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