1 /* 2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved. 3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved. 4 * 5 * This code is derived from software contributed to The DragonFly Project 6 * by Jeffrey M. Hsu. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of The DragonFly Project nor the names of its 17 * contributors may be used to endorse or promote products derived 18 * from this software without specific, prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 /* 35 * Copyright (c) 1982, 1986, 1988, 1993 36 * The Regents of the University of California. All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. Neither the name of the University nor the names of its contributors 47 * may be used to endorse or promote products derived from this software 48 * without specific prior written permission. 49 * 50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 60 * SUCH DAMAGE. 61 * 62 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 63 * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $ 64 */ 65 66 #define _IP_VHL 67 68 #include "opt_bootp.h" 69 #include "opt_ipdn.h" 70 #include "opt_ipdivert.h" 71 #include "opt_ipstealth.h" 72 #include "opt_ipsec.h" 73 #include "opt_rss.h" 74 75 #include <sys/param.h> 76 #include <sys/systm.h> 77 #include <sys/mbuf.h> 78 #include <sys/malloc.h> 79 #include <sys/mpipe.h> 80 #include <sys/domain.h> 81 #include <sys/protosw.h> 82 #include <sys/socket.h> 83 #include <sys/time.h> 84 #include <sys/globaldata.h> 85 #include <sys/thread.h> 86 #include <sys/kernel.h> 87 #include <sys/syslog.h> 88 #include <sys/sysctl.h> 89 #include <sys/in_cksum.h> 90 #include <sys/lock.h> 91 92 #include <sys/mplock2.h> 93 94 #include <machine/stdarg.h> 95 96 #include <net/if.h> 97 #include <net/if_types.h> 98 #include <net/if_var.h> 99 #include <net/if_dl.h> 100 #include <net/pfil.h> 101 #include <net/route.h> 102 #include <net/netisr2.h> 103 104 #include <netinet/in.h> 105 #include <netinet/in_systm.h> 106 #include <netinet/in_var.h> 107 #include <netinet/ip.h> 108 #include <netinet/in_pcb.h> 109 #include <netinet/ip_var.h> 110 #include <netinet/ip_icmp.h> 111 #include <netinet/ip_divert.h> 112 #include <netinet/ip_flow.h> 113 114 #include <sys/thread2.h> 115 #include <sys/msgport2.h> 116 #include <net/netmsg2.h> 117 118 #include <sys/socketvar.h> 119 120 #include <net/ipfw/ip_fw.h> 121 #include <net/dummynet/ip_dummynet.h> 122 123 #ifdef IPSEC 124 #include <netinet6/ipsec.h> 125 #include <netproto/key/key.h> 126 #endif 127 128 #ifdef FAST_IPSEC 129 #include <netproto/ipsec/ipsec.h> 130 #include <netproto/ipsec/key.h> 131 #endif 132 133 int rsvp_on = 0; 134 static int ip_rsvp_on; 135 struct socket *ip_rsvpd; 136 137 int ipforwarding = 0; 138 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW, 139 &ipforwarding, 0, "Enable IP forwarding between interfaces"); 140 141 static int ipsendredirects = 1; /* XXX */ 142 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW, 143 &ipsendredirects, 0, "Enable sending IP redirects"); 144 145 int ip_defttl = IPDEFTTL; 146 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 147 &ip_defttl, 0, "Maximum TTL on IP packets"); 148 149 static int ip_dosourceroute = 0; 150 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW, 151 &ip_dosourceroute, 0, "Enable forwarding source routed IP packets"); 152 153 static int ip_acceptsourceroute = 0; 154 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute, 155 CTLFLAG_RW, &ip_acceptsourceroute, 0, 156 "Enable accepting source routed IP packets"); 157 158 static int ip_keepfaith = 0; 159 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW, 160 &ip_keepfaith, 0, 161 "Enable packet capture for FAITH IPv4->IPv6 translator daemon"); 162 163 static int maxnipq; 164 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW, 165 &maxnipq, 0, 166 "Maximum number of IPv4 fragment reassembly queue entries"); 167 168 static int maxfragsperpacket; 169 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW, 170 &maxfragsperpacket, 0, 171 "Maximum number of IPv4 fragments allowed per packet"); 172 173 static int ip_sendsourcequench = 0; 174 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW, 175 &ip_sendsourcequench, 0, 176 "Enable the transmission of source quench packets"); 177 178 int ip_do_randomid = 1; 179 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW, 180 &ip_do_randomid, 0, 181 "Assign random ip_id values"); 182 /* 183 * XXX - Setting ip_checkinterface mostly implements the receive side of 184 * the Strong ES model described in RFC 1122, but since the routing table 185 * and transmit implementation do not implement the Strong ES model, 186 * setting this to 1 results in an odd hybrid. 187 * 188 * XXX - ip_checkinterface currently must be disabled if you use ipnat 189 * to translate the destination address to another local interface. 190 * 191 * XXX - ip_checkinterface must be disabled if you add IP aliases 192 * to the loopback interface instead of the interface where the 193 * packets for those addresses are received. 194 */ 195 static int ip_checkinterface = 0; 196 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW, 197 &ip_checkinterface, 0, "Verify packet arrives on correct interface"); 198 199 static u_long ip_hash_count = 0; 200 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD, 201 &ip_hash_count, 0, "Number of packets hashed by IP"); 202 203 #ifdef RSS_DEBUG 204 static u_long ip_rehash_count = 0; 205 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD, 206 &ip_rehash_count, 0, "Number of packets rehashed by IP"); 207 208 static u_long ip_dispatch_fast = 0; 209 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD, 210 &ip_dispatch_fast, 0, "Number of packets handled on current CPU"); 211 212 static u_long ip_dispatch_slow = 0; 213 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD, 214 &ip_dispatch_slow, 0, "Number of packets messaged to another CPU"); 215 #endif 216 217 #ifdef DIAGNOSTIC 218 static int ipprintfs = 0; 219 #endif 220 221 extern struct domain inetdomain; 222 extern struct protosw inetsw[]; 223 u_char ip_protox[IPPROTO_MAX]; 224 struct in_ifaddrhead in_ifaddrheads[MAXCPU]; /* first inet address */ 225 struct in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU]; 226 /* inet addr hash table */ 227 u_long in_ifaddrhmask; /* mask for hash table */ 228 229 static struct mbuf *ipforward_mtemp[MAXCPU]; 230 231 struct ip_stats ipstats_percpu[MAXCPU] __cachealign; 232 233 static int 234 sysctl_ipstats(SYSCTL_HANDLER_ARGS) 235 { 236 int cpu, error = 0; 237 238 for (cpu = 0; cpu < netisr_ncpus; ++cpu) { 239 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu], 240 sizeof(struct ip_stats)))) 241 break; 242 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu], 243 sizeof(struct ip_stats)))) 244 break; 245 } 246 247 return (error); 248 } 249 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW), 250 0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics"); 251 252 /* Packet reassembly stuff */ 253 #define IPREASS_NHASH_LOG2 6 254 #define IPREASS_NHASH (1 << IPREASS_NHASH_LOG2) 255 #define IPREASS_HMASK (IPREASS_NHASH - 1) 256 #define IPREASS_HASH(x,y) \ 257 (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK) 258 259 TAILQ_HEAD(ipqhead, ipq); 260 struct ipfrag_queue { 261 int nipq; 262 volatile int draining; 263 struct netmsg_base timeo_netmsg; 264 struct callout timeo_ch; 265 struct netmsg_base drain_netmsg; 266 struct ipqhead ipq[IPREASS_NHASH]; 267 } __cachealign; 268 269 static struct ipfrag_queue ipfrag_queue_pcpu[MAXCPU]; 270 271 #ifdef IPCTL_DEFMTU 272 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 273 &ip_mtu, 0, "Default MTU"); 274 #endif 275 276 #ifdef IPSTEALTH 277 static int ipstealth = 0; 278 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, ""); 279 #else 280 static const int ipstealth = 0; 281 #endif 282 283 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int); 284 285 struct pfil_head inet_pfil_hook; 286 287 /* 288 * struct ip_srcrt_opt is used to store packet state while it travels 289 * through the stack. 290 * 291 * XXX Note that the code even makes assumptions on the size and 292 * alignment of fields inside struct ip_srcrt so e.g. adding some 293 * fields will break the code. This needs to be fixed. 294 * 295 * We need to save the IP options in case a protocol wants to respond 296 * to an incoming packet over the same route if the packet got here 297 * using IP source routing. This allows connection establishment and 298 * maintenance when the remote end is on a network that is not known 299 * to us. 300 */ 301 struct ip_srcrt { 302 struct in_addr dst; /* final destination */ 303 char nop; /* one NOP to align */ 304 char srcopt[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN and OFFSET */ 305 struct in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)]; 306 }; 307 308 struct ip_srcrt_opt { 309 int ip_nhops; 310 struct ip_srcrt ip_srcrt; 311 }; 312 313 #define IPFRAG_MPIPE_MAX 4096 314 #define MAXIPFRAG_MIN ((IPFRAG_MPIPE_MAX * 2) / 256) 315 316 #define IPFRAG_TIMEO (hz / PR_SLOWHZ) 317 318 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management"); 319 static struct malloc_pipe ipq_mpipe; 320 321 static void save_rte(struct mbuf *, u_char *, struct in_addr); 322 static int ip_dooptions(struct mbuf *m, int, struct sockaddr_in *); 323 static void ip_freef(struct ipfrag_queue *, struct ipqhead *, 324 struct ipq *); 325 static void ip_input_handler(netmsg_t); 326 327 static void ipfrag_timeo_dispatch(netmsg_t); 328 static void ipfrag_timeo(void *); 329 static void ipfrag_drain_dispatch(netmsg_t); 330 331 /* 332 * IP initialization: fill in IP protocol switch table. 333 * All protocols not implemented in kernel go to raw IP protocol handler. 334 */ 335 void 336 ip_init(void) 337 { 338 struct ipfrag_queue *fragq; 339 struct protosw *pr; 340 int cpu, i; 341 342 /* 343 * Make sure we can handle a reasonable number of fragments but 344 * cap it at IPFRAG_MPIPE_MAX. 345 */ 346 mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq), 347 IFQ_MAXLEN, IPFRAG_MPIPE_MAX, 0, NULL, NULL, NULL); 348 349 /* 350 * Make in_ifaddrhead and in_ifaddrhashtbl available on all CPUs, 351 * since they could be accessed by any threads. 352 */ 353 for (cpu = 0; cpu < ncpus; ++cpu) { 354 TAILQ_INIT(&in_ifaddrheads[cpu]); 355 in_ifaddrhashtbls[cpu] = 356 hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask); 357 } 358 359 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 360 if (pr == NULL) 361 panic("ip_init"); 362 for (i = 0; i < IPPROTO_MAX; i++) 363 ip_protox[i] = pr - inetsw; 364 for (pr = inetdomain.dom_protosw; 365 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 366 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) { 367 if (pr->pr_protocol != IPPROTO_RAW) 368 ip_protox[pr->pr_protocol] = pr - inetsw; 369 } 370 } 371 372 inet_pfil_hook.ph_type = PFIL_TYPE_AF; 373 inet_pfil_hook.ph_af = AF_INET; 374 if ((i = pfil_head_register(&inet_pfil_hook)) != 0) { 375 kprintf("%s: WARNING: unable to register pfil hook, " 376 "error %d\n", __func__, i); 377 } 378 379 maxnipq = (nmbclusters / 32) / netisr_ncpus; 380 if (maxnipq < MAXIPFRAG_MIN) 381 maxnipq = MAXIPFRAG_MIN; 382 maxfragsperpacket = 16; 383 384 ip_id = time_second & 0xffff; /* time_second survives reboots */ 385 386 for (cpu = 0; cpu < netisr_ncpus; ++cpu) { 387 /* 388 * Initialize IP statistics counters for each CPU. 389 */ 390 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats)); 391 392 /* 393 * Preallocate mbuf template for forwarding 394 */ 395 MGETHDR(ipforward_mtemp[cpu], M_WAITOK, MT_DATA); 396 397 /* 398 * Initialize per-cpu ip fragments queues 399 */ 400 fragq = &ipfrag_queue_pcpu[cpu]; 401 for (i = 0; i < IPREASS_NHASH; i++) 402 TAILQ_INIT(&fragq->ipq[i]); 403 404 callout_init_mp(&fragq->timeo_ch); 405 netmsg_init(&fragq->timeo_netmsg, NULL, &netisr_adone_rport, 406 MSGF_PRIORITY, ipfrag_timeo_dispatch); 407 netmsg_init(&fragq->drain_netmsg, NULL, &netisr_adone_rport, 408 MSGF_PRIORITY, ipfrag_drain_dispatch); 409 } 410 411 netisr_register(NETISR_IP, ip_input_handler, ip_hashfn); 412 netisr_register_hashcheck(NETISR_IP, ip_hashcheck); 413 414 for (cpu = 0; cpu < netisr_ncpus; ++cpu) { 415 fragq = &ipfrag_queue_pcpu[cpu]; 416 callout_reset_bycpu(&fragq->timeo_ch, IPFRAG_TIMEO, 417 ipfrag_timeo, NULL, cpu); 418 } 419 } 420 421 /* Do transport protocol processing. */ 422 static void 423 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip) 424 { 425 const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]]; 426 427 /* 428 * Switch out to protocol's input routine. 429 */ 430 PR_GET_MPLOCK(pr); 431 pr->pr_input(&m, &hlen, ip->ip_p); 432 PR_REL_MPLOCK(pr); 433 } 434 435 static void 436 transport_processing_handler(netmsg_t msg) 437 { 438 struct netmsg_packet *pmsg = &msg->packet; 439 struct ip *ip; 440 int hlen; 441 442 ip = mtod(pmsg->nm_packet, struct ip *); 443 hlen = pmsg->base.lmsg.u.ms_result; 444 445 transport_processing_oncpu(pmsg->nm_packet, hlen, ip); 446 /* msg was embedded in the mbuf, do not reply! */ 447 } 448 449 static void 450 ip_input_handler(netmsg_t msg) 451 { 452 ip_input(msg->packet.nm_packet); 453 /* msg was embedded in the mbuf, do not reply! */ 454 } 455 456 /* 457 * IP input routine. Checksum and byte swap header. If fragmented 458 * try to reassemble. Process options. Pass to next level. 459 */ 460 void 461 ip_input(struct mbuf *m) 462 { 463 struct ip *ip; 464 struct in_ifaddr *ia = NULL; 465 struct in_ifaddr_container *iac; 466 int hlen, checkif; 467 u_short sum; 468 struct in_addr pkt_dst; 469 boolean_t using_srcrt = FALSE; /* forward (by PFIL_HOOKS) */ 470 struct in_addr odst; /* original dst address(NAT) */ 471 struct m_tag *mtag; 472 struct sockaddr_in *next_hop = NULL; 473 lwkt_port_t port; 474 #ifdef FAST_IPSEC 475 struct tdb_ident *tdbi; 476 struct secpolicy *sp; 477 int error; 478 #endif 479 480 ASSERT_NETISR_NCPUS(mycpuid); 481 M_ASSERTPKTHDR(m); 482 483 /* length checks already done in ip_hashfn() */ 484 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf")); 485 486 /* 487 * This routine is called from numerous places which may not have 488 * characterized the packet. 489 */ 490 ip = mtod(m, struct ip *); 491 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || 492 (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK))) { 493 /* 494 * Force hash recalculation for fragments and multicast 495 * packets; hardware may not do it correctly. 496 * XXX add flag to indicate the hash is from hardware 497 */ 498 m->m_flags &= ~M_HASH; 499 } 500 if ((m->m_flags & M_HASH) == 0) { 501 ip_hashfn(&m, 0); 502 if (m == NULL) 503 return; 504 KKASSERT(m->m_flags & M_HASH); 505 506 if (&curthread->td_msgport != 507 netisr_hashport(m->m_pkthdr.hash)) { 508 netisr_queue(NETISR_IP, m); 509 /* Requeued to other netisr msgport; done */ 510 return; 511 } 512 513 /* mbuf could have been changed */ 514 ip = mtod(m, struct ip *); 515 } 516 517 /* 518 * Pull out certain tags 519 */ 520 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 521 /* Next hop */ 522 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 523 KKASSERT(mtag != NULL); 524 next_hop = m_tag_data(mtag); 525 } 526 527 if (m->m_pkthdr.fw_flags & 528 (DUMMYNET_MBUF_TAGGED | IPFW_MBUF_CONTINUE)) { 529 /* 530 * - Dummynet already filtered this packet. 531 * - This packet was processed by ipfw on another 532 * cpu, and the rest of the ipfw processing should 533 * be carried out on this cpu. 534 */ 535 ip = mtod(m, struct ip *); 536 ip->ip_len = ntohs(ip->ip_len); 537 ip->ip_off = ntohs(ip->ip_off); 538 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 539 goto iphack; 540 } 541 542 ipstat.ips_total++; 543 544 if (IP_VHL_V(ip->ip_vhl) != IPVERSION) { 545 ipstat.ips_badvers++; 546 goto bad; 547 } 548 549 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 550 /* length checks already done in ip_hashfn() */ 551 KASSERT(hlen >= sizeof(struct ip), ("IP header len too small")); 552 KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf")); 553 554 /* 127/8 must not appear on wire - RFC1122 */ 555 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 556 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 557 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) { 558 ipstat.ips_badaddr++; 559 goto bad; 560 } 561 } 562 563 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 564 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 565 } else { 566 if (hlen == sizeof(struct ip)) 567 sum = in_cksum_hdr(ip); 568 else 569 sum = in_cksum(m, hlen); 570 } 571 if (sum != 0) { 572 ipstat.ips_badsum++; 573 goto bad; 574 } 575 576 #ifdef ALTQ 577 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) { 578 /* packet is dropped by traffic conditioner */ 579 return; 580 } 581 #endif 582 /* 583 * Convert fields to host representation. 584 */ 585 ip->ip_len = ntohs(ip->ip_len); 586 ip->ip_off = ntohs(ip->ip_off); 587 588 /* length checks already done in ip_hashfn() */ 589 KASSERT(ip->ip_len >= hlen, ("total length less then header length")); 590 KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short")); 591 592 /* 593 * Trim mbufs if longer than the IP header would have us expect. 594 */ 595 if (m->m_pkthdr.len > ip->ip_len) { 596 if (m->m_len == m->m_pkthdr.len) { 597 m->m_len = ip->ip_len; 598 m->m_pkthdr.len = ip->ip_len; 599 } else { 600 m_adj(m, ip->ip_len - m->m_pkthdr.len); 601 } 602 } 603 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF) 604 /* 605 * Bypass packet filtering for packets from a tunnel (gif). 606 */ 607 if (ipsec_gethist(m, NULL)) 608 goto pass; 609 #endif 610 611 /* 612 * IpHack's section. 613 * Right now when no processing on packet has done 614 * and it is still fresh out of network we do our black 615 * deals with it. 616 * - Firewall: deny/allow/divert 617 * - Xlate: translate packet's addr/port (NAT). 618 * - Pipe: pass pkt through dummynet. 619 * - Wrap: fake packet's addr/port <unimpl.> 620 * - Encapsulate: put it in another IP and send out. <unimp.> 621 */ 622 623 iphack: 624 /* 625 * If we've been forwarded from the output side, then 626 * skip the firewall a second time 627 */ 628 if (next_hop != NULL) 629 goto ours; 630 631 /* No pfil hooks */ 632 if (!pfil_has_hooks(&inet_pfil_hook)) { 633 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 634 /* 635 * Strip dummynet tags from stranded packets 636 */ 637 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 638 KKASSERT(mtag != NULL); 639 m_tag_delete(m, mtag); 640 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 641 } 642 goto pass; 643 } 644 645 /* 646 * Run through list of hooks for input packets. 647 * 648 * NOTE! If the packet is rewritten pf/ipfw/whoever must 649 * clear M_HASH. 650 */ 651 odst = ip->ip_dst; 652 if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN)) 653 return; 654 if (m == NULL) /* consumed by filter */ 655 return; 656 ip = mtod(m, struct ip *); 657 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 658 using_srcrt = (odst.s_addr != ip->ip_dst.s_addr); 659 660 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 661 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 662 KKASSERT(mtag != NULL); 663 next_hop = m_tag_data(mtag); 664 } 665 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 666 ip_dn_queue(m); 667 return; 668 } 669 if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) 670 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH; 671 if (m->m_pkthdr.fw_flags & IPFW_MBUF_CONTINUE) { 672 /* ipfw was disabled/unloaded. */ 673 goto bad; 674 } 675 pass: 676 /* 677 * Process options and, if not destined for us, 678 * ship it on. ip_dooptions returns 1 when an 679 * error was detected (causing an icmp message 680 * to be sent and the original packet to be freed). 681 */ 682 if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop)) 683 return; 684 685 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 686 * matter if it is destined to another node, or whether it is 687 * a multicast one, RSVP wants it! and prevents it from being forwarded 688 * anywhere else. Also checks if the rsvp daemon is running before 689 * grabbing the packet. 690 */ 691 if (rsvp_on && ip->ip_p == IPPROTO_RSVP) 692 goto ours; 693 694 /* 695 * Check our list of addresses, to see if the packet is for us. 696 * If we don't have any addresses, assume any unicast packet 697 * we receive might be for us (and let the upper layers deal 698 * with it). 699 */ 700 if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) && 701 !(m->m_flags & (M_MCAST | M_BCAST))) 702 goto ours; 703 704 /* 705 * Cache the destination address of the packet; this may be 706 * changed by use of 'ipfw fwd'. 707 */ 708 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 709 710 /* 711 * Enable a consistency check between the destination address 712 * and the arrival interface for a unicast packet (the RFC 1122 713 * strong ES model) if IP forwarding is disabled and the packet 714 * is not locally generated and the packet is not subject to 715 * 'ipfw fwd'. 716 * 717 * XXX - Checking also should be disabled if the destination 718 * address is ipnat'ed to a different interface. 719 * 720 * XXX - Checking is incompatible with IP aliases added 721 * to the loopback interface instead of the interface where 722 * the packets are received. 723 */ 724 checkif = ip_checkinterface && 725 !ipforwarding && 726 m->m_pkthdr.rcvif != NULL && 727 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) && 728 next_hop == NULL; 729 730 /* 731 * Check for exact addresses in the hash bucket. 732 */ 733 LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) { 734 ia = iac->ia; 735 736 /* 737 * If the address matches, verify that the packet 738 * arrived via the correct interface if checking is 739 * enabled. 740 */ 741 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr && 742 (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif)) 743 goto ours; 744 } 745 ia = NULL; 746 747 /* 748 * Check for broadcast addresses. 749 * 750 * Only accept broadcast packets that arrive via the matching 751 * interface. Reception of forwarded directed broadcasts would 752 * be handled via ip_forward() and ether_output() with the loopback 753 * into the stack for SIMPLEX interfaces handled by ether_output(). 754 */ 755 if (m->m_pkthdr.rcvif != NULL && 756 m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) { 757 struct ifaddr_container *ifac; 758 759 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid], 760 ifa_link) { 761 struct ifaddr *ifa = ifac->ifa; 762 763 if (ifa->ifa_addr == NULL) /* shutdown/startup race */ 764 continue; 765 if (ifa->ifa_addr->sa_family != AF_INET) 766 continue; 767 ia = ifatoia(ifa); 768 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 769 pkt_dst.s_addr) 770 goto ours; 771 if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr) 772 goto ours; 773 #ifdef BOOTP_COMPAT 774 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) 775 goto ours; 776 #endif 777 } 778 } 779 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 780 struct in_multi *inm; 781 782 if (ip_mrouter != NULL) { 783 /* XXX Multicast routing is not MPSAFE yet */ 784 get_mplock(); 785 786 /* 787 * If we are acting as a multicast router, all 788 * incoming multicast packets are passed to the 789 * kernel-level multicast forwarding function. 790 * The packet is returned (relatively) intact; if 791 * ip_mforward() returns a non-zero value, the packet 792 * must be discarded, else it may be accepted below. 793 */ 794 if (ip_mforward != NULL && 795 ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) { 796 rel_mplock(); 797 ipstat.ips_cantforward++; 798 m_freem(m); 799 return; 800 } 801 802 rel_mplock(); 803 804 /* 805 * The process-level routing daemon needs to receive 806 * all multicast IGMP packets, whether or not this 807 * host belongs to their destination groups. 808 */ 809 if (ip->ip_p == IPPROTO_IGMP) 810 goto ours; 811 ipstat.ips_forward++; 812 } 813 /* 814 * See if we belong to the destination multicast group on the 815 * arrival interface. 816 */ 817 inm = IN_LOOKUP_MULTI(&ip->ip_dst, m->m_pkthdr.rcvif); 818 if (inm == NULL) { 819 ipstat.ips_notmember++; 820 m_freem(m); 821 return; 822 } 823 goto ours; 824 } 825 if (ip->ip_dst.s_addr == INADDR_BROADCAST) 826 goto ours; 827 if (ip->ip_dst.s_addr == INADDR_ANY) 828 goto ours; 829 830 /* 831 * FAITH(Firewall Aided Internet Translator) 832 */ 833 if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) { 834 if (ip_keepfaith) { 835 if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 836 goto ours; 837 } 838 m_freem(m); 839 return; 840 } 841 842 /* 843 * Not for us; forward if possible and desirable. 844 */ 845 if (!ipforwarding) { 846 ipstat.ips_cantforward++; 847 m_freem(m); 848 } else { 849 #ifdef IPSEC 850 /* 851 * Enforce inbound IPsec SPD. 852 */ 853 if (ipsec4_in_reject(m, NULL)) { 854 ipsecstat.in_polvio++; 855 goto bad; 856 } 857 #endif 858 #ifdef FAST_IPSEC 859 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 860 crit_enter(); 861 if (mtag != NULL) { 862 tdbi = (struct tdb_ident *)m_tag_data(mtag); 863 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 864 } else { 865 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 866 IP_FORWARDING, &error); 867 } 868 if (sp == NULL) { /* NB: can happen if error */ 869 crit_exit(); 870 /*XXX error stat???*/ 871 DPRINTF(("ip_input: no SP for forwarding\n")); /*XXX*/ 872 goto bad; 873 } 874 875 /* 876 * Check security policy against packet attributes. 877 */ 878 error = ipsec_in_reject(sp, m); 879 KEY_FREESP(&sp); 880 crit_exit(); 881 if (error) { 882 ipstat.ips_cantforward++; 883 goto bad; 884 } 885 #endif 886 ip_forward(m, using_srcrt, next_hop); 887 } 888 return; 889 890 ours: 891 892 /* 893 * IPSTEALTH: Process non-routing options only 894 * if the packet is destined for us. 895 */ 896 if (ipstealth && 897 hlen > sizeof(struct ip) && 898 ip_dooptions(m, 1, next_hop)) 899 return; 900 901 /* Count the packet in the ip address stats */ 902 if (ia != NULL) { 903 IFA_STAT_INC(&ia->ia_ifa, ipackets, 1); 904 IFA_STAT_INC(&ia->ia_ifa, ibytes, m->m_pkthdr.len); 905 } 906 907 /* 908 * If offset or IP_MF are set, must reassemble. 909 * Otherwise, nothing need be done. 910 * (We could look in the reassembly queue to see 911 * if the packet was previously fragmented, 912 * but it's not worth the time; just let them time out.) 913 */ 914 if (ip->ip_off & (IP_MF | IP_OFFMASK)) { 915 /* 916 * Attempt reassembly; if it succeeds, proceed. ip_reass() 917 * will return a different mbuf. 918 * 919 * NOTE: ip_reass() returns m with M_HASH cleared to force 920 * us to recharacterize the packet. 921 */ 922 m = ip_reass(m); 923 if (m == NULL) 924 return; 925 ip = mtod(m, struct ip *); 926 927 /* Get the header length of the reassembled packet */ 928 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 929 } else { 930 ip->ip_len -= hlen; 931 } 932 933 #ifdef IPSEC 934 /* 935 * enforce IPsec policy checking if we are seeing last header. 936 * note that we do not visit this with protocols with pcb layer 937 * code - like udp/tcp/raw ip. 938 */ 939 if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) && 940 ipsec4_in_reject(m, NULL)) { 941 ipsecstat.in_polvio++; 942 goto bad; 943 } 944 #endif 945 #ifdef FAST_IPSEC 946 /* 947 * enforce IPsec policy checking if we are seeing last header. 948 * note that we do not visit this with protocols with pcb layer 949 * code - like udp/tcp/raw ip. 950 */ 951 if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) { 952 /* 953 * Check if the packet has already had IPsec processing 954 * done. If so, then just pass it along. This tag gets 955 * set during AH, ESP, etc. input handling, before the 956 * packet is returned to the ip input queue for delivery. 957 */ 958 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL); 959 crit_enter(); 960 if (mtag != NULL) { 961 tdbi = (struct tdb_ident *)m_tag_data(mtag); 962 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND); 963 } else { 964 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, 965 IP_FORWARDING, &error); 966 } 967 if (sp != NULL) { 968 /* 969 * Check security policy against packet attributes. 970 */ 971 error = ipsec_in_reject(sp, m); 972 KEY_FREESP(&sp); 973 } else { 974 /* XXX error stat??? */ 975 error = EINVAL; 976 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/ 977 crit_exit(); 978 goto bad; 979 } 980 crit_exit(); 981 if (error) 982 goto bad; 983 } 984 #endif /* FAST_IPSEC */ 985 986 /* 987 * We must forward the packet to the correct protocol thread if 988 * we are not already in it. 989 * 990 * NOTE: ip_len is now in host form. ip_len is not adjusted 991 * further for protocol processing, instead we pass hlen 992 * to the protosw and let it deal with it. 993 */ 994 ipstat.ips_delivered++; 995 996 if ((m->m_flags & M_HASH) == 0) { 997 #ifdef RSS_DEBUG 998 atomic_add_long(&ip_rehash_count, 1); 999 #endif 1000 ip->ip_len = htons(ip->ip_len + hlen); 1001 ip->ip_off = htons(ip->ip_off); 1002 1003 ip_hashfn(&m, 0); 1004 if (m == NULL) 1005 return; 1006 1007 ip = mtod(m, struct ip *); 1008 ip->ip_len = ntohs(ip->ip_len) - hlen; 1009 ip->ip_off = ntohs(ip->ip_off); 1010 KKASSERT(m->m_flags & M_HASH); 1011 } 1012 port = netisr_hashport(m->m_pkthdr.hash); 1013 1014 if (port != &curthread->td_msgport) { 1015 struct netmsg_packet *pmsg; 1016 1017 #ifdef RSS_DEBUG 1018 atomic_add_long(&ip_dispatch_slow, 1); 1019 #endif 1020 1021 pmsg = &m->m_hdr.mh_netmsg; 1022 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 1023 0, transport_processing_handler); 1024 pmsg->nm_packet = m; 1025 pmsg->base.lmsg.u.ms_result = hlen; 1026 lwkt_sendmsg(port, &pmsg->base.lmsg); 1027 } else { 1028 #ifdef RSS_DEBUG 1029 atomic_add_long(&ip_dispatch_fast, 1); 1030 #endif 1031 transport_processing_oncpu(m, hlen, ip); 1032 } 1033 return; 1034 1035 bad: 1036 m_freem(m); 1037 } 1038 1039 /* 1040 * Take incoming datagram fragment and try to reassemble it into 1041 * whole datagram. If a chain for reassembly of this datagram already 1042 * exists, then it is given as fp; otherwise have to make a chain. 1043 */ 1044 struct mbuf * 1045 ip_reass(struct mbuf *m) 1046 { 1047 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid]; 1048 struct ip *ip = mtod(m, struct ip *); 1049 struct mbuf *p = NULL, *q, *nq; 1050 struct mbuf *n; 1051 struct ipq *fp = NULL; 1052 struct ipqhead *head; 1053 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1054 int i, next; 1055 u_short sum; 1056 1057 /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */ 1058 if (maxnipq == 0 || maxfragsperpacket == 0) { 1059 ipstat.ips_fragments++; 1060 ipstat.ips_fragdropped++; 1061 m_freem(m); 1062 return NULL; 1063 } 1064 1065 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id); 1066 /* 1067 * Look for queue of fragments of this datagram. 1068 */ 1069 head = &fragq->ipq[sum]; 1070 TAILQ_FOREACH(fp, head, ipq_list) { 1071 if (ip->ip_id == fp->ipq_id && 1072 ip->ip_src.s_addr == fp->ipq_src.s_addr && 1073 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 1074 ip->ip_p == fp->ipq_p) 1075 goto found; 1076 } 1077 1078 fp = NULL; 1079 1080 /* 1081 * Enforce upper bound on number of fragmented packets 1082 * for which we attempt reassembly; 1083 * If maxnipq is -1, accept all fragments without limitation. 1084 */ 1085 if (fragq->nipq > maxnipq && maxnipq > 0) { 1086 /* 1087 * drop something from the tail of the current queue 1088 * before proceeding further 1089 */ 1090 struct ipq *q = TAILQ_LAST(head, ipqhead); 1091 if (q == NULL) { 1092 /* 1093 * The current queue is empty, 1094 * so drop from one of the others. 1095 */ 1096 for (i = 0; i < IPREASS_NHASH; i++) { 1097 struct ipq *r = TAILQ_LAST(&fragq->ipq[i], 1098 ipqhead); 1099 if (r) { 1100 ipstat.ips_fragtimeout += r->ipq_nfrags; 1101 ip_freef(fragq, &fragq->ipq[i], r); 1102 break; 1103 } 1104 } 1105 } else { 1106 ipstat.ips_fragtimeout += q->ipq_nfrags; 1107 ip_freef(fragq, head, q); 1108 } 1109 } 1110 found: 1111 /* 1112 * Adjust ip_len to not reflect header, 1113 * convert offset of this to bytes. 1114 */ 1115 ip->ip_len -= hlen; 1116 if (ip->ip_off & IP_MF) { 1117 /* 1118 * Make sure that fragments have a data length 1119 * that's a non-zero multiple of 8 bytes. 1120 */ 1121 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) { 1122 ipstat.ips_toosmall++; /* XXX */ 1123 m_freem(m); 1124 goto done; 1125 } 1126 m->m_flags |= M_FRAG; 1127 } else { 1128 m->m_flags &= ~M_FRAG; 1129 } 1130 ip->ip_off <<= 3; 1131 1132 ipstat.ips_fragments++; 1133 m->m_pkthdr.header = ip; 1134 1135 /* 1136 * If the hardware has not done csum over this fragment 1137 * then csum_data is not valid at all. 1138 */ 1139 if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) 1140 == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) { 1141 m->m_pkthdr.csum_data = 0; 1142 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR); 1143 } 1144 1145 /* 1146 * Presence of header sizes in mbufs 1147 * would confuse code below. 1148 */ 1149 m->m_data += hlen; 1150 m->m_len -= hlen; 1151 1152 /* 1153 * If first fragment to arrive, create a reassembly queue. 1154 */ 1155 if (fp == NULL) { 1156 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL) 1157 goto dropfrag; 1158 TAILQ_INSERT_HEAD(head, fp, ipq_list); 1159 fragq->nipq++; 1160 fp->ipq_nfrags = 1; 1161 fp->ipq_ttl = IPFRAGTTL; 1162 fp->ipq_p = ip->ip_p; 1163 fp->ipq_id = ip->ip_id; 1164 fp->ipq_src = ip->ip_src; 1165 fp->ipq_dst = ip->ip_dst; 1166 fp->ipq_frags = m; 1167 m->m_nextpkt = NULL; 1168 goto inserted; 1169 } 1170 fp->ipq_nfrags++; 1171 1172 #define GETIP(m) ((struct ip*)((m)->m_pkthdr.header)) 1173 1174 /* 1175 * Find a segment which begins after this one does. 1176 */ 1177 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1178 if (GETIP(q)->ip_off > ip->ip_off) 1179 break; 1180 } 1181 1182 /* 1183 * If there is a preceding segment, it may provide some of 1184 * our data already. If so, drop the data from the incoming 1185 * segment. If it provides all of our data, drop us, otherwise 1186 * stick new segment in the proper place. 1187 * 1188 * If some of the data is dropped from the the preceding 1189 * segment, then it's checksum is invalidated. 1190 */ 1191 if (p) { 1192 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off; 1193 if (i > 0) { 1194 if (i >= ip->ip_len) 1195 goto dropfrag; 1196 m_adj(m, i); 1197 m->m_pkthdr.csum_flags = 0; 1198 ip->ip_off += i; 1199 ip->ip_len -= i; 1200 } 1201 m->m_nextpkt = p->m_nextpkt; 1202 p->m_nextpkt = m; 1203 } else { 1204 m->m_nextpkt = fp->ipq_frags; 1205 fp->ipq_frags = m; 1206 } 1207 1208 /* 1209 * While we overlap succeeding segments trim them or, 1210 * if they are completely covered, dequeue them. 1211 */ 1212 for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off; 1213 q = nq) { 1214 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off; 1215 if (i < GETIP(q)->ip_len) { 1216 GETIP(q)->ip_len -= i; 1217 GETIP(q)->ip_off += i; 1218 m_adj(q, i); 1219 q->m_pkthdr.csum_flags = 0; 1220 break; 1221 } 1222 nq = q->m_nextpkt; 1223 m->m_nextpkt = nq; 1224 ipstat.ips_fragdropped++; 1225 fp->ipq_nfrags--; 1226 q->m_nextpkt = NULL; 1227 m_freem(q); 1228 } 1229 1230 inserted: 1231 /* 1232 * Check for complete reassembly and perform frag per packet 1233 * limiting. 1234 * 1235 * Frag limiting is performed here so that the nth frag has 1236 * a chance to complete the packet before we drop the packet. 1237 * As a result, n+1 frags are actually allowed per packet, but 1238 * only n will ever be stored. (n = maxfragsperpacket.) 1239 * 1240 */ 1241 next = 0; 1242 for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) { 1243 if (GETIP(q)->ip_off != next) { 1244 if (fp->ipq_nfrags > maxfragsperpacket) { 1245 ipstat.ips_fragdropped += fp->ipq_nfrags; 1246 ip_freef(fragq, head, fp); 1247 } 1248 goto done; 1249 } 1250 next += GETIP(q)->ip_len; 1251 } 1252 /* Make sure the last packet didn't have the IP_MF flag */ 1253 if (p->m_flags & M_FRAG) { 1254 if (fp->ipq_nfrags > maxfragsperpacket) { 1255 ipstat.ips_fragdropped += fp->ipq_nfrags; 1256 ip_freef(fragq, head, fp); 1257 } 1258 goto done; 1259 } 1260 1261 /* 1262 * Reassembly is complete. Make sure the packet is a sane size. 1263 */ 1264 q = fp->ipq_frags; 1265 ip = GETIP(q); 1266 if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) { 1267 ipstat.ips_toolong++; 1268 ipstat.ips_fragdropped += fp->ipq_nfrags; 1269 ip_freef(fragq, head, fp); 1270 goto done; 1271 } 1272 1273 /* 1274 * Concatenate fragments. 1275 */ 1276 m = q; 1277 n = m->m_next; 1278 m->m_next = NULL; 1279 m_cat(m, n); 1280 nq = q->m_nextpkt; 1281 q->m_nextpkt = NULL; 1282 for (q = nq; q != NULL; q = nq) { 1283 nq = q->m_nextpkt; 1284 q->m_nextpkt = NULL; 1285 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags; 1286 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data; 1287 m_cat(m, q); 1288 } 1289 1290 /* 1291 * Clean up the 1's complement checksum. Carry over 16 bits must 1292 * be added back. This assumes no more then 65535 packet fragments 1293 * were reassembled. A second carry can also occur (but not a third). 1294 */ 1295 m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) + 1296 (m->m_pkthdr.csum_data >> 16); 1297 if (m->m_pkthdr.csum_data > 0xFFFF) 1298 m->m_pkthdr.csum_data -= 0xFFFF; 1299 1300 /* 1301 * Create header for new ip packet by 1302 * modifying header of first packet; 1303 * dequeue and discard fragment reassembly header. 1304 * Make header visible. 1305 */ 1306 ip->ip_len = next; 1307 ip->ip_src = fp->ipq_src; 1308 ip->ip_dst = fp->ipq_dst; 1309 TAILQ_REMOVE(head, fp, ipq_list); 1310 fragq->nipq--; 1311 mpipe_free(&ipq_mpipe, fp); 1312 m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2); 1313 m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2); 1314 /* some debugging cruft by sklower, below, will go away soon */ 1315 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 1316 int plen = 0; 1317 1318 for (n = m; n; n = n->m_next) 1319 plen += n->m_len; 1320 m->m_pkthdr.len = plen; 1321 } 1322 1323 /* 1324 * Reassembly complete, return the next protocol. 1325 * 1326 * Be sure to clear M_HASH to force the packet 1327 * to be re-characterized. 1328 * 1329 * Clear M_FRAG, we are no longer a fragment. 1330 */ 1331 m->m_flags &= ~(M_HASH | M_FRAG); 1332 1333 ipstat.ips_reassembled++; 1334 return (m); 1335 1336 dropfrag: 1337 ipstat.ips_fragdropped++; 1338 if (fp != NULL) 1339 fp->ipq_nfrags--; 1340 m_freem(m); 1341 done: 1342 return (NULL); 1343 1344 #undef GETIP 1345 } 1346 1347 /* 1348 * Free a fragment reassembly header and all 1349 * associated datagrams. 1350 */ 1351 static void 1352 ip_freef(struct ipfrag_queue *fragq, struct ipqhead *fhp, struct ipq *fp) 1353 { 1354 struct mbuf *q; 1355 1356 /* 1357 * Remove first to protect against blocking 1358 */ 1359 TAILQ_REMOVE(fhp, fp, ipq_list); 1360 1361 /* 1362 * Clean out at our leisure 1363 */ 1364 while (fp->ipq_frags) { 1365 q = fp->ipq_frags; 1366 fp->ipq_frags = q->m_nextpkt; 1367 q->m_nextpkt = NULL; 1368 m_freem(q); 1369 } 1370 mpipe_free(&ipq_mpipe, fp); 1371 fragq->nipq--; 1372 } 1373 1374 /* 1375 * If a timer expires on a reassembly queue, discard it. 1376 */ 1377 static void 1378 ipfrag_timeo_dispatch(netmsg_t nmsg) 1379 { 1380 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid]; 1381 struct ipq *fp, *fp_temp; 1382 struct ipqhead *head; 1383 int i; 1384 1385 crit_enter(); 1386 netisr_replymsg(&nmsg->base, 0); /* reply ASAP */ 1387 crit_exit(); 1388 1389 if (fragq->nipq == 0) 1390 goto done; 1391 1392 for (i = 0; i < IPREASS_NHASH; i++) { 1393 head = &fragq->ipq[i]; 1394 TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) { 1395 if (--fp->ipq_ttl == 0) { 1396 ipstat.ips_fragtimeout += fp->ipq_nfrags; 1397 ip_freef(fragq, head, fp); 1398 } 1399 } 1400 } 1401 /* 1402 * If we are over the maximum number of fragments 1403 * (due to the limit being lowered), drain off 1404 * enough to get down to the new limit. 1405 */ 1406 if (maxnipq >= 0 && fragq->nipq > maxnipq) { 1407 for (i = 0; i < IPREASS_NHASH; i++) { 1408 head = &fragq->ipq[i]; 1409 while (fragq->nipq > maxnipq && !TAILQ_EMPTY(head)) { 1410 ipstat.ips_fragdropped += 1411 TAILQ_FIRST(head)->ipq_nfrags; 1412 ip_freef(fragq, head, TAILQ_FIRST(head)); 1413 } 1414 } 1415 } 1416 done: 1417 callout_reset(&fragq->timeo_ch, IPFRAG_TIMEO, ipfrag_timeo, NULL); 1418 } 1419 1420 static void 1421 ipfrag_timeo(void *dummy __unused) 1422 { 1423 struct netmsg_base *msg = &ipfrag_queue_pcpu[mycpuid].timeo_netmsg; 1424 1425 crit_enter(); 1426 if (msg->lmsg.ms_flags & MSGF_DONE) 1427 netisr_sendmsg_oncpu(msg); 1428 crit_exit(); 1429 } 1430 1431 /* 1432 * Drain off all datagram fragments. 1433 */ 1434 static void 1435 ipfrag_drain_oncpu(struct ipfrag_queue *fragq) 1436 { 1437 struct ipqhead *head; 1438 int i; 1439 1440 for (i = 0; i < IPREASS_NHASH; i++) { 1441 head = &fragq->ipq[i]; 1442 while (!TAILQ_EMPTY(head)) { 1443 ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags; 1444 ip_freef(fragq, head, TAILQ_FIRST(head)); 1445 } 1446 } 1447 } 1448 1449 static void 1450 ipfrag_drain_dispatch(netmsg_t nmsg) 1451 { 1452 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid]; 1453 1454 crit_enter(); 1455 lwkt_replymsg(&nmsg->lmsg, 0); /* reply ASAP */ 1456 crit_exit(); 1457 1458 ipfrag_drain_oncpu(fragq); 1459 fragq->draining = 0; 1460 } 1461 1462 static void 1463 ipfrag_drain_ipi(void *arg __unused) 1464 { 1465 int cpu = mycpuid; 1466 struct lwkt_msg *msg = &ipfrag_queue_pcpu[cpu].drain_netmsg.lmsg; 1467 1468 crit_enter(); 1469 if (msg->ms_flags & MSGF_DONE) 1470 lwkt_sendmsg_oncpu(netisr_cpuport(cpu), msg); 1471 crit_exit(); 1472 } 1473 1474 static void 1475 ipfrag_drain(void) 1476 { 1477 cpumask_t mask; 1478 int cpu; 1479 1480 CPUMASK_ASSBMASK(mask, netisr_ncpus); 1481 CPUMASK_ANDMASK(mask, smp_active_mask); 1482 1483 if (IN_NETISR_NCPUS(mycpuid)) { 1484 ipfrag_drain_oncpu(&ipfrag_queue_pcpu[mycpuid]); 1485 CPUMASK_NANDBIT(mask, mycpuid); 1486 } 1487 1488 for (cpu = 0; cpu < netisr_ncpus; ++cpu) { 1489 struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[cpu]; 1490 1491 if (!CPUMASK_TESTBIT(mask, cpu)) 1492 continue; 1493 1494 if (fragq->nipq == 0 || fragq->draining) { 1495 /* No fragments or is draining; skip this cpu. */ 1496 CPUMASK_NANDBIT(mask, cpu); 1497 continue; 1498 } 1499 fragq->draining = 1; 1500 } 1501 1502 if (CPUMASK_TESTNZERO(mask)) 1503 lwkt_send_ipiq_mask(mask, ipfrag_drain_ipi, NULL); 1504 } 1505 1506 void 1507 ip_drain(void) 1508 { 1509 ipfrag_drain(); 1510 in_rtqdrain(); 1511 } 1512 1513 /* 1514 * Do option processing on a datagram, 1515 * possibly discarding it if bad options are encountered, 1516 * or forwarding it if source-routed. 1517 * The pass argument is used when operating in the IPSTEALTH 1518 * mode to tell what options to process: 1519 * [LS]SRR (pass 0) or the others (pass 1). 1520 * The reason for as many as two passes is that when doing IPSTEALTH, 1521 * non-routing options should be processed only if the packet is for us. 1522 * Returns 1 if packet has been forwarded/freed, 1523 * 0 if the packet should be processed further. 1524 */ 1525 static int 1526 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop) 1527 { 1528 struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET }; 1529 struct ip *ip = mtod(m, struct ip *); 1530 u_char *cp; 1531 struct in_ifaddr *ia; 1532 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB; 1533 boolean_t forward = FALSE; 1534 struct in_addr *sin, dst; 1535 n_time ntime; 1536 1537 dst = ip->ip_dst; 1538 cp = (u_char *)(ip + 1); 1539 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1540 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1541 opt = cp[IPOPT_OPTVAL]; 1542 if (opt == IPOPT_EOL) 1543 break; 1544 if (opt == IPOPT_NOP) 1545 optlen = 1; 1546 else { 1547 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1548 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1549 goto bad; 1550 } 1551 optlen = cp[IPOPT_OLEN]; 1552 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1553 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1554 goto bad; 1555 } 1556 } 1557 switch (opt) { 1558 1559 default: 1560 break; 1561 1562 /* 1563 * Source routing with record. 1564 * Find interface with current destination address. 1565 * If none on this machine then drop if strictly routed, 1566 * or do nothing if loosely routed. 1567 * Record interface address and bring up next address 1568 * component. If strictly routed make sure next 1569 * address is on directly accessible net. 1570 */ 1571 case IPOPT_LSRR: 1572 case IPOPT_SSRR: 1573 if (ipstealth && pass > 0) 1574 break; 1575 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1576 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1577 goto bad; 1578 } 1579 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1580 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1581 goto bad; 1582 } 1583 ipaddr.sin_addr = ip->ip_dst; 1584 ia = (struct in_ifaddr *) 1585 ifa_ifwithaddr((struct sockaddr *)&ipaddr); 1586 if (ia == NULL) { 1587 if (opt == IPOPT_SSRR) { 1588 type = ICMP_UNREACH; 1589 code = ICMP_UNREACH_SRCFAIL; 1590 goto bad; 1591 } 1592 if (!ip_dosourceroute) 1593 goto nosourcerouting; 1594 /* 1595 * Loose routing, and not at next destination 1596 * yet; nothing to do except forward. 1597 */ 1598 break; 1599 } 1600 off--; /* 0 origin */ 1601 if (off > optlen - (int)sizeof(struct in_addr)) { 1602 /* 1603 * End of source route. Should be for us. 1604 */ 1605 if (!ip_acceptsourceroute) 1606 goto nosourcerouting; 1607 save_rte(m, cp, ip->ip_src); 1608 break; 1609 } 1610 if (ipstealth) 1611 goto dropit; 1612 if (!ip_dosourceroute) { 1613 if (ipforwarding) { 1614 char sbuf[INET_ADDRSTRLEN]; 1615 char dbuf[INET_ADDRSTRLEN]; 1616 1617 /* 1618 * Acting as a router, so generate ICMP 1619 */ 1620 nosourcerouting: 1621 log(LOG_WARNING, 1622 "attempted source route from %s to %s\n", 1623 kinet_ntoa(ip->ip_src, sbuf), 1624 kinet_ntoa(ip->ip_dst, dbuf)); 1625 type = ICMP_UNREACH; 1626 code = ICMP_UNREACH_SRCFAIL; 1627 goto bad; 1628 } else { 1629 /* 1630 * Not acting as a router, 1631 * so silently drop. 1632 */ 1633 dropit: 1634 ipstat.ips_cantforward++; 1635 m_freem(m); 1636 return (1); 1637 } 1638 } 1639 1640 /* 1641 * locate outgoing interface 1642 */ 1643 memcpy(&ipaddr.sin_addr, cp + off, 1644 sizeof ipaddr.sin_addr); 1645 1646 if (opt == IPOPT_SSRR) { 1647 #define INA struct in_ifaddr * 1648 #define SA struct sockaddr * 1649 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) 1650 == NULL) 1651 ia = (INA)ifa_ifwithnet((SA)&ipaddr); 1652 } else { 1653 ia = ip_rtaddr(ipaddr.sin_addr, NULL); 1654 } 1655 if (ia == NULL) { 1656 type = ICMP_UNREACH; 1657 code = ICMP_UNREACH_SRCFAIL; 1658 goto bad; 1659 } 1660 ip->ip_dst = ipaddr.sin_addr; 1661 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1662 sizeof(struct in_addr)); 1663 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1664 /* 1665 * Let ip_intr's mcast routing check handle mcast pkts 1666 */ 1667 forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr)); 1668 break; 1669 1670 case IPOPT_RR: 1671 if (ipstealth && pass == 0) 1672 break; 1673 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1674 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1675 goto bad; 1676 } 1677 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1678 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1679 goto bad; 1680 } 1681 /* 1682 * If no space remains, ignore. 1683 */ 1684 off--; /* 0 origin */ 1685 if (off > optlen - (int)sizeof(struct in_addr)) 1686 break; 1687 memcpy(&ipaddr.sin_addr, &ip->ip_dst, 1688 sizeof ipaddr.sin_addr); 1689 /* 1690 * locate outgoing interface; if we're the destination, 1691 * use the incoming interface (should be same). 1692 */ 1693 if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL && 1694 (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) { 1695 type = ICMP_UNREACH; 1696 code = ICMP_UNREACH_HOST; 1697 goto bad; 1698 } 1699 memcpy(cp + off, &IA_SIN(ia)->sin_addr, 1700 sizeof(struct in_addr)); 1701 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1702 break; 1703 1704 case IPOPT_TS: 1705 if (ipstealth && pass == 0) 1706 break; 1707 code = cp - (u_char *)ip; 1708 if (optlen < 4 || optlen > 40) { 1709 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1710 goto bad; 1711 } 1712 if ((off = cp[IPOPT_OFFSET]) < 5) { 1713 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1714 goto bad; 1715 } 1716 if (off > optlen - (int)sizeof(int32_t)) { 1717 cp[IPOPT_OFFSET + 1] += (1 << 4); 1718 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) { 1719 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1720 goto bad; 1721 } 1722 break; 1723 } 1724 off--; /* 0 origin */ 1725 sin = (struct in_addr *)(cp + off); 1726 switch (cp[IPOPT_OFFSET + 1] & 0x0f) { 1727 1728 case IPOPT_TS_TSONLY: 1729 break; 1730 1731 case IPOPT_TS_TSANDADDR: 1732 if (off + sizeof(n_time) + 1733 sizeof(struct in_addr) > optlen) { 1734 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1735 goto bad; 1736 } 1737 ipaddr.sin_addr = dst; 1738 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr, 1739 m->m_pkthdr.rcvif); 1740 if (ia == NULL) 1741 continue; 1742 memcpy(sin, &IA_SIN(ia)->sin_addr, 1743 sizeof(struct in_addr)); 1744 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1745 off += sizeof(struct in_addr); 1746 break; 1747 1748 case IPOPT_TS_PRESPEC: 1749 if (off + sizeof(n_time) + 1750 sizeof(struct in_addr) > optlen) { 1751 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1752 goto bad; 1753 } 1754 memcpy(&ipaddr.sin_addr, sin, 1755 sizeof(struct in_addr)); 1756 if (ifa_ifwithaddr((SA)&ipaddr) == NULL) 1757 continue; 1758 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1759 off += sizeof(struct in_addr); 1760 break; 1761 1762 default: 1763 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip; 1764 goto bad; 1765 } 1766 ntime = iptime(); 1767 memcpy(cp + off, &ntime, sizeof(n_time)); 1768 cp[IPOPT_OFFSET] += sizeof(n_time); 1769 } 1770 } 1771 if (forward && ipforwarding) { 1772 ip_forward(m, TRUE, next_hop); 1773 return (1); 1774 } 1775 return (0); 1776 bad: 1777 icmp_error(m, type, code, 0, 0); 1778 ipstat.ips_badoptions++; 1779 return (1); 1780 } 1781 1782 /* 1783 * Given address of next destination (final or next hop), 1784 * return internet address info of interface to be used to get there. 1785 */ 1786 struct in_ifaddr * 1787 ip_rtaddr(struct in_addr dst, struct route *ro0) 1788 { 1789 struct route sro, *ro; 1790 struct sockaddr_in *sin; 1791 struct in_ifaddr *ia; 1792 1793 if (ro0 != NULL) { 1794 ro = ro0; 1795 } else { 1796 bzero(&sro, sizeof(sro)); 1797 ro = &sro; 1798 } 1799 1800 sin = (struct sockaddr_in *)&ro->ro_dst; 1801 1802 if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) { 1803 if (ro->ro_rt != NULL) { 1804 RTFREE(ro->ro_rt); 1805 ro->ro_rt = NULL; 1806 } 1807 sin->sin_family = AF_INET; 1808 sin->sin_len = sizeof *sin; 1809 sin->sin_addr = dst; 1810 rtalloc_ign(ro, RTF_PRCLONING); 1811 } 1812 1813 if (ro->ro_rt == NULL) 1814 return (NULL); 1815 1816 ia = ifatoia(ro->ro_rt->rt_ifa); 1817 1818 if (ro == &sro) 1819 RTFREE(ro->ro_rt); 1820 return ia; 1821 } 1822 1823 /* 1824 * Save incoming source route for use in replies, 1825 * to be picked up later by ip_srcroute if the receiver is interested. 1826 */ 1827 static void 1828 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1829 { 1830 struct m_tag *mtag; 1831 struct ip_srcrt_opt *opt; 1832 unsigned olen; 1833 1834 mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), M_NOWAIT); 1835 if (mtag == NULL) 1836 return; 1837 opt = m_tag_data(mtag); 1838 1839 olen = option[IPOPT_OLEN]; 1840 #ifdef DIAGNOSTIC 1841 if (ipprintfs) 1842 kprintf("save_rte: olen %d\n", olen); 1843 #endif 1844 if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) { 1845 m_tag_free(mtag); 1846 return; 1847 } 1848 bcopy(option, opt->ip_srcrt.srcopt, olen); 1849 opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1850 opt->ip_srcrt.dst = dst; 1851 m_tag_prepend(m, mtag); 1852 } 1853 1854 /* 1855 * Retrieve incoming source route for use in replies, 1856 * in the same form used by setsockopt. 1857 * The first hop is placed before the options, will be removed later. 1858 */ 1859 struct mbuf * 1860 ip_srcroute(struct mbuf *m0) 1861 { 1862 struct in_addr *p, *q; 1863 struct mbuf *m; 1864 struct m_tag *mtag; 1865 struct ip_srcrt_opt *opt; 1866 1867 if (m0 == NULL) 1868 return NULL; 1869 1870 mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL); 1871 if (mtag == NULL) 1872 return NULL; 1873 opt = m_tag_data(mtag); 1874 1875 if (opt->ip_nhops == 0) 1876 return (NULL); 1877 m = m_get(M_NOWAIT, MT_HEADER); 1878 if (m == NULL) 1879 return (NULL); 1880 1881 #define OPTSIZ (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt)) 1882 1883 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */ 1884 m->m_len = opt->ip_nhops * sizeof(struct in_addr) + 1885 sizeof(struct in_addr) + OPTSIZ; 1886 #ifdef DIAGNOSTIC 1887 if (ipprintfs) { 1888 kprintf("ip_srcroute: nhops %d mlen %d", 1889 opt->ip_nhops, m->m_len); 1890 } 1891 #endif 1892 1893 /* 1894 * First save first hop for return route 1895 */ 1896 p = &opt->ip_srcrt.route[opt->ip_nhops - 1]; 1897 *(mtod(m, struct in_addr *)) = *p--; 1898 #ifdef DIAGNOSTIC 1899 if (ipprintfs) 1900 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr)); 1901 #endif 1902 1903 /* 1904 * Copy option fields and padding (nop) to mbuf. 1905 */ 1906 opt->ip_srcrt.nop = IPOPT_NOP; 1907 opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF; 1908 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop, 1909 OPTSIZ); 1910 q = (struct in_addr *)(mtod(m, caddr_t) + 1911 sizeof(struct in_addr) + OPTSIZ); 1912 #undef OPTSIZ 1913 /* 1914 * Record return path as an IP source route, 1915 * reversing the path (pointers are now aligned). 1916 */ 1917 while (p >= opt->ip_srcrt.route) { 1918 #ifdef DIAGNOSTIC 1919 if (ipprintfs) 1920 kprintf(" %x", ntohl(q->s_addr)); 1921 #endif 1922 *q++ = *p--; 1923 } 1924 /* 1925 * Last hop goes to final destination. 1926 */ 1927 *q = opt->ip_srcrt.dst; 1928 m_tag_delete(m0, mtag); 1929 #ifdef DIAGNOSTIC 1930 if (ipprintfs) 1931 kprintf(" %x\n", ntohl(q->s_addr)); 1932 #endif 1933 return (m); 1934 } 1935 1936 /* 1937 * Strip out IP options. 1938 */ 1939 void 1940 ip_stripoptions(struct mbuf *m) 1941 { 1942 int datalen; 1943 struct ip *ip = mtod(m, struct ip *); 1944 caddr_t opts; 1945 int optlen; 1946 1947 optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1948 opts = (caddr_t)(ip + 1); 1949 datalen = m->m_len - (sizeof(struct ip) + optlen); 1950 bcopy(opts + optlen, opts, datalen); 1951 m->m_len -= optlen; 1952 if (m->m_flags & M_PKTHDR) 1953 m->m_pkthdr.len -= optlen; 1954 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2); 1955 } 1956 1957 u_char inetctlerrmap[PRC_NCMDS] = { 1958 0, 0, 0, 0, 1959 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1960 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1961 EMSGSIZE, EHOSTUNREACH, 0, 0, 1962 0, 0, 0, 0, 1963 ENOPROTOOPT, ECONNREFUSED 1964 }; 1965 1966 /* 1967 * Forward a packet. If some error occurs return the sender 1968 * an icmp packet. Note we can't always generate a meaningful 1969 * icmp message because icmp doesn't have a large enough repertoire 1970 * of codes and types. 1971 * 1972 * If not forwarding, just drop the packet. This could be confusing 1973 * if ipforwarding was zero but some routing protocol was advancing 1974 * us as a gateway to somewhere. However, we must let the routing 1975 * protocol deal with that. 1976 * 1977 * The using_srcrt parameter indicates whether the packet is being forwarded 1978 * via a source route. 1979 */ 1980 void 1981 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop) 1982 { 1983 struct ip *ip = mtod(m, struct ip *); 1984 struct rtentry *rt; 1985 struct route fwd_ro; 1986 int error, type = 0, code = 0, destmtu = 0; 1987 struct mbuf *mcopy, *mtemp = NULL; 1988 n_long dest; 1989 struct in_addr pkt_dst; 1990 1991 dest = INADDR_ANY; 1992 /* 1993 * Cache the destination address of the packet; this may be 1994 * changed by use of 'ipfw fwd'. 1995 */ 1996 pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst; 1997 1998 #ifdef DIAGNOSTIC 1999 if (ipprintfs) 2000 kprintf("forward: src %x dst %x ttl %x\n", 2001 ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl); 2002 #endif 2003 2004 if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) { 2005 ipstat.ips_cantforward++; 2006 m_freem(m); 2007 return; 2008 } 2009 if (!ipstealth && ip->ip_ttl <= IPTTLDEC) { 2010 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 2011 return; 2012 } 2013 2014 bzero(&fwd_ro, sizeof(fwd_ro)); 2015 ip_rtaddr(pkt_dst, &fwd_ro); 2016 if (fwd_ro.ro_rt == NULL) { 2017 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 2018 return; 2019 } 2020 rt = fwd_ro.ro_rt; 2021 2022 if (curthread->td_type == TD_TYPE_NETISR) { 2023 /* 2024 * Save the IP header and at most 8 bytes of the payload, 2025 * in case we need to generate an ICMP message to the src. 2026 */ 2027 mtemp = ipforward_mtemp[mycpuid]; 2028 KASSERT((mtemp->m_flags & M_EXT) == 0 && 2029 mtemp->m_data == mtemp->m_pktdat && 2030 m_tag_first(mtemp) == NULL, 2031 ("ip_forward invalid mtemp1")); 2032 2033 if (!m_dup_pkthdr(mtemp, m, M_NOWAIT)) { 2034 /* 2035 * It's probably ok if the pkthdr dup fails (because 2036 * the deep copy of the tag chain failed), but for now 2037 * be conservative and just discard the copy since 2038 * code below may some day want the tags. 2039 */ 2040 mtemp = NULL; 2041 } else { 2042 mtemp->m_type = m->m_type; 2043 mtemp->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8, 2044 (int)ip->ip_len); 2045 mtemp->m_pkthdr.len = mtemp->m_len; 2046 m_copydata(m, 0, mtemp->m_len, mtod(mtemp, caddr_t)); 2047 } 2048 } 2049 2050 if (!ipstealth) 2051 ip->ip_ttl -= IPTTLDEC; 2052 2053 /* 2054 * If forwarding packet using same interface that it came in on, 2055 * perhaps should send a redirect to sender to shortcut a hop. 2056 * Only send redirect if source is sending directly to us, 2057 * and if packet was not source routed (or has any options). 2058 * Also, don't send redirect if forwarding using a default route 2059 * or a route modified by a redirect. 2060 */ 2061 if (rt->rt_ifp == m->m_pkthdr.rcvif && 2062 !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) && 2063 satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY && 2064 ipsendredirects && !using_srcrt && next_hop == NULL) { 2065 u_long src = ntohl(ip->ip_src.s_addr); 2066 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa; 2067 2068 if (rt_ifa != NULL && 2069 (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) { 2070 if (rt->rt_flags & RTF_GATEWAY) 2071 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 2072 else 2073 dest = pkt_dst.s_addr; 2074 /* 2075 * Router requirements says to only send 2076 * host redirects. 2077 */ 2078 type = ICMP_REDIRECT; 2079 code = ICMP_REDIRECT_HOST; 2080 #ifdef DIAGNOSTIC 2081 if (ipprintfs) 2082 kprintf("redirect (%d) to %x\n", code, dest); 2083 #endif 2084 } 2085 } 2086 2087 error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL); 2088 if (error == 0) { 2089 ipstat.ips_forward++; 2090 if (type == 0) { 2091 if (mtemp) 2092 ipflow_create(&fwd_ro, mtemp); 2093 goto done; 2094 } 2095 ipstat.ips_redirectsent++; 2096 } else { 2097 ipstat.ips_cantforward++; 2098 } 2099 2100 if (mtemp == NULL) 2101 goto done; 2102 2103 /* 2104 * Errors that do not require generating ICMP message 2105 */ 2106 switch (error) { 2107 case ENOBUFS: 2108 /* 2109 * A router should not generate ICMP_SOURCEQUENCH as 2110 * required in RFC1812 Requirements for IP Version 4 Routers. 2111 * Source quench could be a big problem under DoS attacks, 2112 * or if the underlying interface is rate-limited. 2113 * Those who need source quench packets may re-enable them 2114 * via the net.inet.ip.sendsourcequench sysctl. 2115 */ 2116 if (!ip_sendsourcequench) 2117 goto done; 2118 break; 2119 2120 case EACCES: /* ipfw denied packet */ 2121 goto done; 2122 } 2123 2124 KASSERT((mtemp->m_flags & M_EXT) == 0 && 2125 mtemp->m_data == mtemp->m_pktdat, 2126 ("ip_forward invalid mtemp2")); 2127 mcopy = m_copym(mtemp, 0, mtemp->m_len, M_NOWAIT); 2128 if (mcopy == NULL) 2129 goto done; 2130 2131 /* 2132 * Send ICMP message. 2133 */ 2134 switch (error) { 2135 case 0: /* forwarded, but need redirect */ 2136 /* type, code set above */ 2137 break; 2138 2139 case ENETUNREACH: /* shouldn't happen, checked above */ 2140 case EHOSTUNREACH: 2141 case ENETDOWN: 2142 case EHOSTDOWN: 2143 default: 2144 type = ICMP_UNREACH; 2145 code = ICMP_UNREACH_HOST; 2146 break; 2147 2148 case EMSGSIZE: 2149 type = ICMP_UNREACH; 2150 code = ICMP_UNREACH_NEEDFRAG; 2151 #ifdef IPSEC 2152 /* 2153 * If the packet is routed over IPsec tunnel, tell the 2154 * originator the tunnel MTU. 2155 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2156 * XXX quickhack!!! 2157 */ 2158 if (fwd_ro.ro_rt != NULL) { 2159 struct secpolicy *sp = NULL; 2160 int ipsecerror; 2161 int ipsechdr; 2162 struct route *ro; 2163 2164 sp = ipsec4_getpolicybyaddr(mcopy, 2165 IPSEC_DIR_OUTBOUND, 2166 IP_FORWARDING, 2167 &ipsecerror); 2168 2169 if (sp == NULL) 2170 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2171 else { 2172 /* count IPsec header size */ 2173 ipsechdr = ipsec4_hdrsiz(mcopy, 2174 IPSEC_DIR_OUTBOUND, 2175 NULL); 2176 2177 /* 2178 * find the correct route for outer IPv4 2179 * header, compute tunnel MTU. 2180 * 2181 */ 2182 if (sp->req != NULL && sp->req->sav != NULL && 2183 sp->req->sav->sah != NULL) { 2184 ro = &sp->req->sav->sah->sa_route; 2185 if (ro->ro_rt != NULL && 2186 ro->ro_rt->rt_ifp != NULL) { 2187 destmtu = 2188 ro->ro_rt->rt_ifp->if_mtu; 2189 destmtu -= ipsechdr; 2190 } 2191 } 2192 2193 key_freesp(sp); 2194 } 2195 } 2196 #elif defined(FAST_IPSEC) 2197 /* 2198 * If the packet is routed over IPsec tunnel, tell the 2199 * originator the tunnel MTU. 2200 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz 2201 * XXX quickhack!!! 2202 */ 2203 if (fwd_ro.ro_rt != NULL) { 2204 struct secpolicy *sp = NULL; 2205 int ipsecerror; 2206 int ipsechdr; 2207 struct route *ro; 2208 2209 sp = ipsec_getpolicybyaddr(mcopy, 2210 IPSEC_DIR_OUTBOUND, 2211 IP_FORWARDING, 2212 &ipsecerror); 2213 2214 if (sp == NULL) 2215 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2216 else { 2217 /* count IPsec header size */ 2218 ipsechdr = ipsec4_hdrsiz(mcopy, 2219 IPSEC_DIR_OUTBOUND, 2220 NULL); 2221 2222 /* 2223 * find the correct route for outer IPv4 2224 * header, compute tunnel MTU. 2225 */ 2226 2227 if (sp->req != NULL && 2228 sp->req->sav != NULL && 2229 sp->req->sav->sah != NULL) { 2230 ro = &sp->req->sav->sah->sa_route; 2231 if (ro->ro_rt != NULL && 2232 ro->ro_rt->rt_ifp != NULL) { 2233 destmtu = 2234 ro->ro_rt->rt_ifp->if_mtu; 2235 destmtu -= ipsechdr; 2236 } 2237 } 2238 2239 KEY_FREESP(&sp); 2240 } 2241 } 2242 #else /* !IPSEC && !FAST_IPSEC */ 2243 if (fwd_ro.ro_rt != NULL) 2244 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu; 2245 #endif /*IPSEC*/ 2246 ipstat.ips_cantfrag++; 2247 break; 2248 2249 case ENOBUFS: 2250 type = ICMP_SOURCEQUENCH; 2251 code = 0; 2252 break; 2253 2254 case EACCES: /* ipfw denied packet */ 2255 panic("ip_forward EACCES should not reach"); 2256 } 2257 icmp_error(mcopy, type, code, dest, destmtu); 2258 done: 2259 if (mtemp != NULL) 2260 m_tag_delete_chain(mtemp); 2261 if (fwd_ro.ro_rt != NULL) 2262 RTFREE(fwd_ro.ro_rt); 2263 } 2264 2265 void 2266 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 2267 struct mbuf *m) 2268 { 2269 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 2270 struct timeval tv; 2271 2272 microtime(&tv); 2273 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 2274 SCM_TIMESTAMP, SOL_SOCKET); 2275 if (*mp) 2276 mp = &(*mp)->m_next; 2277 } 2278 if (inp->inp_flags & INP_RECVDSTADDR) { 2279 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 2280 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 2281 if (*mp) 2282 mp = &(*mp)->m_next; 2283 } 2284 if (inp->inp_flags & INP_RECVTTL) { 2285 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 2286 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 2287 if (*mp) 2288 mp = &(*mp)->m_next; 2289 } 2290 #ifdef notyet 2291 /* XXX 2292 * Moving these out of udp_input() made them even more broken 2293 * than they already were. 2294 */ 2295 /* options were tossed already */ 2296 if (inp->inp_flags & INP_RECVOPTS) { 2297 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 2298 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 2299 if (*mp) 2300 mp = &(*mp)->m_next; 2301 } 2302 /* ip_srcroute doesn't do what we want here, need to fix */ 2303 if (inp->inp_flags & INP_RECVRETOPTS) { 2304 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 2305 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 2306 if (*mp) 2307 mp = &(*mp)->m_next; 2308 } 2309 #endif 2310 if (inp->inp_flags & INP_RECVIF) { 2311 struct ifnet *ifp; 2312 struct sdlbuf { 2313 struct sockaddr_dl sdl; 2314 u_char pad[32]; 2315 } sdlbuf; 2316 struct sockaddr_dl *sdp; 2317 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 2318 2319 if (((ifp = m->m_pkthdr.rcvif)) && 2320 ((ifp->if_index != 0) && (ifp->if_index <= if_index))) { 2321 sdp = IF_LLSOCKADDR(ifp); 2322 /* 2323 * Change our mind and don't try copy. 2324 */ 2325 if ((sdp->sdl_family != AF_LINK) || 2326 (sdp->sdl_len > sizeof(sdlbuf))) { 2327 goto makedummy; 2328 } 2329 bcopy(sdp, sdl2, sdp->sdl_len); 2330 } else { 2331 makedummy: 2332 sdl2->sdl_len = 2333 offsetof(struct sockaddr_dl, sdl_data[0]); 2334 sdl2->sdl_family = AF_LINK; 2335 sdl2->sdl_index = 0; 2336 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 2337 } 2338 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len, 2339 IP_RECVIF, IPPROTO_IP); 2340 if (*mp) 2341 mp = &(*mp)->m_next; 2342 } 2343 } 2344 2345 /* 2346 * XXX these routines are called from the upper part of the kernel. 2347 * 2348 * They could also be moved to ip_mroute.c, since all the RSVP 2349 * handling is done there already. 2350 */ 2351 int 2352 ip_rsvp_init(struct socket *so) 2353 { 2354 if (so->so_type != SOCK_RAW || 2355 so->so_proto->pr_protocol != IPPROTO_RSVP) 2356 return EOPNOTSUPP; 2357 2358 if (ip_rsvpd != NULL) 2359 return EADDRINUSE; 2360 2361 ip_rsvpd = so; 2362 /* 2363 * This may seem silly, but we need to be sure we don't over-increment 2364 * the RSVP counter, in case something slips up. 2365 */ 2366 if (!ip_rsvp_on) { 2367 ip_rsvp_on = 1; 2368 rsvp_on++; 2369 } 2370 2371 return 0; 2372 } 2373 2374 int 2375 ip_rsvp_done(void) 2376 { 2377 ip_rsvpd = NULL; 2378 /* 2379 * This may seem silly, but we need to be sure we don't over-decrement 2380 * the RSVP counter, in case something slips up. 2381 */ 2382 if (ip_rsvp_on) { 2383 ip_rsvp_on = 0; 2384 rsvp_on--; 2385 } 2386 return 0; 2387 } 2388 2389 int 2390 rsvp_input(struct mbuf **mp, int *offp, int proto) 2391 { 2392 struct mbuf *m = *mp; 2393 2394 *mp = NULL; 2395 2396 if (rsvp_input_p) { /* call the real one if loaded */ 2397 *mp = m; 2398 rsvp_input_p(mp, offp, proto); 2399 return(IPPROTO_DONE); 2400 } 2401 2402 /* Can still get packets with rsvp_on = 0 if there is a local member 2403 * of the group to which the RSVP packet is addressed. But in this 2404 * case we want to throw the packet away. 2405 */ 2406 2407 if (!rsvp_on) { 2408 m_freem(m); 2409 return(IPPROTO_DONE); 2410 } 2411 2412 if (ip_rsvpd != NULL) { 2413 *mp = m; 2414 rip_input(mp, offp, proto); 2415 return(IPPROTO_DONE); 2416 } 2417 /* Drop the packet */ 2418 m_freem(m); 2419 return(IPPROTO_DONE); 2420 } 2421