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 /* 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 extern struct protosw inetsw[]; 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 /* 407 * Process a received ICMP message. 408 */ 409 int 410 icmp_input(struct mbuf **mp, int *offp, int proto) 411 { 412 struct icmp *icp; 413 struct in_ifaddr *ia; 414 struct mbuf *m = *mp; 415 struct ip *ip = mtod(m, struct ip *); 416 struct sockaddr_in icmpsrc, icmpdst, icmpgw; 417 int hlen = *offp; 418 int icmplen = ntohs(ip->ip_len) - *offp; 419 int i, code; 420 void (*ctlfunc)(int, struct sockaddr *, void *); 421 int fibnum; 422 423 NET_EPOCH_ASSERT(); 424 425 *mp = NULL; 426 427 /* 428 * Locate icmp structure in mbuf, and check 429 * that not corrupted and of at least minimum length. 430 */ 431 #ifdef ICMPPRINTFS 432 if (icmpprintfs) { 433 char srcbuf[INET_ADDRSTRLEN]; 434 char dstbuf[INET_ADDRSTRLEN]; 435 436 printf("icmp_input from %s to %s, len %d\n", 437 inet_ntoa_r(ip->ip_src, srcbuf), 438 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen); 439 } 440 #endif 441 if (icmplen < ICMP_MINLEN) { 442 ICMPSTAT_INC(icps_tooshort); 443 goto freeit; 444 } 445 i = hlen + min(icmplen, ICMP_ADVLENMIN); 446 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 447 ICMPSTAT_INC(icps_tooshort); 448 return (IPPROTO_DONE); 449 } 450 ip = mtod(m, struct ip *); 451 m->m_len -= hlen; 452 m->m_data += hlen; 453 icp = mtod(m, struct icmp *); 454 if (in_cksum(m, icmplen)) { 455 ICMPSTAT_INC(icps_checksum); 456 goto freeit; 457 } 458 m->m_len += hlen; 459 m->m_data -= hlen; 460 461 #ifdef ICMPPRINTFS 462 if (icmpprintfs) 463 printf("icmp_input, type %d code %d\n", icp->icmp_type, 464 icp->icmp_code); 465 #endif 466 467 /* 468 * Message type specific processing. 469 */ 470 if (icp->icmp_type > ICMP_MAXTYPE) 471 goto raw; 472 473 /* Initialize */ 474 bzero(&icmpsrc, sizeof(icmpsrc)); 475 icmpsrc.sin_len = sizeof(struct sockaddr_in); 476 icmpsrc.sin_family = AF_INET; 477 bzero(&icmpdst, sizeof(icmpdst)); 478 icmpdst.sin_len = sizeof(struct sockaddr_in); 479 icmpdst.sin_family = AF_INET; 480 bzero(&icmpgw, sizeof(icmpgw)); 481 icmpgw.sin_len = sizeof(struct sockaddr_in); 482 icmpgw.sin_family = AF_INET; 483 484 ICMPSTAT_INC(icps_inhist[icp->icmp_type]); 485 code = icp->icmp_code; 486 switch (icp->icmp_type) { 487 case ICMP_UNREACH: 488 switch (code) { 489 case ICMP_UNREACH_NET: 490 case ICMP_UNREACH_HOST: 491 case ICMP_UNREACH_SRCFAIL: 492 case ICMP_UNREACH_NET_UNKNOWN: 493 case ICMP_UNREACH_HOST_UNKNOWN: 494 case ICMP_UNREACH_ISOLATED: 495 case ICMP_UNREACH_TOSNET: 496 case ICMP_UNREACH_TOSHOST: 497 case ICMP_UNREACH_HOST_PRECEDENCE: 498 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 499 code = PRC_UNREACH_NET; 500 break; 501 502 case ICMP_UNREACH_NEEDFRAG: 503 code = PRC_MSGSIZE; 504 break; 505 506 /* 507 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. 508 * Treat subcodes 2,3 as immediate RST 509 */ 510 case ICMP_UNREACH_PROTOCOL: 511 code = PRC_UNREACH_PROTOCOL; 512 break; 513 case ICMP_UNREACH_PORT: 514 code = PRC_UNREACH_PORT; 515 break; 516 517 case ICMP_UNREACH_NET_PROHIB: 518 case ICMP_UNREACH_HOST_PROHIB: 519 case ICMP_UNREACH_FILTER_PROHIB: 520 code = PRC_UNREACH_ADMIN_PROHIB; 521 break; 522 523 default: 524 goto badcode; 525 } 526 goto deliver; 527 528 case ICMP_TIMXCEED: 529 if (code > 1) 530 goto badcode; 531 code += PRC_TIMXCEED_INTRANS; 532 goto deliver; 533 534 case ICMP_PARAMPROB: 535 if (code > 1) 536 goto badcode; 537 code = PRC_PARAMPROB; 538 deliver: 539 /* 540 * Problem with datagram; advise higher level routines. 541 */ 542 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 543 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 544 ICMPSTAT_INC(icps_badlen); 545 goto freeit; 546 } 547 /* Discard ICMP's in response to multicast packets */ 548 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 549 goto badcode; 550 #ifdef ICMPPRINTFS 551 if (icmpprintfs) 552 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 553 #endif 554 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 555 /* 556 * XXX if the packet contains [IPv4 AH TCP], we can't make a 557 * notification to TCP layer. 558 */ 559 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp)); 560 ip_stripoptions(m); 561 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 562 /* This should actually not happen */ 563 ICMPSTAT_INC(icps_tooshort); 564 return (IPPROTO_DONE); 565 } 566 ip = mtod(m, struct ip *); 567 icp = (struct icmp *)(ip + 1); 568 /* 569 * The upper layer handler can rely on: 570 * - The outer IP header has no options. 571 * - The outer IP header, the ICMP header, the inner IP header, 572 * and the first n bytes of the inner payload are contiguous. 573 * n is at least 8, but might be larger based on 574 * ICMP_ADVLENPREF. See its definition in ip_icmp.h. 575 */ 576 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 577 if (ctlfunc) 578 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc, 579 (void *)&icp->icmp_ip); 580 break; 581 582 badcode: 583 ICMPSTAT_INC(icps_badcode); 584 break; 585 586 case ICMP_ECHO: 587 if (!V_icmpbmcastecho 588 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 589 ICMPSTAT_INC(icps_bmcastecho); 590 break; 591 } 592 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 593 goto freeit; 594 icp->icmp_type = ICMP_ECHOREPLY; 595 goto reflect; 596 597 case ICMP_TSTAMP: 598 if (V_icmptstamprepl == 0) 599 break; 600 if (!V_icmpbmcastecho 601 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 602 ICMPSTAT_INC(icps_bmcasttstamp); 603 break; 604 } 605 if (icmplen < ICMP_TSLEN) { 606 ICMPSTAT_INC(icps_badlen); 607 break; 608 } 609 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 610 goto freeit; 611 icp->icmp_type = ICMP_TSTAMPREPLY; 612 icp->icmp_rtime = iptime(); 613 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 614 goto reflect; 615 616 case ICMP_MASKREQ: 617 if (V_icmpmaskrepl == 0) 618 break; 619 /* 620 * We are not able to respond with all ones broadcast 621 * unless we receive it over a point-to-point interface. 622 */ 623 if (icmplen < ICMP_MASKLEN) 624 break; 625 switch (ip->ip_dst.s_addr) { 626 case INADDR_BROADCAST: 627 case INADDR_ANY: 628 icmpdst.sin_addr = ip->ip_src; 629 break; 630 631 default: 632 icmpdst.sin_addr = ip->ip_dst; 633 } 634 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 635 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 636 if (ia == NULL) 637 break; 638 if (ia->ia_ifp == NULL) 639 break; 640 icp->icmp_type = ICMP_MASKREPLY; 641 if (V_icmpmaskfake == 0) 642 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 643 else 644 icp->icmp_mask = V_icmpmaskfake; 645 if (ip->ip_src.s_addr == 0) { 646 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 647 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 648 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 649 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 650 } 651 reflect: 652 ICMPSTAT_INC(icps_reflect); 653 ICMPSTAT_INC(icps_outhist[icp->icmp_type]); 654 icmp_reflect(m); 655 return (IPPROTO_DONE); 656 657 case ICMP_REDIRECT: 658 if (V_log_redirect) { 659 u_long src, dst, gw; 660 661 src = ntohl(ip->ip_src.s_addr); 662 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 663 gw = ntohl(icp->icmp_gwaddr.s_addr); 664 printf("icmp redirect from %d.%d.%d.%d: " 665 "%d.%d.%d.%d => %d.%d.%d.%d\n", 666 (int)(src >> 24), (int)((src >> 16) & 0xff), 667 (int)((src >> 8) & 0xff), (int)(src & 0xff), 668 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 669 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 670 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 671 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 672 } 673 /* 674 * RFC1812 says we must ignore ICMP redirects if we 675 * are acting as router. 676 */ 677 if (V_drop_redirect || V_ipforwarding) 678 break; 679 if (code > 3) 680 goto badcode; 681 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 682 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 683 ICMPSTAT_INC(icps_badlen); 684 break; 685 } 686 /* 687 * Short circuit routing redirects to force 688 * immediate change in the kernel's routing 689 * tables. The message is also handed to anyone 690 * listening on a raw socket (e.g. the routing 691 * daemon for use in updating its tables). 692 */ 693 icmpgw.sin_addr = ip->ip_src; 694 icmpdst.sin_addr = icp->icmp_gwaddr; 695 #ifdef ICMPPRINTFS 696 if (icmpprintfs) { 697 char dstbuf[INET_ADDRSTRLEN]; 698 char gwbuf[INET_ADDRSTRLEN]; 699 700 printf("redirect dst %s to %s\n", 701 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf), 702 inet_ntoa_r(icp->icmp_gwaddr, gwbuf)); 703 } 704 #endif 705 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 706 707 /* 708 * RFC 1122 says network (code 0,2) redirects SHOULD 709 * be treated identically to the host redirects. 710 * Given that, ignore network masks. 711 */ 712 713 /* 714 * Variable values: 715 * icmpsrc: route destination 716 * icmpdst: route gateway 717 * icmpgw: message source 718 */ 719 720 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst, 721 M_GETFIB(m)) != 0) { 722 /* TODO: increment bad redirects here */ 723 break; 724 } 725 726 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 727 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc, 728 (struct sockaddr *)&icmpdst, 729 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif, 730 RTF_GATEWAY, V_redirtimeout); 731 } 732 break; 733 734 /* 735 * No kernel processing for the following; 736 * just fall through to send to raw listener. 737 */ 738 case ICMP_ECHOREPLY: 739 case ICMP_ROUTERADVERT: 740 case ICMP_ROUTERSOLICIT: 741 case ICMP_TSTAMPREPLY: 742 case ICMP_IREQREPLY: 743 case ICMP_MASKREPLY: 744 case ICMP_SOURCEQUENCH: 745 default: 746 break; 747 } 748 749 raw: 750 *mp = m; 751 rip_input(mp, offp, proto); 752 return (IPPROTO_DONE); 753 754 freeit: 755 m_freem(m); 756 return (IPPROTO_DONE); 757 } 758 759 /* 760 * Reflect the ip packet back to the source 761 */ 762 static void 763 icmp_reflect(struct mbuf *m) 764 { 765 struct ip *ip = mtod(m, struct ip *); 766 struct ifaddr *ifa; 767 struct ifnet *ifp; 768 struct in_ifaddr *ia; 769 struct in_addr t; 770 struct nhop_object *nh; 771 struct mbuf *opts = NULL; 772 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 773 774 NET_EPOCH_ASSERT(); 775 776 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 777 (IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net240) || 778 (IN_ZERONET(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net0) ) { 779 m_freem(m); /* Bad return address */ 780 ICMPSTAT_INC(icps_badaddr); 781 goto done; /* Ip_output() will check for broadcast */ 782 } 783 784 t = ip->ip_dst; 785 ip->ip_dst = ip->ip_src; 786 787 /* 788 * Source selection for ICMP replies: 789 * 790 * If the incoming packet was addressed directly to one of our 791 * own addresses, use dst as the src for the reply. 792 */ 793 CK_LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { 794 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { 795 t = IA_SIN(ia)->sin_addr; 796 goto match; 797 } 798 } 799 800 /* 801 * If the incoming packet was addressed to one of our broadcast 802 * addresses, use the first non-broadcast address which corresponds 803 * to the incoming interface. 804 */ 805 ifp = m->m_pkthdr.rcvif; 806 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 807 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 808 if (ifa->ifa_addr->sa_family != AF_INET) 809 continue; 810 ia = ifatoia(ifa); 811 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 812 t.s_addr) { 813 t = IA_SIN(ia)->sin_addr; 814 goto match; 815 } 816 } 817 } 818 /* 819 * If the packet was transiting through us, use the address of 820 * the interface the packet came through in. If that interface 821 * doesn't have a suitable IP address, the normal selection 822 * criteria apply. 823 */ 824 if (V_icmp_rfi && ifp != NULL) { 825 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 826 if (ifa->ifa_addr->sa_family != AF_INET) 827 continue; 828 ia = ifatoia(ifa); 829 t = IA_SIN(ia)->sin_addr; 830 goto match; 831 } 832 } 833 /* 834 * If the incoming packet was not addressed directly to us, use 835 * designated interface for icmp replies specified by sysctl 836 * net.inet.icmp.reply_src (default not set). Otherwise continue 837 * with normal source selection. 838 */ 839 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) { 840 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 841 if (ifa->ifa_addr->sa_family != AF_INET) 842 continue; 843 ia = ifatoia(ifa); 844 t = IA_SIN(ia)->sin_addr; 845 goto match; 846 } 847 } 848 /* 849 * If the packet was transiting through us, use the address of 850 * the interface that is the closest to the packet source. 851 * When we don't have a route back to the packet source, stop here 852 * and drop the packet. 853 */ 854 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0); 855 if (nh == NULL) { 856 m_freem(m); 857 ICMPSTAT_INC(icps_noroute); 858 goto done; 859 } 860 t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr; 861 match: 862 #ifdef MAC 863 mac_netinet_icmp_replyinplace(m); 864 #endif 865 ip->ip_src = t; 866 ip->ip_ttl = V_ip_defttl; 867 868 if (optlen > 0) { 869 u_char *cp; 870 int opt, cnt; 871 u_int len; 872 873 /* 874 * Retrieve any source routing from the incoming packet; 875 * add on any record-route or timestamp options. 876 */ 877 cp = (u_char *) (ip + 1); 878 if ((opts = ip_srcroute(m)) == NULL && 879 (opts = m_gethdr(M_NOWAIT, MT_DATA))) { 880 opts->m_len = sizeof(struct in_addr); 881 mtod(opts, struct in_addr *)->s_addr = 0; 882 } 883 if (opts) { 884 #ifdef ICMPPRINTFS 885 if (icmpprintfs) 886 printf("icmp_reflect optlen %d rt %d => ", 887 optlen, opts->m_len); 888 #endif 889 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 890 opt = cp[IPOPT_OPTVAL]; 891 if (opt == IPOPT_EOL) 892 break; 893 if (opt == IPOPT_NOP) 894 len = 1; 895 else { 896 if (cnt < IPOPT_OLEN + sizeof(*cp)) 897 break; 898 len = cp[IPOPT_OLEN]; 899 if (len < IPOPT_OLEN + sizeof(*cp) || 900 len > cnt) 901 break; 902 } 903 /* 904 * Should check for overflow, but it "can't happen" 905 */ 906 if (opt == IPOPT_RR || opt == IPOPT_TS || 907 opt == IPOPT_SECURITY) { 908 bcopy((caddr_t)cp, 909 mtod(opts, caddr_t) + opts->m_len, len); 910 opts->m_len += len; 911 } 912 } 913 /* Terminate & pad, if necessary */ 914 cnt = opts->m_len % 4; 915 if (cnt) { 916 for (; cnt < 4; cnt++) { 917 *(mtod(opts, caddr_t) + opts->m_len) = 918 IPOPT_EOL; 919 opts->m_len++; 920 } 921 } 922 #ifdef ICMPPRINTFS 923 if (icmpprintfs) 924 printf("%d\n", opts->m_len); 925 #endif 926 } 927 ip_stripoptions(m); 928 } 929 m_tag_delete_nonpersistent(m); 930 m->m_flags &= ~(M_BCAST|M_MCAST); 931 icmp_send(m, opts); 932 done: 933 if (opts) 934 (void)m_free(opts); 935 } 936 937 /* 938 * Verifies if redirect message is valid, according to RFC 1122 939 * 940 * @src: sockaddr with address of redirect originator 941 * @dst: sockaddr with destination in question 942 * @gateway: new proposed gateway 943 * 944 * Returns 0 on success. 945 */ 946 static int 947 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst, 948 struct sockaddr_in *gateway, u_int fibnum) 949 { 950 struct nhop_object *nh; 951 struct ifaddr *ifa; 952 953 NET_EPOCH_ASSERT(); 954 955 /* Verify the gateway is directly reachable. */ 956 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL) 957 return (ENETUNREACH); 958 959 /* TODO: fib-aware. */ 960 if (ifa_ifwithaddr_check((struct sockaddr *)gateway)) 961 return (EHOSTUNREACH); 962 963 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0); 964 if (nh == NULL) 965 return (EINVAL); 966 967 /* 968 * If the redirect isn't from our current router for this dst, 969 * it's either old or wrong. If it redirects us to ourselves, 970 * we have a routing loop, perhaps as a result of an interface 971 * going down recently. 972 */ 973 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa)) 974 return (EINVAL); 975 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) 976 return (EINVAL); 977 978 /* If host route already exists, ignore redirect. */ 979 if (nh->nh_flags & NHF_HOST) 980 return (EEXIST); 981 982 /* If the prefix is directly reachable, ignore redirect. */ 983 if (!(nh->nh_flags & NHF_GATEWAY)) 984 return (EEXIST); 985 986 return (0); 987 } 988 989 /* 990 * Send an icmp packet back to the ip level, 991 * after supplying a checksum. 992 */ 993 static void 994 icmp_send(struct mbuf *m, struct mbuf *opts) 995 { 996 struct ip *ip = mtod(m, struct ip *); 997 int hlen; 998 struct icmp *icp; 999 1000 hlen = ip->ip_hl << 2; 1001 m->m_data += hlen; 1002 m->m_len -= hlen; 1003 icp = mtod(m, struct icmp *); 1004 icp->icmp_cksum = 0; 1005 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 1006 m->m_data -= hlen; 1007 m->m_len += hlen; 1008 m->m_pkthdr.rcvif = (struct ifnet *)0; 1009 #ifdef ICMPPRINTFS 1010 if (icmpprintfs) { 1011 char dstbuf[INET_ADDRSTRLEN]; 1012 char srcbuf[INET_ADDRSTRLEN]; 1013 1014 printf("icmp_send dst %s src %s\n", 1015 inet_ntoa_r(ip->ip_dst, dstbuf), 1016 inet_ntoa_r(ip->ip_src, srcbuf)); 1017 } 1018 #endif 1019 (void) ip_output(m, opts, NULL, 0, NULL, NULL); 1020 } 1021 1022 /* 1023 * Return milliseconds since 00:00 UTC in network format. 1024 */ 1025 uint32_t 1026 iptime(void) 1027 { 1028 struct timeval atv; 1029 u_long t; 1030 1031 getmicrotime(&atv); 1032 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1033 return (htonl(t)); 1034 } 1035 1036 /* 1037 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1038 * given current value MTU. If DIR is less than zero, a larger plateau 1039 * is returned; otherwise, a smaller value is returned. 1040 */ 1041 int 1042 ip_next_mtu(int mtu, int dir) 1043 { 1044 static int mtutab[] = { 1045 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 1046 296, 68, 0 1047 }; 1048 int i, size; 1049 1050 size = (sizeof mtutab) / (sizeof mtutab[0]); 1051 if (dir >= 0) { 1052 for (i = 0; i < size; i++) 1053 if (mtu > mtutab[i]) 1054 return mtutab[i]; 1055 } else { 1056 for (i = size - 1; i >= 0; i--) 1057 if (mtu < mtutab[i]) 1058 return mtutab[i]; 1059 if (mtu == mtutab[0]) 1060 return mtutab[0]; 1061 } 1062 return 0; 1063 } 1064 #endif /* INET */ 1065 1066 /* 1067 * badport_bandlim() - check for ICMP bandwidth limit 1068 * 1069 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1070 * hit our bandwidth limit and it is not ok. 1071 * 1072 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1073 * 1074 * For now we separate the TCP and UDP subsystems w/ different 'which' 1075 * values. We may eventually remove this separation (and simplify the 1076 * code further). 1077 * 1078 * Note that the printing of the error message is delayed so we can 1079 * properly print the icmp error rate that the system was trying to do 1080 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1081 * the 'final' error, but it doesn't make sense to solve the printing 1082 * delay with more complex code. 1083 */ 1084 struct icmp_rate { 1085 const char *descr; 1086 struct counter_rate cr; 1087 }; 1088 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = { 1089 { "icmp unreach response" }, 1090 { "icmp ping response" }, 1091 { "icmp tstamp response" }, 1092 { "closed port RST response" }, 1093 { "open port RST response" }, 1094 { "icmp6 unreach response" }, 1095 { "sctp ootb response" } 1096 }; 1097 #define V_icmp_rates VNET(icmp_rates) 1098 1099 static void 1100 icmp_bandlimit_init(void) 1101 { 1102 1103 for (int i = 0; i < BANDLIM_MAX; i++) { 1104 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK); 1105 V_icmp_rates[i].cr.cr_ticks = ticks; 1106 } 1107 } 1108 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, 1109 icmp_bandlimit_init, NULL); 1110 1111 static void 1112 icmp_bandlimit_uninit(void) 1113 { 1114 1115 for (int i = 0; i < BANDLIM_MAX; i++) 1116 counter_u64_free(V_icmp_rates[i].cr.cr_rate); 1117 } 1118 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1119 icmp_bandlimit_uninit, NULL); 1120 1121 int 1122 badport_bandlim(int which) 1123 { 1124 int64_t pps; 1125 1126 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED) 1127 return (0); 1128 1129 KASSERT(which >= 0 && which < BANDLIM_MAX, 1130 ("%s: which %d", __func__, which)); 1131 1132 if ((V_icmplim + V_icmplim_curr_jitter) <= 0) 1133 V_icmplim_curr_jitter = -V_icmplim + 1; 1134 1135 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim + 1136 V_icmplim_curr_jitter); 1137 if (pps > 0) { 1138 /* 1139 * Adjust limit +/- to jitter the measurement to deny a 1140 * side-channel port scan as in CVE-2020-25705 1141 */ 1142 if (V_icmplim_jitter > 0) { 1143 int32_t inc = 1144 arc4random_uniform(V_icmplim_jitter * 2 +1) 1145 - V_icmplim_jitter; 1146 1147 V_icmplim_curr_jitter = inc; 1148 } 1149 } 1150 if (pps == -1) 1151 return (-1); 1152 if (pps > 0 && V_icmplim_output) 1153 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n", 1154 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim + 1155 V_icmplim_curr_jitter); 1156 return (0); 1157 } 1158