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 * 3. 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 * $FreeBSD: src/sys/netinet/ip_icmp.c,v 1.39.2.19 2003/01/24 05:11:34 sam Exp $ 31 */ 32 33 #include "opt_icmpprintfs.h" 34 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/malloc.h> /* for M_NOWAIT */ 38 #include <sys/mbuf.h> 39 #include <sys/protosw.h> 40 #include <sys/socket.h> 41 #include <sys/socketops.h> 42 #include <sys/time.h> 43 #include <sys/kernel.h> 44 #include <sys/sysctl.h> 45 #include <sys/in_cksum.h> 46 47 #include <machine/stdarg.h> 48 49 #include <net/if.h> 50 #include <net/if_types.h> 51 #include <net/netisr2.h> 52 #include <net/netmsg2.h> 53 #include <net/route.h> 54 55 #define _IP_VHL 56 #include <netinet/in.h> 57 #include <netinet/in_systm.h> 58 #include <netinet/in_var.h> 59 #include <netinet/ip.h> 60 #include <netinet/ip_icmp.h> 61 #include <netinet/ip_var.h> 62 #include <netinet/icmp_var.h> 63 64 /* 65 * ICMP routines: error generation, receive packet processing, and 66 * routines to turnaround packets back to the originator, and 67 * host table maintenance routines. 68 */ 69 70 struct icmpstat icmpstat; 71 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW, 72 &icmpstat, icmpstat, "ICMP statistics"); 73 74 static int icmpmaskrepl = 0; 75 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW, 76 &icmpmaskrepl, 0, "Allow replies to netmask requests"); 77 78 static int drop_redirect = 0; 79 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW, 80 &drop_redirect, 0, "Ignore ICMP redirects"); 81 82 static int log_redirect = 0; 83 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW, 84 &log_redirect, 0, "Enable output about ICMP redirects"); 85 86 static int discard_sourcequench = 1; 87 SYSCTL_INT(_net_inet_icmp, OID_AUTO, discard_sourcequench, CTLFLAG_RW, 88 &discard_sourcequench, 0, "Discard ICMP Source Quench"); 89 90 #ifdef ICMP_BANDLIM 91 92 /* 93 * ICMP error-response bandwidth limiting sysctl. If not enabled, sysctl 94 * variable content is -1 and read-only. 95 */ 96 97 static int icmplim = 200; 98 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW, 99 &icmplim, 0, "ICMP bandwidth limit"); 100 #else 101 102 static int icmplim = -1; 103 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD, 104 &icmplim, 0, "ICMP bandwidth limit"); 105 106 #endif 107 108 static int icmplim_output = 0; 109 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW, 110 &icmplim_output, 0, "Enable output about ICMP bandwidth limits"); 111 112 /* 113 * ICMP broadcast echo sysctl 114 */ 115 116 static int icmpbmcastecho = 0; 117 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW, 118 &icmpbmcastecho, 0, ""); 119 120 static char icmp_reply_src[IFNAMSIZ]; 121 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW, 122 icmp_reply_src, IFNAMSIZ, "icmp reply source for non-local packets."); 123 124 static int icmp_rfi; 125 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW, 126 &icmp_rfi, 0, "ICMP reply from incoming interface for " 127 "non-local packets"); 128 129 #ifdef ICMPPRINTFS 130 static int icmpprintfs = 0; 131 SYSCTL_INT(_net_inet_icmp, OID_AUTO, debug_prints, CTLFLAG_RW, 132 &icmpprintfs, 0, "extra ICMP debug prints"); 133 #endif 134 135 static void icmp_reflect (struct mbuf *); 136 static void icmp_send (struct mbuf *, struct mbuf *, struct route *); 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 destmtu) 146 { 147 struct ip *oip = mtod(n, struct ip *), *nip; 148 unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2; 149 struct icmp *icp; 150 struct mbuf *m; 151 unsigned icmplen; 152 153 #ifdef ICMPPRINTFS 154 if (icmpprintfs) 155 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code); 156 #endif 157 if (type != ICMP_REDIRECT) 158 icmpstat.icps_error++; 159 /* 160 * Don't send error if the original packet was encrypted. 161 * Don't send error if not the first fragment of message. 162 * Don't error if the old packet protocol was ICMP 163 * error message, only known informational types. 164 */ 165 if (n->m_flags & M_DECRYPTED) 166 goto freeit; 167 if (oip->ip_off & htons(~(IP_MF|IP_DF))) 168 goto freeit; 169 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 170 n->m_len >= oiplen + ICMP_MINLEN && 171 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) { 172 icmpstat.icps_oldicmp++; 173 goto freeit; 174 } 175 /* Don't send error in response to a multicast or broadcast packet */ 176 if (n->m_flags & (M_BCAST|M_MCAST)) 177 goto freeit; 178 /* 179 * First, formulate icmp message 180 */ 181 m = m_gethdr(M_NOWAIT, MT_HEADER); 182 if (m == NULL) 183 goto freeit; 184 icmplen = min(oiplen + 8, ntohs(oip->ip_len)); 185 if (icmplen < sizeof(struct ip)) 186 panic("icmp_error: bad length"); 187 m->m_len = icmplen + ICMP_MINLEN; 188 MH_ALIGN(m, m->m_len); 189 icp = mtod(m, struct icmp *); 190 if ((u_int)type > ICMP_MAXTYPE) 191 panic("icmp_error"); 192 icmpstat.icps_outhist[type]++; 193 icp->icmp_type = type; 194 if (type == ICMP_REDIRECT) 195 icp->icmp_gwaddr.s_addr = dest; 196 else { 197 icp->icmp_void = 0; 198 /* 199 * The following assignments assume an overlay with the 200 * zeroed icmp_void field. 201 */ 202 if (type == ICMP_PARAMPROB) { 203 icp->icmp_pptr = code; 204 code = 0; 205 } else if (type == ICMP_UNREACH && 206 code == ICMP_UNREACH_NEEDFRAG && destmtu) { 207 icp->icmp_nextmtu = htons(destmtu); 208 } 209 } 210 211 icp->icmp_code = code; 212 m_copydata(n, 0, icmplen, &icp->icmp_ip); 213 nip = &icp->icmp_ip; 214 215 /* 216 * Now, copy old ip header (without options) 217 * in front of icmp message. 218 */ 219 if (m->m_data - sizeof(struct ip) < m->m_pktdat) 220 panic("icmp len"); 221 m->m_data -= sizeof(struct ip); 222 m->m_len += sizeof(struct ip); 223 m->m_pkthdr.len = m->m_len; 224 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 225 nip = mtod(m, struct ip *); 226 bcopy(oip, nip, sizeof(struct ip)); 227 nip->ip_len = htons(m->m_len); 228 nip->ip_vhl = IP_VHL_BORING; 229 nip->ip_p = IPPROTO_ICMP; 230 nip->ip_tos = 0; 231 m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED; 232 icmp_reflect(m); 233 234 freeit: 235 m_freem(n); 236 } 237 238 static void 239 icmp_ctlinput_done_handler(netmsg_t nmsg) 240 { 241 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 242 struct mbuf *m = msg->m; 243 int hlen = msg->hlen; 244 245 rip_input(&m, &hlen, msg->proto); 246 } 247 248 static void 249 icmp_ctlinput_done(struct mbuf *m) 250 { 251 struct netmsg_ctlinput *msg = &m->m_hdr.mh_ctlmsg; 252 253 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 254 icmp_ctlinput_done_handler); 255 lwkt_sendmsg(netisr_cpuport(0), &msg->base.lmsg); 256 } 257 258 static void 259 icmp_mtudisc(struct mbuf *m, int hlen) 260 { 261 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 262 struct rtentry *rt; 263 struct icmp *icp; 264 265 KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr")); 266 267 icp = mtodoff(m, struct icmp *, hlen); 268 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 269 270 /* 271 * MTU discovery: 272 * If we got a needfrag and there is a host route to the original 273 * destination, and the MTU is not locked, then set the MTU in the 274 * route to the suggested new value (if given) and then notify as 275 * usual. The ULPs will notice that the MTU has changed and adapt 276 * accordingly. If no new MTU was suggested, then we guess a new 277 * one less than the current value. If the new MTU is unreasonably 278 * small (arbitrarily set at 296), then we reset the MTU to the 279 * interface value and enable the lock bit, indicating that we are 280 * no longer doing MTU discovery. 281 */ 282 rt = rtpurelookup((struct sockaddr *)&icmpsrc); 283 if (rt != NULL && (rt->rt_flags & RTF_HOST) && 284 !(rt->rt_rmx.rmx_locks & RTV_MTU)) { 285 #ifdef DEBUG_MTUDISC 286 char src_buf[INET_ADDRSTRLEN]; 287 #endif 288 int mtu; 289 290 mtu = ntohs(icp->icmp_nextmtu); 291 if (!mtu) 292 mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu, 1); 293 #ifdef DEBUG_MTUDISC 294 kprintf("MTU for %s reduced to %d\n", 295 inet_ntop(AF_INET, &icmpsrc.sin_addr, 296 src_buf, INET_ADDRSTRLEN), mtu); 297 #endif 298 if (mtu < 296) { 299 /* rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu; */ 300 rt->rt_rmx.rmx_locks |= RTV_MTU; 301 } else if (rt->rt_rmx.rmx_mtu > mtu) { 302 rt->rt_rmx.rmx_mtu = mtu; 303 } 304 } 305 if (rt != NULL) 306 --rt->rt_refcnt; 307 308 /* 309 * XXX if the packet contains [IPv4 AH TCP], we can't make a 310 * notification to TCP layer. 311 */ 312 so_pr_ctlinput_direct(&inetsw[ip_protox[icp->icmp_ip.ip_p]], 313 PRC_MSGSIZE, (struct sockaddr *)&icmpsrc, &icp->icmp_ip); 314 } 315 316 static void 317 icmp_mtudisc_handler(netmsg_t nmsg) 318 { 319 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 320 int nextcpu; 321 322 ASSERT_NETISR_NCPUS(mycpuid); 323 324 icmp_mtudisc(msg->m, msg->hlen); 325 326 nextcpu = mycpuid + 1; 327 if (nextcpu < netisr_ncpus) 328 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg); 329 else 330 icmp_ctlinput_done(msg->m); 331 } 332 333 static boolean_t 334 icmp_mtudisc_start(struct mbuf *m, int hlen, int proto) 335 { 336 struct netmsg_ctlinput *msg; 337 338 ASSERT_NETISR0; 339 340 icmp_mtudisc(m, hlen); 341 342 if (netisr_ncpus == 1) { 343 /* There is only one netisr; done */ 344 return FALSE; 345 } 346 347 msg = &m->m_hdr.mh_ctlmsg; 348 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 349 icmp_mtudisc_handler); 350 msg->m = m; 351 msg->cmd = PRC_MSGSIZE; 352 msg->hlen = hlen; 353 msg->proto = proto; 354 355 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg); 356 return TRUE; 357 } 358 359 static void 360 icmp_ctlinput(struct mbuf *m, int cmd, int hlen) 361 { 362 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 363 struct icmp *icp; 364 365 KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr")); 366 367 icp = mtodoff(m, struct icmp *, hlen); 368 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 369 370 /* 371 * XXX if the packet contains [IPv4 AH TCP], we can't make a 372 * notification to TCP layer. 373 */ 374 so_pr_ctlinput_direct(&inetsw[ip_protox[icp->icmp_ip.ip_p]], 375 cmd, (struct sockaddr *)&icmpsrc, &icp->icmp_ip); 376 } 377 378 static void 379 icmp_ctlinput_handler(netmsg_t nmsg) 380 { 381 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 382 383 ASSERT_NETISR_NCPUS(mycpuid); 384 385 icmp_ctlinput(msg->m, msg->cmd, msg->hlen); 386 icmp_ctlinput_done(msg->m); 387 } 388 389 static void 390 icmp_ctlinput_start(struct mbuf *m, struct lwkt_port *port, 391 int cmd, int hlen, int proto) 392 { 393 struct netmsg_ctlinput *msg; 394 395 KASSERT(&curthread->td_msgport != port, 396 ("send icmp ctlinput to the current netisr")); 397 398 msg = &m->m_hdr.mh_ctlmsg; 399 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 400 icmp_ctlinput_handler); 401 msg->m = m; 402 msg->cmd = cmd; 403 msg->hlen = hlen; 404 msg->proto = proto; 405 406 lwkt_sendmsg(port, &msg->base.lmsg); 407 } 408 409 static void 410 icmp_ctlinput_global_handler(netmsg_t nmsg) 411 { 412 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 413 int nextcpu; 414 415 ASSERT_NETISR_NCPUS(mycpuid); 416 417 icmp_ctlinput(msg->m, msg->cmd, msg->hlen); 418 419 nextcpu = mycpuid + 1; 420 if (nextcpu < netisr_ncpus) 421 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg); 422 else 423 icmp_ctlinput_done(msg->m); 424 } 425 426 static void 427 icmp_ctlinput_global_start(struct mbuf *m, int cmd, int hlen, int proto) 428 { 429 struct netmsg_ctlinput *msg; 430 431 ASSERT_NETISR0; 432 KASSERT(netisr_ncpus > 1, ("there is only 1 netisr cpu")); 433 434 icmp_ctlinput(m, cmd, hlen); 435 436 msg = &m->m_hdr.mh_ctlmsg; 437 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 438 icmp_ctlinput_global_handler); 439 msg->m = m; 440 msg->cmd = cmd; 441 msg->hlen = hlen; 442 msg->proto = proto; 443 444 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg); 445 } 446 447 #define ICMP_RTREDIRECT_FLAGS (RTF_GATEWAY | RTF_HOST) 448 449 static void 450 icmp_redirect(struct mbuf *m, int hlen, boolean_t prt) 451 { 452 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 453 struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET }; 454 struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET }; 455 struct icmp *icp; 456 struct ip *ip; 457 458 KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr")); 459 460 ip = mtod(m, struct ip *); 461 icp = mtodoff(m, struct icmp *, hlen); 462 463 /* 464 * Short circuit routing redirects to force immediate change 465 * in the kernel's routing tables. The message is also handed 466 * to anyone listening on a raw socket (e.g. the routing daemon 467 * for use in updating its tables). 468 */ 469 #ifdef ICMPPRINTFS 470 if (icmpprintfs && prt) { 471 char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN]; 472 473 kprintf("redirect dst %s to %s\n", 474 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, 475 dst_buf, INET_ADDRSTRLEN), 476 inet_ntop(AF_INET, &icp->icmp_gwaddr, 477 gw_buf, INET_ADDRSTRLEN)); 478 } 479 #endif 480 icmpgw.sin_addr = ip->ip_src; 481 icmpdst.sin_addr = icp->icmp_gwaddr; 482 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 483 rtredirect_oncpu((struct sockaddr *)&icmpsrc, 484 (struct sockaddr *)&icmpdst, NULL, ICMP_RTREDIRECT_FLAGS, 485 (struct sockaddr *)&icmpgw); 486 kpfctlinput_direct(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); 487 } 488 489 static void 490 icmp_redirect_done_handler(netmsg_t nmsg) 491 { 492 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 493 struct mbuf *m = msg->m; 494 int hlen = msg->hlen; 495 496 rip_input(&m, &hlen, msg->proto); 497 } 498 499 static void 500 icmp_redirect_done(struct mbuf *m, int hlen, boolean_t dispatch_rip) 501 { 502 struct rt_addrinfo rtinfo; 503 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 504 struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET }; 505 struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET }; 506 struct icmp *icp; 507 struct ip *ip; 508 509 ip = mtod(m, struct ip *); 510 icp = mtodoff(m, struct icmp *, hlen); 511 512 icmpgw.sin_addr = ip->ip_src; 513 icmpdst.sin_addr = icp->icmp_gwaddr; 514 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 515 516 bzero(&rtinfo, sizeof(struct rt_addrinfo)); 517 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&icmpsrc; 518 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&icmpdst; 519 rtinfo.rti_info[RTAX_NETMASK] = NULL; 520 rtinfo.rti_info[RTAX_AUTHOR] = (struct sockaddr *)&icmpgw; 521 rt_missmsg(RTM_REDIRECT, &rtinfo, ICMP_RTREDIRECT_FLAGS, 0); 522 523 if (dispatch_rip) { 524 struct netmsg_ctlinput *msg = &m->m_hdr.mh_ctlmsg; 525 526 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 527 icmp_redirect_done_handler); 528 lwkt_sendmsg(netisr_cpuport(0), &msg->base.lmsg); 529 } 530 } 531 532 static void 533 icmp_redirect_handler(netmsg_t nmsg) 534 { 535 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 536 int nextcpu; 537 538 ASSERT_NETISR_NCPUS(mycpuid); 539 540 icmp_redirect(msg->m, msg->hlen, FALSE); 541 542 nextcpu = mycpuid + 1; 543 if (nextcpu < netisr_ncpus) 544 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg); 545 else 546 icmp_redirect_done(msg->m, msg->hlen, TRUE); 547 } 548 549 static boolean_t 550 icmp_redirect_start(struct mbuf *m, int hlen, int proto) 551 { 552 struct netmsg_ctlinput *msg; 553 554 ASSERT_NETISR0; 555 556 icmp_redirect(m, hlen, TRUE); 557 558 if (netisr_ncpus == 1) { 559 /* There is only one netisr; done */ 560 icmp_redirect_done(m, hlen, FALSE); 561 return FALSE; 562 } 563 564 msg = &m->m_hdr.mh_ctlmsg; 565 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 566 icmp_redirect_handler); 567 msg->m = m; 568 msg->cmd = PRC_REDIRECT_HOST; 569 msg->hlen = hlen; 570 msg->proto = proto; 571 572 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg); 573 return TRUE; 574 } 575 576 /* 577 * Process a received ICMP message. 578 */ 579 int 580 icmp_input(struct mbuf **mp, int *offp, int proto) 581 { 582 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 583 struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET }; 584 struct icmp *icp; 585 struct in_ifaddr *ia; 586 struct mbuf *m = *mp; 587 struct ip *ip; 588 int icmplen; 589 int i, hlen; 590 int code; 591 592 ASSERT_NETISR0; 593 594 *mp = NULL; 595 hlen = *offp; 596 ip = mtod(m, struct ip *); 597 icmplen = ntohs(ip->ip_len) - hlen; 598 599 /* 600 * Locate icmp structure in mbuf, and check 601 * that not corrupted and of at least minimum length. 602 */ 603 #ifdef ICMPPRINTFS 604 if (icmpprintfs) { 605 char src_buf[INET_ADDRSTRLEN], dst_buf[INET_ADDRSTRLEN]; 606 607 kprintf("icmp_input from %s to %s, len %d\n", 608 inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN), 609 inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN), 610 icmplen); 611 } 612 #endif 613 if (icmplen < ICMP_MINLEN) { 614 icmpstat.icps_tooshort++; 615 goto freeit; 616 } 617 i = hlen + min(icmplen, ICMP_ADVLENMIN); 618 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 619 icmpstat.icps_tooshort++; 620 return(IPPROTO_DONE); 621 } 622 ip = mtod(m, struct ip *); 623 624 if (in_cksum_skip(m, hlen + icmplen, hlen)) { 625 icmpstat.icps_checksum++; 626 goto freeit; 627 } 628 icp = (struct icmp *)((caddr_t)ip + hlen); 629 630 #ifdef ICMPPRINTFS 631 if (icmpprintfs) 632 kprintf("icmp_input, type %d code %d\n", icp->icmp_type, 633 icp->icmp_code); 634 #endif 635 636 /* 637 * Message type specific processing. 638 */ 639 if (icp->icmp_type > ICMP_MAXTYPE) 640 goto raw; 641 icmpstat.icps_inhist[icp->icmp_type]++; 642 code = icp->icmp_code; 643 switch (icp->icmp_type) { 644 645 case ICMP_UNREACH: 646 switch (code) { 647 case ICMP_UNREACH_NET: 648 case ICMP_UNREACH_HOST: 649 case ICMP_UNREACH_SRCFAIL: 650 case ICMP_UNREACH_NET_UNKNOWN: 651 case ICMP_UNREACH_HOST_UNKNOWN: 652 case ICMP_UNREACH_ISOLATED: 653 case ICMP_UNREACH_TOSNET: 654 case ICMP_UNREACH_TOSHOST: 655 case ICMP_UNREACH_HOST_PRECEDENCE: 656 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 657 code = PRC_UNREACH_NET; 658 break; 659 660 case ICMP_UNREACH_NEEDFRAG: 661 code = PRC_MSGSIZE; 662 break; 663 664 /* 665 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. 666 * Treat subcodes 2,3 as immediate RST 667 */ 668 case ICMP_UNREACH_PROTOCOL: 669 case ICMP_UNREACH_PORT: 670 code = PRC_UNREACH_PORT; 671 break; 672 673 case ICMP_UNREACH_NET_PROHIB: 674 case ICMP_UNREACH_HOST_PROHIB: 675 case ICMP_UNREACH_FILTER_PROHIB: 676 code = PRC_UNREACH_ADMIN_PROHIB; 677 break; 678 679 default: 680 goto badcode; 681 } 682 goto deliver; 683 684 case ICMP_TIMXCEED: 685 if (code > 1) 686 goto badcode; 687 code += PRC_TIMXCEED_INTRANS; 688 goto deliver; 689 690 case ICMP_PARAMPROB: 691 if (code > 1) 692 goto badcode; 693 code = PRC_PARAMPROB; 694 goto deliver; 695 696 case ICMP_SOURCEQUENCH: 697 if (code) 698 goto badcode; 699 if (discard_sourcequench) 700 break; 701 code = PRC_QUENCH; 702 deliver: 703 /* 704 * Problem with datagram; advise higher level routines. 705 */ 706 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 707 IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) { 708 icmpstat.icps_badlen++; 709 goto freeit; 710 } 711 /* Discard ICMP's in response to multicast packets */ 712 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 713 goto badcode; 714 #ifdef ICMPPRINTFS 715 if (icmpprintfs) 716 kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 717 #endif 718 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 719 720 /* 721 * MTU discovery 722 */ 723 if (code == PRC_MSGSIZE) { 724 /* Run MTU discovery in all netisrs */ 725 if (icmp_mtudisc_start(m, hlen, proto)) { 726 /* Forwarded; done */ 727 return IPPROTO_DONE; 728 } 729 /* Move on; run rip_input() directly */ 730 } else { 731 struct protosw *pr; 732 struct lwkt_port *port; 733 int cpu; 734 735 pr = &inetsw[ip_protox[icp->icmp_ip.ip_p]]; 736 port = so_pr_ctlport(pr, code, 737 (struct sockaddr *)&icmpsrc, &icp->icmp_ip, &cpu); 738 if (port != NULL) { 739 if (cpu == netisr_ncpus) { 740 if (netisr_ncpus > 1) { 741 /* 742 * Run pr_ctlinput in all 743 * netisrs 744 */ 745 icmp_ctlinput_global_start(m, 746 code, hlen, proto); 747 return IPPROTO_DONE; 748 } 749 /* 750 * There is only one netisr; run 751 * pr_ctlinput directly. 752 */ 753 } else if (cpu != mycpuid) { 754 /* 755 * Send to the target netisr to run 756 * pr_ctlinput. 757 */ 758 icmp_ctlinput_start(m, port, 759 code, hlen, proto); 760 return IPPROTO_DONE; 761 } 762 763 /* 764 * The target netisr is this netisr. 765 * 766 * XXX if the packet contains [IPv4 AH TCP], 767 * we can't make a notification to TCP layer. 768 */ 769 so_pr_ctlinput_direct(pr, code, 770 (struct sockaddr *)&icmpsrc, &icp->icmp_ip); 771 } 772 /* Move on; run rip_input() directly */ 773 } 774 break; 775 badcode: 776 icmpstat.icps_badcode++; 777 break; 778 779 case ICMP_ECHO: 780 if (!icmpbmcastecho 781 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 782 icmpstat.icps_bmcastecho++; 783 break; 784 } 785 icp->icmp_type = ICMP_ECHOREPLY; 786 #ifdef ICMP_BANDLIM 787 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 788 goto freeit; 789 else 790 #endif 791 goto reflect; 792 793 case ICMP_TSTAMP: 794 if (!icmpbmcastecho 795 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 796 icmpstat.icps_bmcasttstamp++; 797 break; 798 } 799 if (icmplen < ICMP_TSLEN) { 800 icmpstat.icps_badlen++; 801 break; 802 } 803 icp->icmp_type = ICMP_TSTAMPREPLY; 804 icp->icmp_rtime = iptime(); 805 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 806 #ifdef ICMP_BANDLIM 807 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 808 goto freeit; 809 else 810 #endif 811 goto reflect; 812 813 case ICMP_MASKREQ: 814 if (icmpmaskrepl == 0) 815 break; 816 /* 817 * We are not able to respond with all ones broadcast 818 * unless we receive it over a point-to-point interface. 819 */ 820 if (icmplen < ICMP_MASKLEN) 821 break; 822 switch (ip->ip_dst.s_addr) { 823 824 case INADDR_BROADCAST: 825 case INADDR_ANY: 826 icmpdst.sin_addr = ip->ip_src; 827 break; 828 829 default: 830 icmpdst.sin_addr = ip->ip_dst; 831 } 832 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 833 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 834 if (ia == NULL) 835 break; 836 if (ia->ia_ifp == 0) 837 break; 838 icp->icmp_type = ICMP_MASKREPLY; 839 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 840 if (ip->ip_src.s_addr == 0) { 841 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 842 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 843 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 844 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 845 } 846 reflect: 847 icmpstat.icps_reflect++; 848 icmpstat.icps_outhist[icp->icmp_type]++; 849 icmp_reflect(m); 850 return(IPPROTO_DONE); 851 852 case ICMP_REDIRECT: 853 if (log_redirect) { 854 char src_buf[INET_ADDRSTRLEN]; 855 char dst_buf[INET_ADDRSTRLEN]; 856 char gwy_buf[INET_ADDRSTRLEN]; 857 858 kprintf("icmp redirect from %s: %s => %s\n", 859 inet_ntop(AF_INET, &ip->ip_src, 860 src_buf, INET_ADDRSTRLEN), 861 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, 862 dst_buf, INET_ADDRSTRLEN), 863 inet_ntop(AF_INET, &icp->icmp_gwaddr, 864 gwy_buf, INET_ADDRSTRLEN)); 865 } 866 if (drop_redirect) 867 break; 868 if (code > 3) 869 goto badcode; 870 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 871 IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) { 872 icmpstat.icps_badlen++; 873 break; 874 } 875 #ifdef ICMPPRINTFS 876 if (icmpprintfs) { 877 char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN]; 878 879 kprintf("redirect dst %s to %s\n", 880 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, 881 dst_buf, INET_ADDRSTRLEN), 882 inet_ntop(AF_INET, &icp->icmp_gwaddr, 883 gw_buf, INET_ADDRSTRLEN)); 884 } 885 #endif 886 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 887 888 /* Run redirect in all netisrs */ 889 if (icmp_redirect_start(m, hlen, proto)) { 890 /* Forwarded; done */ 891 return IPPROTO_DONE; 892 } 893 /* Move on; run rip_input() directly */ 894 break; 895 896 /* 897 * No kernel processing for the following; 898 * just fall through to send to raw listener. 899 */ 900 case ICMP_ECHOREPLY: 901 case ICMP_ROUTERADVERT: 902 case ICMP_ROUTERSOLICIT: 903 case ICMP_TSTAMPREPLY: 904 case ICMP_IREQREPLY: 905 case ICMP_MASKREPLY: 906 default: 907 break; 908 } 909 910 raw: 911 *mp = m; 912 rip_input(mp, offp, proto); 913 return(IPPROTO_DONE); 914 915 freeit: 916 m_freem(m); 917 return(IPPROTO_DONE); 918 } 919 920 /* 921 * Reflect the ip packet back to the source 922 */ 923 static void 924 icmp_reflect(struct mbuf *m) 925 { 926 struct ip *ip = mtod(m, struct ip *); 927 struct in_ifaddr *ia; 928 struct in_ifaddr_container *iac; 929 struct ifaddr_container *ifac; 930 struct ifnet *ifp; 931 struct in_addr t; 932 struct mbuf *opts = NULL; 933 int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 934 struct route *ro = NULL, rt; 935 936 if (!in_canforward(ip->ip_src) && 937 ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) != 938 (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { 939 m_freem(m); /* Bad return address */ 940 icmpstat.icps_badaddr++; 941 goto done; /* Ip_output() will check for broadcast */ 942 } 943 t = ip->ip_dst; 944 ip->ip_dst = ip->ip_src; 945 946 ro = &rt; 947 bzero(ro, sizeof *ro); 948 949 /* 950 * If the incoming packet was addressed directly to us, 951 * use dst as the src for the reply. Otherwise (broadcast 952 * or anonymous), use the address which corresponds 953 * to the incoming interface. 954 */ 955 ia = NULL; 956 LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) { 957 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) { 958 ia = iac->ia; 959 goto match; 960 } 961 } 962 ifp = m->m_pkthdr.rcvif; 963 if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) { 964 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 965 struct ifaddr *ifa = ifac->ifa; 966 967 if (ifa->ifa_addr->sa_family != AF_INET) 968 continue; 969 ia = ifatoia(ifa); 970 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 971 t.s_addr) 972 goto match; 973 } 974 } 975 /* 976 * If the packet was transiting through us, use the address of 977 * the interface the packet came through in. If that interface 978 * doesn't have a suitable IP address, the normal selection 979 * criteria apply. 980 */ 981 if (icmp_rfi && ifp != NULL) { 982 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 983 struct ifaddr *ifa = ifac->ifa; 984 985 if (ifa->ifa_addr->sa_family != AF_INET) 986 continue; 987 ia = ifatoia(ifa); 988 goto match; 989 } 990 } 991 /* 992 * If the incoming packet was not addressed directly to us, use 993 * designated interface for icmp replies specified by sysctl 994 * net.inet.icmp.reply_src (default not set). Otherwise continue 995 * with normal source selection. 996 */ 997 if (icmp_reply_src[0] != '\0' && 998 (ifp = ifunit_netisr(icmp_reply_src))) { 999 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 1000 struct ifaddr *ifa = ifac->ifa; 1001 1002 if (ifa->ifa_addr->sa_family != AF_INET) 1003 continue; 1004 ia = ifatoia(ifa); 1005 goto match; 1006 } 1007 } 1008 /* 1009 * If the packet was transiting through us, use the address of 1010 * the interface that is the closest to the packet source. 1011 * When we don't have a route back to the packet source, stop here 1012 * and drop the packet. 1013 */ 1014 ia = ip_rtaddr(ip->ip_dst, ro); 1015 if (ia == NULL) { 1016 m_freem(m); 1017 icmpstat.icps_noroute++; 1018 goto done; 1019 } 1020 match: 1021 t = IA_SIN(ia)->sin_addr; 1022 ip->ip_src = t; 1023 ip->ip_ttl = ip_defttl; 1024 1025 if (optlen > 0) { 1026 u_char *cp; 1027 int opt, cnt; 1028 u_int len; 1029 1030 /* 1031 * Retrieve any source routing from the incoming packet; 1032 * add on any record-route or timestamp options. 1033 */ 1034 cp = (u_char *) (ip + 1); 1035 if ((opts = ip_srcroute(m)) == NULL && 1036 (opts = m_gethdr(M_NOWAIT, MT_HEADER))) { 1037 opts->m_len = sizeof(struct in_addr); 1038 mtod(opts, struct in_addr *)->s_addr = 0; 1039 } 1040 if (opts) { 1041 #ifdef ICMPPRINTFS 1042 if (icmpprintfs) 1043 kprintf("icmp_reflect optlen %d rt %d => ", 1044 optlen, opts->m_len); 1045 #endif 1046 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 1047 opt = cp[IPOPT_OPTVAL]; 1048 if (opt == IPOPT_EOL) 1049 break; 1050 if (opt == IPOPT_NOP) 1051 len = 1; 1052 else { 1053 if (cnt < IPOPT_OLEN + sizeof *cp) 1054 break; 1055 len = cp[IPOPT_OLEN]; 1056 if (len < IPOPT_OLEN + sizeof *cp || 1057 len > cnt) 1058 break; 1059 } 1060 /* 1061 * Should check for overflow, but it 1062 * "can't happen". 1063 */ 1064 if (opt == IPOPT_RR || opt == IPOPT_TS || 1065 opt == IPOPT_SECURITY) { 1066 bcopy(cp, 1067 mtod(opts, caddr_t) + opts->m_len, 1068 len); 1069 opts->m_len += len; 1070 } 1071 } 1072 /* Terminate & pad, if necessary */ 1073 cnt = opts->m_len % 4; 1074 if (cnt) { 1075 for (; cnt < 4; cnt++) { 1076 *(mtod(opts, caddr_t) + opts->m_len) = 1077 IPOPT_EOL; 1078 opts->m_len++; 1079 } 1080 } 1081 #ifdef ICMPPRINTFS 1082 if (icmpprintfs) 1083 kprintf("%d\n", opts->m_len); 1084 #endif 1085 } 1086 /* 1087 * Now strip out original options by copying rest of first 1088 * mbuf's data back, and adjust the IP length. 1089 */ 1090 ip->ip_len = htons(ntohs(ip->ip_len) - optlen); 1091 ip->ip_vhl = IP_VHL_BORING; 1092 m->m_len -= optlen; 1093 if (m->m_flags & M_PKTHDR) 1094 m->m_pkthdr.len -= optlen; 1095 optlen += sizeof(struct ip); 1096 bcopy((caddr_t)ip + optlen, ip + 1, 1097 m->m_len - sizeof(struct ip)); 1098 } 1099 m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED; 1100 m->m_flags &= ~(M_BCAST|M_MCAST); 1101 icmp_send(m, opts, ro); 1102 done: 1103 if (opts) 1104 m_free(opts); 1105 if (ro && ro->ro_rt) 1106 RTFREE(ro->ro_rt); 1107 } 1108 1109 /* 1110 * Send an icmp packet back to the ip level, 1111 * after supplying a checksum. 1112 */ 1113 static void 1114 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt) 1115 { 1116 struct ip *ip = mtod(m, struct ip *); 1117 struct icmp *icp; 1118 int hlen; 1119 1120 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1121 m->m_data += hlen; 1122 m->m_len -= hlen; 1123 icp = mtod(m, struct icmp *); 1124 icp->icmp_cksum = 0; 1125 icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen); 1126 m->m_data -= hlen; 1127 m->m_len += hlen; 1128 m->m_pkthdr.rcvif = NULL; 1129 #ifdef ICMPPRINTFS 1130 if (icmpprintfs) { 1131 char dst_buf[INET_ADDRSTRLEN], src_buf[INET_ADDRSTRLEN]; 1132 1133 kprintf("icmp_send dst %s src %s\n", 1134 inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN), 1135 inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN)); 1136 } 1137 #endif 1138 ip_output(m, opts, rt, 0, NULL, NULL); 1139 } 1140 1141 n_time 1142 iptime(void) 1143 { 1144 struct timeval atv; 1145 u_long t; 1146 1147 getmicrotime(&atv); 1148 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1149 return (htonl(t)); 1150 } 1151 1152 #if 1 1153 /* 1154 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1155 * given current value MTU. If DIR is less than zero, a larger plateau 1156 * is returned; otherwise, a smaller value is returned. 1157 */ 1158 int 1159 ip_next_mtu(int mtu, int dir) 1160 { 1161 static int mtutab[] = { 1162 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296, 1163 68, 0 1164 }; 1165 int i; 1166 1167 for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) { 1168 if (mtu >= mtutab[i]) 1169 break; 1170 } 1171 1172 if (dir < 0) { 1173 if (i == 0) { 1174 return 0; 1175 } else { 1176 return mtutab[i - 1]; 1177 } 1178 } else { 1179 if (mtutab[i] == 0) { 1180 return 0; 1181 } else if(mtu > mtutab[i]) { 1182 return mtutab[i]; 1183 } else { 1184 return mtutab[i + 1]; 1185 } 1186 } 1187 } 1188 #endif 1189 1190 #ifdef ICMP_BANDLIM 1191 /* 1192 * badport_bandlim() - check for ICMP bandwidth limit 1193 * 1194 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1195 * hit our bandwidth limit and it is not ok. 1196 * 1197 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1198 * 1199 * For now we separate the TCP and UDP subsystems w/ different 'which' 1200 * values. We may eventually remove this separation (and simplify the 1201 * code further). 1202 * 1203 * Note that the printing of the error message is delayed so we can 1204 * properly print the icmp error rate that the system was trying to do 1205 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1206 * the 'final' error, but it doesn't make sense to solve the printing 1207 * delay with more complex code. 1208 */ 1209 int 1210 badport_bandlim(int which) 1211 { 1212 static int lticks[BANDLIM_MAX + 1]; 1213 static int lpackets[BANDLIM_MAX + 1]; 1214 int dticks; 1215 const char *bandlimittype[] = { 1216 "Limiting icmp unreach response", 1217 "Limiting icmp ping response", 1218 "Limiting icmp tstamp response", 1219 "Limiting closed port RST response", 1220 "Limiting open port RST response" 1221 }; 1222 1223 /* 1224 * Return ok status if feature disabled or argument out of 1225 * ranage. 1226 */ 1227 1228 if (icmplim <= 0 || which > BANDLIM_MAX || which < 0) 1229 return(0); 1230 dticks = ticks - lticks[which]; 1231 1232 /* 1233 * reset stats when cumulative dt exceeds one second. 1234 */ 1235 1236 if ((unsigned int)dticks > hz) { 1237 if (lpackets[which] > icmplim && icmplim_output) { 1238 kprintf("%s from %d to %d packets per second\n", 1239 bandlimittype[which], 1240 lpackets[which], 1241 icmplim 1242 ); 1243 } 1244 lticks[which] = ticks; 1245 lpackets[which] = 0; 1246 } 1247 1248 /* 1249 * bump packet count 1250 */ 1251 1252 if (++lpackets[which] > icmplim) { 1253 return(-1); 1254 } 1255 return(0); 1256 } 1257 #endif 1258