1 /* $OpenBSD: ip_input.c,v 1.322 2017/09/07 10:54:49 bluhm Exp $ */ 2 /* $NetBSD: ip_input.c,v 1.30 1996/03/16 23:53:58 christos Exp $ */ 3 4 /* 5 * Copyright (c) 1982, 1986, 1988, 1993 6 * The Regents of the University of California. All rights reserved. 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 University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 33 */ 34 35 #include "pf.h" 36 #include "carp.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/mbuf.h> 41 #include <sys/domain.h> 42 #include <sys/mutex.h> 43 #include <sys/protosw.h> 44 #include <sys/socket.h> 45 #include <sys/socketvar.h> 46 #include <sys/sysctl.h> 47 #include <sys/pool.h> 48 #include <sys/task.h> 49 50 #include <net/if.h> 51 #include <net/if_var.h> 52 #include <net/if_dl.h> 53 #include <net/route.h> 54 #include <net/netisr.h> 55 56 #include <netinet/in.h> 57 #include <netinet/in_systm.h> 58 #include <netinet/if_ether.h> 59 #include <netinet/ip.h> 60 #include <netinet/in_pcb.h> 61 #include <netinet/in_var.h> 62 #include <netinet/ip_var.h> 63 #include <netinet/ip_icmp.h> 64 65 #ifdef INET6 66 #include <netinet6/ip6protosw.h> 67 #include <netinet6/ip6_var.h> 68 #endif 69 70 #if NPF > 0 71 #include <net/pfvar.h> 72 #endif 73 74 #ifdef MROUTING 75 #include <netinet/ip_mroute.h> 76 #endif 77 78 #ifdef IPSEC 79 #include <netinet/ip_ipsp.h> 80 #endif /* IPSEC */ 81 82 #if NCARP > 0 83 #include <net/if_types.h> 84 #include <netinet/ip_carp.h> 85 #endif 86 87 int encdebug = 0; 88 int ipsec_keep_invalid = IPSEC_DEFAULT_EMBRYONIC_SA_TIMEOUT; 89 int ipsec_require_pfs = IPSEC_DEFAULT_PFS; 90 int ipsec_soft_allocations = IPSEC_DEFAULT_SOFT_ALLOCATIONS; 91 int ipsec_exp_allocations = IPSEC_DEFAULT_EXP_ALLOCATIONS; 92 int ipsec_soft_bytes = IPSEC_DEFAULT_SOFT_BYTES; 93 int ipsec_exp_bytes = IPSEC_DEFAULT_EXP_BYTES; 94 int ipsec_soft_timeout = IPSEC_DEFAULT_SOFT_TIMEOUT; 95 int ipsec_exp_timeout = IPSEC_DEFAULT_EXP_TIMEOUT; 96 int ipsec_soft_first_use = IPSEC_DEFAULT_SOFT_FIRST_USE; 97 int ipsec_exp_first_use = IPSEC_DEFAULT_EXP_FIRST_USE; 98 int ipsec_expire_acquire = IPSEC_DEFAULT_EXPIRE_ACQUIRE; 99 char ipsec_def_enc[20]; 100 char ipsec_def_auth[20]; 101 char ipsec_def_comp[20]; 102 103 /* values controllable via sysctl */ 104 int ipforwarding = 0; 105 int ipmforwarding = 0; 106 int ipmultipath = 0; 107 int ipsendredirects = 1; 108 int ip_dosourceroute = 0; 109 int ip_defttl = IPDEFTTL; 110 int ip_mtudisc = 1; 111 u_int ip_mtudisc_timeout = IPMTUDISCTIMEOUT; 112 int ip_directedbcast = 0; 113 114 struct rttimer_queue *ip_mtudisc_timeout_q = NULL; 115 116 /* Protects `ipq' and `ip_frags'. */ 117 struct mutex ipq_mutex = MUTEX_INITIALIZER(IPL_SOFTNET); 118 119 /* IP reassembly queue */ 120 LIST_HEAD(, ipq) ipq; 121 122 /* Keep track of memory used for reassembly */ 123 int ip_maxqueue = 300; 124 int ip_frags = 0; 125 126 int *ipctl_vars[IPCTL_MAXID] = IPCTL_VARS; 127 128 struct niqueue ipintrq = NIQUEUE_INITIALIZER(IPQ_MAXLEN, NETISR_IP); 129 130 struct pool ipqent_pool; 131 struct pool ipq_pool; 132 133 struct cpumem *ipcounters; 134 135 int ip_sysctl_ipstat(void *, size_t *, void *); 136 137 static struct mbuf_queue ipsend_mq; 138 139 int ip_ours(struct mbuf **, int *, int, int); 140 int ip_local(struct mbuf **, int *, int, int); 141 int ip_dooptions(struct mbuf *, struct ifnet *); 142 int in_ouraddr(struct mbuf *, struct ifnet *, struct rtentry **); 143 144 static void ip_send_dispatch(void *); 145 static struct task ipsend_task = TASK_INITIALIZER(ip_send_dispatch, &ipsend_mq); 146 /* 147 * Used to save the IP options in case a protocol wants to respond 148 * to an incoming packet over the same route if the packet got here 149 * using IP source routing. This allows connection establishment and 150 * maintenance when the remote end is on a network that is not known 151 * to us. 152 */ 153 struct ip_srcrt { 154 int isr_nhops; /* number of hops */ 155 struct in_addr isr_dst; /* final destination */ 156 char isr_nop; /* one NOP to align */ 157 char isr_hdr[IPOPT_OFFSET + 1]; /* OPTVAL, OLEN & OFFSET */ 158 struct in_addr isr_routes[MAX_IPOPTLEN/sizeof(struct in_addr)]; 159 }; 160 161 void save_rte(struct mbuf *, u_char *, struct in_addr); 162 163 /* 164 * IP initialization: fill in IP protocol switch table. 165 * All protocols not implemented in kernel go to raw IP protocol handler. 166 */ 167 void 168 ip_init(void) 169 { 170 struct protosw *pr; 171 int i; 172 const u_int16_t defbaddynamicports_tcp[] = DEFBADDYNAMICPORTS_TCP; 173 const u_int16_t defbaddynamicports_udp[] = DEFBADDYNAMICPORTS_UDP; 174 const u_int16_t defrootonlyports_tcp[] = DEFROOTONLYPORTS_TCP; 175 const u_int16_t defrootonlyports_udp[] = DEFROOTONLYPORTS_UDP; 176 177 ipcounters = counters_alloc(ips_ncounters); 178 179 pool_init(&ipqent_pool, sizeof(struct ipqent), 0, 180 IPL_SOFTNET, 0, "ipqe", NULL); 181 pool_init(&ipq_pool, sizeof(struct ipq), 0, 182 IPL_SOFTNET, 0, "ipq", NULL); 183 184 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 185 if (pr == NULL) 186 panic("ip_init"); 187 for (i = 0; i < IPPROTO_MAX; i++) 188 ip_protox[i] = pr - inetsw; 189 for (pr = inetdomain.dom_protosw; 190 pr < inetdomain.dom_protoswNPROTOSW; pr++) 191 if (pr->pr_domain->dom_family == PF_INET && 192 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW && 193 pr->pr_protocol < IPPROTO_MAX) 194 ip_protox[pr->pr_protocol] = pr - inetsw; 195 LIST_INIT(&ipq); 196 if (ip_mtudisc != 0) 197 ip_mtudisc_timeout_q = 198 rt_timer_queue_create(ip_mtudisc_timeout); 199 200 /* Fill in list of ports not to allocate dynamically. */ 201 memset(&baddynamicports, 0, sizeof(baddynamicports)); 202 for (i = 0; defbaddynamicports_tcp[i] != 0; i++) 203 DP_SET(baddynamicports.tcp, defbaddynamicports_tcp[i]); 204 for (i = 0; defbaddynamicports_udp[i] != 0; i++) 205 DP_SET(baddynamicports.udp, defbaddynamicports_udp[i]); 206 207 /* Fill in list of ports only root can bind to. */ 208 memset(&rootonlyports, 0, sizeof(rootonlyports)); 209 for (i = 0; defrootonlyports_tcp[i] != 0; i++) 210 DP_SET(rootonlyports.tcp, defrootonlyports_tcp[i]); 211 for (i = 0; defrootonlyports_udp[i] != 0; i++) 212 DP_SET(rootonlyports.udp, defrootonlyports_udp[i]); 213 214 strlcpy(ipsec_def_enc, IPSEC_DEFAULT_DEF_ENC, sizeof(ipsec_def_enc)); 215 strlcpy(ipsec_def_auth, IPSEC_DEFAULT_DEF_AUTH, sizeof(ipsec_def_auth)); 216 strlcpy(ipsec_def_comp, IPSEC_DEFAULT_DEF_COMP, sizeof(ipsec_def_comp)); 217 218 mq_init(&ipsend_mq, 64, IPL_SOFTNET); 219 } 220 221 /* 222 * Enqueue packet for local delivery. Queuing is used as a boundary 223 * between the network layer (input/forward path) running without 224 * KERNEL_LOCK() and the transport layer still needing it. 225 */ 226 int 227 ip_ours(struct mbuf **mp, int *offp, int nxt, int af) 228 { 229 /* We are already in a IPv4/IPv6 local deliver loop. */ 230 if (af != AF_UNSPEC) 231 return ip_local(mp, offp, nxt, af); 232 233 niq_enqueue(&ipintrq, *mp); 234 *mp = NULL; 235 return IPPROTO_DONE; 236 } 237 238 /* 239 * Dequeue and process locally delivered packets. 240 */ 241 void 242 ipintr(void) 243 { 244 struct mbuf *m; 245 int off, nxt; 246 247 while ((m = niq_dequeue(&ipintrq)) != NULL) { 248 #ifdef DIAGNOSTIC 249 if ((m->m_flags & M_PKTHDR) == 0) 250 panic("ipintr no HDR"); 251 #endif 252 off = 0; 253 nxt = ip_local(&m, &off, IPPROTO_IPV4, AF_UNSPEC); 254 KASSERT(nxt == IPPROTO_DONE); 255 } 256 } 257 258 /* 259 * IPv4 input routine. 260 * 261 * Checksum and byte swap header. Process options. Forward or deliver. 262 */ 263 void 264 ipv4_input(struct ifnet *ifp, struct mbuf *m) 265 { 266 int off, nxt; 267 268 off = 0; 269 nxt = ip_input_if(&m, &off, IPPROTO_IPV4, AF_UNSPEC, ifp); 270 KASSERT(nxt == IPPROTO_DONE); 271 } 272 273 int 274 ip_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp) 275 { 276 struct mbuf *m = *mp; 277 struct rtentry *rt = NULL; 278 struct ip *ip; 279 int hlen, len; 280 in_addr_t pfrdr = 0; 281 282 KASSERT(*offp == 0); 283 284 ipstat_inc(ips_total); 285 if (m->m_len < sizeof (struct ip) && 286 (m = *mp = m_pullup(m, sizeof (struct ip))) == NULL) { 287 ipstat_inc(ips_toosmall); 288 goto bad; 289 } 290 ip = mtod(m, struct ip *); 291 if (ip->ip_v != IPVERSION) { 292 ipstat_inc(ips_badvers); 293 goto bad; 294 } 295 hlen = ip->ip_hl << 2; 296 if (hlen < sizeof(struct ip)) { /* minimum header length */ 297 ipstat_inc(ips_badhlen); 298 goto bad; 299 } 300 if (hlen > m->m_len) { 301 if ((m = *mp = m_pullup(m, hlen)) == NULL) { 302 ipstat_inc(ips_badhlen); 303 goto bad; 304 } 305 ip = mtod(m, struct ip *); 306 } 307 308 /* 127/8 must not appear on wire - RFC1122 */ 309 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 310 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 311 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 312 ipstat_inc(ips_badaddr); 313 goto bad; 314 } 315 } 316 317 if ((m->m_pkthdr.csum_flags & M_IPV4_CSUM_IN_OK) == 0) { 318 if (m->m_pkthdr.csum_flags & M_IPV4_CSUM_IN_BAD) { 319 ipstat_inc(ips_badsum); 320 goto bad; 321 } 322 323 ipstat_inc(ips_inswcsum); 324 if (in_cksum(m, hlen) != 0) { 325 ipstat_inc(ips_badsum); 326 goto bad; 327 } 328 } 329 330 /* Retrieve the packet length. */ 331 len = ntohs(ip->ip_len); 332 333 /* 334 * Convert fields to host representation. 335 */ 336 if (len < hlen) { 337 ipstat_inc(ips_badlen); 338 goto bad; 339 } 340 341 /* 342 * Check that the amount of data in the buffers 343 * is at least as much as the IP header would have us expect. 344 * Trim mbufs if longer than we expect. 345 * Drop packet if shorter than we expect. 346 */ 347 if (m->m_pkthdr.len < len) { 348 ipstat_inc(ips_tooshort); 349 goto bad; 350 } 351 if (m->m_pkthdr.len > len) { 352 if (m->m_len == m->m_pkthdr.len) { 353 m->m_len = len; 354 m->m_pkthdr.len = len; 355 } else 356 m_adj(m, len - m->m_pkthdr.len); 357 } 358 359 #if NCARP > 0 360 if (ifp->if_type == IFT_CARP && 361 carp_lsdrop(m, AF_INET, &ip->ip_src.s_addr, &ip->ip_dst.s_addr, 362 (ip->ip_p == IPPROTO_ICMP ? 0 : 1))) 363 goto bad; 364 #endif 365 366 #if NPF > 0 367 /* 368 * Packet filter 369 */ 370 pfrdr = ip->ip_dst.s_addr; 371 if (pf_test(AF_INET, PF_IN, ifp, mp) != PF_PASS) 372 goto bad; 373 m = *mp; 374 if (m == NULL) 375 goto bad; 376 377 ip = mtod(m, struct ip *); 378 hlen = ip->ip_hl << 2; 379 pfrdr = (pfrdr != ip->ip_dst.s_addr); 380 #endif 381 382 /* 383 * Process options and, if not destined for us, 384 * ship it on. ip_dooptions returns 1 when an 385 * error was detected (causing an icmp message 386 * to be sent and the original packet to be freed). 387 */ 388 if (hlen > sizeof (struct ip) && ip_dooptions(m, ifp)) { 389 m = *mp = NULL; 390 goto bad; 391 } 392 393 if (ip->ip_dst.s_addr == INADDR_BROADCAST || 394 ip->ip_dst.s_addr == INADDR_ANY) { 395 nxt = ip_ours(mp, offp, nxt, af); 396 goto out; 397 } 398 399 if (in_ouraddr(m, ifp, &rt)) { 400 nxt = ip_ours(mp, offp, nxt, af); 401 goto out; 402 } 403 404 if (IN_MULTICAST(ip->ip_dst.s_addr)) { 405 /* 406 * Make sure M_MCAST is set. It should theoretically 407 * already be there, but let's play safe because upper 408 * layers check for this flag. 409 */ 410 m->m_flags |= M_MCAST; 411 412 #ifdef MROUTING 413 if (ipmforwarding && ip_mrouter[ifp->if_rdomain]) { 414 int error; 415 416 if (m->m_flags & M_EXT) { 417 if ((m = *mp = m_pullup(m, hlen)) == NULL) { 418 ipstat_inc(ips_toosmall); 419 goto bad; 420 } 421 ip = mtod(m, struct ip *); 422 } 423 /* 424 * If we are acting as a multicast router, all 425 * incoming multicast packets are passed to the 426 * kernel-level multicast forwarding function. 427 * The packet is returned (relatively) intact; if 428 * ip_mforward() returns a non-zero value, the packet 429 * must be discarded, else it may be accepted below. 430 * 431 * (The IP ident field is put in the same byte order 432 * as expected when ip_mforward() is called from 433 * ip_output().) 434 */ 435 KERNEL_LOCK(); 436 error = ip_mforward(m, ifp); 437 KERNEL_UNLOCK(); 438 if (error) { 439 ipstat_inc(ips_cantforward); 440 goto bad; 441 } 442 443 /* 444 * The process-level routing daemon needs to receive 445 * all multicast IGMP packets, whether or not this 446 * host belongs to their destination groups. 447 */ 448 if (ip->ip_p == IPPROTO_IGMP) { 449 nxt = ip_ours(mp, offp, nxt, af); 450 goto out; 451 } 452 ipstat_inc(ips_forward); 453 } 454 #endif 455 /* 456 * See if we belong to the destination multicast group on the 457 * arrival interface. 458 */ 459 if (!in_hasmulti(&ip->ip_dst, ifp)) { 460 ipstat_inc(ips_notmember); 461 if (!IN_LOCAL_GROUP(ip->ip_dst.s_addr)) 462 ipstat_inc(ips_cantforward); 463 goto bad; 464 } 465 nxt = ip_ours(mp, offp, nxt, af); 466 goto out; 467 } 468 469 #if NCARP > 0 470 if (ifp->if_type == IFT_CARP && ip->ip_p == IPPROTO_ICMP && 471 carp_lsdrop(m, AF_INET, &ip->ip_src.s_addr, &ip->ip_dst.s_addr, 1)) 472 goto bad; 473 #endif 474 /* 475 * Not for us; forward if possible and desirable. 476 */ 477 if (ipforwarding == 0) { 478 ipstat_inc(ips_cantforward); 479 goto bad; 480 } 481 #ifdef IPSEC 482 if (ipsec_in_use) { 483 int rv; 484 485 KERNEL_ASSERT_LOCKED(); 486 487 rv = ipsec_forward_check(m, hlen, AF_INET); 488 if (rv != 0) { 489 ipstat_inc(ips_cantforward); 490 goto bad; 491 } 492 /* 493 * Fall through, forward packet. Outbound IPsec policy 494 * checking will occur in ip_output(). 495 */ 496 } 497 #endif /* IPSEC */ 498 499 ip_forward(m, ifp, rt, pfrdr); 500 *mp = NULL; 501 return IPPROTO_DONE; 502 bad: 503 nxt = IPPROTO_DONE; 504 m_freemp(mp); 505 out: 506 rtfree(rt); 507 return nxt; 508 } 509 510 /* 511 * IPv4 local-delivery routine. 512 * 513 * If fragmented try to reassemble. Pass to next level. 514 */ 515 int 516 ip_local(struct mbuf **mp, int *offp, int nxt, int af) 517 { 518 struct mbuf *m = *mp; 519 struct ip *ip = mtod(m, struct ip *); 520 struct ipq *fp; 521 struct ipqent *ipqe; 522 int mff, hlen; 523 524 hlen = ip->ip_hl << 2; 525 526 /* 527 * If offset or IP_MF are set, must reassemble. 528 * Otherwise, nothing need be done. 529 * (We could look in the reassembly queue to see 530 * if the packet was previously fragmented, 531 * but it's not worth the time; just let them time out.) 532 */ 533 if (ip->ip_off &~ htons(IP_DF | IP_RF)) { 534 if (m->m_flags & M_EXT) { /* XXX */ 535 if ((m = *mp = m_pullup(m, hlen)) == NULL) { 536 ipstat_inc(ips_toosmall); 537 return IPPROTO_DONE; 538 } 539 ip = mtod(m, struct ip *); 540 } 541 542 mtx_enter(&ipq_mutex); 543 544 /* 545 * Look for queue of fragments 546 * of this datagram. 547 */ 548 LIST_FOREACH(fp, &ipq, ipq_q) { 549 if (ip->ip_id == fp->ipq_id && 550 ip->ip_src.s_addr == fp->ipq_src.s_addr && 551 ip->ip_dst.s_addr == fp->ipq_dst.s_addr && 552 ip->ip_p == fp->ipq_p) 553 break; 554 } 555 556 /* 557 * Adjust ip_len to not reflect header, 558 * set ipqe_mff if more fragments are expected, 559 * convert offset of this to bytes. 560 */ 561 ip->ip_len = htons(ntohs(ip->ip_len) - hlen); 562 mff = (ip->ip_off & htons(IP_MF)) != 0; 563 if (mff) { 564 /* 565 * Make sure that fragments have a data length 566 * that's a non-zero multiple of 8 bytes. 567 */ 568 if (ntohs(ip->ip_len) == 0 || 569 (ntohs(ip->ip_len) & 0x7) != 0) { 570 ipstat_inc(ips_badfrags); 571 goto bad; 572 } 573 } 574 ip->ip_off = htons(ntohs(ip->ip_off) << 3); 575 576 /* 577 * If datagram marked as having more fragments 578 * or if this is not the first fragment, 579 * attempt reassembly; if it succeeds, proceed. 580 */ 581 if (mff || ip->ip_off) { 582 ipstat_inc(ips_fragments); 583 if (ip_frags + 1 > ip_maxqueue) { 584 ip_flush(); 585 ipstat_inc(ips_rcvmemdrop); 586 goto bad; 587 } 588 589 ipqe = pool_get(&ipqent_pool, PR_NOWAIT); 590 if (ipqe == NULL) { 591 ipstat_inc(ips_rcvmemdrop); 592 goto bad; 593 } 594 ip_frags++; 595 ipqe->ipqe_mff = mff; 596 ipqe->ipqe_m = m; 597 ipqe->ipqe_ip = ip; 598 m = *mp = ip_reass(ipqe, fp); 599 if (m == NULL) 600 goto bad; 601 ipstat_inc(ips_reassembled); 602 ip = mtod(m, struct ip *); 603 hlen = ip->ip_hl << 2; 604 ip->ip_len = htons(ntohs(ip->ip_len) + hlen); 605 } else 606 if (fp) 607 ip_freef(fp); 608 609 mtx_leave(&ipq_mutex); 610 } 611 612 *offp = hlen; 613 nxt = ip->ip_p; 614 /* Check wheter we are already in a IPv4/IPv6 local deliver loop. */ 615 if (af == AF_UNSPEC) 616 nxt = ip_deliver(mp, offp, nxt, AF_INET); 617 return nxt; 618 bad: 619 mtx_leave(&ipq_mutex); 620 m_freemp(mp); 621 return IPPROTO_DONE; 622 } 623 624 #ifndef INET6 625 #define IPSTAT_INC(name) ipstat_inc(ips_##name) 626 #else 627 #define IPSTAT_INC(name) (af == AF_INET ? \ 628 ipstat_inc(ips_##name) : ip6stat_inc(ip6s_##name)) 629 #endif 630 631 int 632 ip_deliver(struct mbuf **mp, int *offp, int nxt, int af) 633 { 634 struct protosw *psw; 635 int naf = af; 636 #ifdef INET6 637 int nest = 0; 638 #endif /* INET6 */ 639 640 KERNEL_ASSERT_LOCKED(); 641 642 /* pf might have modified stuff, might have to chksum */ 643 switch (af) { 644 case AF_INET: 645 in_proto_cksum_out(*mp, NULL); 646 break; 647 #ifdef INET6 648 case AF_INET6: 649 in6_proto_cksum_out(*mp, NULL); 650 break; 651 #endif /* INET6 */ 652 } 653 654 /* 655 * Tell launch routine the next header 656 */ 657 IPSTAT_INC(delivered); 658 659 while (nxt != IPPROTO_DONE) { 660 #ifdef INET6 661 if (af == AF_INET6 && 662 ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) { 663 ip6stat_inc(ip6s_toomanyhdr); 664 goto bad; 665 } 666 #endif /* INET6 */ 667 668 /* 669 * protection against faulty packet - there should be 670 * more sanity checks in header chain processing. 671 */ 672 if ((*mp)->m_pkthdr.len < *offp) { 673 IPSTAT_INC(tooshort); 674 goto bad; 675 } 676 677 #ifdef INET6 678 /* draft-itojun-ipv6-tcp-to-anycast */ 679 if (af == AF_INET6 && 680 ISSET((*mp)->m_flags, M_ACAST) && (nxt == IPPROTO_TCP)) { 681 if ((*mp)->m_len >= sizeof(struct ip6_hdr)) { 682 icmp6_error(*mp, ICMP6_DST_UNREACH, 683 ICMP6_DST_UNREACH_ADDR, 684 offsetof(struct ip6_hdr, ip6_dst)); 685 *mp = NULL; 686 } 687 goto bad; 688 } 689 #endif /* INET6 */ 690 691 #ifdef IPSEC 692 if (ipsec_in_use) { 693 if (ipsec_local_check(*mp, *offp, nxt, af) != 0) { 694 IPSTAT_INC(cantforward); 695 goto bad; 696 } 697 } 698 /* Otherwise, just fall through and deliver the packet */ 699 #endif /* IPSEC */ 700 701 switch (nxt) { 702 case IPPROTO_IPV4: 703 naf = AF_INET; 704 ipstat_inc(ips_delivered); 705 break; 706 #ifdef INET6 707 case IPPROTO_IPV6: 708 naf = AF_INET6; 709 ip6stat_inc(ip6s_delivered); 710 break; 711 #endif /* INET6 */ 712 } 713 switch (af) { 714 case AF_INET: 715 psw = &inetsw[ip_protox[nxt]]; 716 break; 717 #ifdef INET6 718 case AF_INET6: 719 psw = &inet6sw[ip6_protox[nxt]]; 720 break; 721 #endif /* INET6 */ 722 } 723 nxt = (*psw->pr_input)(mp, offp, nxt, af); 724 af = naf; 725 } 726 return nxt; 727 bad: 728 m_freemp(mp); 729 return IPPROTO_DONE; 730 } 731 #undef IPSTAT_INC 732 733 int 734 in_ouraddr(struct mbuf *m, struct ifnet *ifp, struct rtentry **prt) 735 { 736 struct rtentry *rt; 737 struct ip *ip; 738 struct sockaddr_in sin; 739 int match = 0; 740 741 #if NPF > 0 742 switch (pf_ouraddr(m)) { 743 case 0: 744 return (0); 745 case 1: 746 return (1); 747 default: 748 /* pf does not know it */ 749 break; 750 } 751 #endif 752 753 ip = mtod(m, struct ip *); 754 755 memset(&sin, 0, sizeof(sin)); 756 sin.sin_len = sizeof(sin); 757 sin.sin_family = AF_INET; 758 sin.sin_addr = ip->ip_dst; 759 rt = rtalloc_mpath(sintosa(&sin), &ip->ip_src.s_addr, 760 m->m_pkthdr.ph_rtableid); 761 if (rtisvalid(rt)) { 762 if (ISSET(rt->rt_flags, RTF_LOCAL)) 763 match = 1; 764 765 /* 766 * If directedbcast is enabled we only consider it local 767 * if it is received on the interface with that address. 768 */ 769 if (ISSET(rt->rt_flags, RTF_BROADCAST) && 770 (!ip_directedbcast || rt->rt_ifidx == ifp->if_index)) { 771 match = 1; 772 773 /* Make sure M_BCAST is set */ 774 m->m_flags |= M_BCAST; 775 } 776 } 777 *prt = rt; 778 779 if (!match) { 780 struct ifaddr *ifa; 781 782 /* 783 * No local address or broadcast address found, so check for 784 * ancient classful broadcast addresses. 785 * It must have been broadcast on the link layer, and for an 786 * address on the interface it was received on. 787 */ 788 if (!ISSET(m->m_flags, M_BCAST) || 789 !IN_CLASSFULBROADCAST(ip->ip_dst.s_addr, ip->ip_dst.s_addr)) 790 return (0); 791 792 if (ifp->if_rdomain != rtable_l2(m->m_pkthdr.ph_rtableid)) 793 return (0); 794 /* 795 * The check in the loop assumes you only rx a packet on an UP 796 * interface, and that M_BCAST will only be set on a BROADCAST 797 * interface. 798 */ 799 NET_ASSERT_LOCKED(); 800 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) { 801 if (ifa->ifa_addr->sa_family != AF_INET) 802 continue; 803 804 if (IN_CLASSFULBROADCAST(ip->ip_dst.s_addr, 805 ifatoia(ifa)->ia_addr.sin_addr.s_addr)) { 806 match = 1; 807 break; 808 } 809 } 810 } 811 812 return (match); 813 } 814 815 /* 816 * Take incoming datagram fragment and try to 817 * reassemble it into whole datagram. If a chain for 818 * reassembly of this datagram already exists, then it 819 * is given as fp; otherwise have to make a chain. 820 */ 821 struct mbuf * 822 ip_reass(struct ipqent *ipqe, struct ipq *fp) 823 { 824 struct mbuf *m = ipqe->ipqe_m; 825 struct ipqent *nq, *p, *q; 826 struct ip *ip; 827 struct mbuf *t; 828 int hlen = ipqe->ipqe_ip->ip_hl << 2; 829 int i, next; 830 u_int8_t ecn, ecn0; 831 832 MUTEX_ASSERT_LOCKED(&ipq_mutex); 833 834 /* 835 * Presence of header sizes in mbufs 836 * would confuse code below. 837 */ 838 m->m_data += hlen; 839 m->m_len -= hlen; 840 841 /* 842 * If first fragment to arrive, create a reassembly queue. 843 */ 844 if (fp == NULL) { 845 fp = pool_get(&ipq_pool, PR_NOWAIT); 846 if (fp == NULL) 847 goto dropfrag; 848 LIST_INSERT_HEAD(&ipq, fp, ipq_q); 849 fp->ipq_ttl = IPFRAGTTL; 850 fp->ipq_p = ipqe->ipqe_ip->ip_p; 851 fp->ipq_id = ipqe->ipqe_ip->ip_id; 852 LIST_INIT(&fp->ipq_fragq); 853 fp->ipq_src = ipqe->ipqe_ip->ip_src; 854 fp->ipq_dst = ipqe->ipqe_ip->ip_dst; 855 p = NULL; 856 goto insert; 857 } 858 859 /* 860 * Handle ECN by comparing this segment with the first one; 861 * if CE is set, do not lose CE. 862 * drop if CE and not-ECT are mixed for the same packet. 863 */ 864 ecn = ipqe->ipqe_ip->ip_tos & IPTOS_ECN_MASK; 865 ecn0 = LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos & IPTOS_ECN_MASK; 866 if (ecn == IPTOS_ECN_CE) { 867 if (ecn0 == IPTOS_ECN_NOTECT) 868 goto dropfrag; 869 if (ecn0 != IPTOS_ECN_CE) 870 LIST_FIRST(&fp->ipq_fragq)->ipqe_ip->ip_tos |= 871 IPTOS_ECN_CE; 872 } 873 if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT) 874 goto dropfrag; 875 876 /* 877 * Find a segment which begins after this one does. 878 */ 879 for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL; 880 p = q, q = LIST_NEXT(q, ipqe_q)) 881 if (ntohs(q->ipqe_ip->ip_off) > ntohs(ipqe->ipqe_ip->ip_off)) 882 break; 883 884 /* 885 * If there is a preceding segment, it may provide some of 886 * our data already. If so, drop the data from the incoming 887 * segment. If it provides all of our data, drop us. 888 */ 889 if (p != NULL) { 890 i = ntohs(p->ipqe_ip->ip_off) + ntohs(p->ipqe_ip->ip_len) - 891 ntohs(ipqe->ipqe_ip->ip_off); 892 if (i > 0) { 893 if (i >= ntohs(ipqe->ipqe_ip->ip_len)) 894 goto dropfrag; 895 m_adj(ipqe->ipqe_m, i); 896 ipqe->ipqe_ip->ip_off = 897 htons(ntohs(ipqe->ipqe_ip->ip_off) + i); 898 ipqe->ipqe_ip->ip_len = 899 htons(ntohs(ipqe->ipqe_ip->ip_len) - i); 900 } 901 } 902 903 /* 904 * While we overlap succeeding segments trim them or, 905 * if they are completely covered, dequeue them. 906 */ 907 for (; q != NULL && 908 ntohs(ipqe->ipqe_ip->ip_off) + ntohs(ipqe->ipqe_ip->ip_len) > 909 ntohs(q->ipqe_ip->ip_off); q = nq) { 910 i = (ntohs(ipqe->ipqe_ip->ip_off) + 911 ntohs(ipqe->ipqe_ip->ip_len)) - ntohs(q->ipqe_ip->ip_off); 912 if (i < ntohs(q->ipqe_ip->ip_len)) { 913 q->ipqe_ip->ip_len = 914 htons(ntohs(q->ipqe_ip->ip_len) - i); 915 q->ipqe_ip->ip_off = 916 htons(ntohs(q->ipqe_ip->ip_off) + i); 917 m_adj(q->ipqe_m, i); 918 break; 919 } 920 nq = LIST_NEXT(q, ipqe_q); 921 m_freem(q->ipqe_m); 922 LIST_REMOVE(q, ipqe_q); 923 pool_put(&ipqent_pool, q); 924 ip_frags--; 925 } 926 927 insert: 928 /* 929 * Stick new segment in its place; 930 * check for complete reassembly. 931 */ 932 if (p == NULL) { 933 LIST_INSERT_HEAD(&fp->ipq_fragq, ipqe, ipqe_q); 934 } else { 935 LIST_INSERT_AFTER(p, ipqe, ipqe_q); 936 } 937 next = 0; 938 for (p = NULL, q = LIST_FIRST(&fp->ipq_fragq); q != NULL; 939 p = q, q = LIST_NEXT(q, ipqe_q)) { 940 if (ntohs(q->ipqe_ip->ip_off) != next) 941 return (0); 942 next += ntohs(q->ipqe_ip->ip_len); 943 } 944 if (p->ipqe_mff) 945 return (0); 946 947 /* 948 * Reassembly is complete. Check for a bogus message size and 949 * concatenate fragments. 950 */ 951 q = LIST_FIRST(&fp->ipq_fragq); 952 ip = q->ipqe_ip; 953 if ((next + (ip->ip_hl << 2)) > IP_MAXPACKET) { 954 ipstat_inc(ips_toolong); 955 ip_freef(fp); 956 return (0); 957 } 958 m = q->ipqe_m; 959 t = m->m_next; 960 m->m_next = 0; 961 m_cat(m, t); 962 nq = LIST_NEXT(q, ipqe_q); 963 pool_put(&ipqent_pool, q); 964 ip_frags--; 965 for (q = nq; q != NULL; q = nq) { 966 t = q->ipqe_m; 967 nq = LIST_NEXT(q, ipqe_q); 968 pool_put(&ipqent_pool, q); 969 ip_frags--; 970 m_cat(m, t); 971 } 972 973 /* 974 * Create header for new ip packet by 975 * modifying header of first packet; 976 * dequeue and discard fragment reassembly header. 977 * Make header visible. 978 */ 979 ip->ip_len = htons(next); 980 ip->ip_src = fp->ipq_src; 981 ip->ip_dst = fp->ipq_dst; 982 LIST_REMOVE(fp, ipq_q); 983 pool_put(&ipq_pool, fp); 984 m->m_len += (ip->ip_hl << 2); 985 m->m_data -= (ip->ip_hl << 2); 986 /* some debugging cruft by sklower, below, will go away soon */ 987 if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */ 988 int plen = 0; 989 for (t = m; t; t = t->m_next) 990 plen += t->m_len; 991 m->m_pkthdr.len = plen; 992 } 993 return (m); 994 995 dropfrag: 996 ipstat_inc(ips_fragdropped); 997 m_freem(m); 998 pool_put(&ipqent_pool, ipqe); 999 ip_frags--; 1000 return (NULL); 1001 } 1002 1003 /* 1004 * Free a fragment reassembly header and all 1005 * associated datagrams. 1006 */ 1007 void 1008 ip_freef(struct ipq *fp) 1009 { 1010 struct ipqent *q; 1011 1012 MUTEX_ASSERT_LOCKED(&ipq_mutex); 1013 1014 while ((q = LIST_FIRST(&fp->ipq_fragq)) != NULL) { 1015 LIST_REMOVE(q, ipqe_q); 1016 m_freem(q->ipqe_m); 1017 pool_put(&ipqent_pool, q); 1018 ip_frags--; 1019 } 1020 LIST_REMOVE(fp, ipq_q); 1021 pool_put(&ipq_pool, fp); 1022 } 1023 1024 /* 1025 * IP timer processing; 1026 * if a timer expires on a reassembly queue, discard it. 1027 */ 1028 void 1029 ip_slowtimo(void) 1030 { 1031 struct ipq *fp, *nfp; 1032 1033 mtx_enter(&ipq_mutex); 1034 LIST_FOREACH_SAFE(fp, &ipq, ipq_q, nfp) { 1035 if (--fp->ipq_ttl == 0) { 1036 ipstat_inc(ips_fragtimeout); 1037 ip_freef(fp); 1038 } 1039 } 1040 mtx_leave(&ipq_mutex); 1041 } 1042 1043 /* 1044 * Drain off all datagram fragments. 1045 */ 1046 void 1047 ip_drain(void) 1048 { 1049 mtx_enter(&ipq_mutex); 1050 while (!LIST_EMPTY(&ipq)) { 1051 ipstat_inc(ips_fragdropped); 1052 ip_freef(LIST_FIRST(&ipq)); 1053 } 1054 mtx_leave(&ipq_mutex); 1055 } 1056 1057 /* 1058 * Flush a bunch of datagram fragments, till we are down to 75%. 1059 */ 1060 void 1061 ip_flush(void) 1062 { 1063 int max = 50; 1064 1065 MUTEX_ASSERT_LOCKED(&ipq_mutex); 1066 1067 while (!LIST_EMPTY(&ipq) && ip_frags > ip_maxqueue * 3 / 4 && --max) { 1068 ipstat_inc(ips_fragdropped); 1069 ip_freef(LIST_FIRST(&ipq)); 1070 } 1071 } 1072 1073 /* 1074 * Do option processing on a datagram, 1075 * possibly discarding it if bad options are encountered, 1076 * or forwarding it if source-routed. 1077 * Returns 1 if packet has been forwarded/freed, 1078 * 0 if the packet should be processed further. 1079 */ 1080 int 1081 ip_dooptions(struct mbuf *m, struct ifnet *ifp) 1082 { 1083 struct ip *ip = mtod(m, struct ip *); 1084 unsigned int rtableid = m->m_pkthdr.ph_rtableid; 1085 struct rtentry *rt; 1086 struct sockaddr_in ipaddr; 1087 u_char *cp; 1088 struct ip_timestamp ipt; 1089 struct in_ifaddr *ia; 1090 int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0; 1091 struct in_addr sin, dst; 1092 u_int32_t ntime; 1093 1094 dst = ip->ip_dst; 1095 cp = (u_char *)(ip + 1); 1096 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1097 1098 KERNEL_LOCK(); 1099 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1100 opt = cp[IPOPT_OPTVAL]; 1101 if (opt == IPOPT_EOL) 1102 break; 1103 if (opt == IPOPT_NOP) 1104 optlen = 1; 1105 else { 1106 if (cnt < IPOPT_OLEN + sizeof(*cp)) { 1107 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1108 goto bad; 1109 } 1110 optlen = cp[IPOPT_OLEN]; 1111 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) { 1112 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1113 goto bad; 1114 } 1115 } 1116 1117 switch (opt) { 1118 1119 default: 1120 break; 1121 1122 /* 1123 * Source routing with record. 1124 * Find interface with current destination address. 1125 * If none on this machine then drop if strictly routed, 1126 * or do nothing if loosely routed. 1127 * Record interface address and bring up next address 1128 * component. If strictly routed make sure next 1129 * address is on directly accessible net. 1130 */ 1131 case IPOPT_LSRR: 1132 case IPOPT_SSRR: 1133 if (!ip_dosourceroute) { 1134 type = ICMP_UNREACH; 1135 code = ICMP_UNREACH_SRCFAIL; 1136 goto bad; 1137 } 1138 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1139 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1140 goto bad; 1141 } 1142 memset(&ipaddr, 0, sizeof(ipaddr)); 1143 ipaddr.sin_family = AF_INET; 1144 ipaddr.sin_len = sizeof(ipaddr); 1145 ipaddr.sin_addr = ip->ip_dst; 1146 ia = ifatoia(ifa_ifwithaddr(sintosa(&ipaddr), 1147 m->m_pkthdr.ph_rtableid)); 1148 if (ia == NULL) { 1149 if (opt == IPOPT_SSRR) { 1150 type = ICMP_UNREACH; 1151 code = ICMP_UNREACH_SRCFAIL; 1152 goto bad; 1153 } 1154 /* 1155 * Loose routing, and not at next destination 1156 * yet; nothing to do except forward. 1157 */ 1158 break; 1159 } 1160 off--; /* 0 origin */ 1161 if ((off + sizeof(struct in_addr)) > optlen) { 1162 /* 1163 * End of source route. Should be for us. 1164 */ 1165 save_rte(m, cp, ip->ip_src); 1166 break; 1167 } 1168 1169 /* 1170 * locate outgoing interface 1171 */ 1172 memset(&ipaddr, 0, sizeof(ipaddr)); 1173 ipaddr.sin_family = AF_INET; 1174 ipaddr.sin_len = sizeof(ipaddr); 1175 memcpy(&ipaddr.sin_addr, cp + off, 1176 sizeof(ipaddr.sin_addr)); 1177 /* keep packet in the virtual instance */ 1178 rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid); 1179 if (!rtisvalid(rt) || ((opt == IPOPT_SSRR) && 1180 ISSET(rt->rt_flags, RTF_GATEWAY))) { 1181 type = ICMP_UNREACH; 1182 code = ICMP_UNREACH_SRCFAIL; 1183 rtfree(rt); 1184 goto bad; 1185 } 1186 ia = ifatoia(rt->rt_ifa); 1187 memcpy(cp + off, &ia->ia_addr.sin_addr, 1188 sizeof(struct in_addr)); 1189 rtfree(rt); 1190 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1191 ip->ip_dst = ipaddr.sin_addr; 1192 /* 1193 * Let ip_intr's mcast routing check handle mcast pkts 1194 */ 1195 forward = !IN_MULTICAST(ip->ip_dst.s_addr); 1196 break; 1197 1198 case IPOPT_RR: 1199 if (optlen < IPOPT_OFFSET + sizeof(*cp)) { 1200 code = &cp[IPOPT_OLEN] - (u_char *)ip; 1201 goto bad; 1202 } 1203 if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) { 1204 code = &cp[IPOPT_OFFSET] - (u_char *)ip; 1205 goto bad; 1206 } 1207 1208 /* 1209 * If no space remains, ignore. 1210 */ 1211 off--; /* 0 origin */ 1212 if ((off + sizeof(struct in_addr)) > optlen) 1213 break; 1214 memset(&ipaddr, 0, sizeof(ipaddr)); 1215 ipaddr.sin_family = AF_INET; 1216 ipaddr.sin_len = sizeof(ipaddr); 1217 ipaddr.sin_addr = ip->ip_dst; 1218 /* 1219 * locate outgoing interface; if we're the destination, 1220 * use the incoming interface (should be same). 1221 * Again keep the packet inside the virtual instance. 1222 */ 1223 rt = rtalloc(sintosa(&ipaddr), RT_RESOLVE, rtableid); 1224 if (!rtisvalid(rt)) { 1225 type = ICMP_UNREACH; 1226 code = ICMP_UNREACH_HOST; 1227 rtfree(rt); 1228 goto bad; 1229 } 1230 ia = ifatoia(rt->rt_ifa); 1231 memcpy(cp + off, &ia->ia_addr.sin_addr, 1232 sizeof(struct in_addr)); 1233 rtfree(rt); 1234 cp[IPOPT_OFFSET] += sizeof(struct in_addr); 1235 break; 1236 1237 case IPOPT_TS: 1238 code = cp - (u_char *)ip; 1239 if (optlen < sizeof(struct ip_timestamp)) 1240 goto bad; 1241 memcpy(&ipt, cp, sizeof(struct ip_timestamp)); 1242 if (ipt.ipt_ptr < 5 || ipt.ipt_len < 5) 1243 goto bad; 1244 if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) > ipt.ipt_len) { 1245 if (++ipt.ipt_oflw == 0) 1246 goto bad; 1247 break; 1248 } 1249 memcpy(&sin, cp + ipt.ipt_ptr - 1, sizeof sin); 1250 switch (ipt.ipt_flg) { 1251 1252 case IPOPT_TS_TSONLY: 1253 break; 1254 1255 case IPOPT_TS_TSANDADDR: 1256 if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) + 1257 sizeof(struct in_addr) > ipt.ipt_len) 1258 goto bad; 1259 memset(&ipaddr, 0, sizeof(ipaddr)); 1260 ipaddr.sin_family = AF_INET; 1261 ipaddr.sin_len = sizeof(ipaddr); 1262 ipaddr.sin_addr = dst; 1263 ia = ifatoia(ifaof_ifpforaddr(sintosa(&ipaddr), 1264 ifp)); 1265 if (ia == NULL) 1266 continue; 1267 memcpy(&sin, &ia->ia_addr.sin_addr, 1268 sizeof(struct in_addr)); 1269 ipt.ipt_ptr += sizeof(struct in_addr); 1270 break; 1271 1272 case IPOPT_TS_PRESPEC: 1273 if (ipt.ipt_ptr - 1 + sizeof(u_int32_t) + 1274 sizeof(struct in_addr) > ipt.ipt_len) 1275 goto bad; 1276 memset(&ipaddr, 0, sizeof(ipaddr)); 1277 ipaddr.sin_family = AF_INET; 1278 ipaddr.sin_len = sizeof(ipaddr); 1279 ipaddr.sin_addr = sin; 1280 if (ifa_ifwithaddr(sintosa(&ipaddr), 1281 m->m_pkthdr.ph_rtableid) == NULL) 1282 continue; 1283 ipt.ipt_ptr += sizeof(struct in_addr); 1284 break; 1285 1286 default: 1287 /* XXX can't take &ipt->ipt_flg */ 1288 code = (u_char *)&ipt.ipt_ptr - 1289 (u_char *)ip + 1; 1290 goto bad; 1291 } 1292 ntime = iptime(); 1293 memcpy(cp + ipt.ipt_ptr - 1, &ntime, sizeof(u_int32_t)); 1294 ipt.ipt_ptr += sizeof(u_int32_t); 1295 } 1296 } 1297 KERNEL_UNLOCK(); 1298 if (forward && ipforwarding) { 1299 ip_forward(m, ifp, NULL, 1); 1300 return (1); 1301 } 1302 return (0); 1303 bad: 1304 KERNEL_UNLOCK(); 1305 icmp_error(m, type, code, 0, 0); 1306 ipstat_inc(ips_badoptions); 1307 return (1); 1308 } 1309 1310 /* 1311 * Save incoming source route for use in replies, 1312 * to be picked up later by ip_srcroute if the receiver is interested. 1313 */ 1314 void 1315 save_rte(struct mbuf *m, u_char *option, struct in_addr dst) 1316 { 1317 struct ip_srcrt *isr; 1318 struct m_tag *mtag; 1319 unsigned olen; 1320 1321 olen = option[IPOPT_OLEN]; 1322 if (olen > sizeof(isr->isr_hdr) + sizeof(isr->isr_routes)) 1323 return; 1324 1325 mtag = m_tag_get(PACKET_TAG_SRCROUTE, sizeof(*isr), M_NOWAIT); 1326 if (mtag == NULL) 1327 return; 1328 isr = (struct ip_srcrt *)(mtag + 1); 1329 1330 memcpy(isr->isr_hdr, option, olen); 1331 isr->isr_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr); 1332 isr->isr_dst = dst; 1333 m_tag_prepend(m, mtag); 1334 } 1335 1336 /* 1337 * Retrieve incoming source route for use in replies, 1338 * in the same form used by setsockopt. 1339 * The first hop is placed before the options, will be removed later. 1340 */ 1341 struct mbuf * 1342 ip_srcroute(struct mbuf *m0) 1343 { 1344 struct in_addr *p, *q; 1345 struct mbuf *m; 1346 struct ip_srcrt *isr; 1347 struct m_tag *mtag; 1348 1349 if (!ip_dosourceroute) 1350 return (NULL); 1351 1352 mtag = m_tag_find(m0, PACKET_TAG_SRCROUTE, NULL); 1353 if (mtag == NULL) 1354 return (NULL); 1355 isr = (struct ip_srcrt *)(mtag + 1); 1356 1357 if (isr->isr_nhops == 0) 1358 return (NULL); 1359 m = m_get(M_DONTWAIT, MT_SOOPTS); 1360 if (m == NULL) 1361 return (NULL); 1362 1363 #define OPTSIZ (sizeof(isr->isr_nop) + sizeof(isr->isr_hdr)) 1364 1365 /* length is (nhops+1)*sizeof(addr) + sizeof(nop + header) */ 1366 m->m_len = (isr->isr_nhops + 1) * sizeof(struct in_addr) + OPTSIZ; 1367 1368 /* 1369 * First save first hop for return route 1370 */ 1371 p = &(isr->isr_routes[isr->isr_nhops - 1]); 1372 *(mtod(m, struct in_addr *)) = *p--; 1373 1374 /* 1375 * Copy option fields and padding (nop) to mbuf. 1376 */ 1377 isr->isr_nop = IPOPT_NOP; 1378 isr->isr_hdr[IPOPT_OFFSET] = IPOPT_MINOFF; 1379 memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &isr->isr_nop, 1380 OPTSIZ); 1381 q = (struct in_addr *)(mtod(m, caddr_t) + 1382 sizeof(struct in_addr) + OPTSIZ); 1383 #undef OPTSIZ 1384 /* 1385 * Record return path as an IP source route, 1386 * reversing the path (pointers are now aligned). 1387 */ 1388 while (p >= isr->isr_routes) { 1389 *q++ = *p--; 1390 } 1391 /* 1392 * Last hop goes to final destination. 1393 */ 1394 *q = isr->isr_dst; 1395 m_tag_delete(m0, (struct m_tag *)isr); 1396 return (m); 1397 } 1398 1399 /* 1400 * Strip out IP options, at higher level protocol in the kernel. 1401 */ 1402 void 1403 ip_stripoptions(struct mbuf *m) 1404 { 1405 int i; 1406 struct ip *ip = mtod(m, struct ip *); 1407 caddr_t opts; 1408 int olen; 1409 1410 olen = (ip->ip_hl<<2) - sizeof (struct ip); 1411 opts = (caddr_t)(ip + 1); 1412 i = m->m_len - (sizeof (struct ip) + olen); 1413 memmove(opts, opts + olen, i); 1414 m->m_len -= olen; 1415 if (m->m_flags & M_PKTHDR) 1416 m->m_pkthdr.len -= olen; 1417 ip->ip_hl = sizeof(struct ip) >> 2; 1418 ip->ip_len = htons(ntohs(ip->ip_len) - olen); 1419 } 1420 1421 int inetctlerrmap[PRC_NCMDS] = { 1422 0, 0, 0, 0, 1423 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 1424 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 1425 EMSGSIZE, EHOSTUNREACH, 0, 0, 1426 0, 0, 0, 0, 1427 ENOPROTOOPT 1428 }; 1429 1430 /* 1431 * Forward a packet. If some error occurs return the sender 1432 * an icmp packet. Note we can't always generate a meaningful 1433 * icmp message because icmp doesn't have a large enough repertoire 1434 * of codes and types. 1435 * 1436 * If not forwarding, just drop the packet. This could be confusing 1437 * if ipforwarding was zero but some routing protocol was advancing 1438 * us as a gateway to somewhere. However, we must let the routing 1439 * protocol deal with that. 1440 * 1441 * The srcrt parameter indicates whether the packet is being forwarded 1442 * via a source route. 1443 */ 1444 void 1445 ip_forward(struct mbuf *m, struct ifnet *ifp, struct rtentry *rt, int srcrt) 1446 { 1447 struct mbuf mfake, *mcopy = NULL; 1448 struct ip *ip = mtod(m, struct ip *); 1449 struct sockaddr_in *sin; 1450 struct route ro; 1451 int error, type = 0, code = 0, destmtu = 0, fake = 0, len; 1452 u_int32_t dest; 1453 1454 dest = 0; 1455 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 1456 ipstat_inc(ips_cantforward); 1457 m_freem(m); 1458 goto freecopy; 1459 } 1460 if (ip->ip_ttl <= IPTTLDEC) { 1461 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0); 1462 goto freecopy; 1463 } 1464 1465 sin = satosin(&ro.ro_dst); 1466 memset(sin, 0, sizeof(*sin)); 1467 sin->sin_family = AF_INET; 1468 sin->sin_len = sizeof(*sin); 1469 sin->sin_addr = ip->ip_dst; 1470 1471 if (!rtisvalid(rt)) { 1472 rtfree(rt); 1473 rt = rtalloc_mpath(sintosa(sin), &ip->ip_src.s_addr, 1474 m->m_pkthdr.ph_rtableid); 1475 if (rt == NULL) { 1476 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0); 1477 return; 1478 } 1479 } 1480 1481 /* 1482 * Save at most 68 bytes of the packet in case 1483 * we need to generate an ICMP message to the src. 1484 * The data is saved in the mbuf on the stack that 1485 * acts as a temporary storage not intended to be 1486 * passed down the IP stack or to the mfree. 1487 */ 1488 memset(&mfake.m_hdr, 0, sizeof(mfake.m_hdr)); 1489 mfake.m_type = m->m_type; 1490 if (m_dup_pkthdr(&mfake, m, M_DONTWAIT) == 0) { 1491 mfake.m_data = mfake.m_pktdat; 1492 len = min(ntohs(ip->ip_len), 68); 1493 m_copydata(m, 0, len, mfake.m_pktdat); 1494 mfake.m_pkthdr.len = mfake.m_len = len; 1495 #if NPF > 0 1496 pf_pkt_unlink_state_key(&mfake); 1497 #endif /* NPF > 0 */ 1498 fake = 1; 1499 } 1500 1501 ip->ip_ttl -= IPTTLDEC; 1502 1503 /* 1504 * If forwarding packet using same interface that it came in on, 1505 * perhaps should send a redirect to sender to shortcut a hop. 1506 * Only send redirect if source is sending directly to us, 1507 * and if packet was not source routed (or has any options). 1508 * Also, don't send redirect if forwarding using a default route 1509 * or a route modified by a redirect. 1510 * Don't send redirect if we advertise destination's arp address 1511 * as ours (proxy arp). 1512 */ 1513 if ((rt->rt_ifidx == ifp->if_index) && 1514 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1515 satosin(rt_key(rt))->sin_addr.s_addr != 0 && 1516 ipsendredirects && !srcrt && 1517 !arpproxy(satosin(rt_key(rt))->sin_addr, m->m_pkthdr.ph_rtableid)) { 1518 if ((ip->ip_src.s_addr & ifatoia(rt->rt_ifa)->ia_netmask) == 1519 ifatoia(rt->rt_ifa)->ia_net) { 1520 if (rt->rt_flags & RTF_GATEWAY) 1521 dest = satosin(rt->rt_gateway)->sin_addr.s_addr; 1522 else 1523 dest = ip->ip_dst.s_addr; 1524 /* Router requirements says to only send host redirects */ 1525 type = ICMP_REDIRECT; 1526 code = ICMP_REDIRECT_HOST; 1527 } 1528 } 1529 1530 ro.ro_rt = rt; 1531 ro.ro_tableid = m->m_pkthdr.ph_rtableid; 1532 error = ip_output(m, NULL, &ro, 1533 (IP_FORWARDING | (ip_directedbcast ? IP_ALLOWBROADCAST : 0)), 1534 NULL, NULL, 0); 1535 rt = ro.ro_rt; 1536 if (error) 1537 ipstat_inc(ips_cantforward); 1538 else { 1539 ipstat_inc(ips_forward); 1540 if (type) 1541 ipstat_inc(ips_redirectsent); 1542 else 1543 goto freecopy; 1544 } 1545 if (!fake) 1546 goto freecopy; 1547 1548 switch (error) { 1549 1550 case 0: /* forwarded, but need redirect */ 1551 /* type, code set above */ 1552 break; 1553 1554 case ENETUNREACH: /* shouldn't happen, checked above */ 1555 case EHOSTUNREACH: 1556 case ENETDOWN: 1557 case EHOSTDOWN: 1558 default: 1559 type = ICMP_UNREACH; 1560 code = ICMP_UNREACH_HOST; 1561 break; 1562 1563 case EMSGSIZE: 1564 type = ICMP_UNREACH; 1565 code = ICMP_UNREACH_NEEDFRAG; 1566 1567 #ifdef IPSEC 1568 if (rt != NULL) { 1569 if (rt->rt_mtu) 1570 destmtu = rt->rt_mtu; 1571 else { 1572 struct ifnet *destifp; 1573 1574 destifp = if_get(rt->rt_ifidx); 1575 if (destifp != NULL) 1576 destmtu = destifp->if_mtu; 1577 if_put(destifp); 1578 } 1579 } 1580 #endif /*IPSEC*/ 1581 ipstat_inc(ips_cantfrag); 1582 break; 1583 1584 case EACCES: 1585 /* 1586 * pf(4) blocked the packet. There is no need to send an ICMP 1587 * packet back since pf(4) takes care of it. 1588 */ 1589 goto freecopy; 1590 case ENOBUFS: 1591 /* 1592 * a router should not generate ICMP_SOURCEQUENCH as 1593 * required in RFC1812 Requirements for IP Version 4 Routers. 1594 * source quench could be a big problem under DoS attacks, 1595 * or the underlying interface is rate-limited. 1596 */ 1597 goto freecopy; 1598 } 1599 1600 mcopy = m_copym(&mfake, 0, len, M_DONTWAIT); 1601 if (mcopy) 1602 icmp_error(mcopy, type, code, dest, destmtu); 1603 1604 freecopy: 1605 if (fake) 1606 m_tag_delete_chain(&mfake); 1607 rtfree(rt); 1608 } 1609 1610 int 1611 ip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp, 1612 size_t newlen) 1613 { 1614 int error; 1615 #ifdef MROUTING 1616 extern int ip_mrtproto; 1617 extern struct mrtstat mrtstat; 1618 #endif 1619 1620 NET_ASSERT_LOCKED(); 1621 1622 /* Almost all sysctl names at this level are terminal. */ 1623 if (namelen != 1 && name[0] != IPCTL_IFQUEUE) 1624 return (ENOTDIR); 1625 1626 switch (name[0]) { 1627 #ifdef notyet 1628 case IPCTL_DEFMTU: 1629 return (sysctl_int(oldp, oldlenp, newp, newlen, &ip_mtu)); 1630 #endif 1631 case IPCTL_SOURCEROUTE: 1632 /* 1633 * Don't allow this to change in a secure environment. 1634 */ 1635 if (newp && securelevel > 0) 1636 return (EPERM); 1637 return (sysctl_int(oldp, oldlenp, newp, newlen, 1638 &ip_dosourceroute)); 1639 case IPCTL_MTUDISC: 1640 error = sysctl_int(oldp, oldlenp, newp, newlen, 1641 &ip_mtudisc); 1642 if (ip_mtudisc != 0 && ip_mtudisc_timeout_q == NULL) { 1643 ip_mtudisc_timeout_q = 1644 rt_timer_queue_create(ip_mtudisc_timeout); 1645 } else if (ip_mtudisc == 0 && ip_mtudisc_timeout_q != NULL) { 1646 rt_timer_queue_destroy(ip_mtudisc_timeout_q); 1647 ip_mtudisc_timeout_q = NULL; 1648 } 1649 return error; 1650 case IPCTL_MTUDISCTIMEOUT: 1651 error = sysctl_int(oldp, oldlenp, newp, newlen, 1652 &ip_mtudisc_timeout); 1653 if (ip_mtudisc_timeout_q != NULL) 1654 rt_timer_queue_change(ip_mtudisc_timeout_q, 1655 ip_mtudisc_timeout); 1656 return (error); 1657 case IPCTL_IPSEC_ENC_ALGORITHM: 1658 return (sysctl_tstring(oldp, oldlenp, newp, newlen, 1659 ipsec_def_enc, sizeof(ipsec_def_enc))); 1660 case IPCTL_IPSEC_AUTH_ALGORITHM: 1661 return (sysctl_tstring(oldp, oldlenp, newp, newlen, 1662 ipsec_def_auth, 1663 sizeof(ipsec_def_auth))); 1664 case IPCTL_IPSEC_IPCOMP_ALGORITHM: 1665 return (sysctl_tstring(oldp, oldlenp, newp, newlen, 1666 ipsec_def_comp, 1667 sizeof(ipsec_def_comp))); 1668 case IPCTL_IFQUEUE: 1669 return (sysctl_niq(name + 1, namelen - 1, 1670 oldp, oldlenp, newp, newlen, &ipintrq)); 1671 case IPCTL_STATS: 1672 return (ip_sysctl_ipstat(oldp, oldlenp, newp)); 1673 #ifdef MROUTING 1674 case IPCTL_MRTSTATS: 1675 return (sysctl_rdstruct(oldp, oldlenp, newp, 1676 &mrtstat, sizeof(mrtstat))); 1677 case IPCTL_MRTPROTO: 1678 return (sysctl_rdint(oldp, oldlenp, newp, ip_mrtproto)); 1679 case IPCTL_MRTMFC: 1680 if (newp) 1681 return (EPERM); 1682 return mrt_sysctl_mfc(oldp, oldlenp); 1683 case IPCTL_MRTVIF: 1684 if (newp) 1685 return (EPERM); 1686 return mrt_sysctl_vif(oldp, oldlenp); 1687 #else 1688 case IPCTL_MRTPROTO: 1689 case IPCTL_MRTSTATS: 1690 case IPCTL_MRTMFC: 1691 case IPCTL_MRTVIF: 1692 return (EOPNOTSUPP); 1693 #endif 1694 default: 1695 if (name[0] < IPCTL_MAXID) 1696 return (sysctl_int_arr(ipctl_vars, name, namelen, 1697 oldp, oldlenp, newp, newlen)); 1698 return (EOPNOTSUPP); 1699 } 1700 /* NOTREACHED */ 1701 } 1702 1703 int 1704 ip_sysctl_ipstat(void *oldp, size_t *oldlenp, void *newp) 1705 { 1706 uint64_t counters[ips_ncounters]; 1707 struct ipstat ipstat; 1708 u_long *words = (u_long *)&ipstat; 1709 int i; 1710 1711 CTASSERT(sizeof(ipstat) == (nitems(counters) * sizeof(u_long))); 1712 memset(&ipstat, 0, sizeof ipstat); 1713 counters_read(ipcounters, counters, nitems(counters)); 1714 1715 for (i = 0; i < nitems(counters); i++) 1716 words[i] = (u_long)counters[i]; 1717 1718 return (sysctl_rdstruct(oldp, oldlenp, newp, &ipstat, sizeof(ipstat))); 1719 } 1720 1721 void 1722 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1723 struct mbuf *m) 1724 { 1725 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1726 struct timeval tv; 1727 1728 microtime(&tv); 1729 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv), 1730 SCM_TIMESTAMP, SOL_SOCKET); 1731 if (*mp) 1732 mp = &(*mp)->m_next; 1733 } 1734 1735 if (inp->inp_flags & INP_RECVDSTADDR) { 1736 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst, 1737 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1738 if (*mp) 1739 mp = &(*mp)->m_next; 1740 } 1741 #ifdef notyet 1742 /* this code is broken and will probably never be fixed. */ 1743 /* options were tossed already */ 1744 if (inp->inp_flags & INP_RECVOPTS) { 1745 *mp = sbcreatecontrol((caddr_t) opts_deleted_above, 1746 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1747 if (*mp) 1748 mp = &(*mp)->m_next; 1749 } 1750 /* ip_srcroute doesn't do what we want here, need to fix */ 1751 if (inp->inp_flags & INP_RECVRETOPTS) { 1752 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m), 1753 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1754 if (*mp) 1755 mp = &(*mp)->m_next; 1756 } 1757 #endif 1758 if (inp->inp_flags & INP_RECVIF) { 1759 struct sockaddr_dl sdl; 1760 struct ifnet *ifp; 1761 1762 ifp = if_get(m->m_pkthdr.ph_ifidx); 1763 if (ifp == NULL || ifp->if_sadl == NULL) { 1764 memset(&sdl, 0, sizeof(sdl)); 1765 sdl.sdl_len = offsetof(struct sockaddr_dl, sdl_data[0]); 1766 sdl.sdl_family = AF_LINK; 1767 sdl.sdl_index = ifp != NULL ? ifp->if_index : 0; 1768 sdl.sdl_nlen = sdl.sdl_alen = sdl.sdl_slen = 0; 1769 *mp = sbcreatecontrol((caddr_t) &sdl, sdl.sdl_len, 1770 IP_RECVIF, IPPROTO_IP); 1771 } else { 1772 *mp = sbcreatecontrol((caddr_t) ifp->if_sadl, 1773 ifp->if_sadl->sdl_len, IP_RECVIF, IPPROTO_IP); 1774 } 1775 if (*mp) 1776 mp = &(*mp)->m_next; 1777 if_put(ifp); 1778 } 1779 if (inp->inp_flags & INP_RECVTTL) { 1780 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl, 1781 sizeof(u_int8_t), IP_RECVTTL, IPPROTO_IP); 1782 if (*mp) 1783 mp = &(*mp)->m_next; 1784 } 1785 if (inp->inp_flags & INP_RECVRTABLE) { 1786 u_int rtableid = inp->inp_rtableid; 1787 #if NPF > 0 1788 struct pf_divert *divert; 1789 1790 if (m && m->m_pkthdr.pf.flags & PF_TAG_DIVERTED && 1791 (divert = pf_find_divert(m)) != NULL) 1792 rtableid = divert->rdomain; 1793 #endif 1794 1795 *mp = sbcreatecontrol((caddr_t) &rtableid, 1796 sizeof(u_int), IP_RECVRTABLE, IPPROTO_IP); 1797 if (*mp) 1798 mp = &(*mp)->m_next; 1799 } 1800 } 1801 1802 void 1803 ip_send_dispatch(void *xmq) 1804 { 1805 struct mbuf_queue *mq = xmq; 1806 struct mbuf *m; 1807 struct mbuf_list ml; 1808 #ifdef IPSEC 1809 int locked = 0; 1810 #endif /* IPSEC */ 1811 1812 mq_delist(mq, &ml); 1813 if (ml_empty(&ml)) 1814 return; 1815 1816 NET_LOCK(); 1817 1818 #ifdef IPSEC 1819 /* 1820 * IPsec is not ready to run without KERNEL_LOCK(). So all 1821 * the traffic on your machine is punished if you have IPsec 1822 * enabled. 1823 */ 1824 extern int ipsec_in_use; 1825 if (ipsec_in_use) { 1826 NET_UNLOCK(); 1827 KERNEL_LOCK(); 1828 NET_LOCK(); 1829 locked = 1; 1830 } 1831 #endif /* IPSEC */ 1832 1833 while ((m = ml_dequeue(&ml)) != NULL) { 1834 ip_output(m, NULL, NULL, 0, NULL, NULL, 0); 1835 } 1836 NET_UNLOCK(); 1837 1838 #ifdef IPSEC 1839 if (locked) 1840 KERNEL_UNLOCK(); 1841 #endif /* IPSEC */ 1842 } 1843 1844 void 1845 ip_send(struct mbuf *m) 1846 { 1847 mq_enqueue(&ipsend_mq, m); 1848 task_add(softnettq, &ipsend_task); 1849 } 1850