ip_output.c revision 138408
1/* 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 30 * $FreeBSD: head/sys/netinet/ip_output.c 138408 2004-12-05 22:08:37Z rwatson $ 31 */ 32 33#include "opt_ipfw.h" 34#include "opt_ipsec.h" 35#include "opt_mac.h" 36#include "opt_mbuf_stress_test.h" 37 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/kernel.h> 41#include <sys/mac.h> 42#include <sys/malloc.h> 43#include <sys/mbuf.h> 44#include <sys/protosw.h> 45#include <sys/socket.h> 46#include <sys/socketvar.h> 47#include <sys/sysctl.h> 48 49#include <net/if.h> 50#include <net/netisr.h> 51#include <net/pfil.h> 52#include <net/route.h> 53 54#include <netinet/in.h> 55#include <netinet/in_systm.h> 56#include <netinet/ip.h> 57#include <netinet/in_pcb.h> 58#include <netinet/in_var.h> 59#include <netinet/ip_var.h> 60 61 62#include <machine/in_cksum.h> 63 64static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 65 66#ifdef IPSEC 67#include <netinet6/ipsec.h> 68#include <netkey/key.h> 69#ifdef IPSEC_DEBUG 70#include <netkey/key_debug.h> 71#else 72#define KEYDEBUG(lev,arg) 73#endif 74#endif /*IPSEC*/ 75 76#ifdef FAST_IPSEC 77#include <netipsec/ipsec.h> 78#include <netipsec/xform.h> 79#include <netipsec/key.h> 80#endif /*FAST_IPSEC*/ 81 82#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 83 x, (ntohl(a.s_addr)>>24)&0xFF,\ 84 (ntohl(a.s_addr)>>16)&0xFF,\ 85 (ntohl(a.s_addr)>>8)&0xFF,\ 86 (ntohl(a.s_addr))&0xFF, y); 87 88u_short ip_id; 89 90#ifdef MBUF_STRESS_TEST 91int mbuf_frag_size = 0; 92SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 93 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 94#endif 95 96static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 97static struct ifnet *ip_multicast_if(struct in_addr *, int *); 98static void ip_mloopback 99 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 100static int ip_getmoptions(struct inpcb *, struct sockopt *); 101static int ip_pcbopts(struct inpcb *, int, struct mbuf *); 102static int ip_setmoptions(struct inpcb *, struct sockopt *); 103 104int ip_optcopy(struct ip *, struct ip *); 105 106 107extern struct protosw inetsw[]; 108 109/* 110 * IP output. The packet in mbuf chain m contains a skeletal IP 111 * header (with len, off, ttl, proto, tos, src, dst). 112 * The mbuf chain containing the packet will be freed. 113 * The mbuf opt, if present, will not be freed. 114 * In the IP forwarding case, the packet will arrive with options already 115 * inserted, so must have a NULL opt pointer. 116 */ 117int 118ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, 119 int flags, struct ip_moptions *imo, struct inpcb *inp) 120{ 121 struct ip *ip; 122 struct ifnet *ifp = NULL; /* keep compiler happy */ 123 struct mbuf *m0; 124 int hlen = sizeof (struct ip); 125 int len, error = 0; 126 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 127 struct in_ifaddr *ia = NULL; 128 int isbroadcast, sw_csum; 129 struct route iproute; 130 struct in_addr odst; 131#ifdef IPFIREWALL_FORWARD 132 struct m_tag *fwd_tag = NULL; 133#endif 134#ifdef IPSEC 135 struct secpolicy *sp = NULL; 136#endif 137#ifdef FAST_IPSEC 138 struct secpolicy *sp = NULL; 139 struct tdb_ident *tdbi; 140 struct m_tag *mtag; 141 int s; 142#endif /* FAST_IPSEC */ 143 144 M_ASSERTPKTHDR(m); 145 146 if (ro == NULL) { 147 ro = &iproute; 148 bzero(ro, sizeof (*ro)); 149 } 150 151 if (inp != NULL) 152 INP_LOCK_ASSERT(inp); 153 154 if (opt) { 155 len = 0; 156 m = ip_insertoptions(m, opt, &len); 157 if (len != 0) 158 hlen = len; 159 } 160 ip = mtod(m, struct ip *); 161 162 /* 163 * Fill in IP header. If we are not allowing fragmentation, 164 * then the ip_id field is meaningless, but we don't set it 165 * to zero. Doing so causes various problems when devices along 166 * the path (routers, load balancers, firewalls, etc.) illegally 167 * disable DF on our packet. Note that a 16-bit counter 168 * will wrap around in less than 10 seconds at 100 Mbit/s on a 169 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 170 * for Counting NATted Hosts", Proc. IMW'02, available at 171 * <http://www.research.att.com/~smb/papers/fnat.pdf>. 172 */ 173 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 174 ip->ip_v = IPVERSION; 175 ip->ip_hl = hlen >> 2; 176 ip->ip_id = ip_newid(); 177 ipstat.ips_localout++; 178 } else { 179 hlen = ip->ip_hl << 2; 180 } 181 182 dst = (struct sockaddr_in *)&ro->ro_dst; 183again: 184 /* 185 * If there is a cached route, 186 * check that it is to the same destination 187 * and is still up. If not, free it and try again. 188 * The address family should also be checked in case of sharing the 189 * cache with IPv6. 190 */ 191 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 192 dst->sin_family != AF_INET || 193 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 194 RTFREE(ro->ro_rt); 195 ro->ro_rt = (struct rtentry *)0; 196 } 197#ifdef IPFIREWALL_FORWARD 198 if (ro->ro_rt == NULL && fwd_tag == NULL) { 199#else 200 if (ro->ro_rt == NULL) { 201#endif 202 bzero(dst, sizeof(*dst)); 203 dst->sin_family = AF_INET; 204 dst->sin_len = sizeof(*dst); 205 dst->sin_addr = ip->ip_dst; 206 } 207 /* 208 * If routing to interface only, 209 * short circuit routing lookup. 210 */ 211 if (flags & IP_ROUTETOIF) { 212 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 213 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) { 214 ipstat.ips_noroute++; 215 error = ENETUNREACH; 216 goto bad; 217 } 218 ifp = ia->ia_ifp; 219 ip->ip_ttl = 1; 220 isbroadcast = in_broadcast(dst->sin_addr, ifp); 221 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 222 imo != NULL && imo->imo_multicast_ifp != NULL) { 223 /* 224 * Bypass the normal routing lookup for multicast 225 * packets if the interface is specified. 226 */ 227 ifp = imo->imo_multicast_ifp; 228 IFP_TO_IA(ifp, ia); 229 isbroadcast = 0; /* fool gcc */ 230 } else { 231 /* 232 * We want to do any cloning requested by the link layer, 233 * as this is probably required in all cases for correct 234 * operation (as it is for ARP). 235 */ 236 if (ro->ro_rt == NULL) 237 rtalloc_ign(ro, 0); 238 if (ro->ro_rt == NULL) { 239 ipstat.ips_noroute++; 240 error = EHOSTUNREACH; 241 goto bad; 242 } 243 ia = ifatoia(ro->ro_rt->rt_ifa); 244 ifp = ro->ro_rt->rt_ifp; 245 ro->ro_rt->rt_rmx.rmx_pksent++; 246 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 247 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 248 if (ro->ro_rt->rt_flags & RTF_HOST) 249 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 250 else 251 isbroadcast = in_broadcast(dst->sin_addr, ifp); 252 } 253 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 254 struct in_multi *inm; 255 256 m->m_flags |= M_MCAST; 257 /* 258 * IP destination address is multicast. Make sure "dst" 259 * still points to the address in "ro". (It may have been 260 * changed to point to a gateway address, above.) 261 */ 262 dst = (struct sockaddr_in *)&ro->ro_dst; 263 /* 264 * See if the caller provided any multicast options 265 */ 266 if (imo != NULL) { 267 ip->ip_ttl = imo->imo_multicast_ttl; 268 if (imo->imo_multicast_vif != -1) 269 ip->ip_src.s_addr = 270 ip_mcast_src ? 271 ip_mcast_src(imo->imo_multicast_vif) : 272 INADDR_ANY; 273 } else 274 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 275 /* 276 * Confirm that the outgoing interface supports multicast. 277 */ 278 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 279 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 280 ipstat.ips_noroute++; 281 error = ENETUNREACH; 282 goto bad; 283 } 284 } 285 /* 286 * If source address not specified yet, use address 287 * of outgoing interface. 288 */ 289 if (ip->ip_src.s_addr == INADDR_ANY) { 290 /* Interface may have no addresses. */ 291 if (ia != NULL) 292 ip->ip_src = IA_SIN(ia)->sin_addr; 293 } 294 295 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 296 if (inm != NULL && 297 (imo == NULL || imo->imo_multicast_loop)) { 298 /* 299 * If we belong to the destination multicast group 300 * on the outgoing interface, and the caller did not 301 * forbid loopback, loop back a copy. 302 */ 303 ip_mloopback(ifp, m, dst, hlen); 304 } 305 else { 306 /* 307 * If we are acting as a multicast router, perform 308 * multicast forwarding as if the packet had just 309 * arrived on the interface to which we are about 310 * to send. The multicast forwarding function 311 * recursively calls this function, using the 312 * IP_FORWARDING flag to prevent infinite recursion. 313 * 314 * Multicasts that are looped back by ip_mloopback(), 315 * above, will be forwarded by the ip_input() routine, 316 * if necessary. 317 */ 318 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 319 /* 320 * If rsvp daemon is not running, do not 321 * set ip_moptions. This ensures that the packet 322 * is multicast and not just sent down one link 323 * as prescribed by rsvpd. 324 */ 325 if (!rsvp_on) 326 imo = NULL; 327 if (ip_mforward && 328 ip_mforward(ip, ifp, m, imo) != 0) { 329 m_freem(m); 330 goto done; 331 } 332 } 333 } 334 335 /* 336 * Multicasts with a time-to-live of zero may be looped- 337 * back, above, but must not be transmitted on a network. 338 * Also, multicasts addressed to the loopback interface 339 * are not sent -- the above call to ip_mloopback() will 340 * loop back a copy if this host actually belongs to the 341 * destination group on the loopback interface. 342 */ 343 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 344 m_freem(m); 345 goto done; 346 } 347 348 goto sendit; 349 } 350#ifndef notdef 351 /* 352 * If the source address is not specified yet, use the address 353 * of the outoing interface. 354 */ 355 if (ip->ip_src.s_addr == INADDR_ANY) { 356 /* Interface may have no addresses. */ 357 if (ia != NULL) { 358 ip->ip_src = IA_SIN(ia)->sin_addr; 359 } 360 } 361#endif /* notdef */ 362 /* 363 * Verify that we have any chance at all of being able to queue the 364 * packet or packet fragments, unless ALTQ is enabled on the given 365 * interface in which case packetdrop should be done by queueing. 366 */ 367#ifdef ALTQ 368 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) && 369 ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 370 ifp->if_snd.ifq_maxlen)) 371#else 372 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 373 ifp->if_snd.ifq_maxlen) 374#endif /* ALTQ */ 375 { 376 error = ENOBUFS; 377 ipstat.ips_odropped++; 378 goto bad; 379 } 380 381 /* 382 * Look for broadcast address and 383 * verify user is allowed to send 384 * such a packet. 385 */ 386 if (isbroadcast) { 387 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 388 error = EADDRNOTAVAIL; 389 goto bad; 390 } 391 if ((flags & IP_ALLOWBROADCAST) == 0) { 392 error = EACCES; 393 goto bad; 394 } 395 /* don't allow broadcast messages to be fragmented */ 396 if (ip->ip_len > ifp->if_mtu) { 397 error = EMSGSIZE; 398 goto bad; 399 } 400 if (flags & IP_SENDONES) 401 ip->ip_dst.s_addr = INADDR_BROADCAST; 402 m->m_flags |= M_BCAST; 403 } else { 404 m->m_flags &= ~M_BCAST; 405 } 406 407sendit: 408#ifdef IPSEC 409 /* get SP for this packet */ 410 if (inp == NULL) 411 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 412 flags, &error); 413 else 414 sp = ipsec4_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error); 415 416 if (sp == NULL) { 417 ipsecstat.out_inval++; 418 goto bad; 419 } 420 421 error = 0; 422 423 /* check policy */ 424 switch (sp->policy) { 425 case IPSEC_POLICY_DISCARD: 426 /* 427 * This packet is just discarded. 428 */ 429 ipsecstat.out_polvio++; 430 goto bad; 431 432 case IPSEC_POLICY_BYPASS: 433 case IPSEC_POLICY_NONE: 434 case IPSEC_POLICY_TCP: 435 /* no need to do IPsec. */ 436 goto skip_ipsec; 437 438 case IPSEC_POLICY_IPSEC: 439 if (sp->req == NULL) { 440 /* acquire a policy */ 441 error = key_spdacquire(sp); 442 goto bad; 443 } 444 break; 445 446 case IPSEC_POLICY_ENTRUST: 447 default: 448 printf("ip_output: Invalid policy found. %d\n", sp->policy); 449 } 450 { 451 struct ipsec_output_state state; 452 bzero(&state, sizeof(state)); 453 state.m = m; 454 if (flags & IP_ROUTETOIF) { 455 state.ro = &iproute; 456 bzero(&iproute, sizeof(iproute)); 457 } else 458 state.ro = ro; 459 state.dst = (struct sockaddr *)dst; 460 461 ip->ip_sum = 0; 462 463 /* 464 * XXX 465 * delayed checksums are not currently compatible with IPsec 466 */ 467 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 468 in_delayed_cksum(m); 469 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 470 } 471 472 ip->ip_len = htons(ip->ip_len); 473 ip->ip_off = htons(ip->ip_off); 474 475 error = ipsec4_output(&state, sp, flags); 476 477 m = state.m; 478 if (flags & IP_ROUTETOIF) { 479 /* 480 * if we have tunnel mode SA, we may need to ignore 481 * IP_ROUTETOIF. 482 */ 483 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 484 flags &= ~IP_ROUTETOIF; 485 ro = state.ro; 486 } 487 } else 488 ro = state.ro; 489 dst = (struct sockaddr_in *)state.dst; 490 if (error) { 491 /* mbuf is already reclaimed in ipsec4_output. */ 492 m = NULL; 493 switch (error) { 494 case EHOSTUNREACH: 495 case ENETUNREACH: 496 case EMSGSIZE: 497 case ENOBUFS: 498 case ENOMEM: 499 break; 500 default: 501 printf("ip4_output (ipsec): error code %d\n", error); 502 /*fall through*/ 503 case ENOENT: 504 /* don't show these error codes to the user */ 505 error = 0; 506 break; 507 } 508 goto bad; 509 } 510 511 /* be sure to update variables that are affected by ipsec4_output() */ 512 ip = mtod(m, struct ip *); 513 hlen = ip->ip_hl << 2; 514 if (ro->ro_rt == NULL) { 515 if ((flags & IP_ROUTETOIF) == 0) { 516 printf("ip_output: " 517 "can't update route after IPsec processing\n"); 518 error = EHOSTUNREACH; /*XXX*/ 519 goto bad; 520 } 521 } else { 522 if (state.encap) { 523 ia = ifatoia(ro->ro_rt->rt_ifa); 524 ifp = ro->ro_rt->rt_ifp; 525 } 526 } 527 } 528 529 /* make it flipped, again. */ 530 ip->ip_len = ntohs(ip->ip_len); 531 ip->ip_off = ntohs(ip->ip_off); 532skip_ipsec: 533#endif /*IPSEC*/ 534#ifdef FAST_IPSEC 535 /* 536 * Check the security policy (SP) for the packet and, if 537 * required, do IPsec-related processing. There are two 538 * cases here; the first time a packet is sent through 539 * it will be untagged and handled by ipsec4_checkpolicy. 540 * If the packet is resubmitted to ip_output (e.g. after 541 * AH, ESP, etc. processing), there will be a tag to bypass 542 * the lookup and related policy checking. 543 */ 544 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 545 s = splnet(); 546 if (mtag != NULL) { 547 tdbi = (struct tdb_ident *)(mtag + 1); 548 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 549 if (sp == NULL) 550 error = -EINVAL; /* force silent drop */ 551 m_tag_delete(m, mtag); 552 } else { 553 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 554 &error, inp); 555 } 556 /* 557 * There are four return cases: 558 * sp != NULL apply IPsec policy 559 * sp == NULL, error == 0 no IPsec handling needed 560 * sp == NULL, error == -EINVAL discard packet w/o error 561 * sp == NULL, error != 0 discard packet, report error 562 */ 563 if (sp != NULL) { 564 /* Loop detection, check if ipsec processing already done */ 565 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 566 for (mtag = m_tag_first(m); mtag != NULL; 567 mtag = m_tag_next(m, mtag)) { 568 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 569 continue; 570 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 571 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 572 continue; 573 /* 574 * Check if policy has an SA associated with it. 575 * This can happen when an SP has yet to acquire 576 * an SA; e.g. on first reference. If it occurs, 577 * then we let ipsec4_process_packet do its thing. 578 */ 579 if (sp->req->sav == NULL) 580 break; 581 tdbi = (struct tdb_ident *)(mtag + 1); 582 if (tdbi->spi == sp->req->sav->spi && 583 tdbi->proto == sp->req->sav->sah->saidx.proto && 584 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 585 sizeof (union sockaddr_union)) == 0) { 586 /* 587 * No IPsec processing is needed, free 588 * reference to SP. 589 * 590 * NB: null pointer to avoid free at 591 * done: below. 592 */ 593 KEY_FREESP(&sp), sp = NULL; 594 splx(s); 595 goto spd_done; 596 } 597 } 598 599 /* 600 * Do delayed checksums now because we send before 601 * this is done in the normal processing path. 602 */ 603 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 604 in_delayed_cksum(m); 605 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 606 } 607 608 ip->ip_len = htons(ip->ip_len); 609 ip->ip_off = htons(ip->ip_off); 610 611 /* NB: callee frees mbuf */ 612 error = ipsec4_process_packet(m, sp->req, flags, 0); 613 /* 614 * Preserve KAME behaviour: ENOENT can be returned 615 * when an SA acquire is in progress. Don't propagate 616 * this to user-level; it confuses applications. 617 * 618 * XXX this will go away when the SADB is redone. 619 */ 620 if (error == ENOENT) 621 error = 0; 622 splx(s); 623 goto done; 624 } else { 625 splx(s); 626 627 if (error != 0) { 628 /* 629 * Hack: -EINVAL is used to signal that a packet 630 * should be silently discarded. This is typically 631 * because we asked key management for an SA and 632 * it was delayed (e.g. kicked up to IKE). 633 */ 634 if (error == -EINVAL) 635 error = 0; 636 goto bad; 637 } else { 638 /* No IPsec processing for this packet. */ 639 } 640#ifdef notyet 641 /* 642 * If deferred crypto processing is needed, check that 643 * the interface supports it. 644 */ 645 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 646 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 647 /* notify IPsec to do its own crypto */ 648 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 649 error = EHOSTUNREACH; 650 goto bad; 651 } 652#endif 653 } 654spd_done: 655#endif /* FAST_IPSEC */ 656 657 /* Jump over all PFIL processing if hooks are not active. */ 658 if (inet_pfil_hook.ph_busy_count == -1) 659 goto passout; 660 661 /* Run through list of hooks for output packets. */ 662 odst.s_addr = ip->ip_dst.s_addr; 663 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 664 if (error != 0 || m == NULL) 665 goto done; 666 667 ip = mtod(m, struct ip *); 668 669 /* See if destination IP address was changed by packet filter. */ 670 if (odst.s_addr != ip->ip_dst.s_addr) { 671 m->m_flags |= M_SKIP_FIREWALL; 672 /* If destination is now ourself drop to ip_input(). */ 673 if (in_localip(ip->ip_dst)) { 674 m->m_flags |= M_FASTFWD_OURS; 675 if (m->m_pkthdr.rcvif == NULL) 676 m->m_pkthdr.rcvif = loif; 677 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 678 m->m_pkthdr.csum_flags |= 679 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 680 m->m_pkthdr.csum_data = 0xffff; 681 } 682 m->m_pkthdr.csum_flags |= 683 CSUM_IP_CHECKED | CSUM_IP_VALID; 684 685 error = netisr_queue(NETISR_IP, m); 686 goto done; 687 } else 688 goto again; /* Redo the routing table lookup. */ 689 } 690 691#ifdef IPFIREWALL_FORWARD 692 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 693 if (m->m_flags & M_FASTFWD_OURS) { 694 if (m->m_pkthdr.rcvif == NULL) 695 m->m_pkthdr.rcvif = loif; 696 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 697 m->m_pkthdr.csum_flags |= 698 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 699 m->m_pkthdr.csum_data = 0xffff; 700 } 701 m->m_pkthdr.csum_flags |= 702 CSUM_IP_CHECKED | CSUM_IP_VALID; 703 704 error = netisr_queue(NETISR_IP, m); 705 goto done; 706 } 707 /* Or forward to some other address? */ 708 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 709 if (fwd_tag) { 710 if (!in_localip(ip->ip_src) && !in_localaddr(ip->ip_dst)) { 711 dst = (struct sockaddr_in *)&ro->ro_dst; 712 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 713 m->m_flags |= M_SKIP_FIREWALL; 714 m_tag_delete(m, fwd_tag); 715 goto again; 716 } else { 717 m_tag_delete(m, fwd_tag); 718 /* Continue. */ 719 } 720 } 721#endif 722 723passout: 724 /* 127/8 must not appear on wire - RFC1122. */ 725 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 726 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 727 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 728 ipstat.ips_badaddr++; 729 error = EADDRNOTAVAIL; 730 goto bad; 731 } 732 } 733 734 m->m_pkthdr.csum_flags |= CSUM_IP; 735 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 736 if (sw_csum & CSUM_DELAY_DATA) { 737 in_delayed_cksum(m); 738 sw_csum &= ~CSUM_DELAY_DATA; 739 } 740 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 741 742 /* 743 * If small enough for interface, or the interface will take 744 * care of the fragmentation for us, can just send directly. 745 */ 746 if (ip->ip_len <= ifp->if_mtu || (ifp->if_hwassist & CSUM_FRAGMENT && 747 ((ip->ip_off & IP_DF) == 0))) { 748 ip->ip_len = htons(ip->ip_len); 749 ip->ip_off = htons(ip->ip_off); 750 ip->ip_sum = 0; 751 if (sw_csum & CSUM_DELAY_IP) 752 ip->ip_sum = in_cksum(m, hlen); 753 754 /* Record statistics for this interface address. */ 755 if (!(flags & IP_FORWARDING) && ia) { 756 ia->ia_ifa.if_opackets++; 757 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 758 } 759 760#ifdef IPSEC 761 /* clean ipsec history once it goes out of the node */ 762 ipsec_delaux(m); 763#endif 764 765#ifdef MBUF_STRESS_TEST 766 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 767 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 768#endif 769 error = (*ifp->if_output)(ifp, m, 770 (struct sockaddr *)dst, ro->ro_rt); 771 goto done; 772 } 773 774 if (ip->ip_off & IP_DF) { 775 error = EMSGSIZE; 776 /* 777 * This case can happen if the user changed the MTU 778 * of an interface after enabling IP on it. Because 779 * most netifs don't keep track of routes pointing to 780 * them, there is no way for one to update all its 781 * routes when the MTU is changed. 782 */ 783 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 784 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 785 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 786 } 787 ipstat.ips_cantfrag++; 788 goto bad; 789 } 790 791 /* 792 * Too large for interface; fragment if possible. If successful, 793 * on return, m will point to a list of packets to be sent. 794 */ 795 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 796 if (error) 797 goto bad; 798 for (; m; m = m0) { 799 m0 = m->m_nextpkt; 800 m->m_nextpkt = 0; 801#ifdef IPSEC 802 /* clean ipsec history once it goes out of the node */ 803 ipsec_delaux(m); 804#endif 805 if (error == 0) { 806 /* Record statistics for this interface address. */ 807 if (ia != NULL) { 808 ia->ia_ifa.if_opackets++; 809 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 810 } 811 812 error = (*ifp->if_output)(ifp, m, 813 (struct sockaddr *)dst, ro->ro_rt); 814 } else 815 m_freem(m); 816 } 817 818 if (error == 0) 819 ipstat.ips_fragmented++; 820 821done: 822 if (ro == &iproute && ro->ro_rt) { 823 RTFREE(ro->ro_rt); 824 } 825#ifdef IPSEC 826 if (sp != NULL) { 827 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 828 printf("DP ip_output call free SP:%p\n", sp)); 829 key_freesp(sp); 830 } 831#endif 832#ifdef FAST_IPSEC 833 if (sp != NULL) 834 KEY_FREESP(&sp); 835#endif 836 return (error); 837bad: 838 m_freem(m); 839 goto done; 840} 841 842/* 843 * Create a chain of fragments which fit the given mtu. m_frag points to the 844 * mbuf to be fragmented; on return it points to the chain with the fragments. 845 * Return 0 if no error. If error, m_frag may contain a partially built 846 * chain of fragments that should be freed by the caller. 847 * 848 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 849 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 850 */ 851int 852ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 853 u_long if_hwassist_flags, int sw_csum) 854{ 855 int error = 0; 856 int hlen = ip->ip_hl << 2; 857 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 858 int off; 859 struct mbuf *m0 = *m_frag; /* the original packet */ 860 int firstlen; 861 struct mbuf **mnext; 862 int nfrags; 863 864 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 865 ipstat.ips_cantfrag++; 866 return EMSGSIZE; 867 } 868 869 /* 870 * Must be able to put at least 8 bytes per fragment. 871 */ 872 if (len < 8) 873 return EMSGSIZE; 874 875 /* 876 * If the interface will not calculate checksums on 877 * fragmented packets, then do it here. 878 */ 879 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 880 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 881 in_delayed_cksum(m0); 882 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 883 } 884 885 if (len > PAGE_SIZE) { 886 /* 887 * Fragment large datagrams such that each segment 888 * contains a multiple of PAGE_SIZE amount of data, 889 * plus headers. This enables a receiver to perform 890 * page-flipping zero-copy optimizations. 891 * 892 * XXX When does this help given that sender and receiver 893 * could have different page sizes, and also mtu could 894 * be less than the receiver's page size ? 895 */ 896 int newlen; 897 struct mbuf *m; 898 899 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 900 off += m->m_len; 901 902 /* 903 * firstlen (off - hlen) must be aligned on an 904 * 8-byte boundary 905 */ 906 if (off < hlen) 907 goto smart_frag_failure; 908 off = ((off - hlen) & ~7) + hlen; 909 newlen = (~PAGE_MASK) & mtu; 910 if ((newlen + sizeof (struct ip)) > mtu) { 911 /* we failed, go back the default */ 912smart_frag_failure: 913 newlen = len; 914 off = hlen + len; 915 } 916 len = newlen; 917 918 } else { 919 off = hlen + len; 920 } 921 922 firstlen = off - hlen; 923 mnext = &m0->m_nextpkt; /* pointer to next packet */ 924 925 /* 926 * Loop through length of segment after first fragment, 927 * make new header and copy data of each part and link onto chain. 928 * Here, m0 is the original packet, m is the fragment being created. 929 * The fragments are linked off the m_nextpkt of the original 930 * packet, which after processing serves as the first fragment. 931 */ 932 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 933 struct ip *mhip; /* ip header on the fragment */ 934 struct mbuf *m; 935 int mhlen = sizeof (struct ip); 936 937 MGETHDR(m, M_DONTWAIT, MT_HEADER); 938 if (m == NULL) { 939 error = ENOBUFS; 940 ipstat.ips_odropped++; 941 goto done; 942 } 943 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 944 /* 945 * In the first mbuf, leave room for the link header, then 946 * copy the original IP header including options. The payload 947 * goes into an additional mbuf chain returned by m_copy(). 948 */ 949 m->m_data += max_linkhdr; 950 mhip = mtod(m, struct ip *); 951 *mhip = *ip; 952 if (hlen > sizeof (struct ip)) { 953 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 954 mhip->ip_v = IPVERSION; 955 mhip->ip_hl = mhlen >> 2; 956 } 957 m->m_len = mhlen; 958 /* XXX do we need to add ip->ip_off below ? */ 959 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 960 if (off + len >= ip->ip_len) { /* last fragment */ 961 len = ip->ip_len - off; 962 m->m_flags |= M_LASTFRAG; 963 } else 964 mhip->ip_off |= IP_MF; 965 mhip->ip_len = htons((u_short)(len + mhlen)); 966 m->m_next = m_copy(m0, off, len); 967 if (m->m_next == NULL) { /* copy failed */ 968 m_free(m); 969 error = ENOBUFS; /* ??? */ 970 ipstat.ips_odropped++; 971 goto done; 972 } 973 m->m_pkthdr.len = mhlen + len; 974 m->m_pkthdr.rcvif = (struct ifnet *)0; 975#ifdef MAC 976 mac_create_fragment(m0, m); 977#endif 978 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 979 mhip->ip_off = htons(mhip->ip_off); 980 mhip->ip_sum = 0; 981 if (sw_csum & CSUM_DELAY_IP) 982 mhip->ip_sum = in_cksum(m, mhlen); 983 *mnext = m; 984 mnext = &m->m_nextpkt; 985 } 986 ipstat.ips_ofragments += nfrags; 987 988 /* set first marker for fragment chain */ 989 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 990 m0->m_pkthdr.csum_data = nfrags; 991 992 /* 993 * Update first fragment by trimming what's been copied out 994 * and updating header. 995 */ 996 m_adj(m0, hlen + firstlen - ip->ip_len); 997 m0->m_pkthdr.len = hlen + firstlen; 998 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 999 ip->ip_off |= IP_MF; 1000 ip->ip_off = htons(ip->ip_off); 1001 ip->ip_sum = 0; 1002 if (sw_csum & CSUM_DELAY_IP) 1003 ip->ip_sum = in_cksum(m0, hlen); 1004 1005done: 1006 *m_frag = m0; 1007 return error; 1008} 1009 1010void 1011in_delayed_cksum(struct mbuf *m) 1012{ 1013 struct ip *ip; 1014 u_short csum, offset; 1015 1016 ip = mtod(m, struct ip *); 1017 offset = ip->ip_hl << 2 ; 1018 csum = in_cksum_skip(m, ip->ip_len, offset); 1019 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1020 csum = 0xffff; 1021 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1022 1023 if (offset + sizeof(u_short) > m->m_len) { 1024 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1025 m->m_len, offset, ip->ip_p); 1026 /* 1027 * XXX 1028 * this shouldn't happen, but if it does, the 1029 * correct behavior may be to insert the checksum 1030 * in the existing chain instead of rearranging it. 1031 */ 1032 m = m_pullup(m, offset + sizeof(u_short)); 1033 } 1034 *(u_short *)(m->m_data + offset) = csum; 1035} 1036 1037/* 1038 * Insert IP options into preformed packet. 1039 * Adjust IP destination as required for IP source routing, 1040 * as indicated by a non-zero in_addr at the start of the options. 1041 * 1042 * XXX This routine assumes that the packet has no options in place. 1043 */ 1044static struct mbuf * 1045ip_insertoptions(m, opt, phlen) 1046 register struct mbuf *m; 1047 struct mbuf *opt; 1048 int *phlen; 1049{ 1050 register struct ipoption *p = mtod(opt, struct ipoption *); 1051 struct mbuf *n; 1052 register struct ip *ip = mtod(m, struct ip *); 1053 unsigned optlen; 1054 1055 optlen = opt->m_len - sizeof(p->ipopt_dst); 1056 if (optlen + ip->ip_len > IP_MAXPACKET) { 1057 *phlen = 0; 1058 return (m); /* XXX should fail */ 1059 } 1060 if (p->ipopt_dst.s_addr) 1061 ip->ip_dst = p->ipopt_dst; 1062 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1063 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1064 if (n == NULL) { 1065 *phlen = 0; 1066 return (m); 1067 } 1068 n->m_pkthdr.rcvif = (struct ifnet *)0; 1069#ifdef MAC 1070 mac_create_mbuf_from_mbuf(m, n); 1071#endif 1072 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1073 m->m_len -= sizeof(struct ip); 1074 m->m_data += sizeof(struct ip); 1075 n->m_next = m; 1076 m = n; 1077 m->m_len = optlen + sizeof(struct ip); 1078 m->m_data += max_linkhdr; 1079 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1080 } else { 1081 m->m_data -= optlen; 1082 m->m_len += optlen; 1083 m->m_pkthdr.len += optlen; 1084 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1085 } 1086 ip = mtod(m, struct ip *); 1087 bcopy(p->ipopt_list, ip + 1, optlen); 1088 *phlen = sizeof(struct ip) + optlen; 1089 ip->ip_v = IPVERSION; 1090 ip->ip_hl = *phlen >> 2; 1091 ip->ip_len += optlen; 1092 return (m); 1093} 1094 1095/* 1096 * Copy options from ip to jp, 1097 * omitting those not copied during fragmentation. 1098 */ 1099int 1100ip_optcopy(ip, jp) 1101 struct ip *ip, *jp; 1102{ 1103 register u_char *cp, *dp; 1104 int opt, optlen, cnt; 1105 1106 cp = (u_char *)(ip + 1); 1107 dp = (u_char *)(jp + 1); 1108 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1109 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1110 opt = cp[0]; 1111 if (opt == IPOPT_EOL) 1112 break; 1113 if (opt == IPOPT_NOP) { 1114 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1115 *dp++ = IPOPT_NOP; 1116 optlen = 1; 1117 continue; 1118 } 1119 1120 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1121 ("ip_optcopy: malformed ipv4 option")); 1122 optlen = cp[IPOPT_OLEN]; 1123 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1124 ("ip_optcopy: malformed ipv4 option")); 1125 1126 /* bogus lengths should have been caught by ip_dooptions */ 1127 if (optlen > cnt) 1128 optlen = cnt; 1129 if (IPOPT_COPIED(opt)) { 1130 bcopy(cp, dp, optlen); 1131 dp += optlen; 1132 } 1133 } 1134 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1135 *dp++ = IPOPT_EOL; 1136 return (optlen); 1137} 1138 1139/* 1140 * IP socket option processing. 1141 */ 1142int 1143ip_ctloutput(so, sopt) 1144 struct socket *so; 1145 struct sockopt *sopt; 1146{ 1147 struct inpcb *inp = sotoinpcb(so); 1148 int error, optval; 1149 1150 error = optval = 0; 1151 if (sopt->sopt_level != IPPROTO_IP) { 1152 return (EINVAL); 1153 } 1154 1155 switch (sopt->sopt_dir) { 1156 case SOPT_SET: 1157 switch (sopt->sopt_name) { 1158 case IP_OPTIONS: 1159#ifdef notyet 1160 case IP_RETOPTS: 1161#endif 1162 { 1163 struct mbuf *m; 1164 if (sopt->sopt_valsize > MLEN) { 1165 error = EMSGSIZE; 1166 break; 1167 } 1168 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1169 if (m == NULL) { 1170 error = ENOBUFS; 1171 break; 1172 } 1173 m->m_len = sopt->sopt_valsize; 1174 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1175 m->m_len); 1176 INP_LOCK(inp); 1177 error = ip_pcbopts(inp, sopt->sopt_name, m); 1178 INP_UNLOCK(inp); 1179 return (error); 1180 } 1181 1182 case IP_TOS: 1183 case IP_TTL: 1184 case IP_RECVOPTS: 1185 case IP_RECVRETOPTS: 1186 case IP_RECVDSTADDR: 1187 case IP_RECVTTL: 1188 case IP_RECVIF: 1189 case IP_FAITH: 1190 case IP_ONESBCAST: 1191 error = sooptcopyin(sopt, &optval, sizeof optval, 1192 sizeof optval); 1193 if (error) 1194 break; 1195 1196 switch (sopt->sopt_name) { 1197 case IP_TOS: 1198 inp->inp_ip_tos = optval; 1199 break; 1200 1201 case IP_TTL: 1202 inp->inp_ip_ttl = optval; 1203 break; 1204#define OPTSET(bit) do { \ 1205 INP_LOCK(inp); \ 1206 if (optval) \ 1207 inp->inp_flags |= bit; \ 1208 else \ 1209 inp->inp_flags &= ~bit; \ 1210 INP_UNLOCK(inp); \ 1211} while (0) 1212 1213 case IP_RECVOPTS: 1214 OPTSET(INP_RECVOPTS); 1215 break; 1216 1217 case IP_RECVRETOPTS: 1218 OPTSET(INP_RECVRETOPTS); 1219 break; 1220 1221 case IP_RECVDSTADDR: 1222 OPTSET(INP_RECVDSTADDR); 1223 break; 1224 1225 case IP_RECVTTL: 1226 OPTSET(INP_RECVTTL); 1227 break; 1228 1229 case IP_RECVIF: 1230 OPTSET(INP_RECVIF); 1231 break; 1232 1233 case IP_FAITH: 1234 OPTSET(INP_FAITH); 1235 break; 1236 1237 case IP_ONESBCAST: 1238 OPTSET(INP_ONESBCAST); 1239 break; 1240 } 1241 break; 1242#undef OPTSET 1243 1244 case IP_MULTICAST_IF: 1245 case IP_MULTICAST_VIF: 1246 case IP_MULTICAST_TTL: 1247 case IP_MULTICAST_LOOP: 1248 case IP_ADD_MEMBERSHIP: 1249 case IP_DROP_MEMBERSHIP: 1250 error = ip_setmoptions(inp, sopt); 1251 break; 1252 1253 case IP_PORTRANGE: 1254 error = sooptcopyin(sopt, &optval, sizeof optval, 1255 sizeof optval); 1256 if (error) 1257 break; 1258 1259 INP_LOCK(inp); 1260 switch (optval) { 1261 case IP_PORTRANGE_DEFAULT: 1262 inp->inp_flags &= ~(INP_LOWPORT); 1263 inp->inp_flags &= ~(INP_HIGHPORT); 1264 break; 1265 1266 case IP_PORTRANGE_HIGH: 1267 inp->inp_flags &= ~(INP_LOWPORT); 1268 inp->inp_flags |= INP_HIGHPORT; 1269 break; 1270 1271 case IP_PORTRANGE_LOW: 1272 inp->inp_flags &= ~(INP_HIGHPORT); 1273 inp->inp_flags |= INP_LOWPORT; 1274 break; 1275 1276 default: 1277 error = EINVAL; 1278 break; 1279 } 1280 INP_UNLOCK(inp); 1281 break; 1282 1283#if defined(IPSEC) || defined(FAST_IPSEC) 1284 case IP_IPSEC_POLICY: 1285 { 1286 caddr_t req; 1287 size_t len = 0; 1288 int priv; 1289 struct mbuf *m; 1290 int optname; 1291 1292 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1293 break; 1294 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1295 break; 1296 priv = (sopt->sopt_td != NULL && 1297 suser(sopt->sopt_td) != 0) ? 0 : 1; 1298 req = mtod(m, caddr_t); 1299 len = m->m_len; 1300 optname = sopt->sopt_name; 1301 error = ipsec4_set_policy(inp, optname, req, len, priv); 1302 m_freem(m); 1303 break; 1304 } 1305#endif /*IPSEC*/ 1306 1307 default: 1308 error = ENOPROTOOPT; 1309 break; 1310 } 1311 break; 1312 1313 case SOPT_GET: 1314 switch (sopt->sopt_name) { 1315 case IP_OPTIONS: 1316 case IP_RETOPTS: 1317 if (inp->inp_options) 1318 error = sooptcopyout(sopt, 1319 mtod(inp->inp_options, 1320 char *), 1321 inp->inp_options->m_len); 1322 else 1323 sopt->sopt_valsize = 0; 1324 break; 1325 1326 case IP_TOS: 1327 case IP_TTL: 1328 case IP_RECVOPTS: 1329 case IP_RECVRETOPTS: 1330 case IP_RECVDSTADDR: 1331 case IP_RECVTTL: 1332 case IP_RECVIF: 1333 case IP_PORTRANGE: 1334 case IP_FAITH: 1335 case IP_ONESBCAST: 1336 switch (sopt->sopt_name) { 1337 1338 case IP_TOS: 1339 optval = inp->inp_ip_tos; 1340 break; 1341 1342 case IP_TTL: 1343 optval = inp->inp_ip_ttl; 1344 break; 1345 1346#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1347 1348 case IP_RECVOPTS: 1349 optval = OPTBIT(INP_RECVOPTS); 1350 break; 1351 1352 case IP_RECVRETOPTS: 1353 optval = OPTBIT(INP_RECVRETOPTS); 1354 break; 1355 1356 case IP_RECVDSTADDR: 1357 optval = OPTBIT(INP_RECVDSTADDR); 1358 break; 1359 1360 case IP_RECVTTL: 1361 optval = OPTBIT(INP_RECVTTL); 1362 break; 1363 1364 case IP_RECVIF: 1365 optval = OPTBIT(INP_RECVIF); 1366 break; 1367 1368 case IP_PORTRANGE: 1369 if (inp->inp_flags & INP_HIGHPORT) 1370 optval = IP_PORTRANGE_HIGH; 1371 else if (inp->inp_flags & INP_LOWPORT) 1372 optval = IP_PORTRANGE_LOW; 1373 else 1374 optval = 0; 1375 break; 1376 1377 case IP_FAITH: 1378 optval = OPTBIT(INP_FAITH); 1379 break; 1380 1381 case IP_ONESBCAST: 1382 optval = OPTBIT(INP_ONESBCAST); 1383 break; 1384 } 1385 error = sooptcopyout(sopt, &optval, sizeof optval); 1386 break; 1387 1388 case IP_MULTICAST_IF: 1389 case IP_MULTICAST_VIF: 1390 case IP_MULTICAST_TTL: 1391 case IP_MULTICAST_LOOP: 1392 case IP_ADD_MEMBERSHIP: 1393 case IP_DROP_MEMBERSHIP: 1394 error = ip_getmoptions(inp, sopt); 1395 break; 1396 1397#if defined(IPSEC) || defined(FAST_IPSEC) 1398 case IP_IPSEC_POLICY: 1399 { 1400 struct mbuf *m = NULL; 1401 caddr_t req = NULL; 1402 size_t len = 0; 1403 1404 if (m != 0) { 1405 req = mtod(m, caddr_t); 1406 len = m->m_len; 1407 } 1408 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1409 if (error == 0) 1410 error = soopt_mcopyout(sopt, m); /* XXX */ 1411 if (error == 0) 1412 m_freem(m); 1413 break; 1414 } 1415#endif /*IPSEC*/ 1416 1417 default: 1418 error = ENOPROTOOPT; 1419 break; 1420 } 1421 break; 1422 } 1423 return (error); 1424} 1425 1426/* 1427 * Set up IP options in pcb for insertion in output packets. 1428 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1429 * with destination address if source routed. 1430 */ 1431static int 1432ip_pcbopts(struct inpcb *inp, int optname, struct mbuf *m) 1433{ 1434 register int cnt, optlen; 1435 register u_char *cp; 1436 struct mbuf **pcbopt; 1437 u_char opt; 1438 1439 INP_LOCK_ASSERT(inp); 1440 1441 pcbopt = &inp->inp_options; 1442 1443 /* turn off any old options */ 1444 if (*pcbopt) 1445 (void)m_free(*pcbopt); 1446 *pcbopt = 0; 1447 if (m == NULL || m->m_len == 0) { 1448 /* 1449 * Only turning off any previous options. 1450 */ 1451 if (m != NULL) 1452 (void)m_free(m); 1453 return (0); 1454 } 1455 1456 if (m->m_len % sizeof(int32_t)) 1457 goto bad; 1458 /* 1459 * IP first-hop destination address will be stored before 1460 * actual options; move other options back 1461 * and clear it when none present. 1462 */ 1463 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1464 goto bad; 1465 cnt = m->m_len; 1466 m->m_len += sizeof(struct in_addr); 1467 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1468 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1469 bzero(mtod(m, void *), sizeof(struct in_addr)); 1470 1471 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1472 opt = cp[IPOPT_OPTVAL]; 1473 if (opt == IPOPT_EOL) 1474 break; 1475 if (opt == IPOPT_NOP) 1476 optlen = 1; 1477 else { 1478 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1479 goto bad; 1480 optlen = cp[IPOPT_OLEN]; 1481 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1482 goto bad; 1483 } 1484 switch (opt) { 1485 1486 default: 1487 break; 1488 1489 case IPOPT_LSRR: 1490 case IPOPT_SSRR: 1491 /* 1492 * user process specifies route as: 1493 * ->A->B->C->D 1494 * D must be our final destination (but we can't 1495 * check that since we may not have connected yet). 1496 * A is first hop destination, which doesn't appear in 1497 * actual IP option, but is stored before the options. 1498 */ 1499 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1500 goto bad; 1501 m->m_len -= sizeof(struct in_addr); 1502 cnt -= sizeof(struct in_addr); 1503 optlen -= sizeof(struct in_addr); 1504 cp[IPOPT_OLEN] = optlen; 1505 /* 1506 * Move first hop before start of options. 1507 */ 1508 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1509 sizeof(struct in_addr)); 1510 /* 1511 * Then copy rest of options back 1512 * to close up the deleted entry. 1513 */ 1514 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1515 &cp[IPOPT_OFFSET+1], 1516 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1517 break; 1518 } 1519 } 1520 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1521 goto bad; 1522 *pcbopt = m; 1523 return (0); 1524 1525bad: 1526 (void)m_free(m); 1527 return (EINVAL); 1528} 1529 1530/* 1531 * XXX 1532 * The whole multicast option thing needs to be re-thought. 1533 * Several of these options are equally applicable to non-multicast 1534 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1535 * standard option (IP_TTL). 1536 */ 1537 1538/* 1539 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1540 */ 1541static struct ifnet * 1542ip_multicast_if(a, ifindexp) 1543 struct in_addr *a; 1544 int *ifindexp; 1545{ 1546 int ifindex; 1547 struct ifnet *ifp; 1548 1549 if (ifindexp) 1550 *ifindexp = 0; 1551 if (ntohl(a->s_addr) >> 24 == 0) { 1552 ifindex = ntohl(a->s_addr) & 0xffffff; 1553 if (ifindex < 0 || if_index < ifindex) 1554 return NULL; 1555 ifp = ifnet_byindex(ifindex); 1556 if (ifindexp) 1557 *ifindexp = ifindex; 1558 } else { 1559 INADDR_TO_IFP(*a, ifp); 1560 } 1561 return ifp; 1562} 1563 1564/* 1565 * Set the IP multicast options in response to user setsockopt(). 1566 */ 1567static int 1568ip_setmoptions(struct inpcb *inp, struct sockopt *sopt) 1569{ 1570 int error = 0; 1571 int i; 1572 struct in_addr addr; 1573 struct ip_mreq mreq; 1574 struct ifnet *ifp; 1575 struct ip_moptions *imo; 1576 struct route ro; 1577 struct sockaddr_in *dst; 1578 int ifindex; 1579 int s; 1580 1581 imo = inp->inp_moptions; 1582 if (imo == NULL) { 1583 /* 1584 * No multicast option buffer attached to the pcb; 1585 * allocate one and initialize to default values. 1586 */ 1587 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1588 M_WAITOK); 1589 1590 if (imo == NULL) 1591 return (ENOBUFS); 1592 inp->inp_moptions = imo; 1593 imo->imo_multicast_ifp = NULL; 1594 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1595 imo->imo_multicast_vif = -1; 1596 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1597 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1598 imo->imo_num_memberships = 0; 1599 } 1600 1601 switch (sopt->sopt_name) { 1602 /* store an index number for the vif you wanna use in the send */ 1603 case IP_MULTICAST_VIF: 1604 if (legal_vif_num == 0) { 1605 error = EOPNOTSUPP; 1606 break; 1607 } 1608 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1609 if (error) 1610 break; 1611 if (!legal_vif_num(i) && (i != -1)) { 1612 error = EINVAL; 1613 break; 1614 } 1615 imo->imo_multicast_vif = i; 1616 break; 1617 1618 case IP_MULTICAST_IF: 1619 /* 1620 * Select the interface for outgoing multicast packets. 1621 */ 1622 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1623 if (error) 1624 break; 1625 /* 1626 * INADDR_ANY is used to remove a previous selection. 1627 * When no interface is selected, a default one is 1628 * chosen every time a multicast packet is sent. 1629 */ 1630 if (addr.s_addr == INADDR_ANY) { 1631 imo->imo_multicast_ifp = NULL; 1632 break; 1633 } 1634 /* 1635 * The selected interface is identified by its local 1636 * IP address. Find the interface and confirm that 1637 * it supports multicasting. 1638 */ 1639 s = splimp(); 1640 ifp = ip_multicast_if(&addr, &ifindex); 1641 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1642 splx(s); 1643 error = EADDRNOTAVAIL; 1644 break; 1645 } 1646 imo->imo_multicast_ifp = ifp; 1647 if (ifindex) 1648 imo->imo_multicast_addr = addr; 1649 else 1650 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1651 splx(s); 1652 break; 1653 1654 case IP_MULTICAST_TTL: 1655 /* 1656 * Set the IP time-to-live for outgoing multicast packets. 1657 * The original multicast API required a char argument, 1658 * which is inconsistent with the rest of the socket API. 1659 * We allow either a char or an int. 1660 */ 1661 if (sopt->sopt_valsize == 1) { 1662 u_char ttl; 1663 error = sooptcopyin(sopt, &ttl, 1, 1); 1664 if (error) 1665 break; 1666 imo->imo_multicast_ttl = ttl; 1667 } else { 1668 u_int ttl; 1669 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1670 sizeof ttl); 1671 if (error) 1672 break; 1673 if (ttl > 255) 1674 error = EINVAL; 1675 else 1676 imo->imo_multicast_ttl = ttl; 1677 } 1678 break; 1679 1680 case IP_MULTICAST_LOOP: 1681 /* 1682 * Set the loopback flag for outgoing multicast packets. 1683 * Must be zero or one. The original multicast API required a 1684 * char argument, which is inconsistent with the rest 1685 * of the socket API. We allow either a char or an int. 1686 */ 1687 if (sopt->sopt_valsize == 1) { 1688 u_char loop; 1689 error = sooptcopyin(sopt, &loop, 1, 1); 1690 if (error) 1691 break; 1692 imo->imo_multicast_loop = !!loop; 1693 } else { 1694 u_int loop; 1695 error = sooptcopyin(sopt, &loop, sizeof loop, 1696 sizeof loop); 1697 if (error) 1698 break; 1699 imo->imo_multicast_loop = !!loop; 1700 } 1701 break; 1702 1703 case IP_ADD_MEMBERSHIP: 1704 /* 1705 * Add a multicast group membership. 1706 * Group must be a valid IP multicast address. 1707 */ 1708 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1709 if (error) 1710 break; 1711 1712 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1713 error = EINVAL; 1714 break; 1715 } 1716 s = splimp(); 1717 /* 1718 * If no interface address was provided, use the interface of 1719 * the route to the given multicast address. 1720 */ 1721 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1722 bzero((caddr_t)&ro, sizeof(ro)); 1723 dst = (struct sockaddr_in *)&ro.ro_dst; 1724 dst->sin_len = sizeof(*dst); 1725 dst->sin_family = AF_INET; 1726 dst->sin_addr = mreq.imr_multiaddr; 1727 rtalloc_ign(&ro, RTF_CLONING); 1728 if (ro.ro_rt == NULL) { 1729 error = EADDRNOTAVAIL; 1730 splx(s); 1731 break; 1732 } 1733 ifp = ro.ro_rt->rt_ifp; 1734 RTFREE(ro.ro_rt); 1735 } 1736 else { 1737 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1738 } 1739 1740 /* 1741 * See if we found an interface, and confirm that it 1742 * supports multicast. 1743 */ 1744 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1745 error = EADDRNOTAVAIL; 1746 splx(s); 1747 break; 1748 } 1749 /* 1750 * See if the membership already exists or if all the 1751 * membership slots are full. 1752 */ 1753 for (i = 0; i < imo->imo_num_memberships; ++i) { 1754 if (imo->imo_membership[i]->inm_ifp == ifp && 1755 imo->imo_membership[i]->inm_addr.s_addr 1756 == mreq.imr_multiaddr.s_addr) 1757 break; 1758 } 1759 if (i < imo->imo_num_memberships) { 1760 error = EADDRINUSE; 1761 splx(s); 1762 break; 1763 } 1764 if (i == IP_MAX_MEMBERSHIPS) { 1765 error = ETOOMANYREFS; 1766 splx(s); 1767 break; 1768 } 1769 /* 1770 * Everything looks good; add a new record to the multicast 1771 * address list for the given interface. 1772 */ 1773 if ((imo->imo_membership[i] = 1774 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1775 error = ENOBUFS; 1776 splx(s); 1777 break; 1778 } 1779 ++imo->imo_num_memberships; 1780 splx(s); 1781 break; 1782 1783 case IP_DROP_MEMBERSHIP: 1784 /* 1785 * Drop a multicast group membership. 1786 * Group must be a valid IP multicast address. 1787 */ 1788 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1789 if (error) 1790 break; 1791 1792 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1793 error = EINVAL; 1794 break; 1795 } 1796 1797 s = splimp(); 1798 /* 1799 * If an interface address was specified, get a pointer 1800 * to its ifnet structure. 1801 */ 1802 if (mreq.imr_interface.s_addr == INADDR_ANY) 1803 ifp = NULL; 1804 else { 1805 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1806 if (ifp == NULL) { 1807 error = EADDRNOTAVAIL; 1808 splx(s); 1809 break; 1810 } 1811 } 1812 /* 1813 * Find the membership in the membership array. 1814 */ 1815 for (i = 0; i < imo->imo_num_memberships; ++i) { 1816 if ((ifp == NULL || 1817 imo->imo_membership[i]->inm_ifp == ifp) && 1818 imo->imo_membership[i]->inm_addr.s_addr == 1819 mreq.imr_multiaddr.s_addr) 1820 break; 1821 } 1822 if (i == imo->imo_num_memberships) { 1823 error = EADDRNOTAVAIL; 1824 splx(s); 1825 break; 1826 } 1827 /* 1828 * Give up the multicast address record to which the 1829 * membership points. 1830 */ 1831 in_delmulti(imo->imo_membership[i]); 1832 /* 1833 * Remove the gap in the membership array. 1834 */ 1835 for (++i; i < imo->imo_num_memberships; ++i) 1836 imo->imo_membership[i-1] = imo->imo_membership[i]; 1837 --imo->imo_num_memberships; 1838 splx(s); 1839 break; 1840 1841 default: 1842 error = EOPNOTSUPP; 1843 break; 1844 } 1845 1846 /* 1847 * If all options have default values, no need to keep the mbuf. 1848 */ 1849 if (imo->imo_multicast_ifp == NULL && 1850 imo->imo_multicast_vif == -1 && 1851 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1852 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1853 imo->imo_num_memberships == 0) { 1854 free(inp->inp_moptions, M_IPMOPTS); 1855 inp->inp_moptions = NULL; 1856 } 1857 1858 return (error); 1859} 1860 1861/* 1862 * Return the IP multicast options in response to user getsockopt(). 1863 */ 1864static int 1865ip_getmoptions(struct inpcb *inp, struct sockopt *sopt) 1866{ 1867 struct ip_moptions *imo; 1868 struct in_addr addr; 1869 struct in_ifaddr *ia; 1870 int error, optval; 1871 u_char coptval; 1872 1873 INP_LOCK(inp); 1874 imo = inp->inp_moptions; 1875 1876 error = 0; 1877 switch (sopt->sopt_name) { 1878 case IP_MULTICAST_VIF: 1879 if (imo != NULL) 1880 optval = imo->imo_multicast_vif; 1881 else 1882 optval = -1; 1883 INP_UNLOCK(inp); 1884 error = sooptcopyout(sopt, &optval, sizeof optval); 1885 break; 1886 1887 case IP_MULTICAST_IF: 1888 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1889 addr.s_addr = INADDR_ANY; 1890 else if (imo->imo_multicast_addr.s_addr) { 1891 /* return the value user has set */ 1892 addr = imo->imo_multicast_addr; 1893 } else { 1894 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1895 addr.s_addr = (ia == NULL) ? INADDR_ANY 1896 : IA_SIN(ia)->sin_addr.s_addr; 1897 } 1898 INP_UNLOCK(inp); 1899 error = sooptcopyout(sopt, &addr, sizeof addr); 1900 break; 1901 1902 case IP_MULTICAST_TTL: 1903 if (imo == 0) 1904 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1905 else 1906 optval = coptval = imo->imo_multicast_ttl; 1907 INP_UNLOCK(inp); 1908 if (sopt->sopt_valsize == 1) 1909 error = sooptcopyout(sopt, &coptval, 1); 1910 else 1911 error = sooptcopyout(sopt, &optval, sizeof optval); 1912 break; 1913 1914 case IP_MULTICAST_LOOP: 1915 if (imo == 0) 1916 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1917 else 1918 optval = coptval = imo->imo_multicast_loop; 1919 INP_UNLOCK(inp); 1920 if (sopt->sopt_valsize == 1) 1921 error = sooptcopyout(sopt, &coptval, 1); 1922 else 1923 error = sooptcopyout(sopt, &optval, sizeof optval); 1924 break; 1925 1926 default: 1927 INP_UNLOCK(inp); 1928 error = ENOPROTOOPT; 1929 break; 1930 } 1931 INP_UNLOCK_ASSERT(inp); 1932 1933 return (error); 1934} 1935 1936/* 1937 * Discard the IP multicast options. 1938 */ 1939void 1940ip_freemoptions(imo) 1941 register struct ip_moptions *imo; 1942{ 1943 register int i; 1944 1945 if (imo != NULL) { 1946 for (i = 0; i < imo->imo_num_memberships; ++i) 1947 in_delmulti(imo->imo_membership[i]); 1948 free(imo, M_IPMOPTS); 1949 } 1950} 1951 1952/* 1953 * Routine called from ip_output() to loop back a copy of an IP multicast 1954 * packet to the input queue of a specified interface. Note that this 1955 * calls the output routine of the loopback "driver", but with an interface 1956 * pointer that might NOT be a loopback interface -- evil, but easier than 1957 * replicating that code here. 1958 */ 1959static void 1960ip_mloopback(ifp, m, dst, hlen) 1961 struct ifnet *ifp; 1962 register struct mbuf *m; 1963 register struct sockaddr_in *dst; 1964 int hlen; 1965{ 1966 register struct ip *ip; 1967 struct mbuf *copym; 1968 1969 copym = m_copy(m, 0, M_COPYALL); 1970 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1971 copym = m_pullup(copym, hlen); 1972 if (copym != NULL) { 1973 /* If needed, compute the checksum and mark it as valid. */ 1974 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1975 in_delayed_cksum(copym); 1976 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1977 copym->m_pkthdr.csum_flags |= 1978 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1979 copym->m_pkthdr.csum_data = 0xffff; 1980 } 1981 /* 1982 * We don't bother to fragment if the IP length is greater 1983 * than the interface's MTU. Can this possibly matter? 1984 */ 1985 ip = mtod(copym, struct ip *); 1986 ip->ip_len = htons(ip->ip_len); 1987 ip->ip_off = htons(ip->ip_off); 1988 ip->ip_sum = 0; 1989 ip->ip_sum = in_cksum(copym, hlen); 1990 /* 1991 * NB: 1992 * It's not clear whether there are any lingering 1993 * reentrancy problems in other areas which might 1994 * be exposed by using ip_input directly (in 1995 * particular, everything which modifies the packet 1996 * in-place). Yet another option is using the 1997 * protosw directly to deliver the looped back 1998 * packet. For the moment, we'll err on the side 1999 * of safety by using if_simloop(). 2000 */ 2001#if 1 /* XXX */ 2002 if (dst->sin_family != AF_INET) { 2003 printf("ip_mloopback: bad address family %d\n", 2004 dst->sin_family); 2005 dst->sin_family = AF_INET; 2006 } 2007#endif 2008 2009#ifdef notdef 2010 copym->m_pkthdr.rcvif = ifp; 2011 ip_input(copym); 2012#else 2013 if_simloop(ifp, copym, dst->sin_family, 0); 2014#endif 2015 } 2016} 2017