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/vnet.h> 56 57 #include <netinet/in.h> 58 #include <netinet/in_fib.h> 59 #include <netinet/in_pcb.h> 60 #include <netinet/in_systm.h> 61 #include <netinet/in_var.h> 62 #include <netinet/ip.h> 63 #include <netinet/ip_icmp.h> 64 #include <netinet/ip_var.h> 65 #include <netinet/ip_options.h> 66 #include <netinet/sctp.h> 67 #include <netinet/tcp.h> 68 #include <netinet/tcp_var.h> 69 #include <netinet/tcpip.h> 70 #include <netinet/icmp_var.h> 71 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_output) = 1; 92 #define V_icmplim_output VNET(icmplim_output) 93 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW, 94 &VNET_NAME(icmplim_output), 0, 95 "Enable logging of ICMP response rate limiting"); 96 97 #ifdef INET 98 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat); 99 VNET_PCPUSTAT_SYSINIT(icmpstat); 100 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat, 101 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)"); 102 103 #ifdef VIMAGE 104 VNET_PCPUSTAT_SYSUNINIT(icmpstat); 105 #endif /* VIMAGE */ 106 107 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0; 108 #define V_icmpmaskrepl VNET(icmpmaskrepl) 109 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW, 110 &VNET_NAME(icmpmaskrepl), 0, 111 "Reply to ICMP Address Mask Request packets"); 112 113 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0; 114 #define V_icmpmaskfake VNET(icmpmaskfake) 115 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW, 116 &VNET_NAME(icmpmaskfake), 0, 117 "Fake reply to ICMP Address Mask Request packets"); 118 119 VNET_DEFINE(int, drop_redirect) = 0; 120 #define V_drop_redirect VNET(drop_redirect) 121 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW, 122 &VNET_NAME(drop_redirect), 0, 123 "Ignore ICMP redirects"); 124 125 VNET_DEFINE_STATIC(int, log_redirect) = 0; 126 #define V_log_redirect VNET(log_redirect) 127 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW, 128 &VNET_NAME(log_redirect), 0, 129 "Log ICMP redirects to the console"); 130 131 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]); 132 #define V_reply_src VNET(reply_src) 133 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW, 134 &VNET_NAME(reply_src), IFNAMSIZ, 135 "ICMP reply source for non-local packets"); 136 137 VNET_DEFINE_STATIC(int, icmp_rfi) = 0; 138 #define V_icmp_rfi VNET(icmp_rfi) 139 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW, 140 &VNET_NAME(icmp_rfi), 0, 141 "ICMP reply from incoming interface for non-local packets"); 142 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */ 143 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548; 144 #define V_icmp_quotelen VNET(icmp_quotelen) 145 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW, 146 &VNET_NAME(icmp_quotelen), 0, 147 "Number of bytes from original packet to quote in ICMP reply"); 148 149 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0; 150 #define V_icmpbmcastecho VNET(icmpbmcastecho) 151 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW, 152 &VNET_NAME(icmpbmcastecho), 0, 153 "Reply to multicast ICMP Echo Request and Timestamp packets"); 154 155 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1; 156 #define V_icmptstamprepl VNET(icmptstamprepl) 157 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_RW, 158 &VNET_NAME(icmptstamprepl), 0, 159 "Respond to ICMP Timestamp packets"); 160 161 VNET_DEFINE_STATIC(int, error_keeptags) = 0; 162 #define V_error_keeptags VNET(error_keeptags) 163 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW, 164 &VNET_NAME(error_keeptags), 0, 165 "ICMP error response keeps copy of mbuf_tags of original packet"); 166 167 #ifdef ICMPPRINTFS 168 int icmpprintfs = 0; 169 #endif 170 171 static void icmp_reflect(struct mbuf *); 172 static void icmp_send(struct mbuf *, struct mbuf *); 173 174 extern struct protosw inetsw[]; 175 176 /* 177 * Kernel module interface for updating icmpstat. The argument is an index 178 * into icmpstat treated as an array of u_long. While this encodes the 179 * general layout of icmpstat into the caller, it doesn't encode its 180 * location, so that future changes to add, for example, per-CPU stats 181 * support won't cause binary compatibility problems for kernel modules. 182 */ 183 void 184 kmod_icmpstat_inc(int statnum) 185 { 186 187 counter_u64_add(VNET(icmpstat)[statnum], 1); 188 } 189 190 /* 191 * Generate an error packet of type error 192 * in response to bad packet ip. 193 */ 194 void 195 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu) 196 { 197 struct ip *oip, *nip; 198 struct icmp *icp; 199 struct mbuf *m; 200 unsigned icmplen, icmpelen, nlen, oiphlen; 201 202 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", 203 __func__)); 204 205 if (type != ICMP_REDIRECT) 206 ICMPSTAT_INC(icps_error); 207 /* 208 * Don't send error: 209 * if the original packet was encrypted. 210 * if not the first fragment of message. 211 * in response to a multicast or broadcast packet. 212 * if the old packet protocol was an ICMP error message. 213 */ 214 if (n->m_flags & M_DECRYPTED) 215 goto freeit; 216 if (n->m_flags & (M_BCAST|M_MCAST)) 217 goto freeit; 218 219 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */ 220 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) 221 goto freeit; 222 oip = mtod(n, struct ip *); 223 oiphlen = oip->ip_hl << 2; 224 if (n->m_len < oiphlen + ICMP_MINLEN) 225 goto freeit; 226 #ifdef ICMPPRINTFS 227 if (icmpprintfs) 228 printf("icmp_error(%p, %x, %d)\n", oip, type, code); 229 #endif 230 if (oip->ip_off & htons(~(IP_MF|IP_DF))) 231 goto freeit; 232 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 233 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + 234 oiphlen))->icmp_type)) { 235 ICMPSTAT_INC(icps_oldicmp); 236 goto freeit; 237 } 238 /* 239 * Calculate length to quote from original packet and 240 * prevent the ICMP mbuf from overflowing. 241 * Unfortunately this is non-trivial since ip_forward() 242 * sends us truncated packets. 243 */ 244 nlen = m_length(n, NULL); 245 if (oip->ip_p == IPPROTO_TCP) { 246 struct tcphdr *th; 247 int tcphlen; 248 249 if (oiphlen + sizeof(struct tcphdr) > n->m_len && 250 n->m_next == NULL) 251 goto stdreply; 252 if (n->m_len < oiphlen + sizeof(struct tcphdr) && 253 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL) 254 goto freeit; 255 oip = mtod(n, struct ip *); 256 th = mtodo(n, oiphlen); 257 tcphlen = th->th_off << 2; 258 if (tcphlen < sizeof(struct tcphdr)) 259 goto freeit; 260 if (ntohs(oip->ip_len) < oiphlen + tcphlen) 261 goto freeit; 262 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL) 263 goto stdreply; 264 if (n->m_len < oiphlen + tcphlen && 265 (n = m_pullup(n, oiphlen + tcphlen)) == NULL) 266 goto freeit; 267 oip = mtod(n, struct ip *); 268 icmpelen = max(tcphlen, min(V_icmp_quotelen, 269 ntohs(oip->ip_len) - oiphlen)); 270 } else if (oip->ip_p == IPPROTO_SCTP) { 271 struct sctphdr *sh; 272 struct sctp_chunkhdr *ch; 273 274 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr)) 275 goto stdreply; 276 if (oiphlen + sizeof(struct sctphdr) > n->m_len && 277 n->m_next == NULL) 278 goto stdreply; 279 if (n->m_len < oiphlen + sizeof(struct sctphdr) && 280 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL) 281 goto freeit; 282 oip = mtod(n, struct ip *); 283 icmpelen = max(sizeof(struct sctphdr), 284 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen)); 285 sh = mtodo(n, oiphlen); 286 if (ntohl(sh->v_tag) == 0 && 287 ntohs(oip->ip_len) >= oiphlen + 288 sizeof(struct sctphdr) + 8 && 289 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 || 290 n->m_next != NULL)) { 291 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 && 292 (n = m_pullup(n, oiphlen + 293 sizeof(struct sctphdr) + 8)) == NULL) 294 goto freeit; 295 oip = mtod(n, struct ip *); 296 sh = mtodo(n, oiphlen); 297 ch = (struct sctp_chunkhdr *)(sh + 1); 298 if (ch->chunk_type == SCTP_INITIATION) { 299 icmpelen = max(sizeof(struct sctphdr) + 8, 300 min(V_icmp_quotelen, ntohs(oip->ip_len) - 301 oiphlen)); 302 } 303 } 304 } else 305 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - 306 oiphlen)); 307 308 icmplen = min(oiphlen + icmpelen, nlen); 309 if (icmplen < sizeof(struct ip)) 310 goto freeit; 311 312 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen) 313 m = m_gethdr(M_NOWAIT, MT_DATA); 314 else 315 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 316 if (m == NULL) 317 goto freeit; 318 #ifdef MAC 319 mac_netinet_icmp_reply(n, m); 320 #endif 321 icmplen = min(icmplen, M_TRAILINGSPACE(m) - 322 sizeof(struct ip) - ICMP_MINLEN); 323 m_align(m, ICMP_MINLEN + icmplen); 324 m->m_len = ICMP_MINLEN + icmplen; 325 326 /* XXX MRT make the outgoing packet use the same FIB 327 * that was associated with the incoming packet 328 */ 329 M_SETFIB(m, M_GETFIB(n)); 330 icp = mtod(m, struct icmp *); 331 ICMPSTAT_INC(icps_outhist[type]); 332 icp->icmp_type = type; 333 if (type == ICMP_REDIRECT) 334 icp->icmp_gwaddr.s_addr = dest; 335 else { 336 icp->icmp_void = 0; 337 /* 338 * The following assignments assume an overlay with the 339 * just zeroed icmp_void field. 340 */ 341 if (type == ICMP_PARAMPROB) { 342 icp->icmp_pptr = code; 343 code = 0; 344 } else if (type == ICMP_UNREACH && 345 code == ICMP_UNREACH_NEEDFRAG && mtu) { 346 icp->icmp_nextmtu = htons(mtu); 347 } 348 } 349 icp->icmp_code = code; 350 351 /* 352 * Copy the quotation into ICMP message and 353 * convert quoted IP header back to network representation. 354 */ 355 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 356 nip = &icp->icmp_ip; 357 358 /* 359 * Set up ICMP message mbuf and copy old IP header (without options 360 * in front of ICMP message. 361 * If the original mbuf was meant to bypass the firewall, the error 362 * reply should bypass as well. 363 */ 364 m->m_flags |= n->m_flags & M_SKIP_FIREWALL; 365 m->m_data -= sizeof(struct ip); 366 m->m_len += sizeof(struct ip); 367 m->m_pkthdr.len = m->m_len; 368 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 369 nip = mtod(m, struct ip *); 370 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); 371 nip->ip_len = htons(m->m_len); 372 nip->ip_v = IPVERSION; 373 nip->ip_hl = 5; 374 nip->ip_p = IPPROTO_ICMP; 375 nip->ip_tos = 0; 376 nip->ip_off = 0; 377 378 if (V_error_keeptags) 379 m_tag_copy_chain(m, n, M_NOWAIT); 380 381 icmp_reflect(m); 382 383 freeit: 384 m_freem(n); 385 } 386 387 /* 388 * Process a received ICMP message. 389 */ 390 int 391 icmp_input(struct mbuf **mp, int *offp, int proto) 392 { 393 struct icmp *icp; 394 struct in_ifaddr *ia; 395 struct mbuf *m = *mp; 396 struct ip *ip = mtod(m, struct ip *); 397 struct sockaddr_in icmpsrc, icmpdst, icmpgw; 398 int hlen = *offp; 399 int icmplen = ntohs(ip->ip_len) - *offp; 400 int i, code; 401 void (*ctlfunc)(int, struct sockaddr *, void *); 402 int fibnum; 403 404 *mp = NULL; 405 406 /* 407 * Locate icmp structure in mbuf, and check 408 * that not corrupted and of at least minimum length. 409 */ 410 #ifdef ICMPPRINTFS 411 if (icmpprintfs) { 412 char srcbuf[INET_ADDRSTRLEN]; 413 char dstbuf[INET_ADDRSTRLEN]; 414 415 printf("icmp_input from %s to %s, len %d\n", 416 inet_ntoa_r(ip->ip_src, srcbuf), 417 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen); 418 } 419 #endif 420 NET_EPOCH_ENTER(); 421 if (icmplen < ICMP_MINLEN) { 422 ICMPSTAT_INC(icps_tooshort); 423 goto freeit; 424 } 425 i = hlen + min(icmplen, ICMP_ADVLENMIN); 426 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 427 ICMPSTAT_INC(icps_tooshort); 428 NET_EPOCH_EXIT(); 429 return (IPPROTO_DONE); 430 } 431 ip = mtod(m, struct ip *); 432 m->m_len -= hlen; 433 m->m_data += hlen; 434 icp = mtod(m, struct icmp *); 435 if (in_cksum(m, icmplen)) { 436 ICMPSTAT_INC(icps_checksum); 437 goto freeit; 438 } 439 m->m_len += hlen; 440 m->m_data -= hlen; 441 442 #ifdef ICMPPRINTFS 443 if (icmpprintfs) 444 printf("icmp_input, type %d code %d\n", icp->icmp_type, 445 icp->icmp_code); 446 #endif 447 448 /* 449 * Message type specific processing. 450 */ 451 if (icp->icmp_type > ICMP_MAXTYPE) 452 goto raw; 453 454 /* Initialize */ 455 bzero(&icmpsrc, sizeof(icmpsrc)); 456 icmpsrc.sin_len = sizeof(struct sockaddr_in); 457 icmpsrc.sin_family = AF_INET; 458 bzero(&icmpdst, sizeof(icmpdst)); 459 icmpdst.sin_len = sizeof(struct sockaddr_in); 460 icmpdst.sin_family = AF_INET; 461 bzero(&icmpgw, sizeof(icmpgw)); 462 icmpgw.sin_len = sizeof(struct sockaddr_in); 463 icmpgw.sin_family = AF_INET; 464 465 ICMPSTAT_INC(icps_inhist[icp->icmp_type]); 466 code = icp->icmp_code; 467 switch (icp->icmp_type) { 468 469 case ICMP_UNREACH: 470 switch (code) { 471 case ICMP_UNREACH_NET: 472 case ICMP_UNREACH_HOST: 473 case ICMP_UNREACH_SRCFAIL: 474 case ICMP_UNREACH_NET_UNKNOWN: 475 case ICMP_UNREACH_HOST_UNKNOWN: 476 case ICMP_UNREACH_ISOLATED: 477 case ICMP_UNREACH_TOSNET: 478 case ICMP_UNREACH_TOSHOST: 479 case ICMP_UNREACH_HOST_PRECEDENCE: 480 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 481 code = PRC_UNREACH_NET; 482 break; 483 484 case ICMP_UNREACH_NEEDFRAG: 485 code = PRC_MSGSIZE; 486 break; 487 488 /* 489 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. 490 * Treat subcodes 2,3 as immediate RST 491 */ 492 case ICMP_UNREACH_PROTOCOL: 493 code = PRC_UNREACH_PROTOCOL; 494 break; 495 case ICMP_UNREACH_PORT: 496 code = PRC_UNREACH_PORT; 497 break; 498 499 case ICMP_UNREACH_NET_PROHIB: 500 case ICMP_UNREACH_HOST_PROHIB: 501 case ICMP_UNREACH_FILTER_PROHIB: 502 code = PRC_UNREACH_ADMIN_PROHIB; 503 break; 504 505 default: 506 goto badcode; 507 } 508 goto deliver; 509 510 case ICMP_TIMXCEED: 511 if (code > 1) 512 goto badcode; 513 code += PRC_TIMXCEED_INTRANS; 514 goto deliver; 515 516 case ICMP_PARAMPROB: 517 if (code > 1) 518 goto badcode; 519 code = PRC_PARAMPROB; 520 deliver: 521 /* 522 * Problem with datagram; advise higher level routines. 523 */ 524 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 525 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 526 ICMPSTAT_INC(icps_badlen); 527 goto freeit; 528 } 529 /* Discard ICMP's in response to multicast packets */ 530 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 531 goto badcode; 532 #ifdef ICMPPRINTFS 533 if (icmpprintfs) 534 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 535 #endif 536 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 537 /* 538 * XXX if the packet contains [IPv4 AH TCP], we can't make a 539 * notification to TCP layer. 540 */ 541 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp)); 542 ip_stripoptions(m); 543 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 544 /* This should actually not happen */ 545 ICMPSTAT_INC(icps_tooshort); 546 NET_EPOCH_EXIT(); 547 return (IPPROTO_DONE); 548 } 549 ip = mtod(m, struct ip *); 550 icp = (struct icmp *)(ip + 1); 551 /* 552 * The upper layer handler can rely on: 553 * - The outer IP header has no options. 554 * - The outer IP header, the ICMP header, the inner IP header, 555 * and the first n bytes of the inner payload are contiguous. 556 * n is at least 8, but might be larger based on 557 * ICMP_ADVLENPREF. See its definition in ip_icmp.h. 558 */ 559 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 560 if (ctlfunc) 561 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc, 562 (void *)&icp->icmp_ip); 563 break; 564 565 badcode: 566 ICMPSTAT_INC(icps_badcode); 567 break; 568 569 case ICMP_ECHO: 570 if (!V_icmpbmcastecho 571 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 572 ICMPSTAT_INC(icps_bmcastecho); 573 break; 574 } 575 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 576 goto freeit; 577 icp->icmp_type = ICMP_ECHOREPLY; 578 goto reflect; 579 580 case ICMP_TSTAMP: 581 if (V_icmptstamprepl == 0) 582 break; 583 if (!V_icmpbmcastecho 584 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 585 ICMPSTAT_INC(icps_bmcasttstamp); 586 break; 587 } 588 if (icmplen < ICMP_TSLEN) { 589 ICMPSTAT_INC(icps_badlen); 590 break; 591 } 592 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 593 goto freeit; 594 icp->icmp_type = ICMP_TSTAMPREPLY; 595 icp->icmp_rtime = iptime(); 596 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 597 goto reflect; 598 599 case ICMP_MASKREQ: 600 if (V_icmpmaskrepl == 0) 601 break; 602 /* 603 * We are not able to respond with all ones broadcast 604 * unless we receive it over a point-to-point interface. 605 */ 606 if (icmplen < ICMP_MASKLEN) 607 break; 608 switch (ip->ip_dst.s_addr) { 609 610 case INADDR_BROADCAST: 611 case INADDR_ANY: 612 icmpdst.sin_addr = ip->ip_src; 613 break; 614 615 default: 616 icmpdst.sin_addr = ip->ip_dst; 617 } 618 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 619 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 620 if (ia == NULL) 621 break; 622 if (ia->ia_ifp == NULL) 623 break; 624 icp->icmp_type = ICMP_MASKREPLY; 625 if (V_icmpmaskfake == 0) 626 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 627 else 628 icp->icmp_mask = V_icmpmaskfake; 629 if (ip->ip_src.s_addr == 0) { 630 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 631 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 632 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 633 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 634 } 635 reflect: 636 ICMPSTAT_INC(icps_reflect); 637 ICMPSTAT_INC(icps_outhist[icp->icmp_type]); 638 icmp_reflect(m); 639 NET_EPOCH_EXIT(); 640 return (IPPROTO_DONE); 641 642 case ICMP_REDIRECT: 643 if (V_log_redirect) { 644 u_long src, dst, gw; 645 646 src = ntohl(ip->ip_src.s_addr); 647 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 648 gw = ntohl(icp->icmp_gwaddr.s_addr); 649 printf("icmp redirect from %d.%d.%d.%d: " 650 "%d.%d.%d.%d => %d.%d.%d.%d\n", 651 (int)(src >> 24), (int)((src >> 16) & 0xff), 652 (int)((src >> 8) & 0xff), (int)(src & 0xff), 653 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 654 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 655 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 656 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 657 } 658 /* 659 * RFC1812 says we must ignore ICMP redirects if we 660 * are acting as router. 661 */ 662 if (V_drop_redirect || V_ipforwarding) 663 break; 664 if (code > 3) 665 goto badcode; 666 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 667 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 668 ICMPSTAT_INC(icps_badlen); 669 break; 670 } 671 /* 672 * Short circuit routing redirects to force 673 * immediate change in the kernel's routing 674 * tables. The message is also handed to anyone 675 * listening on a raw socket (e.g. the routing 676 * daemon for use in updating its tables). 677 */ 678 icmpgw.sin_addr = ip->ip_src; 679 icmpdst.sin_addr = icp->icmp_gwaddr; 680 #ifdef ICMPPRINTFS 681 if (icmpprintfs) { 682 char dstbuf[INET_ADDRSTRLEN]; 683 char gwbuf[INET_ADDRSTRLEN]; 684 685 printf("redirect dst %s to %s\n", 686 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf), 687 inet_ntoa_r(icp->icmp_gwaddr, gwbuf)); 688 } 689 #endif 690 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 691 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 692 in_rtredirect((struct sockaddr *)&icmpsrc, 693 (struct sockaddr *)&icmpdst, 694 (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST, 695 (struct sockaddr *)&icmpgw, fibnum); 696 } 697 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); 698 break; 699 700 /* 701 * No kernel processing for the following; 702 * just fall through to send to raw listener. 703 */ 704 case ICMP_ECHOREPLY: 705 case ICMP_ROUTERADVERT: 706 case ICMP_ROUTERSOLICIT: 707 case ICMP_TSTAMPREPLY: 708 case ICMP_IREQREPLY: 709 case ICMP_MASKREPLY: 710 case ICMP_SOURCEQUENCH: 711 default: 712 break; 713 } 714 715 raw: 716 NET_EPOCH_EXIT(); 717 *mp = m; 718 rip_input(mp, offp, proto); 719 return (IPPROTO_DONE); 720 721 freeit: 722 NET_EPOCH_EXIT(); 723 m_freem(m); 724 return (IPPROTO_DONE); 725 } 726 727 /* 728 * Reflect the ip packet back to the source 729 */ 730 static void 731 icmp_reflect(struct mbuf *m) 732 { 733 struct rm_priotracker in_ifa_tracker; 734 struct ip *ip = mtod(m, struct ip *); 735 struct ifaddr *ifa; 736 struct ifnet *ifp; 737 struct in_ifaddr *ia; 738 struct in_addr t; 739 struct nhop4_extended nh_ext; 740 struct mbuf *opts = NULL; 741 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 742 743 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 744 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) || 745 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) { 746 m_freem(m); /* Bad return address */ 747 ICMPSTAT_INC(icps_badaddr); 748 goto done; /* Ip_output() will check for broadcast */ 749 } 750 751 t = ip->ip_dst; 752 ip->ip_dst = ip->ip_src; 753 754 /* 755 * Source selection for ICMP replies: 756 * 757 * If the incoming packet was addressed directly to one of our 758 * own addresses, use dst as the src for the reply. 759 */ 760 IN_IFADDR_RLOCK(&in_ifa_tracker); 761 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { 762 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { 763 t = IA_SIN(ia)->sin_addr; 764 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 765 goto match; 766 } 767 } 768 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 769 770 /* 771 * If the incoming packet was addressed to one of our broadcast 772 * addresses, use the first non-broadcast address which corresponds 773 * to the incoming interface. 774 */ 775 ifp = m->m_pkthdr.rcvif; 776 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 777 IF_ADDR_RLOCK(ifp); 778 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 779 if (ifa->ifa_addr->sa_family != AF_INET) 780 continue; 781 ia = ifatoia(ifa); 782 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 783 t.s_addr) { 784 t = IA_SIN(ia)->sin_addr; 785 IF_ADDR_RUNLOCK(ifp); 786 goto match; 787 } 788 } 789 IF_ADDR_RUNLOCK(ifp); 790 } 791 /* 792 * If the packet was transiting through us, use the address of 793 * the interface the packet came through in. If that interface 794 * doesn't have a suitable IP address, the normal selection 795 * criteria apply. 796 */ 797 if (V_icmp_rfi && ifp != NULL) { 798 IF_ADDR_RLOCK(ifp); 799 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 800 if (ifa->ifa_addr->sa_family != AF_INET) 801 continue; 802 ia = ifatoia(ifa); 803 t = IA_SIN(ia)->sin_addr; 804 IF_ADDR_RUNLOCK(ifp); 805 goto match; 806 } 807 IF_ADDR_RUNLOCK(ifp); 808 } 809 /* 810 * If the incoming packet was not addressed directly to us, use 811 * designated interface for icmp replies specified by sysctl 812 * net.inet.icmp.reply_src (default not set). Otherwise continue 813 * with normal source selection. 814 */ 815 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) { 816 IF_ADDR_RLOCK(ifp); 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 t = IA_SIN(ia)->sin_addr; 822 IF_ADDR_RUNLOCK(ifp); 823 goto match; 824 } 825 IF_ADDR_RUNLOCK(ifp); 826 } 827 /* 828 * If the packet was transiting through us, use the address of 829 * the interface that is the closest to the packet source. 830 * When we don't have a route back to the packet source, stop here 831 * and drop the packet. 832 */ 833 if (fib4_lookup_nh_ext(M_GETFIB(m), ip->ip_dst, 0, 0, &nh_ext) != 0) { 834 m_freem(m); 835 ICMPSTAT_INC(icps_noroute); 836 goto done; 837 } 838 t = nh_ext.nh_src; 839 match: 840 #ifdef MAC 841 mac_netinet_icmp_replyinplace(m); 842 #endif 843 ip->ip_src = t; 844 ip->ip_ttl = V_ip_defttl; 845 846 if (optlen > 0) { 847 u_char *cp; 848 int opt, cnt; 849 u_int len; 850 851 /* 852 * Retrieve any source routing from the incoming packet; 853 * add on any record-route or timestamp options. 854 */ 855 cp = (u_char *) (ip + 1); 856 if ((opts = ip_srcroute(m)) == NULL && 857 (opts = m_gethdr(M_NOWAIT, MT_DATA))) { 858 opts->m_len = sizeof(struct in_addr); 859 mtod(opts, struct in_addr *)->s_addr = 0; 860 } 861 if (opts) { 862 #ifdef ICMPPRINTFS 863 if (icmpprintfs) 864 printf("icmp_reflect optlen %d rt %d => ", 865 optlen, opts->m_len); 866 #endif 867 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 868 opt = cp[IPOPT_OPTVAL]; 869 if (opt == IPOPT_EOL) 870 break; 871 if (opt == IPOPT_NOP) 872 len = 1; 873 else { 874 if (cnt < IPOPT_OLEN + sizeof(*cp)) 875 break; 876 len = cp[IPOPT_OLEN]; 877 if (len < IPOPT_OLEN + sizeof(*cp) || 878 len > cnt) 879 break; 880 } 881 /* 882 * Should check for overflow, but it "can't happen" 883 */ 884 if (opt == IPOPT_RR || opt == IPOPT_TS || 885 opt == IPOPT_SECURITY) { 886 bcopy((caddr_t)cp, 887 mtod(opts, caddr_t) + opts->m_len, len); 888 opts->m_len += len; 889 } 890 } 891 /* Terminate & pad, if necessary */ 892 cnt = opts->m_len % 4; 893 if (cnt) { 894 for (; cnt < 4; cnt++) { 895 *(mtod(opts, caddr_t) + opts->m_len) = 896 IPOPT_EOL; 897 opts->m_len++; 898 } 899 } 900 #ifdef ICMPPRINTFS 901 if (icmpprintfs) 902 printf("%d\n", opts->m_len); 903 #endif 904 } 905 ip_stripoptions(m); 906 } 907 m_tag_delete_nonpersistent(m); 908 m->m_flags &= ~(M_BCAST|M_MCAST); 909 icmp_send(m, opts); 910 done: 911 if (opts) 912 (void)m_free(opts); 913 } 914 915 /* 916 * Send an icmp packet back to the ip level, 917 * after supplying a checksum. 918 */ 919 static void 920 icmp_send(struct mbuf *m, struct mbuf *opts) 921 { 922 struct ip *ip = mtod(m, struct ip *); 923 int hlen; 924 struct icmp *icp; 925 926 hlen = ip->ip_hl << 2; 927 m->m_data += hlen; 928 m->m_len -= hlen; 929 icp = mtod(m, struct icmp *); 930 icp->icmp_cksum = 0; 931 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 932 m->m_data -= hlen; 933 m->m_len += hlen; 934 m->m_pkthdr.rcvif = (struct ifnet *)0; 935 #ifdef ICMPPRINTFS 936 if (icmpprintfs) { 937 char dstbuf[INET_ADDRSTRLEN]; 938 char srcbuf[INET_ADDRSTRLEN]; 939 940 printf("icmp_send dst %s src %s\n", 941 inet_ntoa_r(ip->ip_dst, dstbuf), 942 inet_ntoa_r(ip->ip_src, srcbuf)); 943 } 944 #endif 945 (void) ip_output(m, opts, NULL, 0, NULL, NULL); 946 } 947 948 /* 949 * Return milliseconds since 00:00 UTC in network format. 950 */ 951 uint32_t 952 iptime(void) 953 { 954 struct timeval atv; 955 u_long t; 956 957 getmicrotime(&atv); 958 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 959 return (htonl(t)); 960 } 961 962 /* 963 * Return the next larger or smaller MTU plateau (table from RFC 1191) 964 * given current value MTU. If DIR is less than zero, a larger plateau 965 * is returned; otherwise, a smaller value is returned. 966 */ 967 int 968 ip_next_mtu(int mtu, int dir) 969 { 970 static int mtutab[] = { 971 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 972 296, 68, 0 973 }; 974 int i, size; 975 976 size = (sizeof mtutab) / (sizeof mtutab[0]); 977 if (dir >= 0) { 978 for (i = 0; i < size; i++) 979 if (mtu > mtutab[i]) 980 return mtutab[i]; 981 } else { 982 for (i = size - 1; i >= 0; i--) 983 if (mtu < mtutab[i]) 984 return mtutab[i]; 985 if (mtu == mtutab[0]) 986 return mtutab[0]; 987 } 988 return 0; 989 } 990 #endif /* INET */ 991 992 993 /* 994 * badport_bandlim() - check for ICMP bandwidth limit 995 * 996 * Return 0 if it is ok to send an ICMP error response, -1 if we have 997 * hit our bandwidth limit and it is not ok. 998 * 999 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1000 * 1001 * For now we separate the TCP and UDP subsystems w/ different 'which' 1002 * values. We may eventually remove this separation (and simplify the 1003 * code further). 1004 * 1005 * Note that the printing of the error message is delayed so we can 1006 * properly print the icmp error rate that the system was trying to do 1007 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1008 * the 'final' error, but it doesn't make sense to solve the printing 1009 * delay with more complex code. 1010 */ 1011 struct icmp_rate { 1012 const char *descr; 1013 struct counter_rate cr; 1014 }; 1015 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = { 1016 { "icmp unreach response" }, 1017 { "icmp ping response" }, 1018 { "icmp tstamp response" }, 1019 { "closed port RST response" }, 1020 { "open port RST response" }, 1021 { "icmp6 unreach response" }, 1022 { "sctp ootb response" } 1023 }; 1024 #define V_icmp_rates VNET(icmp_rates) 1025 1026 static void 1027 icmp_bandlimit_init(void) 1028 { 1029 1030 for (int i = 0; i < BANDLIM_MAX; i++) { 1031 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK); 1032 V_icmp_rates[i].cr.cr_ticks = ticks; 1033 } 1034 } 1035 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, 1036 icmp_bandlimit_init, NULL); 1037 1038 static void 1039 icmp_bandlimit_uninit(void) 1040 { 1041 1042 for (int i = 0; i < BANDLIM_MAX; i++) 1043 counter_u64_free(V_icmp_rates[i].cr.cr_rate); 1044 } 1045 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1046 icmp_bandlimit_uninit, NULL); 1047 1048 int 1049 badport_bandlim(int which) 1050 { 1051 int64_t pps; 1052 1053 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED) 1054 return (0); 1055 1056 KASSERT(which >= 0 && which < BANDLIM_MAX, 1057 ("%s: which %d", __func__, which)); 1058 1059 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim); 1060 if (pps == -1) 1061 return (-1); 1062 if (pps > 0 && V_icmplim_output) 1063 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n", 1064 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim); 1065 return (0); 1066 } 1067