ip_icmp.c revision 1.167
1 /* $NetBSD: ip_icmp.c,v 1.167 2018/02/05 08:38:06 maxv Exp $ */ 2 3 /* 4 * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Public Access Networks Corporation ("Panix"). It was developed under 9 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 10 * 11 * This code is derived from software contributed to The NetBSD Foundation 12 * by Jason R. Thorpe of Zembu Labs, Inc. 13 * 14 * Redistribution and use in source and binary forms, with or without 15 * modification, are permitted provided that the following conditions 16 * are met: 17 * 1. Redistributions of source code must retain the above copyright 18 * notice, this list of conditions and the following disclaimer. 19 * 2. Redistributions in binary form must reproduce the above copyright 20 * notice, this list of conditions and the following disclaimer in the 21 * documentation and/or other materials provided with the distribution. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 24 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 25 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 26 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 27 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 30 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 31 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 32 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 33 * POSSIBILITY OF SUCH DAMAGE. 34 */ 35 36 /* 37 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 38 * All rights reserved. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. Neither the name of the project nor the names of its contributors 49 * may be used to endorse or promote products derived from this software 50 * without specific prior written permission. 51 * 52 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 55 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 62 * SUCH DAMAGE. 63 */ 64 65 /* 66 * Copyright (c) 1982, 1986, 1988, 1993 67 * The Regents of the University of California. All rights reserved. 68 * 69 * Redistribution and use in source and binary forms, with or without 70 * modification, are permitted provided that the following conditions 71 * are met: 72 * 1. Redistributions of source code must retain the above copyright 73 * notice, this list of conditions and the following disclaimer. 74 * 2. Redistributions in binary form must reproduce the above copyright 75 * notice, this list of conditions and the following disclaimer in the 76 * documentation and/or other materials provided with the distribution. 77 * 3. Neither the name of the University nor the names of its contributors 78 * may be used to endorse or promote products derived from this software 79 * without specific prior written permission. 80 * 81 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 82 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 83 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 84 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 85 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 86 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 87 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 88 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 89 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 90 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 91 * SUCH DAMAGE. 92 * 93 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 94 */ 95 96 #include <sys/cdefs.h> 97 __KERNEL_RCSID(0, "$NetBSD: ip_icmp.c,v 1.167 2018/02/05 08:38:06 maxv Exp $"); 98 99 #ifdef _KERNEL_OPT 100 #include "opt_ipsec.h" 101 #endif 102 103 #include <sys/param.h> 104 #include <sys/systm.h> 105 #include <sys/mbuf.h> 106 #include <sys/protosw.h> 107 #include <sys/socket.h> 108 #include <sys/socketvar.h> /* For softnet_lock */ 109 #include <sys/kmem.h> 110 #include <sys/time.h> 111 #include <sys/kernel.h> 112 #include <sys/syslog.h> 113 #include <sys/sysctl.h> 114 115 #include <net/if.h> 116 #include <net/route.h> 117 118 #include <netinet/in.h> 119 #include <netinet/in_systm.h> 120 #include <netinet/in_var.h> 121 #include <netinet/ip.h> 122 #include <netinet/ip_icmp.h> 123 #include <netinet/ip_var.h> 124 #include <netinet/in_pcb.h> 125 #include <netinet/in_proto.h> 126 #include <netinet/icmp_var.h> 127 #include <netinet/icmp_private.h> 128 #include <netinet/wqinput.h> 129 130 #ifdef IPSEC 131 #include <netipsec/ipsec.h> 132 #include <netipsec/key.h> 133 #endif 134 135 /* 136 * ICMP routines: error generation, receive packet processing, and 137 * routines to turnaround packets back to the originator, and 138 * host table maintenance routines. 139 */ 140 141 int icmpmaskrepl = 0; 142 int icmpbmcastecho = 0; 143 #ifdef ICMPPRINTFS 144 int icmpprintfs = 0; 145 #endif 146 int icmpreturndatabytes = 8; 147 148 percpu_t *icmpstat_percpu; 149 150 /* 151 * List of callbacks to notify when Path MTU changes are made. 152 */ 153 struct icmp_mtudisc_callback { 154 LIST_ENTRY(icmp_mtudisc_callback) mc_list; 155 void (*mc_func)(struct in_addr); 156 }; 157 158 LIST_HEAD(, icmp_mtudisc_callback) icmp_mtudisc_callbacks = 159 LIST_HEAD_INITIALIZER(&icmp_mtudisc_callbacks); 160 161 /* unused... */ 162 u_int ip_next_mtu(u_int, int); 163 164 static int icmperrppslim = 100; /* 100pps */ 165 static int icmperrpps_count = 0; 166 static struct timeval icmperrppslim_last; 167 static int icmp_rediraccept = 1; 168 static int icmp_redirtimeout = 600; 169 static struct rttimer_queue *icmp_redirect_timeout_q = NULL; 170 171 /* Protect mtudisc and redirect stuff */ 172 static kmutex_t icmp_mtx __cacheline_aligned; 173 174 static void icmp_send(struct mbuf *, struct mbuf *); 175 static void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *); 176 static void icmp_redirect_timeout(struct rtentry *, struct rttimer *); 177 178 static void sysctl_netinet_icmp_setup(struct sysctllog **); 179 180 /* workqueue-based pr_input */ 181 static struct wqinput *icmp_wqinput; 182 static void _icmp_input(struct mbuf *, int, int); 183 184 void 185 icmp_init(void) 186 { 187 188 sysctl_netinet_icmp_setup(NULL); 189 190 mutex_init(&icmp_mtx, MUTEX_DEFAULT, IPL_NONE); 191 /* 192 * This is only useful if the user initializes redirtimeout to 193 * something other than zero. 194 */ 195 mutex_enter(&icmp_mtx); 196 icmp_redirect_timeout_q = rt_timer_queue_create(icmp_redirtimeout); 197 mutex_exit(&icmp_mtx); 198 199 icmpstat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP_NSTATS); 200 icmp_wqinput = wqinput_create("icmp", _icmp_input); 201 } 202 203 void 204 icmp_mtudisc_lock(void) 205 { 206 207 mutex_enter(&icmp_mtx); 208 } 209 210 void 211 icmp_mtudisc_unlock(void) 212 { 213 214 mutex_exit(&icmp_mtx); 215 } 216 217 /* 218 * Register a Path MTU Discovery callback. 219 */ 220 void 221 icmp_mtudisc_callback_register(void (*func)(struct in_addr)) 222 { 223 struct icmp_mtudisc_callback *mc, *new; 224 225 new = kmem_alloc(sizeof(*mc), KM_SLEEP); 226 227 mutex_enter(&icmp_mtx); 228 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL; 229 mc = LIST_NEXT(mc, mc_list)) { 230 if (mc->mc_func == func) { 231 mutex_exit(&icmp_mtx); 232 kmem_free(new, sizeof(*mc)); 233 return; 234 } 235 } 236 237 new->mc_func = func; 238 LIST_INSERT_HEAD(&icmp_mtudisc_callbacks, new, mc_list); 239 mutex_exit(&icmp_mtx); 240 } 241 242 /* 243 * Generate an error packet of type error in response to a bad IP packet. 'n' 244 * contains this packet. We create 'm' and send it. 245 * 246 * As we are not required to return everything we have, we return whatever 247 * we can return at ease. 248 * 249 * Note that ICMP datagrams longer than 576 octets are out of spec according 250 * to RFC1812; the limit on icmpreturndatabytes will keep things below that 251 * limit. 252 */ 253 void 254 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu) 255 { 256 struct ip *oip = mtod(n, struct ip *), *nip; 257 const unsigned oiphlen = oip->ip_hl << 2; 258 struct icmp *icp; 259 struct mbuf *m; 260 struct m_tag *mtag; 261 unsigned datalen, mblen; 262 int totlen; 263 264 #ifdef ICMPPRINTFS 265 if (icmpprintfs) 266 printf("icmp_error(%p, type:%d, code:%d)\n", oip, type, code); 267 #endif 268 269 if (type != ICMP_REDIRECT) 270 ICMP_STATINC(ICMP_STAT_ERROR); 271 272 /* 273 * Don't send error if: 274 * - The original packet was encrypted. 275 * - The packet is multicast or broadcast. 276 * - The packet is not the first fragment of the message. 277 * - The packet is an ICMP message with an unknown type. 278 */ 279 if (n->m_flags & M_DECRYPTED) 280 goto freeit; 281 if (n->m_flags & (M_BCAST|M_MCAST)) 282 goto freeit; 283 if (oip->ip_off &~ htons(IP_MF|IP_DF)) 284 goto freeit; 285 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 286 n->m_len >= oiphlen + ICMP_MINLEN) { 287 struct icmp *oicp = (struct icmp *)((char *)oip + oiphlen); 288 if (!ICMP_INFOTYPE(oicp->icmp_type)) { 289 ICMP_STATINC(ICMP_STAT_OLDICMP); 290 goto freeit; 291 } 292 } 293 294 /* 295 * First, do a rate limitation check. 296 */ 297 if (icmp_ratelimit(&oip->ip_src, type, code)) { 298 /* XXX stat */ 299 goto freeit; 300 } 301 302 /* 303 * Compute the number of bytes we will put in 'icmp_ip'. Truncate 304 * it to the size of the mbuf, if it's too big. 305 */ 306 datalen = oiphlen + min(icmpreturndatabytes, 307 ntohs(oip->ip_len) - oiphlen); 308 mblen = 0; 309 for (m = n; m && (mblen < datalen); m = m->m_next) 310 mblen += m->m_len; 311 datalen = min(mblen, datalen); 312 313 /* 314 * Compute the total length of the new packet. Truncate it if it's 315 * bigger than the size of a cluster. 316 */ 317 CTASSERT(ICMP_MINLEN + sizeof(struct ip) <= MCLBYTES); 318 totlen = sizeof(struct ip) + ICMP_MINLEN + datalen; 319 if (totlen > MCLBYTES) { 320 datalen = MCLBYTES - ICMP_MINLEN - sizeof(struct ip); 321 totlen = MCLBYTES; 322 } 323 324 /* 325 * Allocate the mbuf for the new packet. 326 */ 327 m = m_gethdr(M_DONTWAIT, MT_HEADER); 328 if (m && (totlen > MHLEN)) { 329 MCLGET(m, M_DONTWAIT); 330 if ((m->m_flags & M_EXT) == 0) { 331 m_freem(m); 332 m = NULL; 333 } 334 } 335 if (m == NULL) 336 goto freeit; 337 MCLAIM(m, n->m_owner); 338 m->m_len = totlen; 339 m->m_pkthdr.len = m->m_len; 340 m_copy_rcvif(m, n); 341 342 if ((u_int)type > ICMP_MAXTYPE) 343 panic("icmp_error"); 344 ICMP_STATINC(ICMP_STAT_OUTHIST + type); 345 346 if ((m->m_flags & M_EXT) == 0) 347 MH_ALIGN(m, m->m_len); 348 349 /* 350 * Get pointers on the IP header and the ICMP header. 351 */ 352 nip = mtod(m, struct ip *); 353 icp = (struct icmp *)(nip + 1); 354 355 /* 356 * Fill in the fields of the ICMP header: icmp_type, icmp_code 357 * and icmp_ip. icmp_cksum gets filled later. 358 */ 359 icp->icmp_type = type; 360 if (type == ICMP_REDIRECT) { 361 icp->icmp_gwaddr.s_addr = dest; 362 } else { 363 icp->icmp_void = 0; 364 /* 365 * The following assignments assume an overlay with the 366 * zeroed icmp_void field. 367 */ 368 if (type == ICMP_PARAMPROB) { 369 icp->icmp_pptr = code; 370 code = 0; 371 } else if (type == ICMP_UNREACH && 372 code == ICMP_UNREACH_NEEDFRAG && destmtu) 373 icp->icmp_nextmtu = htons(destmtu); 374 } 375 icp->icmp_code = code; 376 m_copydata(n, 0, datalen, (void *)&icp->icmp_ip); 377 378 /* 379 * Now, copy the old IP header (without options) in front of the 380 * ICMP message. The src/dst fields will be swapped in icmp_reflect. 381 */ 382 /* ip_v set in ip_output */ 383 nip->ip_hl = sizeof(struct ip) >> 2; 384 nip->ip_tos = 0; 385 nip->ip_len = htons(m->m_len); 386 /* ip_id set in ip_output */ 387 nip->ip_off = htons(0); 388 /* ip_ttl set in icmp_reflect */ 389 nip->ip_p = IPPROTO_ICMP; 390 nip->ip_src = oip->ip_src; 391 nip->ip_dst = oip->ip_dst; 392 /* move PF m_tag to new packet, if it exists */ 393 mtag = m_tag_find(n, PACKET_TAG_PF, NULL); 394 if (mtag != NULL) { 395 m_tag_unlink(n, mtag); 396 m_tag_prepend(m, mtag); 397 } 398 399 icmp_reflect(m); 400 401 freeit: 402 m_freem(n); 403 } 404 405 struct sockaddr_in icmpsrc = { 406 .sin_len = sizeof(struct sockaddr_in), 407 .sin_family = AF_INET, 408 }; 409 410 /* 411 * Process a received ICMP message. 412 */ 413 static void 414 _icmp_input(struct mbuf *m, int hlen, int proto) 415 { 416 struct icmp *icp; 417 struct ip *ip = mtod(m, struct ip *); 418 int icmplen; 419 int i; 420 struct in_ifaddr *ia; 421 void *(*ctlfunc)(int, const struct sockaddr *, void *); 422 int code; 423 struct rtentry *rt; 424 struct sockaddr_in icmpdst = { 425 .sin_len = sizeof(struct sockaddr_in), 426 .sin_family = AF_INET, 427 }; 428 struct sockaddr_in icmpgw = { 429 .sin_len = sizeof(struct sockaddr_in), 430 .sin_family = AF_INET, 431 }; 432 433 /* 434 * Locate icmp structure in mbuf, and check 435 * that not corrupted and of at least minimum length. 436 */ 437 icmplen = ntohs(ip->ip_len) - hlen; 438 #ifdef ICMPPRINTFS 439 if (icmpprintfs) { 440 char sbuf[INET_ADDRSTRLEN], dbuf[INET_ADDRSTRLEN]; 441 printf("icmp_input from `%s' to `%s', len %d\n", 442 IN_PRINT(sbuf, &ip->ip_src), IN_PRINT(dbuf, &ip->ip_dst), 443 icmplen); 444 } 445 #endif 446 if (icmplen < ICMP_MINLEN) { 447 ICMP_STATINC(ICMP_STAT_TOOSHORT); 448 goto freeit; 449 } 450 i = hlen + min(icmplen, ICMP_ADVLENMIN); 451 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == NULL) { 452 ICMP_STATINC(ICMP_STAT_TOOSHORT); 453 return; 454 } 455 ip = mtod(m, struct ip *); 456 m->m_len -= hlen; 457 m->m_data += hlen; 458 icp = mtod(m, struct icmp *); 459 /* Don't need to assert alignment, here. */ 460 if (in_cksum(m, icmplen)) { 461 ICMP_STATINC(ICMP_STAT_CHECKSUM); 462 goto freeit; 463 } 464 m->m_len += hlen; 465 m->m_data -= hlen; 466 467 #ifdef ICMPPRINTFS 468 /* 469 * Message type specific processing. 470 */ 471 if (icmpprintfs) 472 printf("icmp_input(type:%d, code:%d)\n", icp->icmp_type, 473 icp->icmp_code); 474 #endif 475 if (icp->icmp_type > ICMP_MAXTYPE) 476 goto raw; 477 ICMP_STATINC(ICMP_STAT_INHIST + icp->icmp_type); 478 code = icp->icmp_code; 479 480 switch (icp->icmp_type) { 481 case ICMP_UNREACH: 482 switch (code) { 483 case ICMP_UNREACH_PROTOCOL: 484 code = PRC_UNREACH_PROTOCOL; 485 break; 486 487 case ICMP_UNREACH_PORT: 488 code = PRC_UNREACH_PORT; 489 break; 490 491 case ICMP_UNREACH_SRCFAIL: 492 code = PRC_UNREACH_SRCFAIL; 493 break; 494 495 case ICMP_UNREACH_NEEDFRAG: 496 code = PRC_MSGSIZE; 497 break; 498 499 case ICMP_UNREACH_NET: 500 case ICMP_UNREACH_NET_UNKNOWN: 501 case ICMP_UNREACH_NET_PROHIB: 502 case ICMP_UNREACH_TOSNET: 503 code = PRC_UNREACH_NET; 504 break; 505 506 case ICMP_UNREACH_HOST: 507 case ICMP_UNREACH_HOST_UNKNOWN: 508 case ICMP_UNREACH_ISOLATED: 509 case ICMP_UNREACH_HOST_PROHIB: 510 case ICMP_UNREACH_TOSHOST: 511 case ICMP_UNREACH_ADMIN_PROHIBIT: 512 case ICMP_UNREACH_HOST_PREC: 513 case ICMP_UNREACH_PREC_CUTOFF: 514 code = PRC_UNREACH_HOST; 515 break; 516 517 default: 518 goto badcode; 519 } 520 goto deliver; 521 522 case ICMP_TIMXCEED: 523 if (code > 1) 524 goto badcode; 525 code += PRC_TIMXCEED_INTRANS; 526 goto deliver; 527 528 case ICMP_PARAMPROB: 529 if (code > 1) 530 goto badcode; 531 code = PRC_PARAMPROB; 532 goto deliver; 533 534 case ICMP_SOURCEQUENCH: 535 if (code) 536 goto badcode; 537 code = PRC_QUENCH; 538 goto deliver; 539 540 deliver: 541 /* 542 * Problem with datagram; advise higher level routines. 543 */ 544 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 545 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 546 ICMP_STATINC(ICMP_STAT_BADLEN); 547 goto freeit; 548 } 549 if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr)) 550 goto badcode; 551 #ifdef ICMPPRINTFS 552 if (icmpprintfs) 553 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 554 #endif 555 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 556 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 557 if (ctlfunc) 558 (void) (*ctlfunc)(code, sintosa(&icmpsrc), 559 &icp->icmp_ip); 560 break; 561 562 badcode: 563 ICMP_STATINC(ICMP_STAT_BADCODE); 564 break; 565 566 case ICMP_ECHO: 567 if (!icmpbmcastecho && 568 (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 569 ICMP_STATINC(ICMP_STAT_BMCASTECHO); 570 break; 571 } 572 icp->icmp_type = ICMP_ECHOREPLY; 573 goto reflect; 574 575 case ICMP_TSTAMP: 576 if (icmplen < ICMP_TSLEN) { 577 ICMP_STATINC(ICMP_STAT_BADLEN); 578 break; 579 } 580 if (!icmpbmcastecho && 581 (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 582 ICMP_STATINC(ICMP_STAT_BMCASTTSTAMP); 583 break; 584 } 585 icp->icmp_type = ICMP_TSTAMPREPLY; 586 icp->icmp_rtime = iptime(); 587 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 588 goto reflect; 589 590 case ICMP_MASKREQ: { 591 struct ifnet *rcvif; 592 int s, ss; 593 struct ifaddr *ifa = NULL; 594 595 if (icmpmaskrepl == 0) 596 break; 597 /* 598 * We are not able to respond with all ones broadcast 599 * unless we receive it over a point-to-point interface. 600 */ 601 if (icmplen < ICMP_MASKLEN) { 602 ICMP_STATINC(ICMP_STAT_BADLEN); 603 break; 604 } 605 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 606 in_nullhost(ip->ip_dst)) 607 icmpdst.sin_addr = ip->ip_src; 608 else 609 icmpdst.sin_addr = ip->ip_dst; 610 ss = pserialize_read_enter(); 611 rcvif = m_get_rcvif(m, &s); 612 if (__predict_true(rcvif != NULL)) 613 ifa = ifaof_ifpforaddr(sintosa(&icmpdst), rcvif); 614 m_put_rcvif(rcvif, &s); 615 if (ifa == NULL) { 616 pserialize_read_exit(ss); 617 break; 618 } 619 ia = ifatoia(ifa); 620 icp->icmp_type = ICMP_MASKREPLY; 621 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 622 if (in_nullhost(ip->ip_src)) { 623 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 624 ip->ip_src = ia->ia_broadaddr.sin_addr; 625 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 626 ip->ip_src = ia->ia_dstaddr.sin_addr; 627 } 628 pserialize_read_exit(ss); 629 reflect: 630 { 631 uint64_t *icps = percpu_getref(icmpstat_percpu); 632 icps[ICMP_STAT_REFLECT]++; 633 icps[ICMP_STAT_OUTHIST + icp->icmp_type]++; 634 percpu_putref(icmpstat_percpu); 635 } 636 icmp_reflect(m); 637 return; 638 } 639 640 case ICMP_REDIRECT: 641 if (code > 3) 642 goto badcode; 643 if (icmp_rediraccept == 0) 644 goto freeit; 645 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 646 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 647 ICMP_STATINC(ICMP_STAT_BADLEN); 648 break; 649 } 650 /* 651 * Short circuit routing redirects to force 652 * immediate change in the kernel's routing 653 * tables. The message is also handed to anyone 654 * listening on a raw socket (e.g. the routing 655 * daemon for use in updating its tables). 656 */ 657 icmpgw.sin_addr = ip->ip_src; 658 icmpdst.sin_addr = icp->icmp_gwaddr; 659 #ifdef ICMPPRINTFS 660 if (icmpprintfs) { 661 char gbuf[INET_ADDRSTRLEN], dbuf[INET_ADDRSTRLEN]; 662 printf("redirect dst `%s' to `%s'\n", 663 IN_PRINT(dbuf, &icp->icmp_ip.ip_dst), 664 IN_PRINT(gbuf, &icp->icmp_gwaddr)); 665 } 666 #endif 667 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 668 rt = NULL; 669 rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst), 670 NULL, RTF_GATEWAY | RTF_HOST, sintosa(&icmpgw), &rt); 671 mutex_enter(&icmp_mtx); 672 if (rt != NULL && icmp_redirtimeout != 0) { 673 i = rt_timer_add(rt, icmp_redirect_timeout, 674 icmp_redirect_timeout_q); 675 if (i) { 676 char buf[INET_ADDRSTRLEN]; 677 log(LOG_ERR, "ICMP: redirect failed to " 678 "register timeout for route to %s, " 679 "code %d\n", 680 IN_PRINT(buf, &icp->icmp_ip.ip_dst), i); 681 } 682 } 683 mutex_exit(&icmp_mtx); 684 if (rt != NULL) 685 rt_unref(rt); 686 687 pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc)); 688 #if defined(IPSEC) 689 if (ipsec_used) 690 key_sa_routechange((struct sockaddr *)&icmpsrc); 691 #endif 692 break; 693 694 /* 695 * No kernel processing for the following; 696 * just fall through to send to raw listener. 697 */ 698 case ICMP_ECHOREPLY: 699 case ICMP_ROUTERADVERT: 700 case ICMP_ROUTERSOLICIT: 701 case ICMP_TSTAMPREPLY: 702 case ICMP_IREQREPLY: 703 case ICMP_MASKREPLY: 704 default: 705 break; 706 } 707 708 raw: 709 rip_input(m, hlen, proto); 710 return; 711 712 freeit: 713 m_freem(m); 714 return; 715 } 716 717 void 718 icmp_input(struct mbuf *m, ...) 719 { 720 int hlen, proto; 721 va_list ap; 722 723 va_start(ap, m); 724 hlen = va_arg(ap, int); 725 proto = va_arg(ap, int); 726 va_end(ap); 727 728 wqinput_input(icmp_wqinput, m, hlen, proto); 729 } 730 731 /* 732 * Reflect the ip packet back to the source 733 */ 734 void 735 icmp_reflect(struct mbuf *m) 736 { 737 struct ip *ip = mtod(m, struct ip *); 738 struct in_ifaddr *ia; 739 struct ifaddr *ifa; 740 struct sockaddr_in *sin; 741 struct in_addr t; 742 struct mbuf *opts = NULL; 743 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 744 struct ifnet *rcvif; 745 struct psref psref, psref_ia; 746 int s; 747 int bound; 748 749 bound = curlwp_bind(); 750 751 if (!in_canforward(ip->ip_src) && 752 ((ip->ip_src.s_addr & IN_CLASSA_NET) != 753 htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { 754 m_freem(m); /* Bad return address */ 755 goto done; /* ip_output() will check for broadcast */ 756 } 757 t = ip->ip_dst; 758 ip->ip_dst = ip->ip_src; 759 760 /* 761 * If the incoming packet was addressed directly to us, use 762 * dst as the src for the reply. Otherwise (broadcast or 763 * anonymous), use an address which corresponds to the 764 * incoming interface, with a preference for the address which 765 * corresponds to the route to the destination of the ICMP. 766 */ 767 768 /* Look for packet addressed to us */ 769 ia = in_get_ia_psref(t, &psref_ia); 770 if (ia && (ia->ia4_flags & IN_IFF_NOTREADY)) { 771 ia4_release(ia, &psref_ia); 772 ia = NULL; 773 } 774 775 rcvif = m_get_rcvif_psref(m, &psref); 776 777 /* look for packet sent to broadcast address */ 778 if (ia == NULL && rcvif && 779 (rcvif->if_flags & IFF_BROADCAST)) { 780 s = pserialize_read_enter(); 781 IFADDR_READER_FOREACH(ifa, rcvif) { 782 if (ifa->ifa_addr->sa_family != AF_INET) 783 continue; 784 if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) { 785 ia = ifatoia(ifa); 786 if ((ia->ia4_flags & IN_IFF_NOTREADY) == 0) 787 break; 788 ia = NULL; 789 } 790 } 791 if (ia != NULL) 792 ia4_acquire(ia, &psref_ia); 793 pserialize_read_exit(s); 794 } 795 796 sin = ia ? &ia->ia_addr : NULL; 797 798 /* 799 * if the packet is addressed somewhere else, compute the 800 * source address for packets routed back to the source, and 801 * use that, if it's an address on the interface which 802 * received the packet 803 */ 804 if (sin == NULL && rcvif) { 805 struct sockaddr_in sin_dst; 806 struct route icmproute; 807 int errornum; 808 809 sockaddr_in_init(&sin_dst, &ip->ip_dst, 0); 810 memset(&icmproute, 0, sizeof(icmproute)); 811 errornum = 0; 812 ia = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum, 813 &psref_ia); 814 /* errornum is never used */ 815 rtcache_free(&icmproute); 816 /* check to make sure sin is a source address on rcvif */ 817 if (ia != NULL) { 818 sin = &ia->ia_addr; 819 t = sin->sin_addr; 820 sin = NULL; 821 ia4_release(ia, &psref_ia); 822 ia = in_get_ia_on_iface_psref(t, rcvif, &psref_ia); 823 if (ia != NULL) 824 sin = &ia->ia_addr; 825 } 826 } 827 828 /* 829 * if it was not addressed to us, but the route doesn't go out 830 * the source interface, pick an address on the source 831 * interface. This can happen when routing is asymmetric, or 832 * when the incoming packet was encapsulated 833 */ 834 if (sin == NULL && rcvif) { 835 KASSERT(ia == NULL); 836 s = pserialize_read_enter(); 837 IFADDR_READER_FOREACH(ifa, rcvif) { 838 if (ifa->ifa_addr->sa_family != AF_INET) 839 continue; 840 sin = &(ifatoia(ifa)->ia_addr); 841 ia = ifatoia(ifa); 842 ia4_acquire(ia, &psref_ia); 843 break; 844 } 845 pserialize_read_exit(s); 846 } 847 848 m_put_rcvif_psref(rcvif, &psref); 849 850 /* 851 * The following happens if the packet was not addressed to us, 852 * and was received on an interface with no IP address: 853 * We find the first AF_INET address on the first non-loopback 854 * interface. 855 */ 856 if (sin == NULL) { 857 KASSERT(ia == NULL); 858 s = pserialize_read_enter(); 859 IN_ADDRLIST_READER_FOREACH(ia) { 860 if (ia->ia_ifp->if_flags & IFF_LOOPBACK) 861 continue; 862 sin = &ia->ia_addr; 863 ia4_acquire(ia, &psref_ia); 864 break; 865 } 866 pserialize_read_exit(s); 867 } 868 869 /* 870 * If we still didn't find an address, punt. We could have an 871 * interface up (and receiving packets) with no address. 872 */ 873 if (sin == NULL) { 874 KASSERT(ia == NULL); 875 m_freem(m); 876 goto done; 877 } 878 879 ip->ip_src = sin->sin_addr; 880 ip->ip_ttl = MAXTTL; 881 882 if (ia != NULL) 883 ia4_release(ia, &psref_ia); 884 885 if (optlen > 0) { 886 u_char *cp; 887 int opt, cnt; 888 u_int len; 889 890 /* 891 * Retrieve any source routing from the incoming packet; 892 * add on any record-route or timestamp options. 893 */ 894 cp = (u_char *)(ip + 1); 895 if ((opts = ip_srcroute(m)) == NULL && 896 (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { 897 MCLAIM(opts, m->m_owner); 898 opts->m_len = sizeof(struct in_addr); 899 *mtod(opts, struct in_addr *) = zeroin_addr; 900 } 901 902 if (opts) { 903 #ifdef ICMPPRINTFS 904 if (icmpprintfs) 905 printf("icmp_reflect optlen %d rt %d => ", 906 optlen, opts->m_len); 907 #endif 908 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 909 opt = cp[IPOPT_OPTVAL]; 910 if (opt == IPOPT_EOL) 911 break; 912 if (opt == IPOPT_NOP) 913 len = 1; 914 else { 915 if (cnt < IPOPT_OLEN + sizeof(*cp)) 916 break; 917 len = cp[IPOPT_OLEN]; 918 if (len < IPOPT_OLEN + sizeof(*cp) || 919 len > cnt) 920 break; 921 } 922 923 /* Overflows can't happen */ 924 KASSERT(opts->m_len + len <= MHLEN); 925 926 if (opt == IPOPT_RR || opt == IPOPT_TS || 927 opt == IPOPT_SECURITY) { 928 memmove(mtod(opts, char *) + 929 opts->m_len, cp, len); 930 opts->m_len += len; 931 } 932 } 933 934 /* Terminate & pad, if necessary */ 935 if ((cnt = opts->m_len % 4) != 0) { 936 for (; cnt < 4; cnt++) { 937 *(mtod(opts, char *) + opts->m_len) = 938 IPOPT_EOL; 939 opts->m_len++; 940 } 941 } 942 #ifdef ICMPPRINTFS 943 if (icmpprintfs) 944 printf("%d\n", opts->m_len); 945 #endif 946 } 947 948 /* 949 * Now strip out original options by copying rest of first 950 * mbuf's data back, and adjust the IP length. 951 */ 952 ip->ip_len = htons(ntohs(ip->ip_len) - optlen); 953 ip->ip_hl = sizeof(struct ip) >> 2; 954 m->m_len -= optlen; 955 if (m->m_flags & M_PKTHDR) 956 m->m_pkthdr.len -= optlen; 957 optlen += sizeof(struct ip); 958 memmove(ip + 1, (char *)ip + optlen, 959 (unsigned)(m->m_len - sizeof(struct ip))); 960 } 961 m_tag_delete_nonpersistent(m); 962 m->m_flags &= ~(M_BCAST|M_MCAST); 963 964 /* 965 * Clear any in-bound checksum flags for this packet. 966 */ 967 if (m->m_flags & M_PKTHDR) 968 m->m_pkthdr.csum_flags = 0; 969 970 icmp_send(m, opts); 971 done: 972 curlwp_bindx(bound); 973 if (opts) 974 (void)m_free(opts); 975 } 976 977 /* 978 * Send an icmp packet back to the ip level, 979 * after supplying a checksum. 980 */ 981 static void 982 icmp_send(struct mbuf *m, struct mbuf *opts) 983 { 984 struct ip *ip = mtod(m, struct ip *); 985 int hlen; 986 struct icmp *icp; 987 988 hlen = ip->ip_hl << 2; 989 m->m_data += hlen; 990 m->m_len -= hlen; 991 icp = mtod(m, struct icmp *); 992 icp->icmp_cksum = 0; 993 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 994 m->m_data -= hlen; 995 m->m_len += hlen; 996 #ifdef ICMPPRINTFS 997 if (icmpprintfs) { 998 char sbuf[INET_ADDRSTRLEN], dbuf[INET_ADDRSTRLEN]; 999 printf("icmp_send to destination `%s' from `%s'\n", 1000 IN_PRINT(dbuf, &ip->ip_dst), IN_PRINT(sbuf, &ip->ip_src)); 1001 } 1002 #endif 1003 (void)ip_output(m, opts, NULL, 0, NULL, NULL); 1004 } 1005 1006 n_time 1007 iptime(void) 1008 { 1009 struct timeval atv; 1010 u_long t; 1011 1012 microtime(&atv); 1013 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1014 return (htonl(t)); 1015 } 1016 1017 /* 1018 * sysctl helper routine for net.inet.icmp.returndatabytes. ensures 1019 * that the new value is in the correct range. 1020 */ 1021 static int 1022 sysctl_net_inet_icmp_returndatabytes(SYSCTLFN_ARGS) 1023 { 1024 int error, t; 1025 struct sysctlnode node; 1026 1027 node = *rnode; 1028 node.sysctl_data = &t; 1029 t = icmpreturndatabytes; 1030 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1031 if (error || newp == NULL) 1032 return error; 1033 1034 if (t < 8 || t > 512) 1035 return EINVAL; 1036 icmpreturndatabytes = t; 1037 1038 return 0; 1039 } 1040 1041 /* 1042 * sysctl helper routine for net.inet.icmp.redirtimeout. ensures that 1043 * the given value is not less than zero and then resets the timeout 1044 * queue. 1045 */ 1046 static int 1047 sysctl_net_inet_icmp_redirtimeout(SYSCTLFN_ARGS) 1048 { 1049 int error, tmp; 1050 struct sysctlnode node; 1051 1052 mutex_enter(&icmp_mtx); 1053 1054 node = *rnode; 1055 node.sysctl_data = &tmp; 1056 tmp = icmp_redirtimeout; 1057 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 1058 if (error || newp == NULL) 1059 goto out; 1060 if (tmp < 0) { 1061 error = EINVAL; 1062 goto out; 1063 } 1064 icmp_redirtimeout = tmp; 1065 1066 /* 1067 * was it a *defined* side-effect that anyone even *reading* 1068 * this value causes these things to happen? 1069 */ 1070 if (icmp_redirect_timeout_q != NULL) { 1071 if (icmp_redirtimeout == 0) { 1072 rt_timer_queue_destroy(icmp_redirect_timeout_q); 1073 icmp_redirect_timeout_q = NULL; 1074 } else { 1075 rt_timer_queue_change(icmp_redirect_timeout_q, 1076 icmp_redirtimeout); 1077 } 1078 } else if (icmp_redirtimeout > 0) { 1079 icmp_redirect_timeout_q = 1080 rt_timer_queue_create(icmp_redirtimeout); 1081 } 1082 error = 0; 1083 out: 1084 mutex_exit(&icmp_mtx); 1085 return error; 1086 } 1087 1088 static int 1089 sysctl_net_inet_icmp_stats(SYSCTLFN_ARGS) 1090 { 1091 1092 return (NETSTAT_SYSCTL(icmpstat_percpu, ICMP_NSTATS)); 1093 } 1094 1095 static void 1096 sysctl_netinet_icmp_setup(struct sysctllog **clog) 1097 { 1098 1099 sysctl_createv(clog, 0, NULL, NULL, 1100 CTLFLAG_PERMANENT, 1101 CTLTYPE_NODE, "inet", NULL, 1102 NULL, 0, NULL, 0, 1103 CTL_NET, PF_INET, CTL_EOL); 1104 sysctl_createv(clog, 0, NULL, NULL, 1105 CTLFLAG_PERMANENT, 1106 CTLTYPE_NODE, "icmp", 1107 SYSCTL_DESCR("ICMPv4 related settings"), 1108 NULL, 0, NULL, 0, 1109 CTL_NET, PF_INET, IPPROTO_ICMP, CTL_EOL); 1110 1111 sysctl_createv(clog, 0, NULL, NULL, 1112 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1113 CTLTYPE_INT, "maskrepl", 1114 SYSCTL_DESCR("Respond to ICMP_MASKREQ messages"), 1115 NULL, 0, &icmpmaskrepl, 0, 1116 CTL_NET, PF_INET, IPPROTO_ICMP, 1117 ICMPCTL_MASKREPL, CTL_EOL); 1118 sysctl_createv(clog, 0, NULL, NULL, 1119 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1120 CTLTYPE_INT, "returndatabytes", 1121 SYSCTL_DESCR("Number of bytes to return in an ICMP " 1122 "error message"), 1123 sysctl_net_inet_icmp_returndatabytes, 0, 1124 &icmpreturndatabytes, 0, 1125 CTL_NET, PF_INET, IPPROTO_ICMP, 1126 ICMPCTL_RETURNDATABYTES, CTL_EOL); 1127 sysctl_createv(clog, 0, NULL, NULL, 1128 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1129 CTLTYPE_INT, "errppslimit", 1130 SYSCTL_DESCR("Maximum number of outgoing ICMP error " 1131 "messages per second"), 1132 NULL, 0, &icmperrppslim, 0, 1133 CTL_NET, PF_INET, IPPROTO_ICMP, 1134 ICMPCTL_ERRPPSLIMIT, CTL_EOL); 1135 sysctl_createv(clog, 0, NULL, NULL, 1136 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1137 CTLTYPE_INT, "rediraccept", 1138 SYSCTL_DESCR("Accept ICMP_REDIRECT messages"), 1139 NULL, 0, &icmp_rediraccept, 0, 1140 CTL_NET, PF_INET, IPPROTO_ICMP, 1141 ICMPCTL_REDIRACCEPT, CTL_EOL); 1142 sysctl_createv(clog, 0, NULL, NULL, 1143 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1144 CTLTYPE_INT, "redirtimeout", 1145 SYSCTL_DESCR("Lifetime of ICMP_REDIRECT generated " 1146 "routes"), 1147 sysctl_net_inet_icmp_redirtimeout, 0, 1148 &icmp_redirtimeout, 0, 1149 CTL_NET, PF_INET, IPPROTO_ICMP, 1150 ICMPCTL_REDIRTIMEOUT, CTL_EOL); 1151 sysctl_createv(clog, 0, NULL, NULL, 1152 CTLFLAG_PERMANENT, 1153 CTLTYPE_STRUCT, "stats", 1154 SYSCTL_DESCR("ICMP statistics"), 1155 sysctl_net_inet_icmp_stats, 0, NULL, 0, 1156 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_STATS, 1157 CTL_EOL); 1158 sysctl_createv(clog, 0, NULL, NULL, 1159 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1160 CTLTYPE_INT, "bmcastecho", 1161 SYSCTL_DESCR("Respond to ICMP_ECHO or ICMP_TIMESTAMP " 1162 "message to the broadcast or multicast"), 1163 NULL, 0, &icmpbmcastecho, 0, 1164 CTL_NET, PF_INET, IPPROTO_ICMP, ICMPCTL_BMCASTECHO, 1165 CTL_EOL); 1166 } 1167 1168 void 1169 icmp_statinc(u_int stat) 1170 { 1171 1172 KASSERT(stat < ICMP_NSTATS); 1173 ICMP_STATINC(stat); 1174 } 1175 1176 /* Table of common MTUs */ 1177 static const u_int mtu_table[] = { 1178 65535, 65280, 32000, 17914, 9180, 8166, 1179 4352, 2002, 1492, 1006, 508, 296, 68, 0 1180 }; 1181 1182 void 1183 icmp_mtudisc(struct icmp *icp, struct in_addr faddr) 1184 { 1185 struct icmp_mtudisc_callback *mc; 1186 struct sockaddr *dst = sintosa(&icmpsrc); 1187 struct rtentry *rt; 1188 u_long mtu = ntohs(icp->icmp_nextmtu); /* Why a long? IPv6 */ 1189 int error; 1190 1191 rt = rtalloc1(dst, 1); 1192 if (rt == NULL) 1193 return; 1194 1195 /* If we didn't get a host route, allocate one */ 1196 if ((rt->rt_flags & RTF_HOST) == 0) { 1197 struct rtentry *nrt; 1198 1199 error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL, 1200 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 1201 if (error) { 1202 rt_unref(rt); 1203 return; 1204 } 1205 nrt->rt_rmx = rt->rt_rmx; 1206 rt_unref(rt); 1207 rt = nrt; 1208 } 1209 1210 mutex_enter(&icmp_mtx); 1211 error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q); 1212 mutex_exit(&icmp_mtx); 1213 if (error) { 1214 rt_unref(rt); 1215 return; 1216 } 1217 1218 if (mtu == 0) { 1219 int i = 0; 1220 1221 mtu = ntohs(icp->icmp_ip.ip_len); 1222 /* Some 4.2BSD-based routers incorrectly adjust the ip_len */ 1223 if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0) 1224 mtu -= (icp->icmp_ip.ip_hl << 2); 1225 1226 /* If we still can't guess a value, try the route */ 1227 if (mtu == 0) { 1228 mtu = rt->rt_rmx.rmx_mtu; 1229 1230 /* If no route mtu, default to the interface mtu */ 1231 if (mtu == 0) 1232 mtu = rt->rt_ifp->if_mtu; 1233 } 1234 1235 for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++) { 1236 if (mtu > mtu_table[i]) { 1237 mtu = mtu_table[i]; 1238 break; 1239 } 1240 } 1241 } 1242 1243 /* 1244 * XXX: RTV_MTU is overloaded, since the admin can set it 1245 * to turn off PMTU for a route, and the kernel can 1246 * set it to indicate a serious problem with PMTU 1247 * on a route. We should be using a separate flag 1248 * for the kernel to indicate this. 1249 */ 1250 1251 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1252 if (mtu < 296 || mtu > rt->rt_ifp->if_mtu) 1253 rt->rt_rmx.rmx_locks |= RTV_MTU; 1254 else if (rt->rt_rmx.rmx_mtu > mtu || 1255 rt->rt_rmx.rmx_mtu == 0) { 1256 ICMP_STATINC(ICMP_STAT_PMTUCHG); 1257 rt->rt_rmx.rmx_mtu = mtu; 1258 } 1259 } 1260 1261 if (rt != NULL) 1262 rt_unref(rt); 1263 1264 /* 1265 * Notify protocols that the MTU for this destination 1266 * has changed. 1267 */ 1268 mutex_enter(&icmp_mtx); 1269 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL; 1270 mc = LIST_NEXT(mc, mc_list)) 1271 (*mc->mc_func)(faddr); 1272 mutex_exit(&icmp_mtx); 1273 } 1274 1275 /* 1276 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1277 * given current value MTU. If DIR is less than zero, a larger plateau 1278 * is returned; otherwise, a smaller value is returned. 1279 */ 1280 u_int 1281 ip_next_mtu(u_int mtu, int dir) /* XXX unused */ 1282 { 1283 int i; 1284 1285 for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) { 1286 if (mtu >= mtu_table[i]) 1287 break; 1288 } 1289 1290 if (dir < 0) { 1291 if (i == 0) { 1292 return 0; 1293 } else { 1294 return mtu_table[i - 1]; 1295 } 1296 } else { 1297 if (mtu_table[i] == 0) { 1298 return 0; 1299 } else if (mtu > mtu_table[i]) { 1300 return mtu_table[i]; 1301 } else { 1302 return mtu_table[i + 1]; 1303 } 1304 } 1305 } 1306 1307 static void 1308 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 1309 { 1310 1311 KASSERT(rt != NULL); 1312 rt_assert_referenced(rt); 1313 1314 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1315 (RTF_DYNAMIC | RTF_HOST)) { 1316 rtrequest(RTM_DELETE, rt_getkey(rt), 1317 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 1318 } else { 1319 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1320 rt->rt_rmx.rmx_mtu = 0; 1321 } 1322 } 1323 } 1324 1325 static void 1326 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r) 1327 { 1328 1329 KASSERT(rt != NULL); 1330 rt_assert_referenced(rt); 1331 1332 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1333 (RTF_DYNAMIC | RTF_HOST)) { 1334 rtrequest(RTM_DELETE, rt_getkey(rt), 1335 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 1336 } 1337 } 1338 1339 /* 1340 * Perform rate limit check. 1341 * Returns 0 if it is okay to send the icmp packet. 1342 * Returns 1 if the router SHOULD NOT send this icmp packet due to rate 1343 * limitation. 1344 * 1345 * XXX per-destination/type check necessary? 1346 */ 1347 int 1348 icmp_ratelimit(const struct in_addr *dst, const int type, 1349 const int code) 1350 { 1351 1352 /* PPS limit */ 1353 if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count, 1354 icmperrppslim)) { 1355 /* The packet is subject to rate limit */ 1356 return 1; 1357 } 1358 1359 /* okay to send */ 1360 return 0; 1361 } 1362
Indexes created Fri Oct 03 02:09:57 GMT 2025