1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_ipsec.h" 36 #include "opt_mac.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/mbuf.h> 41 #include <sys/protosw.h> 42 #include <sys/socket.h> 43 #include <sys/time.h> 44 #include <sys/kernel.h> 45 #include <sys/sysctl.h> 46 #include <sys/vimage.h> 47 48 #include <net/if.h> 49 #include <net/if_types.h> 50 #include <net/route.h> 51 52 #include <netinet/in.h> 53 #include <netinet/in_pcb.h> 54 #include <netinet/in_systm.h> 55 #include <netinet/in_var.h> 56 #include <netinet/ip.h> 57 #include <netinet/ip_icmp.h> 58 #include <netinet/ip_var.h> 59 #include <netinet/ip_options.h> 60 #include <netinet/tcp.h> 61 #include <netinet/tcp_var.h> 62 #include <netinet/tcpip.h> 63 #include <netinet/icmp_var.h> 64 65 #ifdef IPSEC 66 #include <netipsec/ipsec.h> 67 #include <netipsec/key.h> 68 #endif 69 70 #include <machine/in_cksum.h> 71 72 #include <security/mac/mac_framework.h> 73 74 /* 75 * ICMP routines: error generation, receive packet processing, and 76 * routines to turnaround packets back to the originator, and 77 * host table maintenance routines. 78 */ 79 80 struct icmpstat icmpstat; 81 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW, 82 &icmpstat, icmpstat, ""); 83 84 static int icmpmaskrepl = 0; 85 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW, 86 &icmpmaskrepl, 0, "Reply to ICMP Address Mask Request packets."); 87 88 static u_int icmpmaskfake = 0; 89 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW, 90 &icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets."); 91 92 static int drop_redirect = 0; 93 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW, 94 &drop_redirect, 0, "Ignore ICMP redirects"); 95 96 static int log_redirect = 0; 97 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW, 98 &log_redirect, 0, "Log ICMP redirects to the console"); 99 100 static int icmplim = 200; 101 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW, 102 &icmplim, 0, "Maximum number of ICMP responses per second"); 103 104 static int icmplim_output = 1; 105 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW, 106 &icmplim_output, 0, "Enable rate limiting of ICMP responses"); 107 108 static char reply_src[IFNAMSIZ]; 109 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW, 110 &reply_src, IFNAMSIZ, "icmp reply source for non-local packets."); 111 112 static int icmp_rfi = 0; 113 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW, 114 &icmp_rfi, 0, "ICMP reply from incoming interface for " 115 "non-local packets"); 116 117 static int icmp_quotelen = 8; 118 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW, 119 &icmp_quotelen, 0, "Number of bytes from original packet to " 120 "quote in ICMP reply"); 121 122 /* 123 * ICMP broadcast echo sysctl 124 */ 125 126 static int icmpbmcastecho = 0; 127 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW, 128 &icmpbmcastecho, 0, ""); 129 130 131 #ifdef ICMPPRINTFS 132 int icmpprintfs = 0; 133 #endif 134 135 static void icmp_reflect(struct mbuf *); 136 static void icmp_send(struct mbuf *, struct mbuf *); 137 138 extern struct protosw inetsw[]; 139 140 /* 141 * Generate an error packet of type error 142 * in response to bad packet ip. 143 */ 144 void 145 icmp_error(struct mbuf *n, int type, int code, n_long dest, int mtu) 146 { 147 register struct ip *oip = mtod(n, struct ip *), *nip; 148 register unsigned oiphlen = oip->ip_hl << 2; 149 register struct icmp *icp; 150 register struct mbuf *m; 151 unsigned icmplen, icmpelen, nlen; 152 153 KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__)); 154 #ifdef ICMPPRINTFS 155 if (icmpprintfs) 156 printf("icmp_error(%p, %x, %d)\n", oip, type, code); 157 #endif 158 if (type != ICMP_REDIRECT) 159 V_icmpstat.icps_error++; 160 /* 161 * Don't send error: 162 * if the original packet was encrypted. 163 * if not the first fragment of message. 164 * in response to a multicast or broadcast packet. 165 * if the old packet protocol was an ICMP error message. 166 */ 167 if (n->m_flags & M_DECRYPTED) 168 goto freeit; 169 if (oip->ip_off & ~(IP_MF|IP_DF)) 170 goto freeit; 171 if (n->m_flags & (M_BCAST|M_MCAST)) 172 goto freeit; 173 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 174 n->m_len >= oiphlen + ICMP_MINLEN && 175 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) { 176 V_icmpstat.icps_oldicmp++; 177 goto freeit; 178 } 179 /* Drop if IP header plus 8 bytes is not contignous in first mbuf. */ 180 if (oiphlen + 8 > n->m_len) 181 goto freeit; 182 /* 183 * Calculate length to quote from original packet and 184 * prevent the ICMP mbuf from overflowing. 185 * Unfortunatly this is non-trivial since ip_forward() 186 * sends us truncated packets. 187 */ 188 nlen = m_length(n, NULL); 189 if (oip->ip_p == IPPROTO_TCP) { 190 struct tcphdr *th; 191 int tcphlen; 192 193 if (oiphlen + sizeof(struct tcphdr) > n->m_len && 194 n->m_next == NULL) 195 goto stdreply; 196 if (n->m_len < oiphlen + sizeof(struct tcphdr) && 197 ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)) 198 goto freeit; 199 th = (struct tcphdr *)((caddr_t)oip + oiphlen); 200 tcphlen = th->th_off << 2; 201 if (tcphlen < sizeof(struct tcphdr)) 202 goto freeit; 203 if (oip->ip_len < oiphlen + tcphlen) 204 goto freeit; 205 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL) 206 goto stdreply; 207 if (n->m_len < oiphlen + tcphlen && 208 ((n = m_pullup(n, oiphlen + tcphlen)) == NULL)) 209 goto freeit; 210 icmpelen = max(tcphlen, min(icmp_quotelen, oip->ip_len - oiphlen)); 211 } else 212 stdreply: icmpelen = max(8, min(icmp_quotelen, oip->ip_len - oiphlen)); 213 214 icmplen = min(oiphlen + icmpelen, nlen); 215 if (icmplen < sizeof(struct ip)) 216 goto freeit; 217 218 if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen) 219 m = m_gethdr(M_DONTWAIT, MT_DATA); 220 else 221 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR); 222 if (m == NULL) 223 goto freeit; 224 #ifdef MAC 225 mac_netinet_icmp_reply(n, m); 226 #endif 227 icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN); 228 m_align(m, ICMP_MINLEN + icmplen); 229 m->m_len = ICMP_MINLEN + icmplen; 230 231 /* XXX MRT make the outgoing packet use the same FIB 232 * that was associated with the incoming packet 233 */ 234 M_SETFIB(m, M_GETFIB(n)); 235 icp = mtod(m, struct icmp *); 236 V_icmpstat.icps_outhist[type]++; 237 icp->icmp_type = type; 238 if (type == ICMP_REDIRECT) 239 icp->icmp_gwaddr.s_addr = dest; 240 else { 241 icp->icmp_void = 0; 242 /* 243 * The following assignments assume an overlay with the 244 * just zeroed icmp_void field. 245 */ 246 if (type == ICMP_PARAMPROB) { 247 icp->icmp_pptr = code; 248 code = 0; 249 } else if (type == ICMP_UNREACH && 250 code == ICMP_UNREACH_NEEDFRAG && mtu) { 251 icp->icmp_nextmtu = htons(mtu); 252 } 253 } 254 icp->icmp_code = code; 255 256 /* 257 * Copy the quotation into ICMP message and 258 * convert quoted IP header back to network representation. 259 */ 260 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 261 nip = &icp->icmp_ip; 262 nip->ip_len = htons(nip->ip_len); 263 nip->ip_off = htons(nip->ip_off); 264 265 /* 266 * Set up ICMP message mbuf and copy old IP header (without options 267 * in front of ICMP message. 268 * If the original mbuf was meant to bypass the firewall, the error 269 * reply should bypass as well. 270 */ 271 m->m_flags |= n->m_flags & M_SKIP_FIREWALL; 272 m->m_data -= sizeof(struct ip); 273 m->m_len += sizeof(struct ip); 274 m->m_pkthdr.len = m->m_len; 275 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 276 nip = mtod(m, struct ip *); 277 bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip)); 278 nip->ip_len = m->m_len; 279 nip->ip_v = IPVERSION; 280 nip->ip_hl = 5; 281 nip->ip_p = IPPROTO_ICMP; 282 nip->ip_tos = 0; 283 icmp_reflect(m); 284 285 freeit: 286 m_freem(n); 287 } 288 289 /* 290 * Process a received ICMP message. 291 */ 292 void 293 icmp_input(struct mbuf *m, int off) 294 { 295 struct icmp *icp; 296 struct in_ifaddr *ia; 297 struct ip *ip = mtod(m, struct ip *); 298 struct sockaddr_in icmpsrc, icmpdst, icmpgw; 299 int hlen = off; 300 int icmplen = ip->ip_len; 301 int i, code; 302 void (*ctlfunc)(int, struct sockaddr *, void *); 303 int fibnum; 304 305 /* 306 * Locate icmp structure in mbuf, and check 307 * that not corrupted and of at least minimum length. 308 */ 309 #ifdef ICMPPRINTFS 310 if (icmpprintfs) { 311 char buf[4 * sizeof "123"]; 312 strcpy(buf, inet_ntoa(ip->ip_src)); 313 printf("icmp_input from %s to %s, len %d\n", 314 buf, inet_ntoa(ip->ip_dst), icmplen); 315 } 316 #endif 317 if (icmplen < ICMP_MINLEN) { 318 V_icmpstat.icps_tooshort++; 319 goto freeit; 320 } 321 i = hlen + min(icmplen, ICMP_ADVLENMIN); 322 if (m->m_len < i && (m = m_pullup(m, i)) == 0) { 323 V_icmpstat.icps_tooshort++; 324 return; 325 } 326 ip = mtod(m, struct ip *); 327 m->m_len -= hlen; 328 m->m_data += hlen; 329 icp = mtod(m, struct icmp *); 330 if (in_cksum(m, icmplen)) { 331 V_icmpstat.icps_checksum++; 332 goto freeit; 333 } 334 m->m_len += hlen; 335 m->m_data -= hlen; 336 337 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 338 /* 339 * Deliver very specific ICMP type only. 340 */ 341 switch (icp->icmp_type) { 342 case ICMP_UNREACH: 343 case ICMP_TIMXCEED: 344 break; 345 default: 346 goto freeit; 347 } 348 } 349 350 #ifdef ICMPPRINTFS 351 if (icmpprintfs) 352 printf("icmp_input, type %d code %d\n", icp->icmp_type, 353 icp->icmp_code); 354 #endif 355 356 /* 357 * Message type specific processing. 358 */ 359 if (icp->icmp_type > ICMP_MAXTYPE) 360 goto raw; 361 362 /* Initialize */ 363 bzero(&icmpsrc, sizeof(icmpsrc)); 364 icmpsrc.sin_len = sizeof(struct sockaddr_in); 365 icmpsrc.sin_family = AF_INET; 366 bzero(&icmpdst, sizeof(icmpdst)); 367 icmpdst.sin_len = sizeof(struct sockaddr_in); 368 icmpdst.sin_family = AF_INET; 369 bzero(&icmpgw, sizeof(icmpgw)); 370 icmpgw.sin_len = sizeof(struct sockaddr_in); 371 icmpgw.sin_family = AF_INET; 372 373 V_icmpstat.icps_inhist[icp->icmp_type]++; 374 code = icp->icmp_code; 375 switch (icp->icmp_type) { 376 377 case ICMP_UNREACH: 378 switch (code) { 379 case ICMP_UNREACH_NET: 380 case ICMP_UNREACH_HOST: 381 case ICMP_UNREACH_SRCFAIL: 382 case ICMP_UNREACH_NET_UNKNOWN: 383 case ICMP_UNREACH_HOST_UNKNOWN: 384 case ICMP_UNREACH_ISOLATED: 385 case ICMP_UNREACH_TOSNET: 386 case ICMP_UNREACH_TOSHOST: 387 case ICMP_UNREACH_HOST_PRECEDENCE: 388 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 389 code = PRC_UNREACH_NET; 390 break; 391 392 case ICMP_UNREACH_NEEDFRAG: 393 code = PRC_MSGSIZE; 394 break; 395 396 /* 397 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. 398 * Treat subcodes 2,3 as immediate RST 399 */ 400 case ICMP_UNREACH_PROTOCOL: 401 case ICMP_UNREACH_PORT: 402 code = PRC_UNREACH_PORT; 403 break; 404 405 case ICMP_UNREACH_NET_PROHIB: 406 case ICMP_UNREACH_HOST_PROHIB: 407 case ICMP_UNREACH_FILTER_PROHIB: 408 code = PRC_UNREACH_ADMIN_PROHIB; 409 break; 410 411 default: 412 goto badcode; 413 } 414 goto deliver; 415 416 case ICMP_TIMXCEED: 417 if (code > 1) 418 goto badcode; 419 code += PRC_TIMXCEED_INTRANS; 420 goto deliver; 421 422 case ICMP_PARAMPROB: 423 if (code > 1) 424 goto badcode; 425 code = PRC_PARAMPROB; 426 goto deliver; 427 428 case ICMP_SOURCEQUENCH: 429 if (code) 430 goto badcode; 431 code = PRC_QUENCH; 432 deliver: 433 /* 434 * Problem with datagram; advise higher level routines. 435 */ 436 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 437 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 438 V_icmpstat.icps_badlen++; 439 goto freeit; 440 } 441 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len); 442 /* Discard ICMP's in response to multicast packets */ 443 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 444 goto badcode; 445 #ifdef ICMPPRINTFS 446 if (icmpprintfs) 447 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 448 #endif 449 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 450 /* 451 * XXX if the packet contains [IPv4 AH TCP], we can't make a 452 * notification to TCP layer. 453 */ 454 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 455 if (ctlfunc) 456 (*ctlfunc)(code, (struct sockaddr *)&icmpsrc, 457 (void *)&icp->icmp_ip); 458 break; 459 460 badcode: 461 V_icmpstat.icps_badcode++; 462 break; 463 464 case ICMP_ECHO: 465 if (!icmpbmcastecho 466 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 467 V_icmpstat.icps_bmcastecho++; 468 break; 469 } 470 icp->icmp_type = ICMP_ECHOREPLY; 471 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 472 goto freeit; 473 else 474 goto reflect; 475 476 case ICMP_TSTAMP: 477 if (!icmpbmcastecho 478 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 479 V_icmpstat.icps_bmcasttstamp++; 480 break; 481 } 482 if (icmplen < ICMP_TSLEN) { 483 V_icmpstat.icps_badlen++; 484 break; 485 } 486 icp->icmp_type = ICMP_TSTAMPREPLY; 487 icp->icmp_rtime = iptime(); 488 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 489 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 490 goto freeit; 491 else 492 goto reflect; 493 494 case ICMP_MASKREQ: 495 if (icmpmaskrepl == 0) 496 break; 497 /* 498 * We are not able to respond with all ones broadcast 499 * unless we receive it over a point-to-point interface. 500 */ 501 if (icmplen < ICMP_MASKLEN) 502 break; 503 switch (ip->ip_dst.s_addr) { 504 505 case INADDR_BROADCAST: 506 case INADDR_ANY: 507 icmpdst.sin_addr = ip->ip_src; 508 break; 509 510 default: 511 icmpdst.sin_addr = ip->ip_dst; 512 } 513 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 514 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 515 if (ia == 0) 516 break; 517 if (ia->ia_ifp == 0) 518 break; 519 icp->icmp_type = ICMP_MASKREPLY; 520 if (icmpmaskfake == 0) 521 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 522 else 523 icp->icmp_mask = icmpmaskfake; 524 if (ip->ip_src.s_addr == 0) { 525 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 526 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 527 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 528 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 529 } 530 reflect: 531 ip->ip_len += hlen; /* since ip_input deducts this */ 532 V_icmpstat.icps_reflect++; 533 V_icmpstat.icps_outhist[icp->icmp_type]++; 534 icmp_reflect(m); 535 return; 536 537 case ICMP_REDIRECT: 538 if (log_redirect) { 539 u_long src, dst, gw; 540 541 src = ntohl(ip->ip_src.s_addr); 542 dst = ntohl(icp->icmp_ip.ip_dst.s_addr); 543 gw = ntohl(icp->icmp_gwaddr.s_addr); 544 printf("icmp redirect from %d.%d.%d.%d: " 545 "%d.%d.%d.%d => %d.%d.%d.%d\n", 546 (int)(src >> 24), (int)((src >> 16) & 0xff), 547 (int)((src >> 8) & 0xff), (int)(src & 0xff), 548 (int)(dst >> 24), (int)((dst >> 16) & 0xff), 549 (int)((dst >> 8) & 0xff), (int)(dst & 0xff), 550 (int)(gw >> 24), (int)((gw >> 16) & 0xff), 551 (int)((gw >> 8) & 0xff), (int)(gw & 0xff)); 552 } 553 /* 554 * RFC1812 says we must ignore ICMP redirects if we 555 * are acting as router. 556 */ 557 if (drop_redirect || V_ipforwarding) 558 break; 559 if (code > 3) 560 goto badcode; 561 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 562 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 563 V_icmpstat.icps_badlen++; 564 break; 565 } 566 /* 567 * Short circuit routing redirects to force 568 * immediate change in the kernel's routing 569 * tables. The message is also handed to anyone 570 * listening on a raw socket (e.g. the routing 571 * daemon for use in updating its tables). 572 */ 573 icmpgw.sin_addr = ip->ip_src; 574 icmpdst.sin_addr = icp->icmp_gwaddr; 575 #ifdef ICMPPRINTFS 576 if (icmpprintfs) { 577 char buf[4 * sizeof "123"]; 578 strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst)); 579 580 printf("redirect dst %s to %s\n", 581 buf, inet_ntoa(icp->icmp_gwaddr)); 582 } 583 #endif 584 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 585 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) { 586 in_rtredirect((struct sockaddr *)&icmpsrc, 587 (struct sockaddr *)&icmpdst, 588 (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST, 589 (struct sockaddr *)&icmpgw, fibnum); 590 } 591 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); 592 #ifdef IPSEC 593 key_sa_routechange((struct sockaddr *)&icmpsrc); 594 #endif 595 break; 596 597 /* 598 * No kernel processing for the following; 599 * just fall through to send to raw listener. 600 */ 601 case ICMP_ECHOREPLY: 602 case ICMP_ROUTERADVERT: 603 case ICMP_ROUTERSOLICIT: 604 case ICMP_TSTAMPREPLY: 605 case ICMP_IREQREPLY: 606 case ICMP_MASKREPLY: 607 default: 608 break; 609 } 610 611 raw: 612 rip_input(m, off); 613 return; 614 615 freeit: 616 m_freem(m); 617 } 618 619 /* 620 * Reflect the ip packet back to the source 621 */ 622 static void 623 icmp_reflect(struct mbuf *m) 624 { 625 struct ip *ip = mtod(m, struct ip *); 626 struct ifaddr *ifa; 627 struct ifnet *ifn; 628 struct in_ifaddr *ia; 629 struct in_addr t; 630 struct mbuf *opts = 0; 631 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 632 633 if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) || 634 IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) || 635 IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) { 636 m_freem(m); /* Bad return address */ 637 V_icmpstat.icps_badaddr++; 638 goto done; /* Ip_output() will check for broadcast */ 639 } 640 641 t = ip->ip_dst; 642 ip->ip_dst = ip->ip_src; 643 644 /* 645 * Source selection for ICMP replies: 646 * 647 * If the incoming packet was addressed directly to one of our 648 * own addresses, use dst as the src for the reply. 649 */ 650 LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) 651 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) 652 goto match; 653 /* 654 * If the incoming packet was addressed to one of our broadcast 655 * addresses, use the first non-broadcast address which corresponds 656 * to the incoming interface. 657 */ 658 if (m->m_pkthdr.rcvif != NULL && 659 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 660 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 661 if (ifa->ifa_addr->sa_family != AF_INET) 662 continue; 663 ia = ifatoia(ifa); 664 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 665 t.s_addr) 666 goto match; 667 } 668 } 669 /* 670 * If the packet was transiting through us, use the address of 671 * the interface the packet came through in. If that interface 672 * doesn't have a suitable IP address, the normal selection 673 * criteria apply. 674 */ 675 if (icmp_rfi && m->m_pkthdr.rcvif != NULL) { 676 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 677 if (ifa->ifa_addr->sa_family != AF_INET) 678 continue; 679 ia = ifatoia(ifa); 680 goto match; 681 } 682 } 683 /* 684 * If the incoming packet was not addressed directly to us, use 685 * designated interface for icmp replies specified by sysctl 686 * net.inet.icmp.reply_src (default not set). Otherwise continue 687 * with normal source selection. 688 */ 689 if (reply_src[0] != '\0' && (ifn = ifunit(reply_src))) { 690 TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) { 691 if (ifa->ifa_addr->sa_family != AF_INET) 692 continue; 693 ia = ifatoia(ifa); 694 goto match; 695 } 696 } 697 /* 698 * If the packet was transiting through us, use the address of 699 * the interface that is the closest to the packet source. 700 * When we don't have a route back to the packet source, stop here 701 * and drop the packet. 702 */ 703 ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m)); 704 if (ia == NULL) { 705 m_freem(m); 706 V_icmpstat.icps_noroute++; 707 goto done; 708 } 709 match: 710 #ifdef MAC 711 mac_netinet_icmp_replyinplace(m); 712 #endif 713 t = IA_SIN(ia)->sin_addr; 714 ip->ip_src = t; 715 ip->ip_ttl = V_ip_defttl; 716 717 if (optlen > 0) { 718 register u_char *cp; 719 int opt, cnt; 720 u_int len; 721 722 /* 723 * Retrieve any source routing from the incoming packet; 724 * add on any record-route or timestamp options. 725 */ 726 cp = (u_char *) (ip + 1); 727 if ((opts = ip_srcroute(m)) == 0 && 728 (opts = m_gethdr(M_DONTWAIT, MT_DATA))) { 729 opts->m_len = sizeof(struct in_addr); 730 mtod(opts, struct in_addr *)->s_addr = 0; 731 } 732 if (opts) { 733 #ifdef ICMPPRINTFS 734 if (icmpprintfs) 735 printf("icmp_reflect optlen %d rt %d => ", 736 optlen, opts->m_len); 737 #endif 738 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 739 opt = cp[IPOPT_OPTVAL]; 740 if (opt == IPOPT_EOL) 741 break; 742 if (opt == IPOPT_NOP) 743 len = 1; 744 else { 745 if (cnt < IPOPT_OLEN + sizeof(*cp)) 746 break; 747 len = cp[IPOPT_OLEN]; 748 if (len < IPOPT_OLEN + sizeof(*cp) || 749 len > cnt) 750 break; 751 } 752 /* 753 * Should check for overflow, but it "can't happen" 754 */ 755 if (opt == IPOPT_RR || opt == IPOPT_TS || 756 opt == IPOPT_SECURITY) { 757 bcopy((caddr_t)cp, 758 mtod(opts, caddr_t) + opts->m_len, len); 759 opts->m_len += len; 760 } 761 } 762 /* Terminate & pad, if necessary */ 763 cnt = opts->m_len % 4; 764 if (cnt) { 765 for (; cnt < 4; cnt++) { 766 *(mtod(opts, caddr_t) + opts->m_len) = 767 IPOPT_EOL; 768 opts->m_len++; 769 } 770 } 771 #ifdef ICMPPRINTFS 772 if (icmpprintfs) 773 printf("%d\n", opts->m_len); 774 #endif 775 } 776 /* 777 * Now strip out original options by copying rest of first 778 * mbuf's data back, and adjust the IP length. 779 */ 780 ip->ip_len -= optlen; 781 ip->ip_v = IPVERSION; 782 ip->ip_hl = 5; 783 m->m_len -= optlen; 784 if (m->m_flags & M_PKTHDR) 785 m->m_pkthdr.len -= optlen; 786 optlen += sizeof(struct ip); 787 bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1), 788 (unsigned)(m->m_len - sizeof(struct ip))); 789 } 790 m_tag_delete_nonpersistent(m); 791 m->m_flags &= ~(M_BCAST|M_MCAST); 792 icmp_send(m, opts); 793 done: 794 if (opts) 795 (void)m_free(opts); 796 } 797 798 /* 799 * Send an icmp packet back to the ip level, 800 * after supplying a checksum. 801 */ 802 static void 803 icmp_send(struct mbuf *m, struct mbuf *opts) 804 { 805 register struct ip *ip = mtod(m, struct ip *); 806 register int hlen; 807 register struct icmp *icp; 808 809 hlen = ip->ip_hl << 2; 810 m->m_data += hlen; 811 m->m_len -= hlen; 812 icp = mtod(m, struct icmp *); 813 icp->icmp_cksum = 0; 814 icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen); 815 m->m_data -= hlen; 816 m->m_len += hlen; 817 m->m_pkthdr.rcvif = (struct ifnet *)0; 818 #ifdef ICMPPRINTFS 819 if (icmpprintfs) { 820 char buf[4 * sizeof "123"]; 821 strcpy(buf, inet_ntoa(ip->ip_dst)); 822 printf("icmp_send dst %s src %s\n", 823 buf, inet_ntoa(ip->ip_src)); 824 } 825 #endif 826 (void) ip_output(m, opts, NULL, 0, NULL, NULL); 827 } 828 829 n_time 830 iptime(void) 831 { 832 struct timeval atv; 833 u_long t; 834 835 getmicrotime(&atv); 836 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 837 return (htonl(t)); 838 } 839 840 /* 841 * Return the next larger or smaller MTU plateau (table from RFC 1191) 842 * given current value MTU. If DIR is less than zero, a larger plateau 843 * is returned; otherwise, a smaller value is returned. 844 */ 845 int 846 ip_next_mtu(int mtu, int dir) 847 { 848 static int mtutab[] = { 849 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508, 850 296, 68, 0 851 }; 852 int i, size; 853 854 size = (sizeof mtutab) / (sizeof mtutab[0]); 855 if (dir >= 0) { 856 for (i = 0; i < size; i++) 857 if (mtu > mtutab[i]) 858 return mtutab[i]; 859 } else { 860 for (i = size - 1; i >= 0; i--) 861 if (mtu < mtutab[i]) 862 return mtutab[i]; 863 if (mtu == mtutab[0]) 864 return mtutab[0]; 865 } 866 return 0; 867 } 868 869 870 /* 871 * badport_bandlim() - check for ICMP bandwidth limit 872 * 873 * Return 0 if it is ok to send an ICMP error response, -1 if we have 874 * hit our bandwidth limit and it is not ok. 875 * 876 * If icmplim is <= 0, the feature is disabled and 0 is returned. 877 * 878 * For now we separate the TCP and UDP subsystems w/ different 'which' 879 * values. We may eventually remove this separation (and simplify the 880 * code further). 881 * 882 * Note that the printing of the error message is delayed so we can 883 * properly print the icmp error rate that the system was trying to do 884 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 885 * the 'final' error, but it doesn't make sense to solve the printing 886 * delay with more complex code. 887 */ 888 889 int 890 badport_bandlim(int which) 891 { 892 #define N(a) (sizeof (a) / sizeof (a[0])) 893 static struct rate { 894 const char *type; 895 struct timeval lasttime; 896 int curpps; 897 } rates[BANDLIM_MAX+1] = { 898 { "icmp unreach response" }, 899 { "icmp ping response" }, 900 { "icmp tstamp response" }, 901 { "closed port RST response" }, 902 { "open port RST response" }, 903 { "icmp6 unreach response" } 904 }; 905 906 /* 907 * Return ok status if feature disabled or argument out of range. 908 */ 909 if (icmplim > 0 && (u_int) which < N(rates)) { 910 struct rate *r = &rates[which]; 911 int opps = r->curpps; 912 913 if (!ppsratecheck(&r->lasttime, &r->curpps, icmplim)) 914 return -1; /* discard packet */ 915 /* 916 * If we've dropped below the threshold after having 917 * rate-limited traffic print the message. This preserves 918 * the previous behaviour at the expense of added complexity. 919 */ 920 if (icmplim_output && opps > icmplim) 921 printf("Limiting %s from %d to %d packets/sec\n", 922 r->type, opps, icmplim); 923 } 924 return 0; /* okay to send packet */ 925 #undef N 926 } 927