ip_icmp.c revision 1.163
1 /* $NetBSD: ip_icmp.c,v 1.163 2018/01/19 13:17:29 maxv Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * 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 project 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 PROJECT 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 PROJECT 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 32 /*- 33 * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * This code is derived from software contributed to The NetBSD Foundation 41 * by Jason R. Thorpe of Zembu Labs, Inc. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 52 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 53 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 54 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 55 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 56 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 57 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 58 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 59 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 60 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 61 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 62 * POSSIBILITY OF 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 1.163 2018/01/19 13:17:29 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 /* IPSEC*/ 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 #if 0 162 static u_int ip_next_mtu(u_int, int); 163 #else 164 /*static*/ u_int ip_next_mtu(u_int, int); 165 #endif 166 167 extern int icmperrppslim; 168 static int icmperrpps_count = 0; 169 static struct timeval icmperrppslim_last; 170 static int icmp_rediraccept = 1; 171 static int icmp_redirtimeout = 600; 172 static struct rttimer_queue *icmp_redirect_timeout_q = NULL; 173 174 /* Protect mtudisc and redirect stuffs */ 175 static kmutex_t icmp_mtx __cacheline_aligned; 176 177 static void icmp_mtudisc_timeout(struct rtentry *, struct rttimer *); 178 static void icmp_redirect_timeout(struct rtentry *, struct rttimer *); 179 180 static void sysctl_netinet_icmp_setup(struct sysctllog **); 181 182 /* workqueue-based pr_input */ 183 static struct wqinput *icmp_wqinput; 184 static void _icmp_input(struct mbuf *, int, int); 185 186 void 187 icmp_init(void) 188 { 189 190 sysctl_netinet_icmp_setup(NULL); 191 192 mutex_init(&icmp_mtx, MUTEX_DEFAULT, IPL_NONE); 193 /* 194 * This is only useful if the user initializes redirtimeout to 195 * something other than zero. 196 */ 197 mutex_enter(&icmp_mtx); 198 icmp_redirect_timeout_q = rt_timer_queue_create(icmp_redirtimeout); 199 mutex_exit(&icmp_mtx); 200 201 icmpstat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP_NSTATS); 202 icmp_wqinput = wqinput_create("icmp", _icmp_input); 203 } 204 205 void 206 icmp_mtudisc_lock(void) 207 { 208 209 mutex_enter(&icmp_mtx); 210 } 211 212 void 213 icmp_mtudisc_unlock(void) 214 { 215 216 mutex_exit(&icmp_mtx); 217 } 218 219 /* 220 * Register a Path MTU Discovery callback. 221 */ 222 void 223 icmp_mtudisc_callback_register(void (*func)(struct in_addr)) 224 { 225 struct icmp_mtudisc_callback *mc, *new; 226 227 new = kmem_alloc(sizeof(*mc), KM_SLEEP); 228 229 mutex_enter(&icmp_mtx); 230 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL; 231 mc = LIST_NEXT(mc, mc_list)) { 232 if (mc->mc_func == func) { 233 mutex_exit(&icmp_mtx); 234 kmem_free(new, sizeof(*mc)); 235 return; 236 } 237 } 238 239 new->mc_func = func; 240 LIST_INSERT_HEAD(&icmp_mtudisc_callbacks, new, mc_list); 241 mutex_exit(&icmp_mtx); 242 } 243 244 /* 245 * Generate an error packet of type error in response to a bad IP packet. 'n' 246 * contains this packet. We create 'm' and send it. 247 * 248 * As we are not required to return everything we have, we return whatever 249 * we can return at ease. 250 * 251 * Note that ICMP datagrams longer than 576 octets are out of spec according 252 * to RFC1812; the limit on icmpreturndatabytes will keep things below that 253 * limit. 254 */ 255 void 256 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu) 257 { 258 struct ip *oip = mtod(n, struct ip *), *nip; 259 const unsigned oiphlen = oip->ip_hl << 2; 260 struct icmp *icp; 261 struct mbuf *m; 262 struct m_tag *mtag; 263 unsigned datalen, mblen; 264 int totlen; 265 266 #ifdef ICMPPRINTFS 267 if (icmpprintfs) 268 printf("icmp_error(%p, type:%d, code:%d)\n", oip, type, code); 269 #endif 270 271 if (type != ICMP_REDIRECT) 272 ICMP_STATINC(ICMP_STAT_ERROR); 273 274 /* 275 * Don't send error if: 276 * - The original packet was encrypted. 277 * - The packet is multicast or broadcast. 278 * - The packet is not the first fragment of the message. 279 * - The packet is an ICMP message with an unknown type. 280 */ 281 if (n->m_flags & M_DECRYPTED) 282 goto freeit; 283 if (n->m_flags & (M_BCAST|M_MCAST)) 284 goto freeit; 285 if (oip->ip_off &~ htons(IP_MF|IP_DF)) 286 goto freeit; 287 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 288 n->m_len >= oiphlen + ICMP_MINLEN) { 289 struct icmp *oicp = (struct icmp *)((char *)oip + oiphlen); 290 if (!ICMP_INFOTYPE(oicp->icmp_type)) { 291 ICMP_STATINC(ICMP_STAT_OLDICMP); 292 goto freeit; 293 } 294 } 295 296 /* 297 * First, do a rate limitation check. 298 */ 299 if (icmp_ratelimit(&oip->ip_src, type, code)) { 300 /* XXX stat */ 301 goto freeit; 302 } 303 304 /* 305 * Compute the number of bytes we will put in 'icmp_ip'. Truncate 306 * it to the size of the mbuf, if it's too big. 307 */ 308 datalen = oiphlen + min(icmpreturndatabytes, 309 ntohs(oip->ip_len) - oiphlen); 310 mblen = 0; 311 for (m = n; m && (mblen < datalen); m = m->m_next) 312 mblen += m->m_len; 313 datalen = min(mblen, datalen); 314 315 /* 316 * Compute the total length of the new packet. Truncate it if it's 317 * bigger than the size of a cluster. 318 */ 319 CTASSERT(ICMP_MINLEN + sizeof(struct ip) <= MCLBYTES); 320 totlen = sizeof(struct ip) + ICMP_MINLEN + datalen; 321 if (totlen > MCLBYTES) { 322 datalen = MCLBYTES - ICMP_MINLEN - sizeof(struct ip); 323 totlen = MCLBYTES; 324 } 325 326 /* 327 * Allocate the mbuf for the new packet. 328 */ 329 m = m_gethdr(M_DONTWAIT, MT_HEADER); 330 if (m && (totlen > MHLEN)) { 331 MCLGET(m, M_DONTWAIT); 332 if ((m->m_flags & M_EXT) == 0) { 333 m_freem(m); 334 m = NULL; 335 } 336 } 337 if (m == NULL) 338 goto freeit; 339 MCLAIM(m, n->m_owner); 340 m->m_len = totlen; 341 m->m_pkthdr.len = m->m_len; 342 m_copy_rcvif(m, n); 343 344 if ((u_int)type > ICMP_MAXTYPE) 345 panic("icmp_error"); 346 ICMP_STATINC(ICMP_STAT_OUTHIST + type); 347 348 /* 349 * Get pointers on the IP header and the ICMP header. 350 */ 351 nip = mtod(m, struct ip *); 352 icp = (struct icmp *)(nip + 1); 353 354 /* 355 * Fill in the fields of the ICMP header: icmp_type, icmp_code 356 * and icmp_ip. icmp_cksum gets filled later. 357 */ 358 icp->icmp_type = type; 359 if (type == ICMP_REDIRECT) { 360 icp->icmp_gwaddr.s_addr = dest; 361 } else { 362 icp->icmp_void = 0; 363 /* 364 * The following assignments assume an overlay with the 365 * zeroed icmp_void field. 366 */ 367 if (type == ICMP_PARAMPROB) { 368 icp->icmp_pptr = code; 369 code = 0; 370 } else if (type == ICMP_UNREACH && 371 code == ICMP_UNREACH_NEEDFRAG && destmtu) 372 icp->icmp_nextmtu = htons(destmtu); 373 } 374 icp->icmp_code = code; 375 m_copydata(n, 0, datalen, (void *)&icp->icmp_ip); 376 377 /* 378 * Now, copy the old IP header (without options) in front of the 379 * ICMP message. The src/dst fields will be swapped in icmp_reflect. 380 */ 381 /* ip_v set in ip_output */ 382 nip->ip_hl = sizeof(struct ip) >> 2; 383 nip->ip_tos = 0; 384 nip->ip_len = htons(m->m_len); 385 /* ip_id set in ip_output */ 386 nip->ip_off = htons(0); 387 /* ip_ttl set in icmp_reflect */ 388 nip->ip_p = IPPROTO_ICMP; 389 nip->ip_src = oip->ip_src; 390 nip->ip_dst = oip->ip_dst; 391 /* move PF m_tag to new packet, if it exists */ 392 mtag = m_tag_find(n, PACKET_TAG_PF, NULL); 393 if (mtag != NULL) { 394 m_tag_unlink(n, mtag); 395 m_tag_prepend(m, mtag); 396 } 397 398 icmp_reflect(m); 399 400 freeit: 401 m_freem(n); 402 } 403 404 struct sockaddr_in icmpsrc = { 405 .sin_len = sizeof (struct sockaddr_in), 406 .sin_family = AF_INET, 407 }; 408 static struct sockaddr_in icmpdst = { 409 .sin_len = sizeof (struct sockaddr_in), 410 .sin_family = AF_INET, 411 }; 412 static struct sockaddr_in icmpgw = { 413 .sin_len = sizeof (struct sockaddr_in), 414 .sin_family = AF_INET, 415 }; 416 struct sockaddr_in icmpmask = { 417 .sin_len = 8, 418 .sin_family = 0, 419 }; 420 421 /* 422 * Process a received ICMP message. 423 */ 424 static void 425 _icmp_input(struct mbuf *m, int hlen, int proto) 426 { 427 struct icmp *icp; 428 struct ip *ip = mtod(m, struct ip *); 429 int icmplen; 430 int i; 431 struct in_ifaddr *ia; 432 void *(*ctlfunc)(int, const struct sockaddr *, void *); 433 int code; 434 struct rtentry *rt; 435 436 /* 437 * Locate icmp structure in mbuf, and check 438 * that not corrupted and of at least minimum length. 439 */ 440 icmplen = ntohs(ip->ip_len) - hlen; 441 #ifdef ICMPPRINTFS 442 if (icmpprintfs) { 443 char sbuf[INET_ADDRSTRLEN], dbuf[INET_ADDRSTRLEN]; 444 printf("icmp_input from `%s' to `%s', len %d\n", 445 IN_PRINT(sbuf, &ip->ip_src), IN_PRINT(dbuf, &ip->ip_dst), 446 icmplen); 447 } 448 #endif 449 if (icmplen < ICMP_MINLEN) { 450 ICMP_STATINC(ICMP_STAT_TOOSHORT); 451 goto freeit; 452 } 453 i = hlen + min(icmplen, ICMP_ADVLENMIN); 454 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == NULL) { 455 ICMP_STATINC(ICMP_STAT_TOOSHORT); 456 return; 457 } 458 ip = mtod(m, struct ip *); 459 m->m_len -= hlen; 460 m->m_data += hlen; 461 icp = mtod(m, struct icmp *); 462 /* Don't need to assert alignment, here. */ 463 if (in_cksum(m, icmplen)) { 464 ICMP_STATINC(ICMP_STAT_CHECKSUM); 465 goto freeit; 466 } 467 m->m_len += hlen; 468 m->m_data -= hlen; 469 470 #ifdef ICMPPRINTFS 471 /* 472 * Message type specific processing. 473 */ 474 if (icmpprintfs) 475 printf("icmp_input(type:%d, code:%d)\n", icp->icmp_type, 476 icp->icmp_code); 477 #endif 478 if (icp->icmp_type > ICMP_MAXTYPE) 479 goto raw; 480 ICMP_STATINC(ICMP_STAT_INHIST + icp->icmp_type); 481 code = icp->icmp_code; 482 switch (icp->icmp_type) { 483 484 case ICMP_UNREACH: 485 switch (code) { 486 case ICMP_UNREACH_PROTOCOL: 487 code = PRC_UNREACH_PROTOCOL; 488 break; 489 490 case ICMP_UNREACH_PORT: 491 code = PRC_UNREACH_PORT; 492 break; 493 494 case ICMP_UNREACH_SRCFAIL: 495 code = PRC_UNREACH_SRCFAIL; 496 break; 497 498 case ICMP_UNREACH_NEEDFRAG: 499 code = PRC_MSGSIZE; 500 break; 501 502 case ICMP_UNREACH_NET: 503 case ICMP_UNREACH_NET_UNKNOWN: 504 case ICMP_UNREACH_NET_PROHIB: 505 case ICMP_UNREACH_TOSNET: 506 code = PRC_UNREACH_NET; 507 break; 508 509 case ICMP_UNREACH_HOST: 510 case ICMP_UNREACH_HOST_UNKNOWN: 511 case ICMP_UNREACH_ISOLATED: 512 case ICMP_UNREACH_HOST_PROHIB: 513 case ICMP_UNREACH_TOSHOST: 514 case ICMP_UNREACH_ADMIN_PROHIBIT: 515 case ICMP_UNREACH_HOST_PREC: 516 case ICMP_UNREACH_PREC_CUTOFF: 517 code = PRC_UNREACH_HOST; 518 break; 519 520 default: 521 goto badcode; 522 } 523 goto deliver; 524 525 case ICMP_TIMXCEED: 526 if (code > 1) 527 goto badcode; 528 code += PRC_TIMXCEED_INTRANS; 529 goto deliver; 530 531 case ICMP_PARAMPROB: 532 if (code > 1) 533 goto badcode; 534 code = PRC_PARAMPROB; 535 goto deliver; 536 537 case ICMP_SOURCEQUENCH: 538 if (code) 539 goto badcode; 540 code = PRC_QUENCH; 541 goto deliver; 542 543 deliver: 544 /* 545 * Problem with datagram; advise higher level routines. 546 */ 547 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 548 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 549 ICMP_STATINC(ICMP_STAT_BADLEN); 550 goto freeit; 551 } 552 if (IN_MULTICAST(icp->icmp_ip.ip_dst.s_addr)) 553 goto badcode; 554 #ifdef ICMPPRINTFS 555 if (icmpprintfs) 556 printf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 557 #endif 558 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 559 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput; 560 if (ctlfunc) 561 (void) (*ctlfunc)(code, sintosa(&icmpsrc), 562 &icp->icmp_ip); 563 break; 564 565 badcode: 566 ICMP_STATINC(ICMP_STAT_BADCODE); 567 break; 568 569 case ICMP_ECHO: 570 if (!icmpbmcastecho && 571 (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 572 ICMP_STATINC(ICMP_STAT_BMCASTECHO); 573 break; 574 } 575 icp->icmp_type = ICMP_ECHOREPLY; 576 goto reflect; 577 578 case ICMP_TSTAMP: 579 if (icmplen < ICMP_TSLEN) { 580 ICMP_STATINC(ICMP_STAT_BADLEN); 581 break; 582 } 583 if (!icmpbmcastecho && 584 (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 585 ICMP_STATINC(ICMP_STAT_BMCASTTSTAMP); 586 break; 587 } 588 icp->icmp_type = ICMP_TSTAMPREPLY; 589 icp->icmp_rtime = iptime(); 590 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 591 goto reflect; 592 593 case ICMP_MASKREQ: { 594 struct ifnet *rcvif; 595 int s, ss; 596 struct ifaddr *ifa = NULL; 597 598 if (icmpmaskrepl == 0) 599 break; 600 /* 601 * We are not able to respond with all ones broadcast 602 * unless we receive it over a point-to-point interface. 603 */ 604 if (icmplen < ICMP_MASKLEN) { 605 ICMP_STATINC(ICMP_STAT_BADLEN); 606 break; 607 } 608 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 609 in_nullhost(ip->ip_dst)) 610 icmpdst.sin_addr = ip->ip_src; 611 else 612 icmpdst.sin_addr = ip->ip_dst; 613 ss = pserialize_read_enter(); 614 rcvif = m_get_rcvif(m, &s); 615 if (__predict_true(rcvif != NULL)) 616 ifa = ifaof_ifpforaddr(sintosa(&icmpdst), rcvif); 617 m_put_rcvif(rcvif, &s); 618 if (ifa == NULL) { 619 pserialize_read_exit(ss); 620 break; 621 } 622 ia = ifatoia(ifa); 623 icp->icmp_type = ICMP_MASKREPLY; 624 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 625 if (in_nullhost(ip->ip_src)) { 626 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 627 ip->ip_src = ia->ia_broadaddr.sin_addr; 628 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 629 ip->ip_src = ia->ia_dstaddr.sin_addr; 630 } 631 pserialize_read_exit(ss); 632 reflect: 633 { 634 uint64_t *icps = percpu_getref(icmpstat_percpu); 635 icps[ICMP_STAT_REFLECT]++; 636 icps[ICMP_STAT_OUTHIST + icp->icmp_type]++; 637 percpu_putref(icmpstat_percpu); 638 } 639 icmp_reflect(m); 640 return; 641 } 642 643 case ICMP_REDIRECT: 644 if (code > 3) 645 goto badcode; 646 if (icmp_rediraccept == 0) 647 goto freeit; 648 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 649 icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) { 650 ICMP_STATINC(ICMP_STAT_BADLEN); 651 break; 652 } 653 /* 654 * Short circuit routing redirects to force 655 * immediate change in the kernel's routing 656 * tables. The message is also handed to anyone 657 * listening on a raw socket (e.g. the routing 658 * daemon for use in updating its tables). 659 */ 660 icmpgw.sin_addr = ip->ip_src; 661 icmpdst.sin_addr = icp->icmp_gwaddr; 662 #ifdef ICMPPRINTFS 663 if (icmpprintfs) { 664 char gbuf[INET_ADDRSTRLEN], dbuf[INET_ADDRSTRLEN]; 665 printf("redirect dst `%s' to `%s'\n", 666 IN_PRINT(dbuf, &icp->icmp_ip.ip_dst), 667 IN_PRINT(gbuf, &icp->icmp_gwaddr)); 668 } 669 #endif 670 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 671 rt = NULL; 672 rtredirect(sintosa(&icmpsrc), sintosa(&icmpdst), 673 NULL, RTF_GATEWAY | RTF_HOST, sintosa(&icmpgw), &rt); 674 mutex_enter(&icmp_mtx); 675 if (rt != NULL && icmp_redirtimeout != 0) { 676 i = rt_timer_add(rt, icmp_redirect_timeout, 677 icmp_redirect_timeout_q); 678 if (i) { 679 char buf[INET_ADDRSTRLEN]; 680 log(LOG_ERR, "ICMP: redirect failed to " 681 "register timeout for route to %s, " 682 "code %d\n", 683 IN_PRINT(buf, &icp->icmp_ip.ip_dst), i); 684 } 685 } 686 mutex_exit(&icmp_mtx); 687 if (rt != NULL) 688 rt_unref(rt); 689 690 pfctlinput(PRC_REDIRECT_HOST, sintosa(&icmpsrc)); 691 #if defined(IPSEC) 692 if (ipsec_used) 693 key_sa_routechange((struct sockaddr *)&icmpsrc); 694 #endif 695 break; 696 697 /* 698 * No kernel processing for the following; 699 * just fall through to send to raw listener. 700 */ 701 case ICMP_ECHOREPLY: 702 case ICMP_ROUTERADVERT: 703 case ICMP_ROUTERSOLICIT: 704 case ICMP_TSTAMPREPLY: 705 case ICMP_IREQREPLY: 706 case ICMP_MASKREPLY: 707 default: 708 break; 709 } 710 711 raw: 712 rip_input(m, hlen, proto); 713 return; 714 715 freeit: 716 m_freem(m); 717 return; 718 } 719 720 void 721 icmp_input(struct mbuf *m, ...) 722 { 723 int hlen, proto; 724 va_list ap; 725 726 va_start(ap, m); 727 hlen = va_arg(ap, int); 728 proto = va_arg(ap, int); 729 va_end(ap); 730 731 wqinput_input(icmp_wqinput, m, hlen, proto); 732 } 733 734 /* 735 * Reflect the ip packet back to the source 736 */ 737 void 738 icmp_reflect(struct mbuf *m) 739 { 740 struct ip *ip = mtod(m, struct ip *); 741 struct in_ifaddr *ia; 742 struct ifaddr *ifa; 743 struct sockaddr_in *sin; 744 struct in_addr t; 745 struct mbuf *opts = NULL; 746 int optlen = (ip->ip_hl << 2) - sizeof(struct ip); 747 struct ifnet *rcvif; 748 struct psref psref, psref_ia; 749 int s; 750 int bound; 751 752 bound = curlwp_bind(); 753 754 if (!in_canforward(ip->ip_src) && 755 ((ip->ip_src.s_addr & IN_CLASSA_NET) != 756 htonl(IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { 757 m_freem(m); /* Bad return address */ 758 goto done; /* ip_output() will check for broadcast */ 759 } 760 t = ip->ip_dst; 761 ip->ip_dst = ip->ip_src; 762 /* 763 * If the incoming packet was addressed directly to us, use 764 * dst as the src for the reply. Otherwise (broadcast or 765 * anonymous), use an address which corresponds to the 766 * incoming interface, with a preference for the address which 767 * corresponds to the route to the destination of the ICMP. 768 */ 769 770 /* Look for packet addressed to us */ 771 ia = in_get_ia_psref(t, &psref_ia); 772 if (ia && (ia->ia4_flags & IN_IFF_NOTREADY)) { 773 ia4_release(ia, &psref_ia); 774 ia = NULL; 775 } 776 777 rcvif = m_get_rcvif_psref(m, &psref); 778 779 /* look for packet sent to broadcast address */ 780 if (ia == NULL && rcvif && 781 (rcvif->if_flags & IFF_BROADCAST)) { 782 s = pserialize_read_enter(); 783 IFADDR_READER_FOREACH(ifa, rcvif) { 784 if (ifa->ifa_addr->sa_family != AF_INET) 785 continue; 786 if (in_hosteq(t,ifatoia(ifa)->ia_broadaddr.sin_addr)) { 787 ia = ifatoia(ifa); 788 if ((ia->ia4_flags & IN_IFF_NOTREADY) == 0) 789 break; 790 ia = NULL; 791 } 792 } 793 if (ia != NULL) 794 ia4_acquire(ia, &psref_ia); 795 pserialize_read_exit(s); 796 } 797 798 sin = ia ? &ia->ia_addr : NULL; 799 800 icmpdst.sin_addr = t; 801 802 /* 803 * if the packet is addressed somewhere else, compute the 804 * source address for packets routed back to the source, and 805 * use that, if it's an address on the interface which 806 * received the packet 807 */ 808 if (sin == NULL && rcvif) { 809 struct sockaddr_in sin_dst; 810 struct route icmproute; 811 int errornum; 812 813 sockaddr_in_init(&sin_dst, &ip->ip_dst, 0); 814 memset(&icmproute, 0, sizeof(icmproute)); 815 errornum = 0; 816 ia = in_selectsrc(&sin_dst, &icmproute, 0, NULL, &errornum, 817 &psref_ia); 818 /* errornum is never used */ 819 rtcache_free(&icmproute); 820 /* check to make sure sin is a source address on rcvif */ 821 if (ia != NULL) { 822 sin = &ia->ia_addr; 823 t = sin->sin_addr; 824 sin = NULL; 825 ia4_release(ia, &psref_ia); 826 ia = in_get_ia_on_iface_psref(t, rcvif, &psref_ia); 827 if (ia != NULL) 828 sin = &ia->ia_addr; 829 } 830 } 831 832 /* 833 * if it was not addressed to us, but the route doesn't go out 834 * the source interface, pick an address on the source 835 * interface. This can happen when routing is asymmetric, or 836 * when the incoming packet was encapsulated 837 */ 838 if (sin == NULL && rcvif) { 839 KASSERT(ia == NULL); 840 s = pserialize_read_enter(); 841 IFADDR_READER_FOREACH(ifa, rcvif) { 842 if (ifa->ifa_addr->sa_family != AF_INET) 843 continue; 844 sin = &(ifatoia(ifa)->ia_addr); 845 ia = ifatoia(ifa); 846 ia4_acquire(ia, &psref_ia); 847 break; 848 } 849 pserialize_read_exit(s); 850 } 851 852 m_put_rcvif_psref(rcvif, &psref); 853 854 /* 855 * The following happens if the packet was not addressed to us, 856 * and was received on an interface with no IP address: 857 * We find the first AF_INET address on the first non-loopback 858 * interface. 859 */ 860 if (sin == NULL) { 861 KASSERT(ia == NULL); 862 s = pserialize_read_enter(); 863 IN_ADDRLIST_READER_FOREACH(ia) { 864 if (ia->ia_ifp->if_flags & IFF_LOOPBACK) 865 continue; 866 sin = &ia->ia_addr; 867 ia4_acquire(ia, &psref_ia); 868 break; 869 } 870 pserialize_read_exit(s); 871 } 872 873 /* 874 * If we still didn't find an address, punt. We could have an 875 * interface up (and receiving packets) with no address. 876 */ 877 if (sin == NULL) { 878 KASSERT(ia == NULL); 879 m_freem(m); 880 goto done; 881 } 882 883 ip->ip_src = sin->sin_addr; 884 ip->ip_ttl = MAXTTL; 885 886 if (ia != NULL) 887 ia4_release(ia, &psref_ia); 888 889 if (optlen > 0) { 890 u_char *cp; 891 int opt, cnt; 892 u_int len; 893 894 /* 895 * Retrieve any source routing from the incoming packet; 896 * add on any record-route or timestamp options. 897 */ 898 cp = (u_char *) (ip + 1); 899 if ((opts = ip_srcroute(m)) == NULL && 900 (opts = m_gethdr(M_DONTWAIT, MT_HEADER))) { 901 MCLAIM(opts, m->m_owner); 902 opts->m_len = sizeof(struct in_addr); 903 *mtod(opts, struct in_addr *) = zeroin_addr; 904 } 905 if (opts) { 906 #ifdef ICMPPRINTFS 907 if (icmpprintfs) 908 printf("icmp_reflect optlen %d rt %d => ", 909 optlen, opts->m_len); 910 #endif 911 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 912 opt = cp[IPOPT_OPTVAL]; 913 if (opt == IPOPT_EOL) 914 break; 915 if (opt == IPOPT_NOP) 916 len = 1; 917 else { 918 if (cnt < IPOPT_OLEN + sizeof(*cp)) 919 break; 920 len = cp[IPOPT_OLEN]; 921 if (len < IPOPT_OLEN + sizeof(*cp) || 922 len > cnt) 923 break; 924 } 925 /* 926 * Should check for overflow, but it "can't happen" 927 */ 928 if (opt == IPOPT_RR || opt == IPOPT_TS || 929 opt == IPOPT_SECURITY) { 930 memmove(mtod(opts, char *) + opts->m_len, 931 cp, len); 932 opts->m_len += len; 933 } 934 } 935 /* Terminate & pad, if necessary */ 936 if ((cnt = opts->m_len % 4) != 0) { 937 for (; cnt < 4; cnt++) { 938 *(mtod(opts, char *) + opts->m_len) = 939 IPOPT_EOL; 940 opts->m_len++; 941 } 942 } 943 #ifdef ICMPPRINTFS 944 if (icmpprintfs) 945 printf("%d\n", opts->m_len); 946 #endif 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 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 1178 static const u_int mtu_table[] = { 1179 65535, 65280, 32000, 17914, 9180, 8166, 1180 4352, 2002, 1492, 1006, 508, 296, 68, 0 1181 }; 1182 1183 void 1184 icmp_mtudisc(struct icmp *icp, struct in_addr faddr) 1185 { 1186 struct icmp_mtudisc_callback *mc; 1187 struct sockaddr *dst = sintosa(&icmpsrc); 1188 struct rtentry *rt; 1189 u_long mtu = ntohs(icp->icmp_nextmtu); /* Why a long? IPv6 */ 1190 int error; 1191 1192 rt = rtalloc1(dst, 1); 1193 if (rt == NULL) 1194 return; 1195 1196 /* If we didn't get a host route, allocate one */ 1197 1198 if ((rt->rt_flags & RTF_HOST) == 0) { 1199 struct rtentry *nrt; 1200 1201 error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL, 1202 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 1203 if (error) { 1204 rt_unref(rt); 1205 return; 1206 } 1207 nrt->rt_rmx = rt->rt_rmx; 1208 rt_unref(rt); 1209 rt = nrt; 1210 } 1211 1212 mutex_enter(&icmp_mtx); 1213 error = rt_timer_add(rt, icmp_mtudisc_timeout, ip_mtudisc_timeout_q); 1214 mutex_exit(&icmp_mtx); 1215 if (error) { 1216 rt_unref(rt); 1217 return; 1218 } 1219 1220 if (mtu == 0) { 1221 int i = 0; 1222 1223 mtu = ntohs(icp->icmp_ip.ip_len); 1224 /* Some 4.2BSD-based routers incorrectly adjust the ip_len */ 1225 if (mtu > rt->rt_rmx.rmx_mtu && rt->rt_rmx.rmx_mtu != 0) 1226 mtu -= (icp->icmp_ip.ip_hl << 2); 1227 1228 /* If we still can't guess a value, try the route */ 1229 1230 if (mtu == 0) { 1231 mtu = rt->rt_rmx.rmx_mtu; 1232 1233 /* If no route mtu, default to the interface mtu */ 1234 1235 if (mtu == 0) 1236 mtu = rt->rt_ifp->if_mtu; 1237 } 1238 1239 for (i = 0; i < sizeof(mtu_table) / sizeof(mtu_table[0]); i++) 1240 if (mtu > mtu_table[i]) { 1241 mtu = mtu_table[i]; 1242 break; 1243 } 1244 } 1245 1246 /* 1247 * XXX: RTV_MTU is overloaded, since the admin can set it 1248 * to turn off PMTU for a route, and the kernel can 1249 * set it to indicate a serious problem with PMTU 1250 * on a route. We should be using a separate flag 1251 * for the kernel to indicate this. 1252 */ 1253 1254 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1255 if (mtu < 296 || mtu > rt->rt_ifp->if_mtu) 1256 rt->rt_rmx.rmx_locks |= RTV_MTU; 1257 else if (rt->rt_rmx.rmx_mtu > mtu || 1258 rt->rt_rmx.rmx_mtu == 0) { 1259 ICMP_STATINC(ICMP_STAT_PMTUCHG); 1260 rt->rt_rmx.rmx_mtu = mtu; 1261 } 1262 } 1263 1264 if (rt != NULL) 1265 rt_unref(rt); 1266 1267 /* 1268 * Notify protocols that the MTU for this destination 1269 * has changed. 1270 */ 1271 mutex_enter(&icmp_mtx); 1272 for (mc = LIST_FIRST(&icmp_mtudisc_callbacks); mc != NULL; 1273 mc = LIST_NEXT(mc, mc_list)) 1274 (*mc->mc_func)(faddr); 1275 mutex_exit(&icmp_mtx); 1276 } 1277 1278 /* 1279 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1280 * given current value MTU. If DIR is less than zero, a larger plateau 1281 * is returned; otherwise, a smaller value is returned. 1282 */ 1283 u_int 1284 ip_next_mtu(u_int mtu, int dir) /* XXX */ 1285 { 1286 int i; 1287 1288 for (i = 0; i < (sizeof mtu_table) / (sizeof mtu_table[0]); i++) { 1289 if (mtu >= mtu_table[i]) 1290 break; 1291 } 1292 1293 if (dir < 0) { 1294 if (i == 0) { 1295 return 0; 1296 } else { 1297 return mtu_table[i - 1]; 1298 } 1299 } else { 1300 if (mtu_table[i] == 0) { 1301 return 0; 1302 } else if (mtu > mtu_table[i]) { 1303 return mtu_table[i]; 1304 } else { 1305 return mtu_table[i + 1]; 1306 } 1307 } 1308 } 1309 1310 static void 1311 icmp_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 1312 { 1313 1314 KASSERT(rt != NULL); 1315 rt_assert_referenced(rt); 1316 1317 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1318 (RTF_DYNAMIC | RTF_HOST)) { 1319 rtrequest(RTM_DELETE, rt_getkey(rt), 1320 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 1321 } else { 1322 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) { 1323 rt->rt_rmx.rmx_mtu = 0; 1324 } 1325 } 1326 } 1327 1328 static void 1329 icmp_redirect_timeout(struct rtentry *rt, struct rttimer *r) 1330 { 1331 1332 KASSERT(rt != NULL); 1333 rt_assert_referenced(rt); 1334 1335 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 1336 (RTF_DYNAMIC | RTF_HOST)) { 1337 rtrequest(RTM_DELETE, rt_getkey(rt), 1338 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 1339 } 1340 } 1341 1342 /* 1343 * Perform rate limit check. 1344 * Returns 0 if it is okay to send the icmp packet. 1345 * Returns 1 if the router SHOULD NOT send this icmp packet due to rate 1346 * limitation. 1347 * 1348 * XXX per-destination/type check necessary? 1349 */ 1350 int 1351 icmp_ratelimit(const struct in_addr *dst, const int type, 1352 const int code) 1353 { 1354 1355 /* PPS limit */ 1356 if (!ppsratecheck(&icmperrppslim_last, &icmperrpps_count, 1357 icmperrppslim)) { 1358 /* The packet is subject to rate limit */ 1359 return 1; 1360 } 1361 1362 /* okay to send */ 1363 return 0; 1364 } 1365
Indexes created Mon Sep 29 21:09:56 GMT 2025