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#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 */ 86VNET_DEFINE_STATIC(int, icmplim) = 200; 87#define V_icmplim VNET(icmplim) 88SYSCTL_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 92VNET_DEFINE_STATIC(int, icmplim_output) = 1; 93#define V_icmplim_output VNET(icmplim_output) 94SYSCTL_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 99VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat); 100VNET_PCPUSTAT_SYSINIT(icmpstat); 101SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat, 102 icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)"); 103 104#ifdef VIMAGE 105VNET_PCPUSTAT_SYSUNINIT(icmpstat); 106#endif /* VIMAGE */ 107 108VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0; 109#define V_icmpmaskrepl VNET(icmpmaskrepl) 110SYSCTL_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 114VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0; 115#define V_icmpmaskfake VNET(icmpmaskfake) 116SYSCTL_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 120VNET_DEFINE(int, drop_redirect) = 0; 121#define V_drop_redirect VNET(drop_redirect) 122SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW, 123 &VNET_NAME(drop_redirect), 0, 124 "Ignore ICMP redirects"); 125 126VNET_DEFINE_STATIC(int, log_redirect) = 0; 127#define V_log_redirect VNET(log_redirect) 128SYSCTL_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 132VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */ 133#define V_redirtimeout VNET(redirtimeout) 134SYSCTL_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 138VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]); 139#define V_reply_src VNET(reply_src) 140SYSCTL_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 144VNET_DEFINE_STATIC(int, icmp_rfi) = 0; 145#define V_icmp_rfi VNET(icmp_rfi) 146SYSCTL_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. */ 150VNET_DEFINE_STATIC(int, icmp_quotelen) = 548; 151#define V_icmp_quotelen VNET(icmp_quotelen) 152SYSCTL_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 156VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0; 157#define V_icmpbmcastecho VNET(icmpbmcastecho) 158SYSCTL_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 162VNET_DEFINE_STATIC(int, icmptstamprepl) = 1; 163#define V_icmptstamprepl VNET(icmptstamprepl) 164SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW, 165 &VNET_NAME(icmptstamprepl), 0, 166 "Respond to ICMP Timestamp packets"); 167 168VNET_DEFINE_STATIC(int, error_keeptags) = 0; 169#define V_error_keeptags VNET(error_keeptags) 170SYSCTL_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 175int icmpprintfs = 0; 176#endif 177 178static void icmp_reflect(struct mbuf *); 179static void icmp_send(struct mbuf *, struct mbuf *); 180static int icmp_verify_redirect_gateway(struct sockaddr_in *, 181 struct sockaddr_in *, struct sockaddr_in *, u_int); 182 183extern 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 */ 192void 193kmod_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 */ 203void 204icmp_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 314stdreply: 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 395freeit: 396 m_freem(n); 397} 398 399/* 400 * Process a received ICMP message. 401 */ 402int 403icmp_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 case ICMP_UNREACH: 481 switch (code) { 482 case ICMP_UNREACH_NET: 483 case ICMP_UNREACH_HOST: 484 case ICMP_UNREACH_SRCFAIL: 485 case ICMP_UNREACH_NET_UNKNOWN: 486 case ICMP_UNREACH_HOST_UNKNOWN: 487 case ICMP_UNREACH_ISOLATED: 488 case ICMP_UNREACH_TOSNET: 489 case ICMP_UNREACH_TOSHOST: 490 case ICMP_UNREACH_HOST_PRECEDENCE: 491 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 492 code = PRC_UNREACH_NET; 493 break; 494 495 case ICMP_UNREACH_NEEDFRAG: 496 code = PRC_MSGSIZE; 497 break; 498 499 /* 500 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. 501 * Treat subcodes 2,3 as immediate RST 502 */ 503 case ICMP_UNREACH_PROTOCOL: 504 code = PRC_UNREACH_PROTOCOL; 505 break; 506 case ICMP_UNREACH_PORT: 507 code = PRC_UNREACH_PORT; 508 break; 509 510 case ICMP_UNREACH_NET_PROHIB: 511 case ICMP_UNREACH_HOST_PROHIB: 512 case ICMP_UNREACH_FILTER_PROHIB: 513 code = PRC_UNREACH_ADMIN_PROHIB; 514 break; 515 516 default: 517 goto badcode; 518 } 519 goto deliver; 520 521 case ICMP_TIMXCEED: 522 if (code > 1) 523 goto badcode; 524 code += PRC_TIMXCEED_INTRANS; 525 goto deliver; 526 527 case ICMP_PARAMPROB: 528 if (code > 1) 529 goto badcode; 530 code = PRC_PARAMPROB; 531 deliver: 532 /* 533 * Problem with datagram; advise higher level routines. 534 */ 535 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 536 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 537 ICMPSTAT_INC(icps_badlen); 538 goto freeit; 539 } 540 /* Discard ICMP's in response to multicast packets */ 541 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 542 goto badcode; 543#ifdef ICMPPRINTFS 544 if (icmpprintfs) 545 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 546#endif 547 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 548 /* 549 * XXX if the packet contains [IPv4 AH TCP], we can't make a 550 * notification to TCP layer. 551 */ 552 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp)); 553 ip_stripoptions(m); 554 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 555 /* This should actually not happen */ 556 ICMPSTAT_INC(icps_tooshort); 557 return (IPPROTO_DONE); 558 } 559 ip = mtod(m, struct ip *); 560 icp = (struct icmp *)(ip + 1); 561 /* 562 * The upper layer handler can rely on: 563 * - The outer IP header has no options. 564 * - The outer IP header, the ICMP header, the inner IP header, 565 * and the first n bytes of the inner payload are contiguous. 566 * n is at least 8, but might be larger based on 567 * ICMP_ADVLENPREF. See its definition in ip_icmp.h. 568 */ 569 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 570 if (ctlfunc) 571 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc, 572 (void *)&icp->icmp_ip); 573 break; 574 575 badcode: 576 ICMPSTAT_INC(icps_badcode); 577 break; 578 579 case ICMP_ECHO: 580 if (!V_icmpbmcastecho 581 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 582 ICMPSTAT_INC(icps_bmcastecho); 583 break; 584 } 585 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 586 goto freeit; 587 icp->icmp_type = ICMP_ECHOREPLY; 588 goto reflect; 589 590 case ICMP_TSTAMP: 591 if (V_icmptstamprepl == 0) 592 break; 593 if (!V_icmpbmcastecho 594 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 595 ICMPSTAT_INC(icps_bmcasttstamp); 596 break; 597 } 598 if (icmplen < ICMP_TSLEN) { 599 ICMPSTAT_INC(icps_badlen); 600 break; 601 } 602 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 603 goto freeit; 604 icp->icmp_type = ICMP_TSTAMPREPLY; 605 icp->icmp_rtime = iptime(); 606 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 607 goto reflect; 608 609 case ICMP_MASKREQ: 610 if (V_icmpmaskrepl == 0) 611 break; 612 /* 613 * We are not able to respond with all ones broadcast 614 * unless we receive it over a point-to-point interface. 615 */ 616 if (icmplen < ICMP_MASKLEN) 617 break; 618 switch (ip->ip_dst.s_addr) { 619 case INADDR_BROADCAST: 620 case INADDR_ANY: 621 icmpdst.sin_addr = ip->ip_src; 622 break; 623 624 default: 625 icmpdst.sin_addr = ip->ip_dst; 626 } 627 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 628 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 629 if (ia == NULL) 630 break; 631 if (ia->ia_ifp == NULL) 632 break; 633 icp->icmp_type = ICMP_MASKREPLY; 634 if (V_icmpmaskfake == 0) 635 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 636 else 637 icp->icmp_mask = V_icmpmaskfake; 638 if (ip->ip_src.s_addr == 0) { 639 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 640 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 641 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 642 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 643 } 644reflect: 645 ICMPSTAT_INC(icps_reflect); 646 ICMPSTAT_INC(icps_outhist[icp->icmp_type]); 647 icmp_reflect(m); 648 return (IPPROTO_DONE); 649 650 case ICMP_REDIRECT: 651 if (V_log_redirect) { 652 u_long src, dst, gw; 653 654 src = ntohl(ip->ip_src.s_addr); 655 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 656 gw = ntohl(icp->icmp_gwaddr.s_addr); 657 printf("icmp redirect from %d.%d.%d.%d: " 658 "%d.%d.%d.%d => %d.%d.%d.%d\n", 659 (int)(src >> 24), (int)((src >> 16) & 0xff), 660 (int)((src >> 8) & 0xff), (int)(src & 0xff), 661 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 662 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 663 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 664 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 665 } 666 /* 667 * RFC1812 says we must ignore ICMP redirects if we 668 * are acting as router. 669 */ 670 if (V_drop_redirect || V_ipforwarding) 671 break; 672 if (code > 3) 673 goto badcode; 674 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 675 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 676 ICMPSTAT_INC(icps_badlen); 677 break; 678 } 679 /* 680 * Short circuit routing redirects to force 681 * immediate change in the kernel's routing 682 * tables. The message is also handed to anyone 683 * listening on a raw socket (e.g. the routing 684 * daemon for use in updating its tables). 685 */ 686 icmpgw.sin_addr = ip->ip_src; 687 icmpdst.sin_addr = icp->icmp_gwaddr; 688#ifdef ICMPPRINTFS 689 if (icmpprintfs) { 690 char dstbuf[INET_ADDRSTRLEN]; 691 char gwbuf[INET_ADDRSTRLEN]; 692 693 printf("redirect dst %s to %s\n", 694 inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf), 695 inet_ntoa_r(icp->icmp_gwaddr, gwbuf)); 696 } 697#endif 698 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 699 700 /* 701 * RFC 1122 says network (code 0,2) redirects SHOULD 702 * be treated identically to the host redirects. 703 * Given that, ignore network masks. 704 */ 705 706 /* 707 * Variable values: 708 * icmpsrc: route destination 709 * icmpdst: route gateway 710 * icmpgw: message source 711 */ 712 713 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst, 714 M_GETFIB(m)) != 0) { 715 /* TODO: increment bad redirects here */ 716 break; 717 } 718 719 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 720 rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc, 721 (struct sockaddr *)&icmpdst, 722 (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif, 723 RTF_GATEWAY, V_redirtimeout); 724 } 725 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); 726 break; 727 728 /* 729 * No kernel processing for the following; 730 * just fall through to send to raw listener. 731 */ 732 case ICMP_ECHOREPLY: 733 case ICMP_ROUTERADVERT: 734 case ICMP_ROUTERSOLICIT: 735 case ICMP_TSTAMPREPLY: 736 case ICMP_IREQREPLY: 737 case ICMP_MASKREPLY: 738 case ICMP_SOURCEQUENCH: 739 default: 740 break; 741 } 742 743raw: 744 *mp = m; 745 rip_input(mp, offp, proto); 746 return (IPPROTO_DONE); 747 748freeit: 749 m_freem(m); 750 return (IPPROTO_DONE); 751} 752 753/* 754 * Reflect the ip packet back to the source 755 */ 756static void 757icmp_reflect(struct mbuf *m) 758{ 759 struct rm_priotracker in_ifa_tracker; 760 struct ip *ip = mtod(m, struct ip *); 761 struct ifaddr *ifa; 762 struct ifnet *ifp; 763 struct in_ifaddr *ia; 764 struct in_addr t; 765 struct nhop_object *nh; 766 struct mbuf *opts = NULL; 767 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 768 769 NET_EPOCH_ASSERT(); 770 771 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 772 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) || 773 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) { 774 m_freem(m); /* Bad return address */ 775 ICMPSTAT_INC(icps_badaddr); 776 goto done; /* Ip_output() will check for broadcast */ 777 } 778 779 t = ip->ip_dst; 780 ip->ip_dst = ip->ip_src; 781 782 /* 783 * Source selection for ICMP replies: 784 * 785 * If the incoming packet was addressed directly to one of our 786 * own addresses, use dst as the src for the reply. 787 */ 788 IN_IFADDR_RLOCK(&in_ifa_tracker); 789 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) { 790 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) { 791 t = IA_SIN(ia)->sin_addr; 792 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 793 goto match; 794 } 795 } 796 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 797 798 /* 799 * If the incoming packet was addressed to one of our broadcast 800 * addresses, use the first non-broadcast address which corresponds 801 * to the incoming interface. 802 */ 803 ifp = m->m_pkthdr.rcvif; 804 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 805 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 806 if (ifa->ifa_addr->sa_family != AF_INET) 807 continue; 808 ia = ifatoia(ifa); 809 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 810 t.s_addr) { 811 t = IA_SIN(ia)->sin_addr; 812 goto match; 813 } 814 } 815 } 816 /* 817 * If the packet was transiting through us, use the address of 818 * the interface the packet came through in. If that interface 819 * doesn't have a suitable IP address, the normal selection 820 * criteria apply. 821 */ 822 if (V_icmp_rfi && ifp != NULL) { 823 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 824 if (ifa->ifa_addr->sa_family != AF_INET) 825 continue; 826 ia = ifatoia(ifa); 827 t = IA_SIN(ia)->sin_addr; 828 goto match; 829 } 830 } 831 /* 832 * If the incoming packet was not addressed directly to us, use 833 * designated interface for icmp replies specified by sysctl 834 * net.inet.icmp.reply_src (default not set). Otherwise continue 835 * with normal source selection. 836 */ 837 if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) { 838 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 839 if (ifa->ifa_addr->sa_family != AF_INET) 840 continue; 841 ia = ifatoia(ifa); 842 t = IA_SIN(ia)->sin_addr; 843 goto match; 844 } 845 } 846 /* 847 * If the packet was transiting through us, use the address of 848 * the interface that is the closest to the packet source. 849 * When we don't have a route back to the packet source, stop here 850 * and drop the packet. 851 */ 852 nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0); 853 if (nh == NULL) { 854 m_freem(m); 855 ICMPSTAT_INC(icps_noroute); 856 goto done; 857 } 858 t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr; 859match: 860#ifdef MAC 861 mac_netinet_icmp_replyinplace(m); 862#endif 863 ip->ip_src = t; 864 ip->ip_ttl = V_ip_defttl; 865 866 if (optlen > 0) { 867 u_char *cp; 868 int opt, cnt; 869 u_int len; 870 871 /* 872 * Retrieve any source routing from the incoming packet; 873 * add on any record-route or timestamp options. 874 */ 875 cp = (u_char *) (ip + 1); 876 if ((opts = ip_srcroute(m)) == NULL && 877 (opts = m_gethdr(M_NOWAIT, MT_DATA))) { 878 opts->m_len = sizeof(struct in_addr); 879 mtod(opts, struct in_addr *)->s_addr = 0; 880 } 881 if (opts) { 882#ifdef ICMPPRINTFS 883 if (icmpprintfs) 884 printf("icmp_reflect optlen %d rt %d => ", 885 optlen, opts->m_len); 886#endif 887 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 888 opt = cp[IPOPT_OPTVAL]; 889 if (opt == IPOPT_EOL) 890 break; 891 if (opt == IPOPT_NOP) 892 len = 1; 893 else { 894 if (cnt < IPOPT_OLEN + sizeof(*cp)) 895 break; 896 len = cp[IPOPT_OLEN]; 897 if (len < IPOPT_OLEN + sizeof(*cp) || 898 len > cnt) 899 break; 900 } 901 /* 902 * Should check for overflow, but it "can't happen" 903 */ 904 if (opt == IPOPT_RR || opt == IPOPT_TS || 905 opt == IPOPT_SECURITY) { 906 bcopy((caddr_t)cp, 907 mtod(opts, caddr_t) + opts->m_len, len); 908 opts->m_len += len; 909 } 910 } 911 /* Terminate & pad, if necessary */ 912 cnt = opts->m_len % 4; 913 if (cnt) { 914 for (; cnt < 4; cnt++) { 915 *(mtod(opts, caddr_t) + opts->m_len) = 916 IPOPT_EOL; 917 opts->m_len++; 918 } 919 } 920#ifdef ICMPPRINTFS 921 if (icmpprintfs) 922 printf("%d\n", opts->m_len); 923#endif 924 } 925 ip_stripoptions(m); 926 } 927 m_tag_delete_nonpersistent(m); 928 m->m_flags &= ~(M_BCAST|M_MCAST); 929 icmp_send(m, opts); 930done: 931 if (opts) 932 (void)m_free(opts); 933} 934 935/* 936 * Verifies if redirect message is valid, according to RFC 1122 937 * 938 * @src: sockaddr with address of redirect originator 939 * @dst: sockaddr with destination in question 940 * @gateway: new proposed gateway 941 * 942 * Returns 0 on success. 943 */ 944static int 945icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst, 946 struct sockaddr_in *gateway, u_int fibnum) 947{ 948 struct nhop_object *nh; 949 struct ifaddr *ifa; 950 951 NET_EPOCH_ASSERT(); 952 953 /* Verify the gateway is directly reachable. */ 954 if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL) 955 return (ENETUNREACH); 956 957 /* TODO: fib-aware. */ 958 if (ifa_ifwithaddr_check((struct sockaddr *)gateway)) 959 return (EHOSTUNREACH); 960 961 nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0); 962 if (nh == NULL) 963 return (EINVAL); 964 965 /* 966 * If the redirect isn't from our current router for this dst, 967 * it's either old or wrong. If it redirects us to ourselves, 968 * we have a routing loop, perhaps as a result of an interface 969 * going down recently. 970 */ 971 if (!sa_equal((struct sockaddr *)src, &nh->gw_sa)) 972 return (EINVAL); 973 if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) 974 return (EINVAL); 975 976 /* If host route already exists, ignore redirect. */ 977 if (nh->nh_flags & NHF_HOST) 978 return (EEXIST); 979 980 /* If the prefix is directly reachable, ignore redirect. */ 981 if (!(nh->nh_flags & NHF_GATEWAY)) 982 return (EEXIST); 983 984 return (0); 985} 986 987/* 988 * Send an icmp packet back to the ip level, 989 * after supplying a checksum. 990 */ 991static void 992icmp_send(struct mbuf *m, struct mbuf *opts) 993{ 994 struct ip *ip = mtod(m, struct ip *); 995 int hlen; 996 struct icmp *icp; 997 998 hlen = ip->ip_hl << 2; 999 m->m_data += hlen; 1000 m->m_len -= hlen; 1001 icp = mtod(m, struct icmp *); 1002 icp->icmp_cksum = 0; 1003 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 1004 m->m_data -= hlen; 1005 m->m_len += hlen; 1006 m->m_pkthdr.rcvif = (struct ifnet *)0; 1007#ifdef ICMPPRINTFS 1008 if (icmpprintfs) { 1009 char dstbuf[INET_ADDRSTRLEN]; 1010 char srcbuf[INET_ADDRSTRLEN]; 1011 1012 printf("icmp_send dst %s src %s\n", 1013 inet_ntoa_r(ip->ip_dst, dstbuf), 1014 inet_ntoa_r(ip->ip_src, srcbuf)); 1015 } 1016#endif 1017 (void) ip_output(m, opts, NULL, 0, NULL, NULL); 1018} 1019 1020/* 1021 * Return milliseconds since 00:00 UTC in network format. 1022 */ 1023uint32_t 1024iptime(void) 1025{ 1026 struct timeval atv; 1027 u_long t; 1028 1029 getmicrotime(&atv); 1030 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1031 return (htonl(t)); 1032} 1033 1034/* 1035 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1036 * given current value MTU. If DIR is less than zero, a larger plateau 1037 * is returned; otherwise, a smaller value is returned. 1038 */ 1039int 1040ip_next_mtu(int mtu, int dir) 1041{ 1042 static int mtutab[] = { 1043 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 1044 296, 68, 0 1045 }; 1046 int i, size; 1047 1048 size = (sizeof mtutab) / (sizeof mtutab[0]); 1049 if (dir >= 0) { 1050 for (i = 0; i < size; i++) 1051 if (mtu > mtutab[i]) 1052 return mtutab[i]; 1053 } else { 1054 for (i = size - 1; i >= 0; i--) 1055 if (mtu < mtutab[i]) 1056 return mtutab[i]; 1057 if (mtu == mtutab[0]) 1058 return mtutab[0]; 1059 } 1060 return 0; 1061} 1062#endif /* INET */ 1063 1064/* 1065 * badport_bandlim() - check for ICMP bandwidth limit 1066 * 1067 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1068 * hit our bandwidth limit and it is not ok. 1069 * 1070 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1071 * 1072 * For now we separate the TCP and UDP subsystems w/ different 'which' 1073 * values. We may eventually remove this separation (and simplify the 1074 * code further). 1075 * 1076 * Note that the printing of the error message is delayed so we can 1077 * properly print the icmp error rate that the system was trying to do 1078 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1079 * the 'final' error, but it doesn't make sense to solve the printing 1080 * delay with more complex code. 1081 */ 1082struct icmp_rate { 1083 const char *descr; 1084 struct counter_rate cr; 1085}; 1086VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = { 1087 { "icmp unreach response" }, 1088 { "icmp ping response" }, 1089 { "icmp tstamp response" }, 1090 { "closed port RST response" }, 1091 { "open port RST response" }, 1092 { "icmp6 unreach response" }, 1093 { "sctp ootb response" } 1094}; 1095#define V_icmp_rates VNET(icmp_rates) 1096 1097static void 1098icmp_bandlimit_init(void) 1099{ 1100 1101 for (int i = 0; i < BANDLIM_MAX; i++) { 1102 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK); 1103 V_icmp_rates[i].cr.cr_ticks = ticks; 1104 } 1105} 1106VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, 1107 icmp_bandlimit_init, NULL); 1108 1109static void 1110icmp_bandlimit_uninit(void) 1111{ 1112 1113 for (int i = 0; i < BANDLIM_MAX; i++) 1114 counter_u64_free(V_icmp_rates[i].cr.cr_rate); 1115} 1116VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1117 icmp_bandlimit_uninit, NULL); 1118 1119int 1120badport_bandlim(int which) 1121{ 1122 int64_t pps; 1123 1124 if (V_icmplim == 0 || which == BANDLIM_UNLIMITED) 1125 return (0); 1126 1127 KASSERT(which >= 0 && which < BANDLIM_MAX, 1128 ("%s: which %d", __func__, which)); 1129 1130 pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim); 1131 if (pps == -1) 1132 return (-1); 1133 if (pps > 0 && V_icmplim_output) 1134 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n", 1135 V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim); 1136 return (0); 1137} 1138