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 32 #include <sys/cdefs.h> 33 #include "opt_inet.h" 34 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/mbuf.h> 38 #include <sys/protosw.h> 39 #include <sys/socket.h> 40 #include <sys/time.h> 41 #include <sys/kernel.h> 42 #include <sys/lock.h> 43 #include <sys/sysctl.h> 44 #include <sys/syslog.h> 45 46 #include <net/if.h> 47 #include <net/if_var.h> 48 #include <net/if_private.h> 49 #include <net/if_types.h> 50 #include <net/route.h> 51 #include <net/route/route_ctl.h> 52 #include <net/route/nhop.h> 53 #include <net/vnet.h> 54 55 #include <netinet/in.h> 56 #include <netinet/in_fib.h> 57 #include <netinet/in_pcb.h> 58 #include <netinet/in_systm.h> 59 #include <netinet/in_var.h> 60 #include <netinet/ip.h> 61 #include <netinet/ip_icmp.h> 62 #include <netinet/ip_var.h> 63 #include <netinet/ip_options.h> 64 #include <netinet/sctp.h> 65 #include <netinet/tcp.h> 66 #include <netinet/tcpip.h> 67 #include <netinet/icmp_var.h> 68 69 #ifdef INET 70 71 #include <machine/in_cksum.h> 72 73 #include <security/mac/mac_framework.h> 74 #endif /* INET */ 75 76 extern ipproto_ctlinput_t *ip_ctlprotox[]; 77 78 /* 79 * ICMP routines: error generation, receive packet processing, and 80 * routines to turnaround packets back to the originator, and 81 * host table maintenance routines. 82 */ 83 VNET_DEFINE_STATIC(int, icmplim) = 200; 84 #define V_icmplim VNET(icmplim) 85 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW, 86 &VNET_NAME(icmplim), 0, 87 "Maximum number of ICMP responses per second"); 88 89 VNET_DEFINE_STATIC(int, icmplim_curr_jitter) = 0; 90 #define V_icmplim_curr_jitter VNET(icmplim_curr_jitter) 91 VNET_DEFINE_STATIC(int, icmplim_jitter) = 16; 92 #define V_icmplim_jitter VNET(icmplim_jitter) 93 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_jitter, CTLFLAG_VNET | CTLFLAG_RW, 94 &VNET_NAME(icmplim_jitter), 0, 95 "Random icmplim jitter adjustment limit"); 96 97 VNET_DEFINE_STATIC(int, icmplim_output) = 1; 98 #define V_icmplim_output VNET(icmplim_output) 99 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW, 100 &VNET_NAME(icmplim_output), 0, 101 "Enable logging of ICMP response rate limiting"); 102 103 #ifdef INET 104 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat); 105 VNET_PCPUSTAT_SYSINIT(icmpstat); 106 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat, 107 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)"); 108 109 #ifdef VIMAGE 110 VNET_PCPUSTAT_SYSUNINIT(icmpstat); 111 #endif /* VIMAGE */ 112 113 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0; 114 #define V_icmpmaskrepl VNET(icmpmaskrepl) 115 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW, 116 &VNET_NAME(icmpmaskrepl), 0, 117 "Reply to ICMP Address Mask Request packets"); 118 119 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0; 120 #define V_icmpmaskfake VNET(icmpmaskfake) 121 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW, 122 &VNET_NAME(icmpmaskfake), 0, 123 "Fake reply to ICMP Address Mask Request packets"); 124 125 VNET_DEFINE(int, drop_redirect) = 0; 126 #define V_drop_redirect VNET(drop_redirect) 127 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW, 128 &VNET_NAME(drop_redirect), 0, 129 "Ignore ICMP redirects"); 130 131 VNET_DEFINE_STATIC(int, log_redirect) = 0; 132 #define V_log_redirect VNET(log_redirect) 133 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW, 134 &VNET_NAME(log_redirect), 0, 135 "Log ICMP redirects to the console"); 136 137 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */ 138 #define V_redirtimeout VNET(redirtimeout) 139 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW, 140 &VNET_NAME(redirtimeout), 0, 141 "Delay in seconds before expiring redirect route"); 142 143 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]); 144 #define V_reply_src VNET(reply_src) 145 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW, 146 &VNET_NAME(reply_src), IFNAMSIZ, 147 "ICMP reply source for non-local packets"); 148 149 VNET_DEFINE_STATIC(int, icmp_rfi) = 0; 150 #define V_icmp_rfi VNET(icmp_rfi) 151 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW, 152 &VNET_NAME(icmp_rfi), 0, 153 "ICMP reply from incoming interface for non-local packets"); 154 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */ 155 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548; 156 #define V_icmp_quotelen VNET(icmp_quotelen) 157 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW, 158 &VNET_NAME(icmp_quotelen), 0, 159 "Number of bytes from original packet to quote in ICMP reply"); 160 161 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0; 162 #define V_icmpbmcastecho VNET(icmpbmcastecho) 163 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW, 164 &VNET_NAME(icmpbmcastecho), 0, 165 "Reply to multicast ICMP Echo Request and Timestamp packets"); 166 167 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1; 168 #define V_icmptstamprepl VNET(icmptstamprepl) 169 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW, 170 &VNET_NAME(icmptstamprepl), 0, 171 "Respond to ICMP Timestamp packets"); 172 173 VNET_DEFINE_STATIC(int, error_keeptags) = 0; 174 #define V_error_keeptags VNET(error_keeptags) 175 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW, 176 &VNET_NAME(error_keeptags), 0, 177 "ICMP error response keeps copy of mbuf_tags of original packet"); 178 179 #ifdef ICMPPRINTFS 180 int icmpprintfs = 0; 181 #endif 182 183 static void icmp_reflect(struct mbuf *); 184 static void icmp_send(struct mbuf *, struct mbuf *); 185 static int icmp_verify_redirect_gateway(struct sockaddr_in *, 186 struct sockaddr_in *, struct sockaddr_in *, u_int); 187 188 /* 189 * Kernel module interface for updating icmpstat. The argument is an index 190 * into icmpstat treated as an array of u_long. While this encodes the 191 * general layout of icmpstat into the caller, it doesn't encode its 192 * location, so that future changes to add, for example, per-CPU stats 193 * support won't cause binary compatibility problems for kernel modules. 194 */ 195 void 196 kmod_icmpstat_inc(int statnum) 197 { 198 199 counter_u64_add(VNET(icmpstat)[statnum], 1); 200 } 201 202 /* 203 * Generate an error packet of type error 204 * in response to bad packet ip. 205 */ 206 void 207 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu) 208 { 209 struct ip *oip, *nip; 210 struct icmp *icp; 211 struct mbuf *m; 212 unsigned icmplen, icmpelen, nlen, oiphlen; 213 214 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", 215 __func__)); 216 217 if (type != ICMP_REDIRECT) 218 ICMPSTAT_INC(icps_error); 219 /* 220 * Don't send error: 221 * if the original packet was encrypted. 222 * if not the first fragment of message. 223 * in response to a multicast or broadcast packet. 224 * if the old packet protocol was an ICMP error message. 225 */ 226 if (n->m_flags & M_DECRYPTED) 227 goto freeit; 228 if (n->m_flags & (M_BCAST|M_MCAST)) 229 goto freeit; 230 231 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */ 232 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) 233 goto freeit; 234 oip = mtod(n, struct ip *); 235 oiphlen = oip->ip_hl << 2; 236 if (n->m_len < oiphlen + ICMP_MINLEN) 237 goto freeit; 238 #ifdef ICMPPRINTFS 239 if (icmpprintfs) 240 printf("icmp_error(%p, %x, %d)\n", oip, type, code); 241 #endif 242 if (oip->ip_off & htons(~(IP_MF|IP_DF))) 243 goto freeit; 244 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 245 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + 246 oiphlen))->icmp_type)) { 247 ICMPSTAT_INC(icps_oldicmp); 248 goto freeit; 249 } 250 /* 251 * Calculate length to quote from original packet and 252 * prevent the ICMP mbuf from overflowing. 253 * Unfortunately this is non-trivial since ip_forward() 254 * sends us truncated packets. 255 */ 256 nlen = m_length(n, NULL); 257 if (oip->ip_p == IPPROTO_TCP) { 258 struct tcphdr *th; 259 int tcphlen; 260 261 if (oiphlen + sizeof(struct tcphdr) > n->m_len && 262 n->m_next == NULL) 263 goto stdreply; 264 if (n->m_len < oiphlen + sizeof(struct tcphdr) && 265 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL) 266 goto freeit; 267 oip = mtod(n, struct ip *); 268 th = mtodo(n, oiphlen); 269 tcphlen = th->th_off << 2; 270 if (tcphlen < sizeof(struct tcphdr)) 271 goto freeit; 272 if (ntohs(oip->ip_len) < oiphlen + tcphlen) 273 goto freeit; 274 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL) 275 goto stdreply; 276 if (n->m_len < oiphlen + tcphlen && 277 (n = m_pullup(n, oiphlen + tcphlen)) == NULL) 278 goto freeit; 279 oip = mtod(n, struct ip *); 280 icmpelen = max(tcphlen, min(V_icmp_quotelen, 281 ntohs(oip->ip_len) - oiphlen)); 282 } else if (oip->ip_p == IPPROTO_SCTP) { 283 struct sctphdr *sh; 284 struct sctp_chunkhdr *ch; 285 286 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr)) 287 goto stdreply; 288 if (oiphlen + sizeof(struct sctphdr) > n->m_len && 289 n->m_next == NULL) 290 goto stdreply; 291 if (n->m_len < oiphlen + sizeof(struct sctphdr) && 292 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL) 293 goto freeit; 294 oip = mtod(n, struct ip *); 295 icmpelen = max(sizeof(struct sctphdr), 296 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen)); 297 sh = mtodo(n, oiphlen); 298 if (ntohl(sh->v_tag) == 0 && 299 ntohs(oip->ip_len) >= oiphlen + 300 sizeof(struct sctphdr) + 8 && 301 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 || 302 n->m_next != NULL)) { 303 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 && 304 (n = m_pullup(n, oiphlen + 305 sizeof(struct sctphdr) + 8)) == NULL) 306 goto freeit; 307 oip = mtod(n, struct ip *); 308 sh = mtodo(n, oiphlen); 309 ch = (struct sctp_chunkhdr *)(sh + 1); 310 if (ch->chunk_type == SCTP_INITIATION) { 311 icmpelen = max(sizeof(struct sctphdr) + 8, 312 min(V_icmp_quotelen, ntohs(oip->ip_len) - 313 oiphlen)); 314 } 315 } 316 } else 317 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - 318 oiphlen)); 319 320 icmplen = min(oiphlen + icmpelen, nlen); 321 if (icmplen < sizeof(struct ip)) 322 goto freeit; 323 324 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen) 325 m = m_gethdr(M_NOWAIT, MT_DATA); 326 else 327 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 328 if (m == NULL) 329 goto freeit; 330 #ifdef MAC 331 mac_netinet_icmp_reply(n, m); 332 #endif 333 icmplen = min(icmplen, M_TRAILINGSPACE(m) - 334 sizeof(struct ip) - ICMP_MINLEN); 335 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen); 336 m->m_data += sizeof(struct ip); 337 m->m_len = ICMP_MINLEN + icmplen; 338 339 /* XXX MRT make the outgoing packet use the same FIB 340 * that was associated with the incoming packet 341 */ 342 M_SETFIB(m, M_GETFIB(n)); 343 icp = mtod(m, struct icmp *); 344 ICMPSTAT_INC(icps_outhist[type]); 345 icp->icmp_type = type; 346 if (type == ICMP_REDIRECT) 347 icp->icmp_gwaddr.s_addr = dest; 348 else { 349 icp->icmp_void = 0; 350 /* 351 * The following assignments assume an overlay with the 352 * just zeroed icmp_void field. 353 */ 354 if (type == ICMP_PARAMPROB) { 355 icp->icmp_pptr = code; 356 code = 0; 357 } else if (type == ICMP_UNREACH && 358 code == ICMP_UNREACH_NEEDFRAG && mtu) { 359 icp->icmp_nextmtu = htons(mtu); 360 } 361 } 362 icp->icmp_code = code; 363 364 /* 365 * Copy the quotation into ICMP message and 366 * convert quoted IP header back to network representation. 367 */ 368 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 369 nip = &icp->icmp_ip; 370 371 /* 372 * Set up ICMP message mbuf and copy old IP header (without options 373 * in front of ICMP message. 374 * If the original mbuf was meant to bypass the firewall, the error 375 * reply should bypass as well. 376 */ 377 m->m_flags |= n->m_flags & M_SKIP_FIREWALL; 378 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip), 379 ("insufficient space for ip header")); 380 m->m_data -= sizeof(struct ip); 381 m->m_len += sizeof(struct ip); 382 m->m_pkthdr.len = m->m_len; 383 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 384 nip = mtod(m, struct ip *); 385 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); 386 nip->ip_len = htons(m->m_len); 387 nip->ip_v = IPVERSION; 388 nip->ip_hl = 5; 389 nip->ip_p = IPPROTO_ICMP; 390 nip->ip_tos = 0; 391 nip->ip_off = 0; 392 393 if (V_error_keeptags) 394 m_tag_copy_chain(m, n, M_NOWAIT); 395 396 icmp_reflect(m); 397 398 freeit: 399 m_freem(n); 400 } 401 402 int 403 icmp_errmap(const struct icmp *icp) 404 { 405 406 switch (icp->icmp_type) { 407 case ICMP_UNREACH: 408 switch (icp->icmp_code) { 409 case ICMP_UNREACH_NET: 410 case ICMP_UNREACH_HOST: 411 case ICMP_UNREACH_SRCFAIL: 412 case ICMP_UNREACH_NET_UNKNOWN: 413 case ICMP_UNREACH_HOST_UNKNOWN: 414 case ICMP_UNREACH_ISOLATED: 415 case ICMP_UNREACH_TOSNET: 416 case ICMP_UNREACH_TOSHOST: 417 case ICMP_UNREACH_HOST_PRECEDENCE: 418 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 419 return (EHOSTUNREACH); 420 case ICMP_UNREACH_NEEDFRAG: 421 return (EMSGSIZE); 422 case ICMP_UNREACH_PROTOCOL: 423 case ICMP_UNREACH_PORT: 424 case ICMP_UNREACH_NET_PROHIB: 425 case ICMP_UNREACH_HOST_PROHIB: 426 case ICMP_UNREACH_FILTER_PROHIB: 427 return (ECONNREFUSED); 428 default: 429 return (0); 430 } 431 case ICMP_TIMXCEED: 432 switch (icp->icmp_code) { 433 case ICMP_TIMXCEED_INTRANS: 434 return (EHOSTUNREACH); 435 default: 436 return (0); 437 } 438 case ICMP_PARAMPROB: 439 switch (icp->icmp_code) { 440 case ICMP_PARAMPROB_ERRATPTR: 441 case ICMP_PARAMPROB_OPTABSENT: 442 return (ENOPROTOOPT); 443 default: 444 return (0); 445 } 446 default: 447 return (0); 448 } 449 } 450 451 /* 452 * Process a received ICMP message. 453 */ 454 int 455 icmp_input(struct mbuf **mp, int *offp, int proto) 456 { 457 struct icmp *icp; 458 struct in_ifaddr *ia; 459 struct mbuf *m = *mp; 460 struct ip *ip = mtod(m, struct ip *); 461 struct sockaddr_in icmpsrc, icmpdst, icmpgw; 462 int hlen = *offp; 463 int icmplen = ntohs(ip->ip_len) - *offp; 464 int i, code; 465 int fibnum; 466 467 NET_EPOCH_ASSERT(); 468 469 *mp = NULL; 470 471 /* 472 * Locate icmp structure in mbuf, and check 473 * that not corrupted and of at least minimum length. 474 */ 475 #ifdef ICMPPRINTFS 476 if (icmpprintfs) { 477 char srcbuf[INET_ADDRSTRLEN]; 478 char dstbuf[INET_ADDRSTRLEN]; 479 480 printf("icmp_input from %s to %s, len %d\n", 481 inet_ntoa_r(ip->ip_src, srcbuf), 482 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen); 483 } 484 #endif 485 if (icmplen < ICMP_MINLEN) { 486 ICMPSTAT_INC(icps_tooshort); 487 goto freeit; 488 } 489 i = hlen + min(icmplen, ICMP_ADVLENMIN); 490 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 491 ICMPSTAT_INC(icps_tooshort); 492 return (IPPROTO_DONE); 493 } 494 ip = mtod(m, struct ip *); 495 m->m_len -= hlen; 496 m->m_data += hlen; 497 icp = mtod(m, struct icmp *); 498 if (in_cksum(m, icmplen)) { 499 ICMPSTAT_INC(icps_checksum); 500 goto freeit; 501 } 502 m->m_len += hlen; 503 m->m_data -= hlen; 504 505 #ifdef ICMPPRINTFS 506 if (icmpprintfs) 507 printf("icmp_input, type %d code %d\n", icp->icmp_type, 508 icp->icmp_code); 509 #endif 510 511 /* 512 * Message type specific processing. 513 */ 514 if (icp->icmp_type > ICMP_MAXTYPE) 515 goto raw; 516 517 /* Initialize */ 518 bzero(&icmpsrc, sizeof(icmpsrc)); 519 icmpsrc.sin_len = sizeof(struct sockaddr_in); 520 icmpsrc.sin_family = AF_INET; 521 bzero(&icmpdst, sizeof(icmpdst)); 522 icmpdst.sin_len = sizeof(struct sockaddr_in); 523 icmpdst.sin_family = AF_INET; 524 bzero(&icmpgw, sizeof(icmpgw)); 525 icmpgw.sin_len = sizeof(struct sockaddr_in); 526 icmpgw.sin_family = AF_INET; 527 528 ICMPSTAT_INC(icps_inhist[icp->icmp_type]); 529 code = icp->icmp_code; 530 switch (icp->icmp_type) { 531 case ICMP_UNREACH: 532 if (code > ICMP_UNREACH_PRECEDENCE_CUTOFF) 533 goto badcode; 534 else 535 goto deliver; 536 537 case ICMP_TIMXCEED: 538 if (code > ICMP_TIMXCEED_REASS) 539 goto badcode; 540 else 541 goto deliver; 542 543 case ICMP_PARAMPROB: 544 if (code > ICMP_PARAMPROB_LENGTH) 545 goto badcode; 546 547 deliver: 548 /* 549 * Problem with datagram; advise higher level routines. 550 */ 551 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 552 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 553 ICMPSTAT_INC(icps_badlen); 554 goto freeit; 555 } 556 /* Discard ICMP's in response to multicast packets */ 557 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 558 goto badcode; 559 /* Filter out responses to INADDR_ANY, protocols ignore it. */ 560 if (icp->icmp_ip.ip_dst.s_addr == INADDR_ANY || 561 icp->icmp_ip.ip_src.s_addr == INADDR_ANY) 562 goto freeit; 563 #ifdef ICMPPRINTFS 564 if (icmpprintfs) 565 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 566 #endif 567 /* 568 * XXX if the packet contains [IPv4 AH TCP], we can't make a 569 * notification to TCP layer. 570 */ 571 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp)); 572 ip_stripoptions(m); 573 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 574 /* This should actually not happen */ 575 ICMPSTAT_INC(icps_tooshort); 576 return (IPPROTO_DONE); 577 } 578 ip = mtod(m, struct ip *); 579 icp = (struct icmp *)(ip + 1); 580 /* 581 * The upper layer handler can rely on: 582 * - The outer IP header has no options. 583 * - The outer IP header, the ICMP header, the inner IP header, 584 * and the first n bytes of the inner payload are contiguous. 585 * n is at least 8, but might be larger based on 586 * ICMP_ADVLENPREF. See its definition in ip_icmp.h. 587 */ 588 if (ip_ctlprotox[icp->icmp_ip.ip_p] != NULL) 589 ip_ctlprotox[icp->icmp_ip.ip_p](icp); 590 break; 591 592 badcode: 593 ICMPSTAT_INC(icps_badcode); 594 break; 595 596 case ICMP_ECHO: 597 if (!V_icmpbmcastecho 598 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 599 ICMPSTAT_INC(icps_bmcastecho); 600 break; 601 } 602 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 603 goto freeit; 604 icp->icmp_type = ICMP_ECHOREPLY; 605 goto reflect; 606 607 case ICMP_TSTAMP: 608 if (V_icmptstamprepl == 0) 609 break; 610 if (!V_icmpbmcastecho 611 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 612 ICMPSTAT_INC(icps_bmcasttstamp); 613 break; 614 } 615 if (icmplen < ICMP_TSLEN) { 616 ICMPSTAT_INC(icps_badlen); 617 break; 618 } 619 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 620 goto freeit; 621 icp->icmp_type = ICMP_TSTAMPREPLY; 622 icp->icmp_rtime = iptime(); 623 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 624 goto reflect; 625 626 case ICMP_MASKREQ: 627 if (V_icmpmaskrepl == 0) 628 break; 629 /* 630 * We are not able to respond with all ones broadcast 631 * unless we receive it over a point-to-point interface. 632 */ 633 if (icmplen < ICMP_MASKLEN) 634 break; 635 switch (ip->ip_dst.s_addr) { 636 case INADDR_BROADCAST: 637 case INADDR_ANY: 638 icmpdst.sin_addr = ip->ip_src; 639 break; 640 641 default: 642 icmpdst.sin_addr = ip->ip_dst; 643 } 644 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 645 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 646 if (ia == NULL) 647 break; 648 if (ia->ia_ifp == NULL) 649 break; 650 icp->icmp_type = ICMP_MASKREPLY; 651 if (V_icmpmaskfake == 0) 652 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 653 else 654 icp->icmp_mask = V_icmpmaskfake; 655 if (ip->ip_src.s_addr == 0) { 656 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 657 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 658 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 659 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 660 } 661 reflect: 662 ICMPSTAT_INC(icps_reflect); 663 ICMPSTAT_INC(icps_outhist[icp->icmp_type]); 664 icmp_reflect(m); 665 return (IPPROTO_DONE); 666 667 case ICMP_REDIRECT: 668 if (V_log_redirect) { 669 u_long src, dst, gw; 670 671 src = ntohl(ip->ip_src.s_addr); 672 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 673 gw = ntohl(icp->icmp_gwaddr.s_addr); 674 printf("icmp redirect from %d.%d.%d.%d: " 675 "%d.%d.%d.%d => %d.%d.%d.%d\n", 676 (int)(src >> 24), (int)((src >> 16) & 0xff), 677 (int)((src >> 8) & 0xff), (int)(src & 0xff), 678 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 679 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 680 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 681 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 682 } 683 /* 684 * RFC1812 says we must ignore ICMP redirects if we 685 * are acting as router. 686 */ 687 if (V_drop_redirect || V_ipforwarding) 688 break; 689 if (code > 3) 690 goto badcode; 691 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 692 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 693 ICMPSTAT_INC(icps_badlen); 694 break; 695 } 696 /* 697 * Short circuit routing redirects to force 698 * immediate change in the kernel's routing 699 * tables. The message is also handed to anyone 700 * listening on a raw socket (e.g. the routing 701 * daemon for use in updating its tables). 702 */ 703 icmpgw.sin_addr = ip->ip_src; 704 icmpdst.sin_addr = icp->icmp_gwaddr; 705 #ifdef ICMPPRINTFS 706 if (icmpprintfs) { 707 char dstbuf[INET_ADDRSTRLEN]; 708 char gwbuf[INET_ADDRSTRLEN]; 709 710 printf("redirect dst %s to %s\n", 711 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf), 712 inet_ntoa_r(icp->icmp_gwaddr, gwbuf)); 713 } 714 #endif 715 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 716 717 /* 718 * RFC 1122 says network (code 0,2) redirects SHOULD 719 * be treated identically to the host redirects. 720 * Given that, ignore network masks. 721 */ 722 723 /* 724 * Variable values: 725 * icmpsrc: route destination 726 * icmpdst: route gateway 727 * icmpgw: message source 728 */ 729 730 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst, 731 M_GETFIB(m)) != 0) { 732 /* TODO: increment bad redirects here */ 733 break; 734 } 735 736 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 737 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc, 738 (struct sockaddr *)&icmpdst, 739 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif, 740 RTF_GATEWAY, V_redirtimeout); 741 } 742 break; 743 744 /* 745 * No kernel processing for the following; 746 * just fall through to send to raw listener. 747 */ 748 case ICMP_ECHOREPLY: 749 case ICMP_ROUTERADVERT: 750 case ICMP_ROUTERSOLICIT: 751 case ICMP_TSTAMPREPLY: 752 case ICMP_IREQREPLY: 753 case ICMP_MASKREPLY: 754 case ICMP_SOURCEQUENCH: 755 default: 756 break; 757 } 758 759 raw: 760 *mp = m; 761 rip_input(mp, offp, proto); 762 return (IPPROTO_DONE); 763 764 freeit: 765 m_freem(m); 766 return (IPPROTO_DONE); 767 } 768 769 /* 770 * Reflect the ip packet back to the source 771 */ 772 static void 773 icmp_reflect(struct mbuf *m) 774 { 775 struct ip *ip = mtod(m, struct ip *); 776 struct ifaddr *ifa; 777 struct ifnet *ifp; 778 struct in_ifaddr *ia; 779 struct in_addr t; 780 struct nhop_object *nh; 781 struct mbuf *opts = NULL; 782 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 783 784 NET_EPOCH_ASSERT(); 785 786 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 787 (IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net240) || 788 (IN_ZERONET(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net0) ) { 789 m_freem(m); /* Bad return address */ 790 ICMPSTAT_INC(icps_badaddr); 791 goto done; /* Ip_output() will check for broadcast */ 792 } 793 794 t = ip->ip_dst; 795 ip->ip_dst = ip->ip_src; 796 797 /* 798 * Source selection for ICMP replies: 799 * 800 * If the incoming packet was addressed directly to one of our 801 * own addresses, use dst as the src for the reply. 802 */ 803 CK_LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { 804 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { 805 t = IA_SIN(ia)->sin_addr; 806 goto match; 807 } 808 } 809 810 /* 811 * If the incoming packet was addressed to one of our broadcast 812 * addresses, use the first non-broadcast address which corresponds 813 * to the incoming interface. 814 */ 815 ifp = m->m_pkthdr.rcvif; 816 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 817 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 818 if (ifa->ifa_addr->sa_family != AF_INET) 819 continue; 820 ia = ifatoia(ifa); 821 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 822 t.s_addr) { 823 t = IA_SIN(ia)->sin_addr; 824 goto match; 825 } 826 } 827 } 828 /* 829 * If the packet was transiting through us, use the address of 830 * the interface the packet came through in. If that interface 831 * doesn't have a suitable IP address, the normal selection 832 * criteria apply. 833 */ 834 if (V_icmp_rfi && ifp != NULL) { 835 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 836 if (ifa->ifa_addr->sa_family != AF_INET) 837 continue; 838 ia = ifatoia(ifa); 839 t = IA_SIN(ia)->sin_addr; 840 goto match; 841 } 842 } 843 /* 844 * If the incoming packet was not addressed directly to us, use 845 * designated interface for icmp replies specified by sysctl 846 * net.inet.icmp.reply_src (default not set). Otherwise continue 847 * with normal source selection. 848 */ 849 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) { 850 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 851 if (ifa->ifa_addr->sa_family != AF_INET) 852 continue; 853 ia = ifatoia(ifa); 854 t = IA_SIN(ia)->sin_addr; 855 goto match; 856 } 857 } 858 /* 859 * If the packet was transiting through us, use the address of 860 * the interface that is the closest to the packet source. 861 * When we don't have a route back to the packet source, stop here 862 * and drop the packet. 863 */ 864 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0); 865 if (nh == NULL) { 866 m_freem(m); 867 ICMPSTAT_INC(icps_noroute); 868 goto done; 869 } 870 t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr; 871 match: 872 #ifdef MAC 873 mac_netinet_icmp_replyinplace(m); 874 #endif 875 ip->ip_src = t; 876 ip->ip_ttl = V_ip_defttl; 877 878 if (optlen > 0) { 879 u_char *cp; 880 int opt, cnt; 881 u_int len; 882 883 /* 884 * Retrieve any source routing from the incoming packet; 885 * add on any record-route or timestamp options. 886 */ 887 cp = (u_char *) (ip + 1); 888 if ((opts = ip_srcroute(m)) == NULL && 889 (opts = m_gethdr(M_NOWAIT, MT_DATA))) { 890 opts->m_len = sizeof(struct in_addr); 891 mtod(opts, struct in_addr *)->s_addr = 0; 892 } 893 if (opts) { 894 #ifdef ICMPPRINTFS 895 if (icmpprintfs) 896 printf("icmp_reflect optlen %d rt %d => ", 897 optlen, opts->m_len); 898 #endif 899 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 900 opt = cp[IPOPT_OPTVAL]; 901 if (opt == IPOPT_EOL) 902 break; 903 if (opt == IPOPT_NOP) 904 len = 1; 905 else { 906 if (cnt < IPOPT_OLEN + sizeof(*cp)) 907 break; 908 len = cp[IPOPT_OLEN]; 909 if (len < IPOPT_OLEN + sizeof(*cp) || 910 len > cnt) 911 break; 912 } 913 /* 914 * Should check for overflow, but it "can't happen" 915 */ 916 if (opt == IPOPT_RR || opt == IPOPT_TS || 917 opt == IPOPT_SECURITY) { 918 bcopy((caddr_t)cp, 919 mtod(opts, caddr_t) + opts->m_len, len); 920 opts->m_len += len; 921 } 922 } 923 /* Terminate & pad, if necessary */ 924 cnt = opts->m_len % 4; 925 if (cnt) { 926 for (; cnt < 4; cnt++) { 927 *(mtod(opts, caddr_t) + opts->m_len) = 928 IPOPT_EOL; 929 opts->m_len++; 930 } 931 } 932 #ifdef ICMPPRINTFS 933 if (icmpprintfs) 934 printf("%d\n", opts->m_len); 935 #endif 936 } 937 ip_stripoptions(m); 938 } 939 m_tag_delete_nonpersistent(m); 940 m->m_flags &= ~(M_BCAST|M_MCAST); 941 icmp_send(m, opts); 942 done: 943 if (opts) 944 (void)m_free(opts); 945 } 946 947 /* 948 * Verifies if redirect message is valid, according to RFC 1122 949 * 950 * @src: sockaddr with address of redirect originator 951 * @dst: sockaddr with destination in question 952 * @gateway: new proposed gateway 953 * 954 * Returns 0 on success. 955 */ 956 static int 957 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst, 958 struct sockaddr_in *gateway, u_int fibnum) 959 { 960 struct nhop_object *nh; 961 struct ifaddr *ifa; 962 963 NET_EPOCH_ASSERT(); 964 965 /* Verify the gateway is directly reachable. */ 966 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL) 967 return (ENETUNREACH); 968 969 /* TODO: fib-aware. */ 970 if (ifa_ifwithaddr_check((struct sockaddr *)gateway)) 971 return (EHOSTUNREACH); 972 973 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0); 974 if (nh == NULL) 975 return (EINVAL); 976 977 /* 978 * If the redirect isn't from our current router for this dst, 979 * it's either old or wrong. If it redirects us to ourselves, 980 * we have a routing loop, perhaps as a result of an interface 981 * going down recently. 982 */ 983 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa)) 984 return (EINVAL); 985 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) 986 return (EINVAL); 987 988 /* If host route already exists, ignore redirect. */ 989 if (nh->nh_flags & NHF_HOST) 990 return (EEXIST); 991 992 /* If the prefix is directly reachable, ignore redirect. */ 993 if (!(nh->nh_flags & NHF_GATEWAY)) 994 return (EEXIST); 995 996 return (0); 997 } 998 999 /* 1000 * Send an icmp packet back to the ip level, 1001 * after supplying a checksum. 1002 */ 1003 static void 1004 icmp_send(struct mbuf *m, struct mbuf *opts) 1005 { 1006 struct ip *ip = mtod(m, struct ip *); 1007 int hlen; 1008 struct icmp *icp; 1009 1010 hlen = ip->ip_hl << 2; 1011 m->m_data += hlen; 1012 m->m_len -= hlen; 1013 icp = mtod(m, struct icmp *); 1014 icp->icmp_cksum = 0; 1015 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 1016 m->m_data -= hlen; 1017 m->m_len += hlen; 1018 m->m_pkthdr.rcvif = (struct ifnet *)0; 1019 #ifdef ICMPPRINTFS 1020 if (icmpprintfs) { 1021 char dstbuf[INET_ADDRSTRLEN]; 1022 char srcbuf[INET_ADDRSTRLEN]; 1023 1024 printf("icmp_send dst %s src %s\n", 1025 inet_ntoa_r(ip->ip_dst, dstbuf), 1026 inet_ntoa_r(ip->ip_src, srcbuf)); 1027 } 1028 #endif 1029 (void) ip_output(m, opts, NULL, 0, NULL, NULL); 1030 } 1031 1032 /* 1033 * Return milliseconds since 00:00 UTC in network format. 1034 */ 1035 uint32_t 1036 iptime(void) 1037 { 1038 struct timeval atv; 1039 u_long t; 1040 1041 getmicrotime(&atv); 1042 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1043 return (htonl(t)); 1044 } 1045 1046 /* 1047 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1048 * given current value MTU. If DIR is less than zero, a larger plateau 1049 * is returned; otherwise, a smaller value is returned. 1050 */ 1051 int 1052 ip_next_mtu(int mtu, int dir) 1053 { 1054 static int mtutab[] = { 1055 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 1056 296, 68, 0 1057 }; 1058 int i, size; 1059 1060 size = (sizeof mtutab) / (sizeof mtutab[0]); 1061 if (dir >= 0) { 1062 for (i = 0; i < size; i++) 1063 if (mtu > mtutab[i]) 1064 return mtutab[i]; 1065 } else { 1066 for (i = size - 1; i >= 0; i--) 1067 if (mtu < mtutab[i]) 1068 return mtutab[i]; 1069 if (mtu == mtutab[0]) 1070 return mtutab[0]; 1071 } 1072 return 0; 1073 } 1074 #endif /* INET */ 1075 1076 /* 1077 * badport_bandlim() - check for ICMP bandwidth limit 1078 * 1079 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1080 * hit our bandwidth limit and it is not ok. 1081 * 1082 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1083 * 1084 * For now we separate the TCP and UDP subsystems w/ different 'which' 1085 * values. We may eventually remove this separation (and simplify the 1086 * code further). 1087 * 1088 * Note that the printing of the error message is delayed so we can 1089 * properly print the icmp error rate that the system was trying to do 1090 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1091 * the 'final' error, but it doesn't make sense to solve the printing 1092 * delay with more complex code. 1093 */ 1094 struct icmp_rate { 1095 const char *descr; 1096 struct counter_rate cr; 1097 }; 1098 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = { 1099 { "icmp unreach response" }, 1100 { "icmp ping response" }, 1101 { "icmp tstamp response" }, 1102 { "closed port RST response" }, 1103 { "open port RST response" }, 1104 { "icmp6 unreach response" }, 1105 { "sctp ootb response" } 1106 }; 1107 #define V_icmp_rates VNET(icmp_rates) 1108 1109 static void 1110 icmp_bandlimit_init(void) 1111 { 1112 1113 for (int i = 0; i < BANDLIM_MAX; i++) { 1114 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK); 1115 V_icmp_rates[i].cr.cr_ticks = ticks; 1116 } 1117 } 1118 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, 1119 icmp_bandlimit_init, NULL); 1120 1121 static void 1122 icmp_bandlimit_uninit(void) 1123 { 1124 1125 for (int i = 0; i < BANDLIM_MAX; i++) 1126 counter_u64_free(V_icmp_rates[i].cr.cr_rate); 1127 } 1128 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1129 icmp_bandlimit_uninit, NULL); 1130 1131 int 1132 badport_bandlim(int which) 1133 { 1134 int64_t pps; 1135 1136 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED) 1137 return (0); 1138 1139 KASSERT(which >= 0 && which < BANDLIM_MAX, 1140 ("%s: which %d", __func__, which)); 1141 1142 if ((V_icmplim + V_icmplim_curr_jitter) <= 0) 1143 V_icmplim_curr_jitter = -V_icmplim + 1; 1144 1145 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim + 1146 V_icmplim_curr_jitter); 1147 if (pps > 0) { 1148 /* 1149 * Adjust limit +/- to jitter the measurement to deny a 1150 * side-channel port scan as in CVE-2020-25705 1151 */ 1152 if (V_icmplim_jitter > 0) { 1153 int32_t inc = 1154 arc4random_uniform(V_icmplim_jitter * 2 +1) 1155 - V_icmplim_jitter; 1156 1157 V_icmplim_curr_jitter = inc; 1158 } 1159 } 1160 if (pps == -1) 1161 return (-1); 1162 if (pps > 0 && V_icmplim_output) 1163 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n", 1164 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim + 1165 V_icmplim_curr_jitter); 1166 return (0); 1167 } 1168