1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_inet.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/mbuf.h> 42 #include <sys/protosw.h> 43 #include <sys/socket.h> 44 #include <sys/time.h> 45 #include <sys/kernel.h> 46 #include <sys/lock.h> 47 #include <sys/rmlock.h> 48 #include <sys/sysctl.h> 49 #include <sys/syslog.h> 50 51 #include <net/if.h> 52 #include <net/if_var.h> 53 #include <net/if_types.h> 54 #include <net/route.h> 55 #include <net/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 74 #ifdef INET 75 76 #include <machine/in_cksum.h> 77 78 #include <security/mac/mac_framework.h> 79 #endif /* INET */ 80 81 /* 82 * ICMP routines: error generation, receive packet processing, and 83 * routines to turnaround packets back to the originator, and 84 * host table maintenance routines. 85 */ 86 VNET_DEFINE_STATIC(int, icmplim) = 200; 87 #define V_icmplim VNET(icmplim) 88 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW, 89 &VNET_NAME(icmplim), 0, 90 "Maximum number of ICMP responses per second"); 91 92 VNET_DEFINE_STATIC(int, icmplim_output) = 1; 93 #define V_icmplim_output VNET(icmplim_output) 94 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW, 95 &VNET_NAME(icmplim_output), 0, 96 "Enable logging of ICMP response rate limiting"); 97 98 #ifdef INET 99 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat); 100 VNET_PCPUSTAT_SYSINIT(icmpstat); 101 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat, 102 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)"); 103 104 #ifdef VIMAGE 105 VNET_PCPUSTAT_SYSUNINIT(icmpstat); 106 #endif /* VIMAGE */ 107 108 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0; 109 #define V_icmpmaskrepl VNET(icmpmaskrepl) 110 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW, 111 &VNET_NAME(icmpmaskrepl), 0, 112 "Reply to ICMP Address Mask Request packets"); 113 114 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0; 115 #define V_icmpmaskfake VNET(icmpmaskfake) 116 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW, 117 &VNET_NAME(icmpmaskfake), 0, 118 "Fake reply to ICMP Address Mask Request packets"); 119 120 VNET_DEFINE(int, drop_redirect) = 0; 121 #define V_drop_redirect VNET(drop_redirect) 122 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW, 123 &VNET_NAME(drop_redirect), 0, 124 "Ignore ICMP redirects"); 125 126 VNET_DEFINE_STATIC(int, log_redirect) = 0; 127 #define V_log_redirect VNET(log_redirect) 128 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW, 129 &VNET_NAME(log_redirect), 0, 130 "Log ICMP redirects to the console"); 131 132 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */ 133 #define V_redirtimeout VNET(redirtimeout) 134 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW, 135 &VNET_NAME(redirtimeout), 0, 136 "Delay in seconds before expiring redirect route"); 137 138 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]); 139 #define V_reply_src VNET(reply_src) 140 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW, 141 &VNET_NAME(reply_src), IFNAMSIZ, 142 "ICMP reply source for non-local packets"); 143 144 VNET_DEFINE_STATIC(int, icmp_rfi) = 0; 145 #define V_icmp_rfi VNET(icmp_rfi) 146 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW, 147 &VNET_NAME(icmp_rfi), 0, 148 "ICMP reply from incoming interface for non-local packets"); 149 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */ 150 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548; 151 #define V_icmp_quotelen VNET(icmp_quotelen) 152 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW, 153 &VNET_NAME(icmp_quotelen), 0, 154 "Number of bytes from original packet to quote in ICMP reply"); 155 156 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0; 157 #define V_icmpbmcastecho VNET(icmpbmcastecho) 158 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW, 159 &VNET_NAME(icmpbmcastecho), 0, 160 "Reply to multicast ICMP Echo Request and Timestamp packets"); 161 162 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1; 163 #define V_icmptstamprepl VNET(icmptstamprepl) 164 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW, 165 &VNET_NAME(icmptstamprepl), 0, 166 "Respond to ICMP Timestamp packets"); 167 168 VNET_DEFINE_STATIC(int, error_keeptags) = 0; 169 #define V_error_keeptags VNET(error_keeptags) 170 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW, 171 &VNET_NAME(error_keeptags), 0, 172 "ICMP error response keeps copy of mbuf_tags of original packet"); 173 174 #ifdef ICMPPRINTFS 175 int icmpprintfs = 0; 176 #endif 177 178 static void icmp_reflect(struct mbuf *); 179 static void icmp_send(struct mbuf *, struct mbuf *); 180 static int icmp_verify_redirect_gateway(struct sockaddr_in *, 181 struct sockaddr_in *, struct sockaddr_in *, u_int); 182 183 extern struct protosw inetsw[]; 184 185 /* 186 * Kernel module interface for updating icmpstat. The argument is an index 187 * into icmpstat treated as an array of u_long. While this encodes the 188 * general layout of icmpstat into the caller, it doesn't encode its 189 * location, so that future changes to add, for example, per-CPU stats 190 * support won't cause binary compatibility problems for kernel modules. 191 */ 192 void 193 kmod_icmpstat_inc(int statnum) 194 { 195 196 counter_u64_add(VNET(icmpstat)[statnum], 1); 197 } 198 199 /* 200 * Generate an error packet of type error 201 * in response to bad packet ip. 202 */ 203 void 204 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu) 205 { 206 struct ip *oip, *nip; 207 struct icmp *icp; 208 struct mbuf *m; 209 unsigned icmplen, icmpelen, nlen, oiphlen; 210 211 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", 212 __func__)); 213 214 if (type != ICMP_REDIRECT) 215 ICMPSTAT_INC(icps_error); 216 /* 217 * Don't send error: 218 * if the original packet was encrypted. 219 * if not the first fragment of message. 220 * in response to a multicast or broadcast packet. 221 * if the old packet protocol was an ICMP error message. 222 */ 223 if (n->m_flags & M_DECRYPTED) 224 goto freeit; 225 if (n->m_flags & (M_BCAST|M_MCAST)) 226 goto freeit; 227 228 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */ 229 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) 230 goto freeit; 231 oip = mtod(n, struct ip *); 232 oiphlen = oip->ip_hl << 2; 233 if (n->m_len < oiphlen + ICMP_MINLEN) 234 goto freeit; 235 #ifdef ICMPPRINTFS 236 if (icmpprintfs) 237 printf("icmp_error(%p, %x, %d)\n", oip, type, code); 238 #endif 239 if (oip->ip_off & htons(~(IP_MF|IP_DF))) 240 goto freeit; 241 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 242 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + 243 oiphlen))->icmp_type)) { 244 ICMPSTAT_INC(icps_oldicmp); 245 goto freeit; 246 } 247 /* 248 * Calculate length to quote from original packet and 249 * prevent the ICMP mbuf from overflowing. 250 * Unfortunately this is non-trivial since ip_forward() 251 * sends us truncated packets. 252 */ 253 nlen = m_length(n, NULL); 254 if (oip->ip_p == IPPROTO_TCP) { 255 struct tcphdr *th; 256 int tcphlen; 257 258 if (oiphlen + sizeof(struct tcphdr) > n->m_len && 259 n->m_next == NULL) 260 goto stdreply; 261 if (n->m_len < oiphlen + sizeof(struct tcphdr) && 262 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL) 263 goto freeit; 264 oip = mtod(n, struct ip *); 265 th = mtodo(n, oiphlen); 266 tcphlen = th->th_off << 2; 267 if (tcphlen < sizeof(struct tcphdr)) 268 goto freeit; 269 if (ntohs(oip->ip_len) < oiphlen + tcphlen) 270 goto freeit; 271 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL) 272 goto stdreply; 273 if (n->m_len < oiphlen + tcphlen && 274 (n = m_pullup(n, oiphlen + tcphlen)) == NULL) 275 goto freeit; 276 oip = mtod(n, struct ip *); 277 icmpelen = max(tcphlen, min(V_icmp_quotelen, 278 ntohs(oip->ip_len) - oiphlen)); 279 } else if (oip->ip_p == IPPROTO_SCTP) { 280 struct sctphdr *sh; 281 struct sctp_chunkhdr *ch; 282 283 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr)) 284 goto stdreply; 285 if (oiphlen + sizeof(struct sctphdr) > n->m_len && 286 n->m_next == NULL) 287 goto stdreply; 288 if (n->m_len < oiphlen + sizeof(struct sctphdr) && 289 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL) 290 goto freeit; 291 oip = mtod(n, struct ip *); 292 icmpelen = max(sizeof(struct sctphdr), 293 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen)); 294 sh = mtodo(n, oiphlen); 295 if (ntohl(sh->v_tag) == 0 && 296 ntohs(oip->ip_len) >= oiphlen + 297 sizeof(struct sctphdr) + 8 && 298 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 || 299 n->m_next != NULL)) { 300 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 && 301 (n = m_pullup(n, oiphlen + 302 sizeof(struct sctphdr) + 8)) == NULL) 303 goto freeit; 304 oip = mtod(n, struct ip *); 305 sh = mtodo(n, oiphlen); 306 ch = (struct sctp_chunkhdr *)(sh + 1); 307 if (ch->chunk_type == SCTP_INITIATION) { 308 icmpelen = max(sizeof(struct sctphdr) + 8, 309 min(V_icmp_quotelen, ntohs(oip->ip_len) - 310 oiphlen)); 311 } 312 } 313 } else 314 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - 315 oiphlen)); 316 317 icmplen = min(oiphlen + icmpelen, nlen); 318 if (icmplen < sizeof(struct ip)) 319 goto freeit; 320 321 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen) 322 m = m_gethdr(M_NOWAIT, MT_DATA); 323 else 324 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 325 if (m == NULL) 326 goto freeit; 327 #ifdef MAC 328 mac_netinet_icmp_reply(n, m); 329 #endif 330 icmplen = min(icmplen, M_TRAILINGSPACE(m) - 331 sizeof(struct ip) - ICMP_MINLEN); 332 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen); 333 m->m_data += sizeof(struct ip); 334 m->m_len = ICMP_MINLEN + icmplen; 335 336 /* XXX MRT make the outgoing packet use the same FIB 337 * that was associated with the incoming packet 338 */ 339 M_SETFIB(m, M_GETFIB(n)); 340 icp = mtod(m, struct icmp *); 341 ICMPSTAT_INC(icps_outhist[type]); 342 icp->icmp_type = type; 343 if (type == ICMP_REDIRECT) 344 icp->icmp_gwaddr.s_addr = dest; 345 else { 346 icp->icmp_void = 0; 347 /* 348 * The following assignments assume an overlay with the 349 * just zeroed icmp_void field. 350 */ 351 if (type == ICMP_PARAMPROB) { 352 icp->icmp_pptr = code; 353 code = 0; 354 } else if (type == ICMP_UNREACH && 355 code == ICMP_UNREACH_NEEDFRAG && mtu) { 356 icp->icmp_nextmtu = htons(mtu); 357 } 358 } 359 icp->icmp_code = code; 360 361 /* 362 * Copy the quotation into ICMP message and 363 * convert quoted IP header back to network representation. 364 */ 365 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 366 nip = &icp->icmp_ip; 367 368 /* 369 * Set up ICMP message mbuf and copy old IP header (without options 370 * in front of ICMP message. 371 * If the original mbuf was meant to bypass the firewall, the error 372 * reply should bypass as well. 373 */ 374 m->m_flags |= n->m_flags & M_SKIP_FIREWALL; 375 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip), 376 ("insufficient space for ip header")); 377 m->m_data -= sizeof(struct ip); 378 m->m_len += sizeof(struct ip); 379 m->m_pkthdr.len = m->m_len; 380 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 381 nip = mtod(m, struct ip *); 382 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); 383 nip->ip_len = htons(m->m_len); 384 nip->ip_v = IPVERSION; 385 nip->ip_hl = 5; 386 nip->ip_p = IPPROTO_ICMP; 387 nip->ip_tos = 0; 388 nip->ip_off = 0; 389 390 if (V_error_keeptags) 391 m_tag_copy_chain(m, n, M_NOWAIT); 392 393 icmp_reflect(m); 394 395 freeit: 396 m_freem(n); 397 } 398 399 /* 400 * Process a received ICMP message. 401 */ 402 int 403 icmp_input(struct mbuf **mp, int *offp, int proto) 404 { 405 struct icmp *icp; 406 struct in_ifaddr *ia; 407 struct mbuf *m = *mp; 408 struct ip *ip = mtod(m, struct ip *); 409 struct sockaddr_in icmpsrc, icmpdst, icmpgw; 410 int hlen = *offp; 411 int icmplen = ntohs(ip->ip_len) - *offp; 412 int i, code; 413 void (*ctlfunc)(int, struct sockaddr *, void *); 414 int fibnum; 415 416 NET_EPOCH_ASSERT(); 417 418 *mp = NULL; 419 420 /* 421 * Locate icmp structure in mbuf, and check 422 * that not corrupted and of at least minimum length. 423 */ 424 #ifdef ICMPPRINTFS 425 if (icmpprintfs) { 426 char srcbuf[INET_ADDRSTRLEN]; 427 char dstbuf[INET_ADDRSTRLEN]; 428 429 printf("icmp_input from %s to %s, len %d\n", 430 inet_ntoa_r(ip->ip_src, srcbuf), 431 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen); 432 } 433 #endif 434 if (icmplen < ICMP_MINLEN) { 435 ICMPSTAT_INC(icps_tooshort); 436 goto freeit; 437 } 438 i = hlen + min(icmplen, ICMP_ADVLENMIN); 439 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 440 ICMPSTAT_INC(icps_tooshort); 441 return (IPPROTO_DONE); 442 } 443 ip = mtod(m, struct ip *); 444 m->m_len -= hlen; 445 m->m_data += hlen; 446 icp = mtod(m, struct icmp *); 447 if (in_cksum(m, icmplen)) { 448 ICMPSTAT_INC(icps_checksum); 449 goto freeit; 450 } 451 m->m_len += hlen; 452 m->m_data -= hlen; 453 454 #ifdef ICMPPRINTFS 455 if (icmpprintfs) 456 printf("icmp_input, type %d code %d\n", icp->icmp_type, 457 icp->icmp_code); 458 #endif 459 460 /* 461 * Message type specific processing. 462 */ 463 if (icp->icmp_type > ICMP_MAXTYPE) 464 goto raw; 465 466 /* Initialize */ 467 bzero(&icmpsrc, sizeof(icmpsrc)); 468 icmpsrc.sin_len = sizeof(struct sockaddr_in); 469 icmpsrc.sin_family = AF_INET; 470 bzero(&icmpdst, sizeof(icmpdst)); 471 icmpdst.sin_len = sizeof(struct sockaddr_in); 472 icmpdst.sin_family = AF_INET; 473 bzero(&icmpgw, sizeof(icmpgw)); 474 icmpgw.sin_len = sizeof(struct sockaddr_in); 475 icmpgw.sin_family = AF_INET; 476 477 ICMPSTAT_INC(icps_inhist[icp->icmp_type]); 478 code = icp->icmp_code; 479 switch (icp->icmp_type) { 480 481 case ICMP_UNREACH: 482 switch (code) { 483 case ICMP_UNREACH_NET: 484 case ICMP_UNREACH_HOST: 485 case ICMP_UNREACH_SRCFAIL: 486 case ICMP_UNREACH_NET_UNKNOWN: 487 case ICMP_UNREACH_HOST_UNKNOWN: 488 case ICMP_UNREACH_ISOLATED: 489 case ICMP_UNREACH_TOSNET: 490 case ICMP_UNREACH_TOSHOST: 491 case ICMP_UNREACH_HOST_PRECEDENCE: 492 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 493 code = PRC_UNREACH_NET; 494 break; 495 496 case ICMP_UNREACH_NEEDFRAG: 497 code = PRC_MSGSIZE; 498 break; 499 500 /* 501 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. 502 * Treat subcodes 2,3 as immediate RST 503 */ 504 case ICMP_UNREACH_PROTOCOL: 505 code = PRC_UNREACH_PROTOCOL; 506 break; 507 case ICMP_UNREACH_PORT: 508 code = PRC_UNREACH_PORT; 509 break; 510 511 case ICMP_UNREACH_NET_PROHIB: 512 case ICMP_UNREACH_HOST_PROHIB: 513 case ICMP_UNREACH_FILTER_PROHIB: 514 code = PRC_UNREACH_ADMIN_PROHIB; 515 break; 516 517 default: 518 goto badcode; 519 } 520 goto deliver; 521 522 case ICMP_TIMXCEED: 523 if (code > 1) 524 goto badcode; 525 code += PRC_TIMXCEED_INTRANS; 526 goto deliver; 527 528 case ICMP_PARAMPROB: 529 if (code > 1) 530 goto badcode; 531 code = PRC_PARAMPROB; 532 deliver: 533 /* 534 * Problem with datagram; advise higher level routines. 535 */ 536 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 537 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 538 ICMPSTAT_INC(icps_badlen); 539 goto freeit; 540 } 541 /* Discard ICMP's in response to multicast packets */ 542 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 543 goto badcode; 544 #ifdef ICMPPRINTFS 545 if (icmpprintfs) 546 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 547 #endif 548 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 549 /* 550 * XXX if the packet contains [IPv4 AH TCP], we can't make a 551 * notification to TCP layer. 552 */ 553 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp)); 554 ip_stripoptions(m); 555 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 556 /* This should actually not happen */ 557 ICMPSTAT_INC(icps_tooshort); 558 return (IPPROTO_DONE); 559 } 560 ip = mtod(m, struct ip *); 561 icp = (struct icmp *)(ip + 1); 562 /* 563 * The upper layer handler can rely on: 564 * - The outer IP header has no options. 565 * - The outer IP header, the ICMP header, the inner IP header, 566 * and the first n bytes of the inner payload are contiguous. 567 * n is at least 8, but might be larger based on 568 * ICMP_ADVLENPREF. See its definition in ip_icmp.h. 569 */ 570 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 571 if (ctlfunc) 572 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc, 573 (void *)&icp->icmp_ip); 574 break; 575 576 badcode: 577 ICMPSTAT_INC(icps_badcode); 578 break; 579 580 case ICMP_ECHO: 581 if (!V_icmpbmcastecho 582 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 583 ICMPSTAT_INC(icps_bmcastecho); 584 break; 585 } 586 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 587 goto freeit; 588 icp->icmp_type = ICMP_ECHOREPLY; 589 goto reflect; 590 591 case ICMP_TSTAMP: 592 if (V_icmptstamprepl == 0) 593 break; 594 if (!V_icmpbmcastecho 595 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 596 ICMPSTAT_INC(icps_bmcasttstamp); 597 break; 598 } 599 if (icmplen < ICMP_TSLEN) { 600 ICMPSTAT_INC(icps_badlen); 601 break; 602 } 603 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 604 goto freeit; 605 icp->icmp_type = ICMP_TSTAMPREPLY; 606 icp->icmp_rtime = iptime(); 607 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 608 goto reflect; 609 610 case ICMP_MASKREQ: 611 if (V_icmpmaskrepl == 0) 612 break; 613 /* 614 * We are not able to respond with all ones broadcast 615 * unless we receive it over a point-to-point interface. 616 */ 617 if (icmplen < ICMP_MASKLEN) 618 break; 619 switch (ip->ip_dst.s_addr) { 620 621 case INADDR_BROADCAST: 622 case INADDR_ANY: 623 icmpdst.sin_addr = ip->ip_src; 624 break; 625 626 default: 627 icmpdst.sin_addr = ip->ip_dst; 628 } 629 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 630 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 631 if (ia == NULL) 632 break; 633 if (ia->ia_ifp == NULL) 634 break; 635 icp->icmp_type = ICMP_MASKREPLY; 636 if (V_icmpmaskfake == 0) 637 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 638 else 639 icp->icmp_mask = V_icmpmaskfake; 640 if (ip->ip_src.s_addr == 0) { 641 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 642 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 643 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 644 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 645 } 646 reflect: 647 ICMPSTAT_INC(icps_reflect); 648 ICMPSTAT_INC(icps_outhist[icp->icmp_type]); 649 icmp_reflect(m); 650 return (IPPROTO_DONE); 651 652 case ICMP_REDIRECT: 653 if (V_log_redirect) { 654 u_long src, dst, gw; 655 656 src = ntohl(ip->ip_src.s_addr); 657 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 658 gw = ntohl(icp->icmp_gwaddr.s_addr); 659 printf("icmp redirect from %d.%d.%d.%d: " 660 "%d.%d.%d.%d => %d.%d.%d.%d\n", 661 (int)(src >> 24), (int)((src >> 16) & 0xff), 662 (int)((src >> 8) & 0xff), (int)(src & 0xff), 663 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 664 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 665 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 666 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 667 } 668 /* 669 * RFC1812 says we must ignore ICMP redirects if we 670 * are acting as router. 671 */ 672 if (V_drop_redirect || V_ipforwarding) 673 break; 674 if (code > 3) 675 goto badcode; 676 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 677 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 678 ICMPSTAT_INC(icps_badlen); 679 break; 680 } 681 /* 682 * Short circuit routing redirects to force 683 * immediate change in the kernel's routing 684 * tables. The message is also handed to anyone 685 * listening on a raw socket (e.g. the routing 686 * daemon for use in updating its tables). 687 */ 688 icmpgw.sin_addr = ip->ip_src; 689 icmpdst.sin_addr = icp->icmp_gwaddr; 690 #ifdef ICMPPRINTFS 691 if (icmpprintfs) { 692 char dstbuf[INET_ADDRSTRLEN]; 693 char gwbuf[INET_ADDRSTRLEN]; 694 695 printf("redirect dst %s to %s\n", 696 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf), 697 inet_ntoa_r(icp->icmp_gwaddr, gwbuf)); 698 } 699 #endif 700 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 701 702 /* 703 * RFC 1122 says network (code 0,2) redirects SHOULD 704 * be treated identically to the host redirects. 705 * Given that, ignore network masks. 706 */ 707 708 /* 709 * Variable values: 710 * icmpsrc: route destination 711 * icmpdst: route gateway 712 * icmpgw: message source 713 */ 714 715 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst, 716 M_GETFIB(m)) != 0) { 717 /* TODO: increment bad redirects here */ 718 break; 719 } 720 721 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 722 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc, 723 (struct sockaddr *)&icmpdst, 724 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif, 725 RTF_GATEWAY, V_redirtimeout); 726 } 727 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); 728 break; 729 730 /* 731 * No kernel processing for the following; 732 * just fall through to send to raw listener. 733 */ 734 case ICMP_ECHOREPLY: 735 case ICMP_ROUTERADVERT: 736 case ICMP_ROUTERSOLICIT: 737 case ICMP_TSTAMPREPLY: 738 case ICMP_IREQREPLY: 739 case ICMP_MASKREPLY: 740 case ICMP_SOURCEQUENCH: 741 default: 742 break; 743 } 744 745 raw: 746 *mp = m; 747 rip_input(mp, offp, proto); 748 return (IPPROTO_DONE); 749 750 freeit: 751 m_freem(m); 752 return (IPPROTO_DONE); 753 } 754 755 /* 756 * Reflect the ip packet back to the source 757 */ 758 static void 759 icmp_reflect(struct mbuf *m) 760 { 761 struct rm_priotracker in_ifa_tracker; 762 struct ip *ip = mtod(m, struct ip *); 763 struct ifaddr *ifa; 764 struct ifnet *ifp; 765 struct in_ifaddr *ia; 766 struct in_addr t; 767 struct nhop4_extended nh_ext; 768 struct mbuf *opts = NULL; 769 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 770 771 NET_EPOCH_ASSERT(); 772 773 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 774 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) || 775 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) { 776 m_freem(m); /* Bad return address */ 777 ICMPSTAT_INC(icps_badaddr); 778 goto done; /* Ip_output() will check for broadcast */ 779 } 780 781 t = ip->ip_dst; 782 ip->ip_dst = ip->ip_src; 783 784 /* 785 * Source selection for ICMP replies: 786 * 787 * If the incoming packet was addressed directly to one of our 788 * own addresses, use dst as the src for the reply. 789 */ 790 IN_IFADDR_RLOCK(&in_ifa_tracker); 791 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { 792 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { 793 t = IA_SIN(ia)->sin_addr; 794 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 795 goto match; 796 } 797 } 798 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 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 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) { 855 m_freem(m); 856 ICMPSTAT_INC(icps_noroute); 857 goto done; 858 } 859 t = nh_ext.nh_src; 860 match: 861 #ifdef MAC 862 mac_netinet_icmp_replyinplace(m); 863 #endif 864 ip->ip_src = t; 865 ip->ip_ttl = V_ip_defttl; 866 867 if (optlen > 0) { 868 u_char *cp; 869 int opt, cnt; 870 u_int len; 871 872 /* 873 * Retrieve any source routing from the incoming packet; 874 * add on any record-route or timestamp options. 875 */ 876 cp = (u_char *) (ip + 1); 877 if ((opts = ip_srcroute(m)) == NULL && 878 (opts = m_gethdr(M_NOWAIT, MT_DATA))) { 879 opts->m_len = sizeof(struct in_addr); 880 mtod(opts, struct in_addr *)->s_addr = 0; 881 } 882 if (opts) { 883 #ifdef ICMPPRINTFS 884 if (icmpprintfs) 885 printf("icmp_reflect optlen %d rt %d => ", 886 optlen, opts->m_len); 887 #endif 888 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 889 opt = cp[IPOPT_OPTVAL]; 890 if (opt == IPOPT_EOL) 891 break; 892 if (opt == IPOPT_NOP) 893 len = 1; 894 else { 895 if (cnt < IPOPT_OLEN + sizeof(*cp)) 896 break; 897 len = cp[IPOPT_OLEN]; 898 if (len < IPOPT_OLEN + sizeof(*cp) || 899 len > cnt) 900 break; 901 } 902 /* 903 * Should check for overflow, but it "can't happen" 904 */ 905 if (opt == IPOPT_RR || opt == IPOPT_TS || 906 opt == IPOPT_SECURITY) { 907 bcopy((caddr_t)cp, 908 mtod(opts, caddr_t) + opts->m_len, len); 909 opts->m_len += len; 910 } 911 } 912 /* Terminate & pad, if necessary */ 913 cnt = opts->m_len % 4; 914 if (cnt) { 915 for (; cnt < 4; cnt++) { 916 *(mtod(opts, caddr_t) + opts->m_len) = 917 IPOPT_EOL; 918 opts->m_len++; 919 } 920 } 921 #ifdef ICMPPRINTFS 922 if (icmpprintfs) 923 printf("%d\n", opts->m_len); 924 #endif 925 } 926 ip_stripoptions(m); 927 } 928 m_tag_delete_nonpersistent(m); 929 m->m_flags &= ~(M_BCAST|M_MCAST); 930 icmp_send(m, opts); 931 done: 932 if (opts) 933 (void)m_free(opts); 934 } 935 936 /* 937 * Verifies if redirect message is valid, according to RFC 1122 938 * 939 * @src: sockaddr with address of redirect originator 940 * @dst: sockaddr with destination in question 941 * @gateway: new proposed gateway 942 * 943 * Returns 0 on success. 944 */ 945 static int 946 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst, 947 struct sockaddr_in *gateway, u_int fibnum) 948 { 949 struct nhop_object *nh; 950 struct ifaddr *ifa; 951 952 NET_EPOCH_ASSERT(); 953 954 /* Verify the gateway is directly reachable. */ 955 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL) 956 return (ENETUNREACH); 957 958 /* TODO: fib-aware. */ 959 if (ifa_ifwithaddr_check((struct sockaddr *)gateway)) 960 return (EHOSTUNREACH); 961 962 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0); 963 if (nh == NULL) 964 return (EINVAL); 965 966 /* 967 * If the redirect isn't from our current router for this dst, 968 * it's either old or wrong. If it redirects us to ourselves, 969 * we have a routing loop, perhaps as a result of an interface 970 * going down recently. 971 */ 972 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa)) 973 return (EINVAL); 974 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) 975 return (EINVAL); 976 977 /* If host route already exists, ignore redirect. */ 978 if (nh->nh_flags & NHF_HOST) 979 return (EEXIST); 980 981 /* If the prefix is directly reachable, ignore redirect. */ 982 if (!(nh->nh_flags & NHF_GATEWAY)) 983 return (EEXIST); 984 985 return (0); 986 } 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 /* 1068 * badport_bandlim() - check for ICMP bandwidth limit 1069 * 1070 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1071 * hit our bandwidth limit and it is not ok. 1072 * 1073 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1074 * 1075 * For now we separate the TCP and UDP subsystems w/ different 'which' 1076 * values. We may eventually remove this separation (and simplify the 1077 * code further). 1078 * 1079 * Note that the printing of the error message is delayed so we can 1080 * properly print the icmp error rate that the system was trying to do 1081 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1082 * the 'final' error, but it doesn't make sense to solve the printing 1083 * delay with more complex code. 1084 */ 1085 struct icmp_rate { 1086 const char *descr; 1087 struct counter_rate cr; 1088 }; 1089 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = { 1090 { "icmp unreach response" }, 1091 { "icmp ping response" }, 1092 { "icmp tstamp response" }, 1093 { "closed port RST response" }, 1094 { "open port RST response" }, 1095 { "icmp6 unreach response" }, 1096 { "sctp ootb response" } 1097 }; 1098 #define V_icmp_rates VNET(icmp_rates) 1099 1100 static void 1101 icmp_bandlimit_init(void) 1102 { 1103 1104 for (int i = 0; i < BANDLIM_MAX; i++) { 1105 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK); 1106 V_icmp_rates[i].cr.cr_ticks = ticks; 1107 } 1108 } 1109 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, 1110 icmp_bandlimit_init, NULL); 1111 1112 static void 1113 icmp_bandlimit_uninit(void) 1114 { 1115 1116 for (int i = 0; i < BANDLIM_MAX; i++) 1117 counter_u64_free(V_icmp_rates[i].cr.cr_rate); 1118 } 1119 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1120 icmp_bandlimit_uninit, NULL); 1121 1122 int 1123 badport_bandlim(int which) 1124 { 1125 int64_t pps; 1126 1127 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED) 1128 return (0); 1129 1130 KASSERT(which >= 0 && which < BANDLIM_MAX, 1131 ("%s: which %d", __func__, which)); 1132 1133 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim); 1134 if (pps == -1) 1135 return (-1); 1136 if (pps > 0 && V_icmplim_output) 1137 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n", 1138 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim); 1139 return (0); 1140 } 1141