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/route_ctl.h> 56 #include <net/route/nhop.h> 57 #include <net/vnet.h> 58 59 #include <netinet/in.h> 60 #include <netinet/in_fib.h> 61 #include <netinet/in_pcb.h> 62 #include <netinet/in_systm.h> 63 #include <netinet/in_var.h> 64 #include <netinet/ip.h> 65 #include <netinet/ip_icmp.h> 66 #include <netinet/ip_var.h> 67 #include <netinet/ip_options.h> 68 #include <netinet/sctp.h> 69 #include <netinet/tcp.h> 70 #include <netinet/tcp_var.h> 71 #include <netinet/tcpip.h> 72 #include <netinet/icmp_var.h> 73 74 75 #ifdef INET 76 77 #include <machine/in_cksum.h> 78 79 #include <security/mac/mac_framework.h> 80 #endif /* INET */ 81 82 /* 83 * ICMP routines: error generation, receive packet processing, and 84 * routines to turnaround packets back to the originator, and 85 * host table maintenance routines. 86 */ 87 VNET_DEFINE_STATIC(int, icmplim) = 200; 88 #define V_icmplim VNET(icmplim) 89 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW, 90 &VNET_NAME(icmplim), 0, 91 "Maximum number of ICMP responses per second"); 92 93 VNET_DEFINE_STATIC(int, icmplim_output) = 1; 94 #define V_icmplim_output VNET(icmplim_output) 95 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW, 96 &VNET_NAME(icmplim_output), 0, 97 "Enable logging of ICMP response rate limiting"); 98 99 #ifdef INET 100 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat); 101 VNET_PCPUSTAT_SYSINIT(icmpstat); 102 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat, 103 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)"); 104 105 #ifdef VIMAGE 106 VNET_PCPUSTAT_SYSUNINIT(icmpstat); 107 #endif /* VIMAGE */ 108 109 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0; 110 #define V_icmpmaskrepl VNET(icmpmaskrepl) 111 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW, 112 &VNET_NAME(icmpmaskrepl), 0, 113 "Reply to ICMP Address Mask Request packets"); 114 115 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0; 116 #define V_icmpmaskfake VNET(icmpmaskfake) 117 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW, 118 &VNET_NAME(icmpmaskfake), 0, 119 "Fake reply to ICMP Address Mask Request packets"); 120 121 VNET_DEFINE(int, drop_redirect) = 0; 122 #define V_drop_redirect VNET(drop_redirect) 123 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW, 124 &VNET_NAME(drop_redirect), 0, 125 "Ignore ICMP redirects"); 126 127 VNET_DEFINE_STATIC(int, log_redirect) = 0; 128 #define V_log_redirect VNET(log_redirect) 129 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW, 130 &VNET_NAME(log_redirect), 0, 131 "Log ICMP redirects to the console"); 132 133 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */ 134 #define V_redirtimeout VNET(redirtimeout) 135 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW, 136 &VNET_NAME(redirtimeout), 0, 137 "Delay in seconds before expiring redirect route"); 138 139 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]); 140 #define V_reply_src VNET(reply_src) 141 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW, 142 &VNET_NAME(reply_src), IFNAMSIZ, 143 "ICMP reply source for non-local packets"); 144 145 VNET_DEFINE_STATIC(int, icmp_rfi) = 0; 146 #define V_icmp_rfi VNET(icmp_rfi) 147 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW, 148 &VNET_NAME(icmp_rfi), 0, 149 "ICMP reply from incoming interface for non-local packets"); 150 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */ 151 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548; 152 #define V_icmp_quotelen VNET(icmp_quotelen) 153 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW, 154 &VNET_NAME(icmp_quotelen), 0, 155 "Number of bytes from original packet to quote in ICMP reply"); 156 157 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0; 158 #define V_icmpbmcastecho VNET(icmpbmcastecho) 159 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW, 160 &VNET_NAME(icmpbmcastecho), 0, 161 "Reply to multicast ICMP Echo Request and Timestamp packets"); 162 163 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1; 164 #define V_icmptstamprepl VNET(icmptstamprepl) 165 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW, 166 &VNET_NAME(icmptstamprepl), 0, 167 "Respond to ICMP Timestamp packets"); 168 169 VNET_DEFINE_STATIC(int, error_keeptags) = 0; 170 #define V_error_keeptags VNET(error_keeptags) 171 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW, 172 &VNET_NAME(error_keeptags), 0, 173 "ICMP error response keeps copy of mbuf_tags of original packet"); 174 175 #ifdef ICMPPRINTFS 176 int icmpprintfs = 0; 177 #endif 178 179 static void icmp_reflect(struct mbuf *); 180 static void icmp_send(struct mbuf *, struct mbuf *); 181 static int icmp_verify_redirect_gateway(struct sockaddr_in *, 182 struct sockaddr_in *, struct sockaddr_in *, u_int); 183 184 extern struct protosw inetsw[]; 185 186 /* 187 * Kernel module interface for updating icmpstat. The argument is an index 188 * into icmpstat treated as an array of u_long. While this encodes the 189 * general layout of icmpstat into the caller, it doesn't encode its 190 * location, so that future changes to add, for example, per-CPU stats 191 * support won't cause binary compatibility problems for kernel modules. 192 */ 193 void 194 kmod_icmpstat_inc(int statnum) 195 { 196 197 counter_u64_add(VNET(icmpstat)[statnum], 1); 198 } 199 200 /* 201 * Generate an error packet of type error 202 * in response to bad packet ip. 203 */ 204 void 205 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu) 206 { 207 struct ip *oip, *nip; 208 struct icmp *icp; 209 struct mbuf *m; 210 unsigned icmplen, icmpelen, nlen, oiphlen; 211 212 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", 213 __func__)); 214 215 if (type != ICMP_REDIRECT) 216 ICMPSTAT_INC(icps_error); 217 /* 218 * Don't send error: 219 * if the original packet was encrypted. 220 * if not the first fragment of message. 221 * in response to a multicast or broadcast packet. 222 * if the old packet protocol was an ICMP error message. 223 */ 224 if (n->m_flags & M_DECRYPTED) 225 goto freeit; 226 if (n->m_flags & (M_BCAST|M_MCAST)) 227 goto freeit; 228 229 /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */ 230 if (n->m_len < sizeof(struct ip) + ICMP_MINLEN) 231 goto freeit; 232 oip = mtod(n, struct ip *); 233 oiphlen = oip->ip_hl << 2; 234 if (n->m_len < oiphlen + ICMP_MINLEN) 235 goto freeit; 236 #ifdef ICMPPRINTFS 237 if (icmpprintfs) 238 printf("icmp_error(%p, %x, %d)\n", oip, type, code); 239 #endif 240 if (oip->ip_off & htons(~(IP_MF|IP_DF))) 241 goto freeit; 242 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 243 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + 244 oiphlen))->icmp_type)) { 245 ICMPSTAT_INC(icps_oldicmp); 246 goto freeit; 247 } 248 /* 249 * Calculate length to quote from original packet and 250 * prevent the ICMP mbuf from overflowing. 251 * Unfortunately this is non-trivial since ip_forward() 252 * sends us truncated packets. 253 */ 254 nlen = m_length(n, NULL); 255 if (oip->ip_p == IPPROTO_TCP) { 256 struct tcphdr *th; 257 int tcphlen; 258 259 if (oiphlen + sizeof(struct tcphdr) > n->m_len && 260 n->m_next == NULL) 261 goto stdreply; 262 if (n->m_len < oiphlen + sizeof(struct tcphdr) && 263 (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL) 264 goto freeit; 265 oip = mtod(n, struct ip *); 266 th = mtodo(n, oiphlen); 267 tcphlen = th->th_off << 2; 268 if (tcphlen < sizeof(struct tcphdr)) 269 goto freeit; 270 if (ntohs(oip->ip_len) < oiphlen + tcphlen) 271 goto freeit; 272 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL) 273 goto stdreply; 274 if (n->m_len < oiphlen + tcphlen && 275 (n = m_pullup(n, oiphlen + tcphlen)) == NULL) 276 goto freeit; 277 oip = mtod(n, struct ip *); 278 icmpelen = max(tcphlen, min(V_icmp_quotelen, 279 ntohs(oip->ip_len) - oiphlen)); 280 } else if (oip->ip_p == IPPROTO_SCTP) { 281 struct sctphdr *sh; 282 struct sctp_chunkhdr *ch; 283 284 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr)) 285 goto stdreply; 286 if (oiphlen + sizeof(struct sctphdr) > n->m_len && 287 n->m_next == NULL) 288 goto stdreply; 289 if (n->m_len < oiphlen + sizeof(struct sctphdr) && 290 (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL) 291 goto freeit; 292 oip = mtod(n, struct ip *); 293 icmpelen = max(sizeof(struct sctphdr), 294 min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen)); 295 sh = mtodo(n, oiphlen); 296 if (ntohl(sh->v_tag) == 0 && 297 ntohs(oip->ip_len) >= oiphlen + 298 sizeof(struct sctphdr) + 8 && 299 (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 || 300 n->m_next != NULL)) { 301 if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 && 302 (n = m_pullup(n, oiphlen + 303 sizeof(struct sctphdr) + 8)) == NULL) 304 goto freeit; 305 oip = mtod(n, struct ip *); 306 sh = mtodo(n, oiphlen); 307 ch = (struct sctp_chunkhdr *)(sh + 1); 308 if (ch->chunk_type == SCTP_INITIATION) { 309 icmpelen = max(sizeof(struct sctphdr) + 8, 310 min(V_icmp_quotelen, ntohs(oip->ip_len) - 311 oiphlen)); 312 } 313 } 314 } else 315 stdreply: icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - 316 oiphlen)); 317 318 icmplen = min(oiphlen + icmpelen, nlen); 319 if (icmplen < sizeof(struct ip)) 320 goto freeit; 321 322 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen) 323 m = m_gethdr(M_NOWAIT, MT_DATA); 324 else 325 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 326 if (m == NULL) 327 goto freeit; 328 #ifdef MAC 329 mac_netinet_icmp_reply(n, m); 330 #endif 331 icmplen = min(icmplen, M_TRAILINGSPACE(m) - 332 sizeof(struct ip) - ICMP_MINLEN); 333 m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen); 334 m->m_data += sizeof(struct ip); 335 m->m_len = ICMP_MINLEN + icmplen; 336 337 /* XXX MRT make the outgoing packet use the same FIB 338 * that was associated with the incoming packet 339 */ 340 M_SETFIB(m, M_GETFIB(n)); 341 icp = mtod(m, struct icmp *); 342 ICMPSTAT_INC(icps_outhist[type]); 343 icp->icmp_type = type; 344 if (type == ICMP_REDIRECT) 345 icp->icmp_gwaddr.s_addr = dest; 346 else { 347 icp->icmp_void = 0; 348 /* 349 * The following assignments assume an overlay with the 350 * just zeroed icmp_void field. 351 */ 352 if (type == ICMP_PARAMPROB) { 353 icp->icmp_pptr = code; 354 code = 0; 355 } else if (type == ICMP_UNREACH && 356 code == ICMP_UNREACH_NEEDFRAG && mtu) { 357 icp->icmp_nextmtu = htons(mtu); 358 } 359 } 360 icp->icmp_code = code; 361 362 /* 363 * Copy the quotation into ICMP message and 364 * convert quoted IP header back to network representation. 365 */ 366 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 367 nip = &icp->icmp_ip; 368 369 /* 370 * Set up ICMP message mbuf and copy old IP header (without options 371 * in front of ICMP message. 372 * If the original mbuf was meant to bypass the firewall, the error 373 * reply should bypass as well. 374 */ 375 m->m_flags |= n->m_flags & M_SKIP_FIREWALL; 376 KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip), 377 ("insufficient space for ip header")); 378 m->m_data -= sizeof(struct ip); 379 m->m_len += sizeof(struct ip); 380 m->m_pkthdr.len = m->m_len; 381 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 382 nip = mtod(m, struct ip *); 383 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); 384 nip->ip_len = htons(m->m_len); 385 nip->ip_v = IPVERSION; 386 nip->ip_hl = 5; 387 nip->ip_p = IPPROTO_ICMP; 388 nip->ip_tos = 0; 389 nip->ip_off = 0; 390 391 if (V_error_keeptags) 392 m_tag_copy_chain(m, n, M_NOWAIT); 393 394 icmp_reflect(m); 395 396 freeit: 397 m_freem(n); 398 } 399 400 /* 401 * Process a received ICMP message. 402 */ 403 int 404 icmp_input(struct mbuf **mp, int *offp, int proto) 405 { 406 struct icmp *icp; 407 struct in_ifaddr *ia; 408 struct mbuf *m = *mp; 409 struct ip *ip = mtod(m, struct ip *); 410 struct sockaddr_in icmpsrc, icmpdst, icmpgw; 411 int hlen = *offp; 412 int icmplen = ntohs(ip->ip_len) - *offp; 413 int i, code; 414 void (*ctlfunc)(int, struct sockaddr *, void *); 415 int fibnum; 416 417 NET_EPOCH_ASSERT(); 418 419 *mp = NULL; 420 421 /* 422 * Locate icmp structure in mbuf, and check 423 * that not corrupted and of at least minimum length. 424 */ 425 #ifdef ICMPPRINTFS 426 if (icmpprintfs) { 427 char srcbuf[INET_ADDRSTRLEN]; 428 char dstbuf[INET_ADDRSTRLEN]; 429 430 printf("icmp_input from %s to %s, len %d\n", 431 inet_ntoa_r(ip->ip_src, srcbuf), 432 inet_ntoa_r(ip->ip_dst, dstbuf), icmplen); 433 } 434 #endif 435 if (icmplen < ICMP_MINLEN) { 436 ICMPSTAT_INC(icps_tooshort); 437 goto freeit; 438 } 439 i = hlen + min(icmplen, ICMP_ADVLENMIN); 440 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 441 ICMPSTAT_INC(icps_tooshort); 442 return (IPPROTO_DONE); 443 } 444 ip = mtod(m, struct ip *); 445 m->m_len -= hlen; 446 m->m_data += hlen; 447 icp = mtod(m, struct icmp *); 448 if (in_cksum(m, icmplen)) { 449 ICMPSTAT_INC(icps_checksum); 450 goto freeit; 451 } 452 m->m_len += hlen; 453 m->m_data -= hlen; 454 455 #ifdef ICMPPRINTFS 456 if (icmpprintfs) 457 printf("icmp_input, type %d code %d\n", icp->icmp_type, 458 icp->icmp_code); 459 #endif 460 461 /* 462 * Message type specific processing. 463 */ 464 if (icp->icmp_type > ICMP_MAXTYPE) 465 goto raw; 466 467 /* Initialize */ 468 bzero(&icmpsrc, sizeof(icmpsrc)); 469 icmpsrc.sin_len = sizeof(struct sockaddr_in); 470 icmpsrc.sin_family = AF_INET; 471 bzero(&icmpdst, sizeof(icmpdst)); 472 icmpdst.sin_len = sizeof(struct sockaddr_in); 473 icmpdst.sin_family = AF_INET; 474 bzero(&icmpgw, sizeof(icmpgw)); 475 icmpgw.sin_len = sizeof(struct sockaddr_in); 476 icmpgw.sin_family = AF_INET; 477 478 ICMPSTAT_INC(icps_inhist[icp->icmp_type]); 479 code = icp->icmp_code; 480 switch (icp->icmp_type) { 481 482 case ICMP_UNREACH: 483 switch (code) { 484 case ICMP_UNREACH_NET: 485 case ICMP_UNREACH_HOST: 486 case ICMP_UNREACH_SRCFAIL: 487 case ICMP_UNREACH_NET_UNKNOWN: 488 case ICMP_UNREACH_HOST_UNKNOWN: 489 case ICMP_UNREACH_ISOLATED: 490 case ICMP_UNREACH_TOSNET: 491 case ICMP_UNREACH_TOSHOST: 492 case ICMP_UNREACH_HOST_PRECEDENCE: 493 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 494 code = PRC_UNREACH_NET; 495 break; 496 497 case ICMP_UNREACH_NEEDFRAG: 498 code = PRC_MSGSIZE; 499 break; 500 501 /* 502 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. 503 * Treat subcodes 2,3 as immediate RST 504 */ 505 case ICMP_UNREACH_PROTOCOL: 506 code = PRC_UNREACH_PROTOCOL; 507 break; 508 case ICMP_UNREACH_PORT: 509 code = PRC_UNREACH_PORT; 510 break; 511 512 case ICMP_UNREACH_NET_PROHIB: 513 case ICMP_UNREACH_HOST_PROHIB: 514 case ICMP_UNREACH_FILTER_PROHIB: 515 code = PRC_UNREACH_ADMIN_PROHIB; 516 break; 517 518 default: 519 goto badcode; 520 } 521 goto deliver; 522 523 case ICMP_TIMXCEED: 524 if (code > 1) 525 goto badcode; 526 code += PRC_TIMXCEED_INTRANS; 527 goto deliver; 528 529 case ICMP_PARAMPROB: 530 if (code > 1) 531 goto badcode; 532 code = PRC_PARAMPROB; 533 deliver: 534 /* 535 * Problem with datagram; advise higher level routines. 536 */ 537 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 538 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 539 ICMPSTAT_INC(icps_badlen); 540 goto freeit; 541 } 542 /* Discard ICMP's in response to multicast packets */ 543 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 544 goto badcode; 545 #ifdef ICMPPRINTFS 546 if (icmpprintfs) 547 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 548 #endif 549 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 550 /* 551 * XXX if the packet contains [IPv4 AH TCP], we can't make a 552 * notification to TCP layer. 553 */ 554 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp)); 555 ip_stripoptions(m); 556 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 557 /* This should actually not happen */ 558 ICMPSTAT_INC(icps_tooshort); 559 return (IPPROTO_DONE); 560 } 561 ip = mtod(m, struct ip *); 562 icp = (struct icmp *)(ip + 1); 563 /* 564 * The upper layer handler can rely on: 565 * - The outer IP header has no options. 566 * - The outer IP header, the ICMP header, the inner IP header, 567 * and the first n bytes of the inner payload are contiguous. 568 * n is at least 8, but might be larger based on 569 * ICMP_ADVLENPREF. See its definition in ip_icmp.h. 570 */ 571 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 572 if (ctlfunc) 573 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc, 574 (void *)&icp->icmp_ip); 575 break; 576 577 badcode: 578 ICMPSTAT_INC(icps_badcode); 579 break; 580 581 case ICMP_ECHO: 582 if (!V_icmpbmcastecho 583 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 584 ICMPSTAT_INC(icps_bmcastecho); 585 break; 586 } 587 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 588 goto freeit; 589 icp->icmp_type = ICMP_ECHOREPLY; 590 goto reflect; 591 592 case ICMP_TSTAMP: 593 if (V_icmptstamprepl == 0) 594 break; 595 if (!V_icmpbmcastecho 596 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 597 ICMPSTAT_INC(icps_bmcasttstamp); 598 break; 599 } 600 if (icmplen < ICMP_TSLEN) { 601 ICMPSTAT_INC(icps_badlen); 602 break; 603 } 604 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 605 goto freeit; 606 icp->icmp_type = ICMP_TSTAMPREPLY; 607 icp->icmp_rtime = iptime(); 608 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 609 goto reflect; 610 611 case ICMP_MASKREQ: 612 if (V_icmpmaskrepl == 0) 613 break; 614 /* 615 * We are not able to respond with all ones broadcast 616 * unless we receive it over a point-to-point interface. 617 */ 618 if (icmplen < ICMP_MASKLEN) 619 break; 620 switch (ip->ip_dst.s_addr) { 621 622 case INADDR_BROADCAST: 623 case INADDR_ANY: 624 icmpdst.sin_addr = ip->ip_src; 625 break; 626 627 default: 628 icmpdst.sin_addr = ip->ip_dst; 629 } 630 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 631 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 632 if (ia == NULL) 633 break; 634 if (ia->ia_ifp == NULL) 635 break; 636 icp->icmp_type = ICMP_MASKREPLY; 637 if (V_icmpmaskfake == 0) 638 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 639 else 640 icp->icmp_mask = V_icmpmaskfake; 641 if (ip->ip_src.s_addr == 0) { 642 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 643 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 644 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 645 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 646 } 647 reflect: 648 ICMPSTAT_INC(icps_reflect); 649 ICMPSTAT_INC(icps_outhist[icp->icmp_type]); 650 icmp_reflect(m); 651 return (IPPROTO_DONE); 652 653 case ICMP_REDIRECT: 654 if (V_log_redirect) { 655 u_long src, dst, gw; 656 657 src = ntohl(ip->ip_src.s_addr); 658 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 659 gw = ntohl(icp->icmp_gwaddr.s_addr); 660 printf("icmp redirect from %d.%d.%d.%d: " 661 "%d.%d.%d.%d => %d.%d.%d.%d\n", 662 (int)(src >> 24), (int)((src >> 16) & 0xff), 663 (int)((src >> 8) & 0xff), (int)(src & 0xff), 664 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 665 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 666 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 667 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 668 } 669 /* 670 * RFC1812 says we must ignore ICMP redirects if we 671 * are acting as router. 672 */ 673 if (V_drop_redirect || V_ipforwarding) 674 break; 675 if (code > 3) 676 goto badcode; 677 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 678 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 679 ICMPSTAT_INC(icps_badlen); 680 break; 681 } 682 /* 683 * Short circuit routing redirects to force 684 * immediate change in the kernel's routing 685 * tables. The message is also handed to anyone 686 * listening on a raw socket (e.g. the routing 687 * daemon for use in updating its tables). 688 */ 689 icmpgw.sin_addr = ip->ip_src; 690 icmpdst.sin_addr = icp->icmp_gwaddr; 691 #ifdef ICMPPRINTFS 692 if (icmpprintfs) { 693 char dstbuf[INET_ADDRSTRLEN]; 694 char gwbuf[INET_ADDRSTRLEN]; 695 696 printf("redirect dst %s to %s\n", 697 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf), 698 inet_ntoa_r(icp->icmp_gwaddr, gwbuf)); 699 } 700 #endif 701 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 702 703 /* 704 * RFC 1122 says network (code 0,2) redirects SHOULD 705 * be treated identically to the host redirects. 706 * Given that, ignore network masks. 707 */ 708 709 /* 710 * Variable values: 711 * icmpsrc: route destination 712 * icmpdst: route gateway 713 * icmpgw: message source 714 */ 715 716 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst, 717 M_GETFIB(m)) != 0) { 718 /* TODO: increment bad redirects here */ 719 break; 720 } 721 722 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 723 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc, 724 (struct sockaddr *)&icmpdst, 725 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif, 726 RTF_GATEWAY, V_redirtimeout); 727 } 728 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); 729 break; 730 731 /* 732 * No kernel processing for the following; 733 * just fall through to send to raw listener. 734 */ 735 case ICMP_ECHOREPLY: 736 case ICMP_ROUTERADVERT: 737 case ICMP_ROUTERSOLICIT: 738 case ICMP_TSTAMPREPLY: 739 case ICMP_IREQREPLY: 740 case ICMP_MASKREPLY: 741 case ICMP_SOURCEQUENCH: 742 default: 743 break; 744 } 745 746 raw: 747 *mp = m; 748 rip_input(mp, offp, proto); 749 return (IPPROTO_DONE); 750 751 freeit: 752 m_freem(m); 753 return (IPPROTO_DONE); 754 } 755 756 /* 757 * Reflect the ip packet back to the source 758 */ 759 static void 760 icmp_reflect(struct mbuf *m) 761 { 762 struct rm_priotracker in_ifa_tracker; 763 struct ip *ip = mtod(m, struct ip *); 764 struct ifaddr *ifa; 765 struct ifnet *ifp; 766 struct in_ifaddr *ia; 767 struct in_addr t; 768 struct nhop_object *nh; 769 struct mbuf *opts = NULL; 770 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 771 772 NET_EPOCH_ASSERT(); 773 774 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 775 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) || 776 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) { 777 m_freem(m); /* Bad return address */ 778 ICMPSTAT_INC(icps_badaddr); 779 goto done; /* Ip_output() will check for broadcast */ 780 } 781 782 t = ip->ip_dst; 783 ip->ip_dst = ip->ip_src; 784 785 /* 786 * Source selection for ICMP replies: 787 * 788 * If the incoming packet was addressed directly to one of our 789 * own addresses, use dst as the src for the reply. 790 */ 791 IN_IFADDR_RLOCK(&in_ifa_tracker); 792 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { 793 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { 794 t = IA_SIN(ia)->sin_addr; 795 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 796 goto match; 797 } 798 } 799 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 800 801 /* 802 * If the incoming packet was addressed to one of our broadcast 803 * addresses, use the first non-broadcast address which corresponds 804 * to the incoming interface. 805 */ 806 ifp = m->m_pkthdr.rcvif; 807 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 808 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 809 if (ifa->ifa_addr->sa_family != AF_INET) 810 continue; 811 ia = ifatoia(ifa); 812 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 813 t.s_addr) { 814 t = IA_SIN(ia)->sin_addr; 815 goto match; 816 } 817 } 818 } 819 /* 820 * If the packet was transiting through us, use the address of 821 * the interface the packet came through in. If that interface 822 * doesn't have a suitable IP address, the normal selection 823 * criteria apply. 824 */ 825 if (V_icmp_rfi && ifp != NULL) { 826 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 827 if (ifa->ifa_addr->sa_family != AF_INET) 828 continue; 829 ia = ifatoia(ifa); 830 t = IA_SIN(ia)->sin_addr; 831 goto match; 832 } 833 } 834 /* 835 * If the incoming packet was not addressed directly to us, use 836 * designated interface for icmp replies specified by sysctl 837 * net.inet.icmp.reply_src (default not set). Otherwise continue 838 * with normal source selection. 839 */ 840 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) { 841 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 842 if (ifa->ifa_addr->sa_family != AF_INET) 843 continue; 844 ia = ifatoia(ifa); 845 t = IA_SIN(ia)->sin_addr; 846 goto match; 847 } 848 } 849 /* 850 * If the packet was transiting through us, use the address of 851 * the interface that is the closest to the packet source. 852 * When we don't have a route back to the packet source, stop here 853 * and drop the packet. 854 */ 855 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0); 856 if (nh == NULL) { 857 m_freem(m); 858 ICMPSTAT_INC(icps_noroute); 859 goto done; 860 } 861 t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr; 862 match: 863 #ifdef MAC 864 mac_netinet_icmp_replyinplace(m); 865 #endif 866 ip->ip_src = t; 867 ip->ip_ttl = V_ip_defttl; 868 869 if (optlen > 0) { 870 u_char *cp; 871 int opt, cnt; 872 u_int len; 873 874 /* 875 * Retrieve any source routing from the incoming packet; 876 * add on any record-route or timestamp options. 877 */ 878 cp = (u_char *) (ip + 1); 879 if ((opts = ip_srcroute(m)) == NULL && 880 (opts = m_gethdr(M_NOWAIT, MT_DATA))) { 881 opts->m_len = sizeof(struct in_addr); 882 mtod(opts, struct in_addr *)->s_addr = 0; 883 } 884 if (opts) { 885 #ifdef ICMPPRINTFS 886 if (icmpprintfs) 887 printf("icmp_reflect optlen %d rt %d => ", 888 optlen, opts->m_len); 889 #endif 890 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 891 opt = cp[IPOPT_OPTVAL]; 892 if (opt == IPOPT_EOL) 893 break; 894 if (opt == IPOPT_NOP) 895 len = 1; 896 else { 897 if (cnt < IPOPT_OLEN + sizeof(*cp)) 898 break; 899 len = cp[IPOPT_OLEN]; 900 if (len < IPOPT_OLEN + sizeof(*cp) || 901 len > cnt) 902 break; 903 } 904 /* 905 * Should check for overflow, but it "can't happen" 906 */ 907 if (opt == IPOPT_RR || opt == IPOPT_TS || 908 opt == IPOPT_SECURITY) { 909 bcopy((caddr_t)cp, 910 mtod(opts, caddr_t) + opts->m_len, len); 911 opts->m_len += len; 912 } 913 } 914 /* Terminate & pad, if necessary */ 915 cnt = opts->m_len % 4; 916 if (cnt) { 917 for (; cnt < 4; cnt++) { 918 *(mtod(opts, caddr_t) + opts->m_len) = 919 IPOPT_EOL; 920 opts->m_len++; 921 } 922 } 923 #ifdef ICMPPRINTFS 924 if (icmpprintfs) 925 printf("%d\n", opts->m_len); 926 #endif 927 } 928 ip_stripoptions(m); 929 } 930 m_tag_delete_nonpersistent(m); 931 m->m_flags &= ~(M_BCAST|M_MCAST); 932 icmp_send(m, opts); 933 done: 934 if (opts) 935 (void)m_free(opts); 936 } 937 938 /* 939 * Verifies if redirect message is valid, according to RFC 1122 940 * 941 * @src: sockaddr with address of redirect originator 942 * @dst: sockaddr with destination in question 943 * @gateway: new proposed gateway 944 * 945 * Returns 0 on success. 946 */ 947 static int 948 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst, 949 struct sockaddr_in *gateway, u_int fibnum) 950 { 951 struct nhop_object *nh; 952 struct ifaddr *ifa; 953 954 NET_EPOCH_ASSERT(); 955 956 /* Verify the gateway is directly reachable. */ 957 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL) 958 return (ENETUNREACH); 959 960 /* TODO: fib-aware. */ 961 if (ifa_ifwithaddr_check((struct sockaddr *)gateway)) 962 return (EHOSTUNREACH); 963 964 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0); 965 if (nh == NULL) 966 return (EINVAL); 967 968 /* 969 * If the redirect isn't from our current router for this dst, 970 * it's either old or wrong. If it redirects us to ourselves, 971 * we have a routing loop, perhaps as a result of an interface 972 * going down recently. 973 */ 974 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa)) 975 return (EINVAL); 976 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) 977 return (EINVAL); 978 979 /* If host route already exists, ignore redirect. */ 980 if (nh->nh_flags & NHF_HOST) 981 return (EEXIST); 982 983 /* If the prefix is directly reachable, ignore redirect. */ 984 if (!(nh->nh_flags & NHF_GATEWAY)) 985 return (EEXIST); 986 987 return (0); 988 } 989 990 991 /* 992 * Send an icmp packet back to the ip level, 993 * after supplying a checksum. 994 */ 995 static void 996 icmp_send(struct mbuf *m, struct mbuf *opts) 997 { 998 struct ip *ip = mtod(m, struct ip *); 999 int hlen; 1000 struct icmp *icp; 1001 1002 hlen = ip->ip_hl << 2; 1003 m->m_data += hlen; 1004 m->m_len -= hlen; 1005 icp = mtod(m, struct icmp *); 1006 icp->icmp_cksum = 0; 1007 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 1008 m->m_data -= hlen; 1009 m->m_len += hlen; 1010 m->m_pkthdr.rcvif = (struct ifnet *)0; 1011 #ifdef ICMPPRINTFS 1012 if (icmpprintfs) { 1013 char dstbuf[INET_ADDRSTRLEN]; 1014 char srcbuf[INET_ADDRSTRLEN]; 1015 1016 printf("icmp_send dst %s src %s\n", 1017 inet_ntoa_r(ip->ip_dst, dstbuf), 1018 inet_ntoa_r(ip->ip_src, srcbuf)); 1019 } 1020 #endif 1021 (void) ip_output(m, opts, NULL, 0, NULL, NULL); 1022 } 1023 1024 /* 1025 * Return milliseconds since 00:00 UTC in network format. 1026 */ 1027 uint32_t 1028 iptime(void) 1029 { 1030 struct timeval atv; 1031 u_long t; 1032 1033 getmicrotime(&atv); 1034 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1035 return (htonl(t)); 1036 } 1037 1038 /* 1039 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1040 * given current value MTU. If DIR is less than zero, a larger plateau 1041 * is returned; otherwise, a smaller value is returned. 1042 */ 1043 int 1044 ip_next_mtu(int mtu, int dir) 1045 { 1046 static int mtutab[] = { 1047 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 1048 296, 68, 0 1049 }; 1050 int i, size; 1051 1052 size = (sizeof mtutab) / (sizeof mtutab[0]); 1053 if (dir >= 0) { 1054 for (i = 0; i < size; i++) 1055 if (mtu > mtutab[i]) 1056 return mtutab[i]; 1057 } else { 1058 for (i = size - 1; i >= 0; i--) 1059 if (mtu < mtutab[i]) 1060 return mtutab[i]; 1061 if (mtu == mtutab[0]) 1062 return mtutab[0]; 1063 } 1064 return 0; 1065 } 1066 #endif /* INET */ 1067 1068 1069 /* 1070 * badport_bandlim() - check for ICMP bandwidth limit 1071 * 1072 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1073 * hit our bandwidth limit and it is not ok. 1074 * 1075 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1076 * 1077 * For now we separate the TCP and UDP subsystems w/ different 'which' 1078 * values. We may eventually remove this separation (and simplify the 1079 * code further). 1080 * 1081 * Note that the printing of the error message is delayed so we can 1082 * properly print the icmp error rate that the system was trying to do 1083 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1084 * the 'final' error, but it doesn't make sense to solve the printing 1085 * delay with more complex code. 1086 */ 1087 struct icmp_rate { 1088 const char *descr; 1089 struct counter_rate cr; 1090 }; 1091 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = { 1092 { "icmp unreach response" }, 1093 { "icmp ping response" }, 1094 { "icmp tstamp response" }, 1095 { "closed port RST response" }, 1096 { "open port RST response" }, 1097 { "icmp6 unreach response" }, 1098 { "sctp ootb response" } 1099 }; 1100 #define V_icmp_rates VNET(icmp_rates) 1101 1102 static void 1103 icmp_bandlimit_init(void) 1104 { 1105 1106 for (int i = 0; i < BANDLIM_MAX; i++) { 1107 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK); 1108 V_icmp_rates[i].cr.cr_ticks = ticks; 1109 } 1110 } 1111 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, 1112 icmp_bandlimit_init, NULL); 1113 1114 static void 1115 icmp_bandlimit_uninit(void) 1116 { 1117 1118 for (int i = 0; i < BANDLIM_MAX; i++) 1119 counter_u64_free(V_icmp_rates[i].cr.cr_rate); 1120 } 1121 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1122 icmp_bandlimit_uninit, NULL); 1123 1124 int 1125 badport_bandlim(int which) 1126 { 1127 int64_t pps; 1128 1129 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED) 1130 return (0); 1131 1132 KASSERT(which >= 0 && which < BANDLIM_MAX, 1133 ("%s: which %d", __func__, which)); 1134 1135 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim); 1136 if (pps == -1) 1137 return (-1); 1138 if (pps > 0 && V_icmplim_output) 1139 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n", 1140 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim); 1141 return (0); 1142 } 1143