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