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 <sys/param.h> 34 #include <sys/systm.h> 35 #include <sys/mbuf.h> 36 #include <sys/protosw.h> 37 #include <sys/socket.h> 38 #include <sys/socketops.h> 39 #include <sys/time.h> 40 #include <sys/kernel.h> 41 #include <sys/sysctl.h> 42 #include <sys/in_cksum.h> 43 44 #include <machine/stdarg.h> 45 46 #include <net/if.h> 47 #include <net/if_types.h> 48 #include <net/netisr2.h> 49 #include <net/netmsg2.h> 50 #include <net/route.h> 51 52 #define _IP_VHL 53 #include <netinet/in.h> 54 #include <netinet/in_systm.h> 55 #include <netinet/in_var.h> 56 #include <netinet/ip.h> 57 #include <netinet/ip_icmp.h> 58 #include <netinet/ip_var.h> 59 #include <netinet/icmp_var.h> 60 61 /* 62 * ICMP routines: error generation, receive packet processing, and 63 * routines to turnaround packets back to the originator, and 64 * host table maintenance routines. 65 */ 66 67 struct icmpstat icmpstat; 68 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW, 69 &icmpstat, icmpstat, "ICMP statistics"); 70 71 static int icmpmaskrepl = 0; 72 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW, 73 &icmpmaskrepl, 0, "Allow replies to netmask requests"); 74 75 static int drop_redirect = 0; 76 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW, 77 &drop_redirect, 0, "Ignore ICMP redirects"); 78 79 static int log_redirect = 0; 80 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW, 81 &log_redirect, 0, "Enable output about ICMP redirects"); 82 83 static int discard_sourcequench = 1; 84 SYSCTL_INT(_net_inet_icmp, OID_AUTO, discard_sourcequench, CTLFLAG_RW, 85 &discard_sourcequench, 0, "Discard ICMP Source Quench"); 86 87 #ifdef ICMP_BANDLIM 88 89 /* 90 * ICMP error-response bandwidth limiting sysctl. If not enabled, sysctl 91 * variable content is -1 and read-only. 92 */ 93 94 static int icmplim = 200; 95 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW, 96 &icmplim, 0, "ICMP bandwidth limit"); 97 #else 98 99 static int icmplim = -1; 100 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD, 101 &icmplim, 0, "ICMP bandwidth limit"); 102 103 #endif 104 105 static int icmplim_output = 0; 106 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW, 107 &icmplim_output, 0, "Enable output about ICMP bandwidth limits"); 108 109 /* 110 * ICMP broadcast echo sysctl 111 */ 112 113 static int icmpbmcastecho = 0; 114 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW, 115 &icmpbmcastecho, 0, ""); 116 117 static char icmp_reply_src[IFNAMSIZ]; 118 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW, 119 icmp_reply_src, IFNAMSIZ, "icmp reply source for non-local packets."); 120 121 static int icmp_rfi; 122 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW, 123 &icmp_rfi, 0, "ICMP reply from incoming interface for " 124 "non-local packets"); 125 126 #ifdef ICMPPRINTFS 127 static int icmpprintfs = 0; 128 SYSCTL_INT(_net_inet_icmp, OID_AUTO, debug_prints, CTLFLAG_RW, 129 &icmpprintfs, 0, "extra ICMP debug prints"); 130 #endif 131 132 static void icmp_reflect (struct mbuf *); 133 static void icmp_send (struct mbuf *, struct mbuf *, struct route *); 134 135 extern struct protosw inetsw[]; 136 137 /* 138 * Generate an error packet of type error 139 * in response to bad packet ip. 140 */ 141 void 142 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu) 143 { 144 struct ip *oip = mtod(n, struct ip *), *nip; 145 unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2; 146 struct icmp *icp; 147 struct mbuf *m; 148 unsigned icmplen; 149 150 #ifdef ICMPPRINTFS 151 if (icmpprintfs) 152 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code); 153 #endif 154 if (type != ICMP_REDIRECT) 155 icmpstat.icps_error++; 156 /* 157 * Don't send error if the original packet was encrypted. 158 * Don't send error if not the first fragment of message. 159 * Don't error if the old packet protocol was ICMP 160 * error message, only known informational types. 161 */ 162 if (n->m_flags & M_DECRYPTED) 163 goto freeit; 164 if (oip->ip_off &~ (IP_MF|IP_DF)) 165 goto freeit; 166 if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT && 167 n->m_len >= oiplen + ICMP_MINLEN && 168 !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) { 169 icmpstat.icps_oldicmp++; 170 goto freeit; 171 } 172 /* Don't send error in response to a multicast or broadcast packet */ 173 if (n->m_flags & (M_BCAST|M_MCAST)) 174 goto freeit; 175 /* 176 * First, formulate icmp message 177 */ 178 m = m_gethdr(M_NOWAIT, MT_HEADER); 179 if (m == NULL) 180 goto freeit; 181 icmplen = min(oiplen + 8, oip->ip_len); 182 if (icmplen < sizeof(struct ip)) 183 panic("icmp_error: bad length"); 184 m->m_len = icmplen + ICMP_MINLEN; 185 MH_ALIGN(m, m->m_len); 186 icp = mtod(m, struct icmp *); 187 if ((u_int)type > ICMP_MAXTYPE) 188 panic("icmp_error"); 189 icmpstat.icps_outhist[type]++; 190 icp->icmp_type = type; 191 if (type == ICMP_REDIRECT) 192 icp->icmp_gwaddr.s_addr = dest; 193 else { 194 icp->icmp_void = 0; 195 /* 196 * The following assignments assume an overlay with the 197 * zeroed icmp_void field. 198 */ 199 if (type == ICMP_PARAMPROB) { 200 icp->icmp_pptr = code; 201 code = 0; 202 } else if (type == ICMP_UNREACH && 203 code == ICMP_UNREACH_NEEDFRAG && destmtu) { 204 icp->icmp_nextmtu = htons(destmtu); 205 } 206 } 207 208 icp->icmp_code = code; 209 m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip); 210 nip = &icp->icmp_ip; 211 212 /* 213 * Convert fields to network representation. 214 */ 215 nip->ip_len = htons(nip->ip_len); 216 nip->ip_off = htons(nip->ip_off); 217 218 /* 219 * Now, copy old ip header (without options) 220 * in front of icmp message. 221 */ 222 if (m->m_data - sizeof(struct ip) < m->m_pktdat) 223 panic("icmp len"); 224 m->m_data -= sizeof(struct ip); 225 m->m_len += sizeof(struct ip); 226 m->m_pkthdr.len = m->m_len; 227 m->m_pkthdr.rcvif = n->m_pkthdr.rcvif; 228 nip = mtod(m, struct ip *); 229 bcopy(oip, nip, sizeof(struct ip)); 230 nip->ip_len = m->m_len; 231 nip->ip_vhl = IP_VHL_BORING; 232 nip->ip_p = IPPROTO_ICMP; 233 nip->ip_tos = 0; 234 m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED; 235 icmp_reflect(m); 236 237 freeit: 238 m_freem(n); 239 } 240 241 static void 242 icmp_ctlinput_done_handler(netmsg_t nmsg) 243 { 244 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 245 struct mbuf *m = msg->m; 246 int hlen = msg->hlen; 247 248 rip_input(&m, &hlen, msg->proto); 249 } 250 251 static void 252 icmp_ctlinput_done(struct mbuf *m) 253 { 254 struct netmsg_ctlinput *msg = &m->m_hdr.mh_ctlmsg; 255 256 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 257 icmp_ctlinput_done_handler); 258 lwkt_sendmsg(netisr_cpuport(0), &msg->base.lmsg); 259 } 260 261 static void 262 icmp_mtudisc(struct mbuf *m, int hlen) 263 { 264 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 265 struct rtentry *rt; 266 struct icmp *icp; 267 268 KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr")); 269 270 icp = mtodoff(m, struct icmp *, hlen); 271 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 272 273 /* 274 * MTU discovery: 275 * If we got a needfrag and there is a host route to the original 276 * destination, and the MTU is not locked, then set the MTU in the 277 * route to the suggested new value (if given) and then notify as 278 * usual. The ULPs will notice that the MTU has changed and adapt 279 * accordingly. If no new MTU was suggested, then we guess a new 280 * one less than the current value. If the new MTU is unreasonably 281 * small (arbitrarily set at 296), then we reset the MTU to the 282 * interface value and enable the lock bit, indicating that we are 283 * no longer doing MTU discovery. 284 */ 285 rt = rtpurelookup((struct sockaddr *)&icmpsrc); 286 if (rt != NULL && (rt->rt_flags & RTF_HOST) && 287 !(rt->rt_rmx.rmx_locks & RTV_MTU)) { 288 #ifdef DEBUG_MTUDISC 289 char src_buf[INET_ADDRSTRLEN]; 290 #endif 291 int mtu; 292 293 mtu = ntohs(icp->icmp_nextmtu); 294 if (!mtu) 295 mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu, 1); 296 #ifdef DEBUG_MTUDISC 297 kprintf("MTU for %s reduced to %d\n", 298 inet_ntop(AF_INET, &icmpsrc.sin_addr, 299 src_buf, INET_ADDRSTRLEN), mtu); 300 #endif 301 if (mtu < 296) { 302 /* rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu; */ 303 rt->rt_rmx.rmx_locks |= RTV_MTU; 304 } else if (rt->rt_rmx.rmx_mtu > mtu) { 305 rt->rt_rmx.rmx_mtu = mtu; 306 } 307 } 308 if (rt != NULL) 309 --rt->rt_refcnt; 310 311 /* 312 * XXX if the packet contains [IPv4 AH TCP], we can't make a 313 * notification to TCP layer. 314 */ 315 so_pr_ctlinput_direct(&inetsw[ip_protox[icp->icmp_ip.ip_p]], 316 PRC_MSGSIZE, (struct sockaddr *)&icmpsrc, &icp->icmp_ip); 317 } 318 319 static void 320 icmp_mtudisc_handler(netmsg_t nmsg) 321 { 322 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 323 int nextcpu; 324 325 ASSERT_NETISR_NCPUS(mycpuid); 326 327 icmp_mtudisc(msg->m, msg->hlen); 328 329 nextcpu = mycpuid + 1; 330 if (nextcpu < netisr_ncpus) 331 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg); 332 else 333 icmp_ctlinput_done(msg->m); 334 } 335 336 static boolean_t 337 icmp_mtudisc_start(struct mbuf *m, int hlen, int proto) 338 { 339 struct netmsg_ctlinput *msg; 340 341 ASSERT_NETISR0; 342 343 icmp_mtudisc(m, hlen); 344 345 if (netisr_ncpus == 1) { 346 /* There is only one netisr; done */ 347 return FALSE; 348 } 349 350 msg = &m->m_hdr.mh_ctlmsg; 351 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 352 icmp_mtudisc_handler); 353 msg->m = m; 354 msg->cmd = PRC_MSGSIZE; 355 msg->hlen = hlen; 356 msg->proto = proto; 357 358 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg); 359 return TRUE; 360 } 361 362 static void 363 icmp_ctlinput(struct mbuf *m, int cmd, int hlen) 364 { 365 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 366 struct icmp *icp; 367 368 KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr")); 369 370 icp = mtodoff(m, struct icmp *, hlen); 371 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 372 373 /* 374 * XXX if the packet contains [IPv4 AH TCP], we can't make a 375 * notification to TCP layer. 376 */ 377 so_pr_ctlinput_direct(&inetsw[ip_protox[icp->icmp_ip.ip_p]], 378 cmd, (struct sockaddr *)&icmpsrc, &icp->icmp_ip); 379 } 380 381 static void 382 icmp_ctlinput_handler(netmsg_t nmsg) 383 { 384 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 385 386 ASSERT_NETISR_NCPUS(mycpuid); 387 388 icmp_ctlinput(msg->m, msg->cmd, msg->hlen); 389 icmp_ctlinput_done(msg->m); 390 } 391 392 static void 393 icmp_ctlinput_start(struct mbuf *m, struct lwkt_port *port, 394 int cmd, int hlen, int proto) 395 { 396 struct netmsg_ctlinput *msg; 397 398 KASSERT(&curthread->td_msgport != port, 399 ("send icmp ctlinput to the current netisr")); 400 401 msg = &m->m_hdr.mh_ctlmsg; 402 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 403 icmp_ctlinput_handler); 404 msg->m = m; 405 msg->cmd = cmd; 406 msg->hlen = hlen; 407 msg->proto = proto; 408 409 lwkt_sendmsg(port, &msg->base.lmsg); 410 } 411 412 static void 413 icmp_ctlinput_global_handler(netmsg_t nmsg) 414 { 415 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 416 int nextcpu; 417 418 ASSERT_NETISR_NCPUS(mycpuid); 419 420 icmp_ctlinput(msg->m, msg->cmd, msg->hlen); 421 422 nextcpu = mycpuid + 1; 423 if (nextcpu < netisr_ncpus) 424 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg); 425 else 426 icmp_ctlinput_done(msg->m); 427 } 428 429 static void 430 icmp_ctlinput_global_start(struct mbuf *m, int cmd, int hlen, int proto) 431 { 432 struct netmsg_ctlinput *msg; 433 434 ASSERT_NETISR0; 435 KASSERT(netisr_ncpus > 1, ("there is only 1 netisr cpu")); 436 437 icmp_ctlinput(m, cmd, hlen); 438 439 msg = &m->m_hdr.mh_ctlmsg; 440 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 441 icmp_ctlinput_global_handler); 442 msg->m = m; 443 msg->cmd = cmd; 444 msg->hlen = hlen; 445 msg->proto = proto; 446 447 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg); 448 } 449 450 #define ICMP_RTREDIRECT_FLAGS (RTF_GATEWAY | RTF_HOST) 451 452 static void 453 icmp_redirect(struct mbuf *m, int hlen, boolean_t prt) 454 { 455 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 456 struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET }; 457 struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET }; 458 struct icmp *icp; 459 struct ip *ip; 460 461 KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr")); 462 463 ip = mtod(m, struct ip *); 464 icp = mtodoff(m, struct icmp *, hlen); 465 466 /* 467 * Short circuit routing redirects to force immediate change 468 * in the kernel's routing tables. The message is also handed 469 * to anyone listening on a raw socket (e.g. the routing daemon 470 * for use in updating its tables). 471 */ 472 #ifdef ICMPPRINTFS 473 if (icmpprintfs && prt) { 474 char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN]; 475 476 kprintf("redirect dst %s to %s\n", 477 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, 478 dst_buf, INET_ADDRSTRLEN), 479 inet_ntop(AF_INET, &icp->icmp_gwaddr, 480 gw_buf, INET_ADDRSTRLEN)); 481 } 482 #endif 483 icmpgw.sin_addr = ip->ip_src; 484 icmpdst.sin_addr = icp->icmp_gwaddr; 485 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 486 rtredirect_oncpu((struct sockaddr *)&icmpsrc, 487 (struct sockaddr *)&icmpdst, NULL, ICMP_RTREDIRECT_FLAGS, 488 (struct sockaddr *)&icmpgw); 489 kpfctlinput_direct(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc); 490 } 491 492 static void 493 icmp_redirect_done_handler(netmsg_t nmsg) 494 { 495 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 496 struct mbuf *m = msg->m; 497 int hlen = msg->hlen; 498 499 rip_input(&m, &hlen, msg->proto); 500 } 501 502 static void 503 icmp_redirect_done(struct mbuf *m, int hlen, boolean_t dispatch_rip) 504 { 505 struct rt_addrinfo rtinfo; 506 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 507 struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET }; 508 struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET }; 509 struct icmp *icp; 510 struct ip *ip; 511 512 ip = mtod(m, struct ip *); 513 icp = mtodoff(m, struct icmp *, hlen); 514 515 icmpgw.sin_addr = ip->ip_src; 516 icmpdst.sin_addr = icp->icmp_gwaddr; 517 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 518 519 bzero(&rtinfo, sizeof(struct rt_addrinfo)); 520 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&icmpsrc; 521 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&icmpdst; 522 rtinfo.rti_info[RTAX_NETMASK] = NULL; 523 rtinfo.rti_info[RTAX_AUTHOR] = (struct sockaddr *)&icmpgw; 524 rt_missmsg(RTM_REDIRECT, &rtinfo, ICMP_RTREDIRECT_FLAGS, 0); 525 526 if (dispatch_rip) { 527 struct netmsg_ctlinput *msg = &m->m_hdr.mh_ctlmsg; 528 529 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 530 icmp_redirect_done_handler); 531 lwkt_sendmsg(netisr_cpuport(0), &msg->base.lmsg); 532 } 533 } 534 535 static void 536 icmp_redirect_handler(netmsg_t nmsg) 537 { 538 struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg; 539 int nextcpu; 540 541 ASSERT_NETISR_NCPUS(mycpuid); 542 543 icmp_redirect(msg->m, msg->hlen, FALSE); 544 545 nextcpu = mycpuid + 1; 546 if (nextcpu < netisr_ncpus) 547 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg); 548 else 549 icmp_redirect_done(msg->m, msg->hlen, TRUE); 550 } 551 552 static boolean_t 553 icmp_redirect_start(struct mbuf *m, int hlen, int proto) 554 { 555 struct netmsg_ctlinput *msg; 556 557 ASSERT_NETISR0; 558 559 icmp_redirect(m, hlen, TRUE); 560 561 if (netisr_ncpus == 1) { 562 /* There is only one netisr; done */ 563 icmp_redirect_done(m, hlen, FALSE); 564 return FALSE; 565 } 566 567 msg = &m->m_hdr.mh_ctlmsg; 568 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0, 569 icmp_redirect_handler); 570 msg->m = m; 571 msg->cmd = PRC_REDIRECT_HOST; 572 msg->hlen = hlen; 573 msg->proto = proto; 574 575 lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg); 576 return TRUE; 577 } 578 579 /* 580 * Process a received ICMP message. 581 */ 582 int 583 icmp_input(struct mbuf **mp, int *offp, int proto) 584 { 585 struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET }; 586 struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET }; 587 struct icmp *icp; 588 struct in_ifaddr *ia; 589 struct mbuf *m = *mp; 590 struct ip *ip = mtod(m, struct ip *); 591 int icmplen = ip->ip_len; 592 int i, hlen; 593 int code; 594 595 ASSERT_NETISR0; 596 597 *mp = NULL; 598 hlen = *offp; 599 600 /* 601 * Locate icmp structure in mbuf, and check 602 * that not corrupted and of at least minimum length. 603 */ 604 #ifdef ICMPPRINTFS 605 if (icmpprintfs) { 606 char src_buf[INET_ADDRSTRLEN], dst_buf[INET_ADDRSTRLEN]; 607 608 kprintf("icmp_input from %s to %s, len %d\n", 609 inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN), 610 inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN), 611 icmplen); 612 } 613 #endif 614 if (icmplen < ICMP_MINLEN) { 615 icmpstat.icps_tooshort++; 616 goto freeit; 617 } 618 i = hlen + min(icmplen, ICMP_ADVLENMIN); 619 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) { 620 icmpstat.icps_tooshort++; 621 return(IPPROTO_DONE); 622 } 623 ip = mtod(m, struct ip *); 624 625 if (in_cksum_skip(m, hlen + icmplen, hlen)) { 626 icmpstat.icps_checksum++; 627 goto freeit; 628 } 629 icp = (struct icmp *)((caddr_t)ip + hlen); 630 631 #ifdef ICMPPRINTFS 632 if (icmpprintfs) 633 kprintf("icmp_input, type %d code %d\n", icp->icmp_type, 634 icp->icmp_code); 635 #endif 636 637 /* 638 * Message type specific processing. 639 */ 640 if (icp->icmp_type > ICMP_MAXTYPE) 641 goto raw; 642 icmpstat.icps_inhist[icp->icmp_type]++; 643 code = icp->icmp_code; 644 switch (icp->icmp_type) { 645 646 case ICMP_UNREACH: 647 switch (code) { 648 case ICMP_UNREACH_NET: 649 case ICMP_UNREACH_HOST: 650 case ICMP_UNREACH_SRCFAIL: 651 case ICMP_UNREACH_NET_UNKNOWN: 652 case ICMP_UNREACH_HOST_UNKNOWN: 653 case ICMP_UNREACH_ISOLATED: 654 case ICMP_UNREACH_TOSNET: 655 case ICMP_UNREACH_TOSHOST: 656 case ICMP_UNREACH_HOST_PRECEDENCE: 657 case ICMP_UNREACH_PRECEDENCE_CUTOFF: 658 code = PRC_UNREACH_NET; 659 break; 660 661 case ICMP_UNREACH_NEEDFRAG: 662 code = PRC_MSGSIZE; 663 break; 664 665 /* 666 * RFC 1122, Sections 3.2.2.1 and 4.2.3.9. 667 * Treat subcodes 2,3 as immediate RST 668 */ 669 case ICMP_UNREACH_PROTOCOL: 670 case ICMP_UNREACH_PORT: 671 code = PRC_UNREACH_PORT; 672 break; 673 674 case ICMP_UNREACH_NET_PROHIB: 675 case ICMP_UNREACH_HOST_PROHIB: 676 case ICMP_UNREACH_FILTER_PROHIB: 677 code = PRC_UNREACH_ADMIN_PROHIB; 678 break; 679 680 default: 681 goto badcode; 682 } 683 goto deliver; 684 685 case ICMP_TIMXCEED: 686 if (code > 1) 687 goto badcode; 688 code += PRC_TIMXCEED_INTRANS; 689 goto deliver; 690 691 case ICMP_PARAMPROB: 692 if (code > 1) 693 goto badcode; 694 code = PRC_PARAMPROB; 695 goto deliver; 696 697 case ICMP_SOURCEQUENCH: 698 if (code) 699 goto badcode; 700 if (discard_sourcequench) 701 break; 702 code = PRC_QUENCH; 703 deliver: 704 /* 705 * Problem with datagram; advise higher level routines. 706 */ 707 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 708 IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) { 709 icmpstat.icps_badlen++; 710 goto freeit; 711 } 712 /* Discard ICMP's in response to multicast packets */ 713 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr))) 714 goto badcode; 715 #ifdef ICMPPRINTFS 716 if (icmpprintfs) 717 kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p); 718 #endif 719 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 720 721 /* 722 * MTU discovery 723 */ 724 if (code == PRC_MSGSIZE) { 725 /* Run MTU discovery in all netisrs */ 726 if (icmp_mtudisc_start(m, hlen, proto)) { 727 /* Forwarded; done */ 728 return IPPROTO_DONE; 729 } 730 /* Move on; run rip_input() directly */ 731 } else { 732 struct protosw *pr; 733 struct lwkt_port *port; 734 int cpu; 735 736 pr = &inetsw[ip_protox[icp->icmp_ip.ip_p]]; 737 port = so_pr_ctlport(pr, code, 738 (struct sockaddr *)&icmpsrc, &icp->icmp_ip, &cpu); 739 if (port != NULL) { 740 if (cpu == netisr_ncpus) { 741 if (netisr_ncpus > 1) { 742 /* 743 * Run pr_ctlinput in all 744 * netisrs 745 */ 746 icmp_ctlinput_global_start(m, 747 code, hlen, proto); 748 return IPPROTO_DONE; 749 } 750 /* 751 * There is only one netisr; run 752 * pr_ctlinput directly. 753 */ 754 } else if (cpu != mycpuid) { 755 /* 756 * Send to the target netisr to run 757 * pr_ctlinput. 758 */ 759 icmp_ctlinput_start(m, port, 760 code, hlen, proto); 761 return IPPROTO_DONE; 762 } 763 764 /* 765 * The target netisr is this netisr. 766 * 767 * XXX if the packet contains [IPv4 AH TCP], 768 * we can't make a notification to TCP layer. 769 */ 770 so_pr_ctlinput_direct(pr, code, 771 (struct sockaddr *)&icmpsrc, &icp->icmp_ip); 772 } 773 /* Move on; run rip_input() directly */ 774 } 775 break; 776 badcode: 777 icmpstat.icps_badcode++; 778 break; 779 780 case ICMP_ECHO: 781 if (!icmpbmcastecho 782 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 783 icmpstat.icps_bmcastecho++; 784 break; 785 } 786 icp->icmp_type = ICMP_ECHOREPLY; 787 #ifdef ICMP_BANDLIM 788 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0) 789 goto freeit; 790 else 791 #endif 792 goto reflect; 793 794 case ICMP_TSTAMP: 795 if (!icmpbmcastecho 796 && (m->m_flags & (M_MCAST | M_BCAST)) != 0) { 797 icmpstat.icps_bmcasttstamp++; 798 break; 799 } 800 if (icmplen < ICMP_TSLEN) { 801 icmpstat.icps_badlen++; 802 break; 803 } 804 icp->icmp_type = ICMP_TSTAMPREPLY; 805 icp->icmp_rtime = iptime(); 806 icp->icmp_ttime = icp->icmp_rtime; /* bogus, do later! */ 807 #ifdef ICMP_BANDLIM 808 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0) 809 goto freeit; 810 else 811 #endif 812 goto reflect; 813 814 case ICMP_MASKREQ: 815 if (icmpmaskrepl == 0) 816 break; 817 /* 818 * We are not able to respond with all ones broadcast 819 * unless we receive it over a point-to-point interface. 820 */ 821 if (icmplen < ICMP_MASKLEN) 822 break; 823 switch (ip->ip_dst.s_addr) { 824 825 case INADDR_BROADCAST: 826 case INADDR_ANY: 827 icmpdst.sin_addr = ip->ip_src; 828 break; 829 830 default: 831 icmpdst.sin_addr = ip->ip_dst; 832 } 833 ia = (struct in_ifaddr *)ifaof_ifpforaddr( 834 (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif); 835 if (ia == NULL) 836 break; 837 if (ia->ia_ifp == 0) 838 break; 839 icp->icmp_type = ICMP_MASKREPLY; 840 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr; 841 if (ip->ip_src.s_addr == 0) { 842 if (ia->ia_ifp->if_flags & IFF_BROADCAST) 843 ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr; 844 else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT) 845 ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr; 846 } 847 reflect: 848 ip->ip_len += hlen; /* since ip_input deducts this */ 849 icmpstat.icps_reflect++; 850 icmpstat.icps_outhist[icp->icmp_type]++; 851 icmp_reflect(m); 852 return(IPPROTO_DONE); 853 854 case ICMP_REDIRECT: 855 if (log_redirect) { 856 char src_buf[INET_ADDRSTRLEN]; 857 char dst_buf[INET_ADDRSTRLEN]; 858 char gwy_buf[INET_ADDRSTRLEN]; 859 860 kprintf("icmp redirect from %s: %s => %s\n", 861 inet_ntop(AF_INET, &ip->ip_src, 862 src_buf, INET_ADDRSTRLEN), 863 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, 864 dst_buf, INET_ADDRSTRLEN), 865 inet_ntop(AF_INET, &icp->icmp_gwaddr, 866 gwy_buf, INET_ADDRSTRLEN)); 867 } 868 if (drop_redirect) 869 break; 870 if (code > 3) 871 goto badcode; 872 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) || 873 IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) { 874 icmpstat.icps_badlen++; 875 break; 876 } 877 #ifdef ICMPPRINTFS 878 if (icmpprintfs) { 879 char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN]; 880 881 kprintf("redirect dst %s to %s\n", 882 inet_ntop(AF_INET, &icp->icmp_ip.ip_dst, 883 dst_buf, INET_ADDRSTRLEN), 884 inet_ntop(AF_INET, &icp->icmp_gwaddr, 885 gw_buf, INET_ADDRSTRLEN)); 886 } 887 #endif 888 icmpsrc.sin_addr = icp->icmp_ip.ip_dst; 889 890 /* Run redirect in all netisrs */ 891 if (icmp_redirect_start(m, hlen, proto)) { 892 /* Forwarded; done */ 893 return IPPROTO_DONE; 894 } 895 /* Move on; run rip_input() directly */ 896 break; 897 898 /* 899 * No kernel processing for the following; 900 * just fall through to send to raw listener. 901 */ 902 case ICMP_ECHOREPLY: 903 case ICMP_ROUTERADVERT: 904 case ICMP_ROUTERSOLICIT: 905 case ICMP_TSTAMPREPLY: 906 case ICMP_IREQREPLY: 907 case ICMP_MASKREPLY: 908 default: 909 break; 910 } 911 912 raw: 913 *mp = m; 914 rip_input(mp, offp, proto); 915 return(IPPROTO_DONE); 916 917 freeit: 918 m_freem(m); 919 return(IPPROTO_DONE); 920 } 921 922 /* 923 * Reflect the ip packet back to the source 924 */ 925 static void 926 icmp_reflect(struct mbuf *m) 927 { 928 struct ip *ip = mtod(m, struct ip *); 929 struct in_ifaddr *ia; 930 struct in_ifaddr_container *iac; 931 struct ifaddr_container *ifac; 932 struct ifnet *ifp; 933 struct in_addr t; 934 struct mbuf *opts = NULL; 935 int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 936 struct route *ro = NULL, rt; 937 938 if (!in_canforward(ip->ip_src) && 939 ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) != 940 (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) { 941 m_freem(m); /* Bad return address */ 942 icmpstat.icps_badaddr++; 943 goto done; /* Ip_output() will check for broadcast */ 944 } 945 t = ip->ip_dst; 946 ip->ip_dst = ip->ip_src; 947 948 ro = &rt; 949 bzero(ro, sizeof *ro); 950 951 /* 952 * If the incoming packet was addressed directly to us, 953 * use dst as the src for the reply. Otherwise (broadcast 954 * or anonymous), use the address which corresponds 955 * to the incoming interface. 956 */ 957 ia = NULL; 958 LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) { 959 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) { 960 ia = iac->ia; 961 goto match; 962 } 963 } 964 ifp = m->m_pkthdr.rcvif; 965 if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) { 966 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 967 struct ifaddr *ifa = ifac->ifa; 968 969 if (ifa->ifa_addr->sa_family != AF_INET) 970 continue; 971 ia = ifatoia(ifa); 972 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 973 t.s_addr) 974 goto match; 975 } 976 } 977 /* 978 * If the packet was transiting through us, use the address of 979 * the interface the packet came through in. If that interface 980 * doesn't have a suitable IP address, the normal selection 981 * criteria apply. 982 */ 983 if (icmp_rfi && ifp != NULL) { 984 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 985 struct ifaddr *ifa = ifac->ifa; 986 987 if (ifa->ifa_addr->sa_family != AF_INET) 988 continue; 989 ia = ifatoia(ifa); 990 goto match; 991 } 992 } 993 /* 994 * If the incoming packet was not addressed directly to us, use 995 * designated interface for icmp replies specified by sysctl 996 * net.inet.icmp.reply_src (default not set). Otherwise continue 997 * with normal source selection. 998 */ 999 if (icmp_reply_src[0] != '\0' && 1000 (ifp = ifunit_netisr(icmp_reply_src))) { 1001 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 1002 struct ifaddr *ifa = ifac->ifa; 1003 1004 if (ifa->ifa_addr->sa_family != AF_INET) 1005 continue; 1006 ia = ifatoia(ifa); 1007 goto match; 1008 } 1009 } 1010 /* 1011 * If the packet was transiting through us, use the address of 1012 * the interface that is the closest to the packet source. 1013 * When we don't have a route back to the packet source, stop here 1014 * and drop the packet. 1015 */ 1016 ia = ip_rtaddr(ip->ip_dst, ro); 1017 if (ia == NULL) { 1018 m_freem(m); 1019 icmpstat.icps_noroute++; 1020 goto done; 1021 } 1022 match: 1023 t = IA_SIN(ia)->sin_addr; 1024 ip->ip_src = t; 1025 ip->ip_ttl = ip_defttl; 1026 1027 if (optlen > 0) { 1028 u_char *cp; 1029 int opt, cnt; 1030 u_int len; 1031 1032 /* 1033 * Retrieve any source routing from the incoming packet; 1034 * add on any record-route or timestamp options. 1035 */ 1036 cp = (u_char *) (ip + 1); 1037 if ((opts = ip_srcroute(m)) == NULL && 1038 (opts = m_gethdr(M_NOWAIT, MT_HEADER))) { 1039 opts->m_len = sizeof(struct in_addr); 1040 mtod(opts, struct in_addr *)->s_addr = 0; 1041 } 1042 if (opts) { 1043 #ifdef ICMPPRINTFS 1044 if (icmpprintfs) 1045 kprintf("icmp_reflect optlen %d rt %d => ", 1046 optlen, opts->m_len); 1047 #endif 1048 for (cnt = optlen; cnt > 0; cnt -= len, cp += len) { 1049 opt = cp[IPOPT_OPTVAL]; 1050 if (opt == IPOPT_EOL) 1051 break; 1052 if (opt == IPOPT_NOP) 1053 len = 1; 1054 else { 1055 if (cnt < IPOPT_OLEN + sizeof *cp) 1056 break; 1057 len = cp[IPOPT_OLEN]; 1058 if (len < IPOPT_OLEN + sizeof *cp || 1059 len > cnt) 1060 break; 1061 } 1062 /* 1063 * Should check for overflow, but it 1064 * "can't happen". 1065 */ 1066 if (opt == IPOPT_RR || opt == IPOPT_TS || 1067 opt == IPOPT_SECURITY) { 1068 bcopy(cp, 1069 mtod(opts, caddr_t) + opts->m_len, 1070 len); 1071 opts->m_len += len; 1072 } 1073 } 1074 /* Terminate & pad, if necessary */ 1075 cnt = opts->m_len % 4; 1076 if (cnt) { 1077 for (; cnt < 4; cnt++) { 1078 *(mtod(opts, caddr_t) + opts->m_len) = 1079 IPOPT_EOL; 1080 opts->m_len++; 1081 } 1082 } 1083 #ifdef ICMPPRINTFS 1084 if (icmpprintfs) 1085 kprintf("%d\n", opts->m_len); 1086 #endif 1087 } 1088 /* 1089 * Now strip out original options by copying rest of first 1090 * mbuf's data back, and adjust the IP length. 1091 */ 1092 ip->ip_len -= optlen; 1093 ip->ip_vhl = IP_VHL_BORING; 1094 m->m_len -= optlen; 1095 if (m->m_flags & M_PKTHDR) 1096 m->m_pkthdr.len -= optlen; 1097 optlen += sizeof(struct ip); 1098 bcopy((caddr_t)ip + optlen, ip + 1, 1099 m->m_len - sizeof(struct ip)); 1100 } 1101 m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED; 1102 m->m_flags &= ~(M_BCAST|M_MCAST); 1103 icmp_send(m, opts, ro); 1104 done: 1105 if (opts) 1106 m_free(opts); 1107 if (ro && ro->ro_rt) 1108 RTFREE(ro->ro_rt); 1109 } 1110 1111 /* 1112 * Send an icmp packet back to the ip level, 1113 * after supplying a checksum. 1114 */ 1115 static void 1116 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt) 1117 { 1118 struct ip *ip = mtod(m, struct ip *); 1119 struct icmp *icp; 1120 int hlen; 1121 1122 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1123 m->m_data += hlen; 1124 m->m_len -= hlen; 1125 icp = mtod(m, struct icmp *); 1126 icp->icmp_cksum = 0; 1127 icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen); 1128 m->m_data -= hlen; 1129 m->m_len += hlen; 1130 m->m_pkthdr.rcvif = NULL; 1131 #ifdef ICMPPRINTFS 1132 if (icmpprintfs) { 1133 char dst_buf[INET_ADDRSTRLEN], src_buf[INET_ADDRSTRLEN]; 1134 1135 kprintf("icmp_send dst %s src %s\n", 1136 inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN), 1137 inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN)); 1138 } 1139 #endif 1140 ip_output(m, opts, rt, 0, NULL, NULL); 1141 } 1142 1143 n_time 1144 iptime(void) 1145 { 1146 struct timeval atv; 1147 u_long t; 1148 1149 getmicrotime(&atv); 1150 t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000; 1151 return (htonl(t)); 1152 } 1153 1154 #if 1 1155 /* 1156 * Return the next larger or smaller MTU plateau (table from RFC 1191) 1157 * given current value MTU. If DIR is less than zero, a larger plateau 1158 * is returned; otherwise, a smaller value is returned. 1159 */ 1160 int 1161 ip_next_mtu(int mtu, int dir) 1162 { 1163 static int mtutab[] = { 1164 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296, 1165 68, 0 1166 }; 1167 int i; 1168 1169 for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) { 1170 if (mtu >= mtutab[i]) 1171 break; 1172 } 1173 1174 if (dir < 0) { 1175 if (i == 0) { 1176 return 0; 1177 } else { 1178 return mtutab[i - 1]; 1179 } 1180 } else { 1181 if (mtutab[i] == 0) { 1182 return 0; 1183 } else if(mtu > mtutab[i]) { 1184 return mtutab[i]; 1185 } else { 1186 return mtutab[i + 1]; 1187 } 1188 } 1189 } 1190 #endif 1191 1192 #ifdef ICMP_BANDLIM 1193 /* 1194 * badport_bandlim() - check for ICMP bandwidth limit 1195 * 1196 * Return 0 if it is ok to send an ICMP error response, -1 if we have 1197 * hit our bandwidth limit and it is not ok. 1198 * 1199 * If icmplim is <= 0, the feature is disabled and 0 is returned. 1200 * 1201 * For now we separate the TCP and UDP subsystems w/ different 'which' 1202 * values. We may eventually remove this separation (and simplify the 1203 * code further). 1204 * 1205 * Note that the printing of the error message is delayed so we can 1206 * properly print the icmp error rate that the system was trying to do 1207 * (i.e. 22000/100 pps, etc...). This can cause long delays in printing 1208 * the 'final' error, but it doesn't make sense to solve the printing 1209 * delay with more complex code. 1210 */ 1211 int 1212 badport_bandlim(int which) 1213 { 1214 static int lticks[BANDLIM_MAX + 1]; 1215 static int lpackets[BANDLIM_MAX + 1]; 1216 int dticks; 1217 const char *bandlimittype[] = { 1218 "Limiting icmp unreach response", 1219 "Limiting icmp ping response", 1220 "Limiting icmp tstamp response", 1221 "Limiting closed port RST response", 1222 "Limiting open port RST response" 1223 }; 1224 1225 /* 1226 * Return ok status if feature disabled or argument out of 1227 * ranage. 1228 */ 1229 1230 if (icmplim <= 0 || which > BANDLIM_MAX || which < 0) 1231 return(0); 1232 dticks = ticks - lticks[which]; 1233 1234 /* 1235 * reset stats when cumulative dt exceeds one second. 1236 */ 1237 1238 if ((unsigned int)dticks > hz) { 1239 if (lpackets[which] > icmplim && icmplim_output) { 1240 kprintf("%s from %d to %d packets per second\n", 1241 bandlimittype[which], 1242 lpackets[which], 1243 icmplim 1244 ); 1245 } 1246 lticks[which] = ticks; 1247 lpackets[which] = 0; 1248 } 1249 1250 /* 1251 * bump packet count 1252 */ 1253 1254 if (++lpackets[which] > icmplim) { 1255 return(-1); 1256 } 1257 return(0); 1258 } 1259 #endif 1260