto done; /* Ip_output() will check for broadcast */ 794 } 795 796 t = ip->ip_dst; 797 ip->ip_dst = ip->ip_src; 798 799 /* 800 * Source selection for ICMP replies: 801 * 802 * If the incoming packet was addressed directly to one of our 803 * own addresses, use dst as the src for the reply. 804 */ 805 CK_LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { 806 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { 807 t = IA_SIN(ia)->sin_addr; 808 goto match; 809 } 810 } 811 812 /* 813 * If the incoming packet was addressed to one of our broadcast 814 * addresses, use the first non-broadcast address which corresponds 815 * to the incoming interface. 816 */ 817 ifp = m->m_pkthdr.rcvif; 818 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 819 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 820 if (ifa->ifa_addr->sa_family != AF_INET) 821 continue; 822 ia = ifatoia(ifa); 823 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 824 t.s_addr) { 825 t = IA_SIN(ia)->sin_addr; 826 goto match; 827 } 828 } 829 } 830 /* 831 * If the packet was transiting through us, use the address of 832 * the interface the packet came through in. If that interface 833 * doesn't have a suitable IP address, the normal selection 834 * criteria apply. 835 */ 836 if (V_icmp_rfi && ifp != NULL) { 837 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 838 if (ifa->ifa_addr->sa_family != AF_INET) 839 continue; 840 ia = ifatoia(ifa); 841 t = IA_SIN(ia)->sin_addr; 842 goto match; 843 } 844 } 845 /* 846 * If the incoming packet was not addressed directly to us, use 847 * designated interface for icmp replies specified by sysctl 848 * net.inet.icmp.reply_src (default not set). Otherwise continue 849 * with normal source selection. 850 */ 851 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) { 852 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 853 if (ifa->ifa_addr->sa_family != AF_INET) 854 continue; 855 ia = ifatoia(ifa); 856 t = IA_SIN(ia)->sin_addr; 857 goto match; 858 } 859 } 860 /* 861 * If the packet was transiting through us, use the address of 862 * the interface that is the closest to the packet source. 863 * When we don't have a route back to the packet source, stop here 864 * and drop the packet. 865 */ 866 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0); 867 if (nh == NULL) { 868 m_freem(m); 869 ICMPSTAT_INC(icps_noroute); 870 goto done; 871 } 872 t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr; 873 match: 874 #ifdef MAC 875 mac_netinet_icmp_replyinplace(m); 876 #endif 877 ip->ip_src = t; 878 ip->ip_ttl = V_ip_defttl; 879 880 if (optlen > 0) { 881 u_char *cp; 882 int opt, cnt; 883 u_int len; 884 885 /* 886 * Retrieve any source routing from the incoming packet; 887 * add on any record-route or timestamp options. 888 */ 889 cp = (u_char *) (ip + 1); 890 if ((opts = ip_srcroute(m)) == NULL && 891 (opts = m_gethdr(M_NOWAIT, MT_DATA))) { 892 opts->m_len = sizeof(struct in_addr); 893 mtod(opts, struct in_addr *)->s_addr = 0; 894 } 895 if (opts) { 896 #ifdef ICMPPRINTFS 897 if (icmpprintfs) 898 printf("icmp_reflect optlen %d rt %d => ", 899 optlen, opts->m_len); 900 #endif 901 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 902 opt = cp[IPOPT_OPTVAL]; 903 if (opt == IPOPT_EOL) 904 break; 905 if (opt == IPOPT_NOP) 906 len = 1; 907 else { 908 if (cnt < IPOPT_OLEN + sizeof(*cp)) 909 break; 910 len = cp[IPOPT_OLEN]; 911 if (len < IPOPT_OLEN + sizeof(*cp) || 912 len > cnt) 913 break; 914 } 915 /* 916 * Should check for overflow, but it "can't happen" 917 */ 918 if (opt == IPOPT_RR || opt == IPOPT_TS || 919 opt == IPOPT_SECURITY) { 920 bcopy((caddr_t)cp, 921 mtod(opts, caddr_t) + opts->m_len, len); 922 opts->m_len += len; 923 } 924 } 925 /* Terminate & pad, if necessary */ 926 cnt = opts->m_len % 4; 927 if (cnt) { 928 for (; cnt < 4; cnt++) { 929 *(mtod(opts, caddr_t) + opts->m_len) = 930 IPOPT_EOL; 931 opts->m_len++; 932 } 933 } 934 #ifdef ICMPPRINTFS 935 if (icmpprintfs) 936 printf("%d\n", opts->m_len); 937 #endif 938 } 939 ip_stripoptions(m); 940 } 941 m_tag_delete_nonpersistent(m); 942 m->m_flags &= ~(M_BCAST|M_MCAST); 943 icmp_send(m, opts); 944 done: 945 if (opts) 946 (void)m_free(opts); 947 } 948 949 /* 950 * Verifies if redirect message is valid, according to RFC 1122 951 * 952 * @src: sockaddr with address of redirect originator 953 * @dst: sockaddr with destination in question 954 * @gateway: new proposed gateway 955 * 956 * Returns 0 on success. 957 */ 958 static int 959 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst, 960 struct sockaddr_in *gateway, u_int fibnum) 961 { 962 struct nhop_object *nh; 963 struct ifaddr *ifa; 964 965 NET_EPOCH_ASSERT(); 966 967 /* Verify the gateway is directly reachable. */ 968 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL) 969 return (ENETUNREACH); 970 971 /* TODO: fib-aware. */ 972 if (ifa_ifwithaddr_check((struct sockaddr *)gateway)) 973 return (EHOSTUNREACH); 974 975 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0); 976 if (nh == NULL) 977 return (EINVAL); 978 979 /* 980 * If the redirect isn't from our current router for this dst, 981 * it's either old or wrong. If it redirects us to ourselves, 982 * we have a routing loop, perhaps as a result of an interface 983 * going down recently. 984 */ 985 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa)) 986 return (EINVAL); 987 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) 988 return (EINVAL); 989 990 /* If host route already exists, ignore redirect. */ 991 if (nh->nh_flags & NHF_HOST) 992 return (EEXIST); 993 994 /* If the prefix is directly reachable, ignore redirect. */ 995 if (!(nh->nh_flags & NHF_GATEWAY)) 996 return (EEXIST); 997 998 return (0); 999 } 1000 1001 /* 1002 * Send an icmp packet back to the ip level, 1003 * after supplying a checksum. 1004 */ 1005 static void 1006 icmp_send(struct mbuf *m, struct mbuf *opts) 1007 { 1008 struct ip *ip = mtod(m, struct ip *); 1009 int hlen; 1010 struct icmp *icp; 1011 1012 hlen = ip->ip_hl << 2; 1013 m->m_data += hlen; 1014 m->m_len -= hlen; 1015 icp = mtod(m, struct icmp *); 1016 icp->icmp_cksum = 0; 1017 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 1018 m->m_data -= hlen; 1019 m->m_len += hlen; 1020 m->m_pkthdr.rcvif = (struct ifnet *)0; 1021 #ifdef ICMPPRINTFS 1022 if (icmpprintfs) { 1023 char dstbuf[INET_ADDRSTRLEN]; 1024 char srcbuf[INET_ADDRSTRLEN]; 1025 1026 printf("icmp_send dst %s src %s\n", 1027 inet_ntoa_r(ip->ip_dst, dstbuf), 1028 inet_ntoa_r(ip->ip_src, srcbuf)); 1029 } 1030 #endif 1031 (void) ip_output(m, opts, NULL, 0, NULL, NULL); 1032 } 1033 1034 /* 1035 * Return milliseconds since 00:00 UTC in network format. 1036 */ 1037 uint32_t 1038 iptime(void) 1039 { 1040 struct timeval atv; 1041 u_long t; 1042 1043 getmicrotime(&atv); 1044 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1045 return (htonl(t)); 1046 } 1047 1048 /* 1049 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1050 * given current value MTU. If DIR is less than zero, a larger plateau 1051 * is returned; otherwise, a smaller value is returned. 1052 */ 1053 int 1054 ip_next_mtu(int mtu, int dir) 1055 { 1056 static int mtutab[] = { 1057 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 1058 296, 68, 0 1059 }; 1060 int i, size; 1061 1062 size = (sizeof mtutab) / (sizeof mtutab[0]); 1063 if (dir >= 0) { 1064 for (i = 0; i < size; i++) 1065 if (mtu > mtutab[i]) 1066 return mtutab[i]; 1067 } else { 1068 for (i = size - 1; i >= 0; i--) 1069 if (mtu < mtutab[i]) 1070 return mtutab[i]; 1071 if (mtu == mtutab[0]) 1072 return mtutab[0]; 1073 } 1074 return 0; 1075 } 1076 #endif /* INET */ 1077 1078 /* 1079 * badport_bandlim() - check for ICMP bandwidth limit 1080 * 1081 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1082 * hit our bandwidth limit and it is not ok. 1083 * 1084 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1085 * 1086 * For now we separate the TCP and UDP subsystems w/ different 'which' 1087 * values. We may eventually remove this separation (and simplify the 1088 * code further). 1089 * 1090 * Note that the printing of the error message is delayed so we can 1091 * properly print the icmp error rate that the system was trying to do 1092 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1093 * the 'final' error, but it doesn't make sense to solve the printing 1094 * delay with more complex code. 1095 */ 1096 struct icmp_rate { 1097 const char *descr; 1098 struct counter_rate cr; 1099 }; 1100 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = { 1101 { "icmp unreach response" }, 1102 { "icmp ping response" }, 1103 { "icmp tstamp response" }, 1104 { "closed port RST response" }, 1105 { "open port RST response" }, 1106 { "icmp6 unreach response" }, 1107 { "sctp ootb response" } 1108 }; 1109 #define V_icmp_rates VNET(icmp_rates) 1110 1111 static void 1112 icmp_bandlimit_init(void) 1113 { 1114 1115 for (int i = 0; i < BANDLIM_MAX; i++) { 1116 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK); 1117 V_icmp_rates[i].cr.cr_ticks = ticks; 1118 } 1119 } 1120 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, 1121 icmp_bandlimit_init, NULL); 1122 1123 static void 1124 icmp_bandlimit_uninit(void) 1125 { 1126 1127 for (int i = 0; i < BANDLIM_MAX; i++) 1128 counter_u64_free(V_icmp_rates[i].cr.cr_rate); 1129 } 1130 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1131 icmp_bandlimit_uninit, NULL); 1132 1133 int 1134 badport_bandlim(int which) 1135 { 1136 int64_t pps; 1137 1138 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED) 1139 return (0); 1140 1141 KASSERT(which >= 0 && which < BANDLIM_MAX, 1142 ("%s: which %d", __func__, which)); 1143 1144 if ((V_icmplim + V_icmplim_curr_jitter) <= 0) 1145 V_icmplim_curr_jitter = -V_icmplim + 1; 1146 1147 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim + 1148 V_icmplim_curr_jitter); 1149 if (pps > 0) { 1150 /* 1151 * Adjust limit +/- to jitter the measurement to deny a 1152 * side-channel port scan as in CVE-2020-25705 1153 */ 1154 if (V_icmplim_jitter > 0) { 1155 int32_t inc = 1156 arc4random_uniform(V_icmplim_jitter * 2 +1) 1157 - V_icmplim_jitter; 1158 1159 V_icmplim_curr_jitter = inc; 1160 } 1161 } 1162 if (pps == -1) 1163 return (-1); 1164 if (pps > 0 && V_icmplim_output) 1165 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n", 1166 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim + 1167 V_icmplim_curr_jitter); 1168 return (0); 1169 } 1170