ip_output.c revision 142248
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 142248 2005-02-22 17:40:40Z andre $ 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#include <machine/in_cksum.h> 62 63static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 64 65#ifdef IPSEC 66#include <netinet6/ipsec.h> 67#include <netkey/key.h> 68#ifdef IPSEC_DEBUG 69#include <netkey/key_debug.h> 70#else 71#define KEYDEBUG(lev,arg) 72#endif 73#endif /*IPSEC*/ 74 75#ifdef FAST_IPSEC 76#include <netipsec/ipsec.h> 77#include <netipsec/xform.h> 78#include <netipsec/key.h> 79#endif /*FAST_IPSEC*/ 80 81#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 82 x, (ntohl(a.s_addr)>>24)&0xFF,\ 83 (ntohl(a.s_addr)>>16)&0xFF,\ 84 (ntohl(a.s_addr)>>8)&0xFF,\ 85 (ntohl(a.s_addr))&0xFF, y); 86 87u_short ip_id; 88 89#ifdef MBUF_STRESS_TEST 90int mbuf_frag_size = 0; 91SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 92 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 93#endif 94 95static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 96static struct ifnet *ip_multicast_if(struct in_addr *, int *); 97static void ip_mloopback 98 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 99static int ip_getmoptions(struct inpcb *, struct sockopt *); 100static int ip_pcbopts(struct inpcb *, int, struct mbuf *); 101static int ip_setmoptions(struct inpcb *, struct sockopt *); 102 103int ip_optcopy(struct ip *, struct ip *); 104 105 106extern struct protosw inetsw[]; 107 108/* 109 * IP output. The packet in mbuf chain m contains a skeletal IP 110 * header (with len, off, ttl, proto, tos, src, dst). 111 * The mbuf chain containing the packet will be freed. 112 * The mbuf opt, if present, will not be freed. 113 * In the IP forwarding case, the packet will arrive with options already 114 * inserted, so must have a NULL opt pointer. 115 */ 116int 117ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, 118 int flags, struct ip_moptions *imo, struct inpcb *inp) 119{ 120 struct ip *ip; 121 struct ifnet *ifp = NULL; /* keep compiler happy */ 122 struct mbuf *m0; 123 int hlen = sizeof (struct ip); 124 int len, error = 0; 125 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 126 struct in_ifaddr *ia = NULL; 127 int isbroadcast, sw_csum; 128 struct route iproute; 129 struct in_addr odst; 130#ifdef IPFIREWALL_FORWARD 131 struct m_tag *fwd_tag = NULL; 132#endif 133#ifdef IPSEC 134 struct secpolicy *sp = NULL; 135#endif 136#ifdef FAST_IPSEC 137 struct secpolicy *sp = NULL; 138 struct tdb_ident *tdbi; 139 struct m_tag *mtag; 140 int s; 141#endif /* FAST_IPSEC */ 142 143 M_ASSERTPKTHDR(m); 144 145 if (ro == NULL) { 146 ro = &iproute; 147 bzero(ro, sizeof (*ro)); 148 } 149 150 if (inp != NULL) 151 INP_LOCK_ASSERT(inp); 152 153 if (opt) { 154 len = 0; 155 m = ip_insertoptions(m, opt, &len); 156 if (len != 0) 157 hlen = len; 158 } 159 ip = mtod(m, struct ip *); 160 161 /* 162 * Fill in IP header. If we are not allowing fragmentation, 163 * then the ip_id field is meaningless, but we don't set it 164 * to zero. Doing so causes various problems when devices along 165 * the path (routers, load balancers, firewalls, etc.) illegally 166 * disable DF on our packet. Note that a 16-bit counter 167 * will wrap around in less than 10 seconds at 100 Mbit/s on a 168 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 169 * for Counting NATted Hosts", Proc. IMW'02, available at 170 * <http://www.research.att.com/~smb/papers/fnat.pdf>. 171 */ 172 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 173 ip->ip_v = IPVERSION; 174 ip->ip_hl = hlen >> 2; 175 ip->ip_id = ip_newid(); 176 ipstat.ips_localout++; 177 } else { 178 hlen = ip->ip_hl << 2; 179 } 180 181 dst = (struct sockaddr_in *)&ro->ro_dst; 182again: 183 /* 184 * If there is a cached route, 185 * check that it is to the same destination 186 * and is still up. If not, free it and try again. 187 * The address family should also be checked in case of sharing the 188 * cache with IPv6. 189 */ 190 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 191 dst->sin_family != AF_INET || 192 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 193 RTFREE(ro->ro_rt); 194 ro->ro_rt = (struct rtentry *)0; 195 } 196#ifdef IPFIREWALL_FORWARD 197 if (ro->ro_rt == NULL && fwd_tag == NULL) { 198#else 199 if (ro->ro_rt == NULL) { 200#endif 201 bzero(dst, sizeof(*dst)); 202 dst->sin_family = AF_INET; 203 dst->sin_len = sizeof(*dst); 204 dst->sin_addr = ip->ip_dst; 205 } 206 /* 207 * If routing to interface only, 208 * short circuit routing lookup. 209 */ 210 if (flags & IP_ROUTETOIF) { 211 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 212 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) { 213 ipstat.ips_noroute++; 214 error = ENETUNREACH; 215 goto bad; 216 } 217 ifp = ia->ia_ifp; 218 ip->ip_ttl = 1; 219 isbroadcast = in_broadcast(dst->sin_addr, ifp); 220 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 221 imo != NULL && imo->imo_multicast_ifp != NULL) { 222 /* 223 * Bypass the normal routing lookup for multicast 224 * packets if the interface is specified. 225 */ 226 ifp = imo->imo_multicast_ifp; 227 IFP_TO_IA(ifp, ia); 228 isbroadcast = 0; /* fool gcc */ 229 } else { 230 /* 231 * We want to do any cloning requested by the link layer, 232 * as this is probably required in all cases for correct 233 * operation (as it is for ARP). 234 */ 235 if (ro->ro_rt == NULL) 236 rtalloc_ign(ro, 0); 237 if (ro->ro_rt == NULL) { 238 ipstat.ips_noroute++; 239 error = EHOSTUNREACH; 240 goto bad; 241 } 242 ia = ifatoia(ro->ro_rt->rt_ifa); 243 ifp = ro->ro_rt->rt_ifp; 244 ro->ro_rt->rt_rmx.rmx_pksent++; 245 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 246 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 247 if (ro->ro_rt->rt_flags & RTF_HOST) 248 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 249 else 250 isbroadcast = in_broadcast(dst->sin_addr, ifp); 251 } 252 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 253 struct in_multi *inm; 254 255 m->m_flags |= M_MCAST; 256 /* 257 * IP destination address is multicast. Make sure "dst" 258 * still points to the address in "ro". (It may have been 259 * changed to point to a gateway address, above.) 260 */ 261 dst = (struct sockaddr_in *)&ro->ro_dst; 262 /* 263 * See if the caller provided any multicast options 264 */ 265 if (imo != NULL) { 266 ip->ip_ttl = imo->imo_multicast_ttl; 267 if (imo->imo_multicast_vif != -1) 268 ip->ip_src.s_addr = 269 ip_mcast_src ? 270 ip_mcast_src(imo->imo_multicast_vif) : 271 INADDR_ANY; 272 } else 273 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 274 /* 275 * Confirm that the outgoing interface supports multicast. 276 */ 277 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 278 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 279 ipstat.ips_noroute++; 280 error = ENETUNREACH; 281 goto bad; 282 } 283 } 284 /* 285 * If source address not specified yet, use address 286 * of outgoing interface. 287 */ 288 if (ip->ip_src.s_addr == INADDR_ANY) { 289 /* Interface may have no addresses. */ 290 if (ia != NULL) 291 ip->ip_src = IA_SIN(ia)->sin_addr; 292 } 293 294 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 295 if (inm != NULL && 296 (imo == NULL || imo->imo_multicast_loop)) { 297 /* 298 * If we belong to the destination multicast group 299 * on the outgoing interface, and the caller did not 300 * forbid loopback, loop back a copy. 301 */ 302 ip_mloopback(ifp, m, dst, hlen); 303 } 304 else { 305 /* 306 * If we are acting as a multicast router, perform 307 * multicast forwarding as if the packet had just 308 * arrived on the interface to which we are about 309 * to send. The multicast forwarding function 310 * recursively calls this function, using the 311 * IP_FORWARDING flag to prevent infinite recursion. 312 * 313 * Multicasts that are looped back by ip_mloopback(), 314 * above, will be forwarded by the ip_input() routine, 315 * if necessary. 316 */ 317 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 318 /* 319 * If rsvp daemon is not running, do not 320 * set ip_moptions. This ensures that the packet 321 * is multicast and not just sent down one link 322 * as prescribed by rsvpd. 323 */ 324 if (!rsvp_on) 325 imo = NULL; 326 if (ip_mforward && 327 ip_mforward(ip, ifp, m, imo) != 0) { 328 m_freem(m); 329 goto done; 330 } 331 } 332 } 333 334 /* 335 * Multicasts with a time-to-live of zero may be looped- 336 * back, above, but must not be transmitted on a network. 337 * Also, multicasts addressed to the loopback interface 338 * are not sent -- the above call to ip_mloopback() will 339 * loop back a copy if this host actually belongs to the 340 * destination group on the loopback interface. 341 */ 342 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 343 m_freem(m); 344 goto done; 345 } 346 347 goto sendit; 348 } 349#ifndef notdef 350 /* 351 * If the source address is not specified yet, use the address 352 * of the outoing interface. 353 */ 354 if (ip->ip_src.s_addr == INADDR_ANY) { 355 /* Interface may have no addresses. */ 356 if (ia != NULL) { 357 ip->ip_src = IA_SIN(ia)->sin_addr; 358 } 359 } 360#endif /* notdef */ 361 /* 362 * Verify that we have any chance at all of being able to queue the 363 * packet or packet fragments, unless ALTQ is enabled on the given 364 * interface in which case packetdrop should be done by queueing. 365 */ 366#ifdef ALTQ 367 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) && 368 ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 369 ifp->if_snd.ifq_maxlen)) 370#else 371 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 372 ifp->if_snd.ifq_maxlen) 373#endif /* ALTQ */ 374 { 375 error = ENOBUFS; 376 ipstat.ips_odropped++; 377 goto bad; 378 } 379 380 /* 381 * Look for broadcast address and 382 * verify user is allowed to send 383 * such a packet. 384 */ 385 if (isbroadcast) { 386 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 387 error = EADDRNOTAVAIL; 388 goto bad; 389 } 390 if ((flags & IP_ALLOWBROADCAST) == 0) { 391 error = EACCES; 392 goto bad; 393 } 394 /* don't allow broadcast messages to be fragmented */ 395 if (ip->ip_len > ifp->if_mtu) { 396 error = EMSGSIZE; 397 goto bad; 398 } 399 if (flags & IP_SENDONES) 400 ip->ip_dst.s_addr = INADDR_BROADCAST; 401 m->m_flags |= M_BCAST; 402 } else { 403 m->m_flags &= ~M_BCAST; 404 } 405 406sendit: 407#ifdef IPSEC 408 /* get SP for this packet */ 409 if (inp == NULL) 410 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 411 flags, &error); 412 else 413 sp = ipsec4_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error); 414 415 if (sp == NULL) { 416 ipsecstat.out_inval++; 417 goto bad; 418 } 419 420 error = 0; 421 422 /* check policy */ 423 switch (sp->policy) { 424 case IPSEC_POLICY_DISCARD: 425 /* 426 * This packet is just discarded. 427 */ 428 ipsecstat.out_polvio++; 429 goto bad; 430 431 case IPSEC_POLICY_BYPASS: 432 case IPSEC_POLICY_NONE: 433 case IPSEC_POLICY_TCP: 434 /* no need to do IPsec. */ 435 goto skip_ipsec; 436 437 case IPSEC_POLICY_IPSEC: 438 if (sp->req == NULL) { 439 /* acquire a policy */ 440 error = key_spdacquire(sp); 441 goto bad; 442 } 443 break; 444 445 case IPSEC_POLICY_ENTRUST: 446 default: 447 printf("ip_output: Invalid policy found. %d\n", sp->policy); 448 } 449 { 450 struct ipsec_output_state state; 451 bzero(&state, sizeof(state)); 452 state.m = m; 453 if (flags & IP_ROUTETOIF) { 454 state.ro = &iproute; 455 bzero(&iproute, sizeof(iproute)); 456 } else 457 state.ro = ro; 458 state.dst = (struct sockaddr *)dst; 459 460 ip->ip_sum = 0; 461 462 /* 463 * XXX 464 * delayed checksums are not currently compatible with IPsec 465 */ 466 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 467 in_delayed_cksum(m); 468 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 469 } 470 471 ip->ip_len = htons(ip->ip_len); 472 ip->ip_off = htons(ip->ip_off); 473 474 error = ipsec4_output(&state, sp, flags); 475 476 m = state.m; 477 if (flags & IP_ROUTETOIF) { 478 /* 479 * if we have tunnel mode SA, we may need to ignore 480 * IP_ROUTETOIF. 481 */ 482 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 483 flags &= ~IP_ROUTETOIF; 484 ro = state.ro; 485 } 486 } else 487 ro = state.ro; 488 dst = (struct sockaddr_in *)state.dst; 489 if (error) { 490 /* mbuf is already reclaimed in ipsec4_output. */ 491 m = NULL; 492 switch (error) { 493 case EHOSTUNREACH: 494 case ENETUNREACH: 495 case EMSGSIZE: 496 case ENOBUFS: 497 case ENOMEM: 498 break; 499 default: 500 printf("ip4_output (ipsec): error code %d\n", error); 501 /*fall through*/ 502 case ENOENT: 503 /* don't show these error codes to the user */ 504 error = 0; 505 break; 506 } 507 goto bad; 508 } 509 510 /* be sure to update variables that are affected by ipsec4_output() */ 511 ip = mtod(m, struct ip *); 512 hlen = ip->ip_hl << 2; 513 if (ro->ro_rt == NULL) { 514 if ((flags & IP_ROUTETOIF) == 0) { 515 printf("ip_output: " 516 "can't update route after IPsec processing\n"); 517 error = EHOSTUNREACH; /*XXX*/ 518 goto bad; 519 } 520 } else { 521 if (state.encap) { 522 ia = ifatoia(ro->ro_rt->rt_ifa); 523 ifp = ro->ro_rt->rt_ifp; 524 } 525 } 526 } 527 528 /* make it flipped, again. */ 529 ip->ip_len = ntohs(ip->ip_len); 530 ip->ip_off = ntohs(ip->ip_off); 531skip_ipsec: 532#endif /*IPSEC*/ 533#ifdef FAST_IPSEC 534 /* 535 * Check the security policy (SP) for the packet and, if 536 * required, do IPsec-related processing. There are two 537 * cases here; the first time a packet is sent through 538 * it will be untagged and handled by ipsec4_checkpolicy. 539 * If the packet is resubmitted to ip_output (e.g. after 540 * AH, ESP, etc. processing), there will be a tag to bypass 541 * the lookup and related policy checking. 542 */ 543 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 544 s = splnet(); 545 if (mtag != NULL) { 546 tdbi = (struct tdb_ident *)(mtag + 1); 547 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 548 if (sp == NULL) 549 error = -EINVAL; /* force silent drop */ 550 m_tag_delete(m, mtag); 551 } else { 552 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 553 &error, inp); 554 } 555 /* 556 * There are four return cases: 557 * sp != NULL apply IPsec policy 558 * sp == NULL, error == 0 no IPsec handling needed 559 * sp == NULL, error == -EINVAL discard packet w/o error 560 * sp == NULL, error != 0 discard packet, report error 561 */ 562 if (sp != NULL) { 563 /* Loop detection, check if ipsec processing already done */ 564 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 565 for (mtag = m_tag_first(m); mtag != NULL; 566 mtag = m_tag_next(m, mtag)) { 567 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 568 continue; 569 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 570 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 571 continue; 572 /* 573 * Check if policy has an SA associated with it. 574 * This can happen when an SP has yet to acquire 575 * an SA; e.g. on first reference. If it occurs, 576 * then we let ipsec4_process_packet do its thing. 577 */ 578 if (sp->req->sav == NULL) 579 break; 580 tdbi = (struct tdb_ident *)(mtag + 1); 581 if (tdbi->spi == sp->req->sav->spi && 582 tdbi->proto == sp->req->sav->sah->saidx.proto && 583 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 584 sizeof (union sockaddr_union)) == 0) { 585 /* 586 * No IPsec processing is needed, free 587 * reference to SP. 588 * 589 * NB: null pointer to avoid free at 590 * done: below. 591 */ 592 KEY_FREESP(&sp), sp = NULL; 593 splx(s); 594 goto spd_done; 595 } 596 } 597 598 /* 599 * Do delayed checksums now because we send before 600 * this is done in the normal processing path. 601 */ 602 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 603 in_delayed_cksum(m); 604 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 605 } 606 607 ip->ip_len = htons(ip->ip_len); 608 ip->ip_off = htons(ip->ip_off); 609 610 /* NB: callee frees mbuf */ 611 error = ipsec4_process_packet(m, sp->req, flags, 0); 612 /* 613 * Preserve KAME behaviour: ENOENT can be returned 614 * when an SA acquire is in progress. Don't propagate 615 * this to user-level; it confuses applications. 616 * 617 * XXX this will go away when the SADB is redone. 618 */ 619 if (error == ENOENT) 620 error = 0; 621 splx(s); 622 goto done; 623 } else { 624 splx(s); 625 626 if (error != 0) { 627 /* 628 * Hack: -EINVAL is used to signal that a packet 629 * should be silently discarded. This is typically 630 * because we asked key management for an SA and 631 * it was delayed (e.g. kicked up to IKE). 632 */ 633 if (error == -EINVAL) 634 error = 0; 635 goto bad; 636 } else { 637 /* No IPsec processing for this packet. */ 638 } 639#ifdef notyet 640 /* 641 * If deferred crypto processing is needed, check that 642 * the interface supports it. 643 */ 644 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 645 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 646 /* notify IPsec to do its own crypto */ 647 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 648 error = EHOSTUNREACH; 649 goto bad; 650 } 651#endif 652 } 653spd_done: 654#endif /* FAST_IPSEC */ 655 656 /* Jump over all PFIL processing if hooks are not active. */ 657 if (inet_pfil_hook.ph_busy_count == -1) 658 goto passout; 659 660 /* Run through list of hooks for output packets. */ 661 odst.s_addr = ip->ip_dst.s_addr; 662 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 663 if (error != 0 || m == NULL) 664 goto done; 665 666 ip = mtod(m, struct ip *); 667 668 /* See if destination IP address was changed by packet filter. */ 669 if (odst.s_addr != ip->ip_dst.s_addr) { 670 m->m_flags |= M_SKIP_FIREWALL; 671 /* If destination is now ourself drop to ip_input(). */ 672 if (in_localip(ip->ip_dst)) { 673 m->m_flags |= M_FASTFWD_OURS; 674 if (m->m_pkthdr.rcvif == NULL) 675 m->m_pkthdr.rcvif = loif; 676 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 677 m->m_pkthdr.csum_flags |= 678 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 679 m->m_pkthdr.csum_data = 0xffff; 680 } 681 m->m_pkthdr.csum_flags |= 682 CSUM_IP_CHECKED | CSUM_IP_VALID; 683 684 error = netisr_queue(NETISR_IP, m); 685 goto done; 686 } else 687 goto again; /* Redo the routing table lookup. */ 688 } 689 690#ifdef IPFIREWALL_FORWARD 691 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 692 if (m->m_flags & M_FASTFWD_OURS) { 693 if (m->m_pkthdr.rcvif == NULL) 694 m->m_pkthdr.rcvif = loif; 695 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 696 m->m_pkthdr.csum_flags |= 697 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 698 m->m_pkthdr.csum_data = 0xffff; 699 } 700 m->m_pkthdr.csum_flags |= 701 CSUM_IP_CHECKED | CSUM_IP_VALID; 702 703 error = netisr_queue(NETISR_IP, m); 704 goto done; 705 } 706 /* Or forward to some other address? */ 707 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 708 if (fwd_tag) { 709#ifndef IPFIREWALL_FORWARD_EXTENDED 710 if (!in_localip(ip->ip_src) && !in_localaddr(ip->ip_dst)) { 711#endif 712 dst = (struct sockaddr_in *)&ro->ro_dst; 713 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 714 m->m_flags |= M_SKIP_FIREWALL; 715 m_tag_delete(m, fwd_tag); 716 goto again; 717#ifndef IPFIREWALL_FORWARD_EXTENDED 718 } else { 719 m_tag_delete(m, fwd_tag); 720 /* Continue. */ 721 } 722#endif 723 } 724#endif /* IPFIREWALL_FORWARD */ 725 726passout: 727 /* 127/8 must not appear on wire - RFC1122. */ 728 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 729 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 730 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 731 ipstat.ips_badaddr++; 732 error = EADDRNOTAVAIL; 733 goto bad; 734 } 735 } 736 737 m->m_pkthdr.csum_flags |= CSUM_IP; 738 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 739 if (sw_csum & CSUM_DELAY_DATA) { 740 in_delayed_cksum(m); 741 sw_csum &= ~CSUM_DELAY_DATA; 742 } 743 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 744 745 /* 746 * If small enough for interface, or the interface will take 747 * care of the fragmentation for us, can just send directly. 748 */ 749 if (ip->ip_len <= ifp->if_mtu || (ifp->if_hwassist & CSUM_FRAGMENT && 750 ((ip->ip_off & IP_DF) == 0))) { 751 ip->ip_len = htons(ip->ip_len); 752 ip->ip_off = htons(ip->ip_off); 753 ip->ip_sum = 0; 754 if (sw_csum & CSUM_DELAY_IP) 755 ip->ip_sum = in_cksum(m, hlen); 756 757 /* Record statistics for this interface address. */ 758 if (!(flags & IP_FORWARDING) && ia) { 759 ia->ia_ifa.if_opackets++; 760 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 761 } 762 763#ifdef IPSEC 764 /* clean ipsec history once it goes out of the node */ 765 ipsec_delaux(m); 766#endif 767 768#ifdef MBUF_STRESS_TEST 769 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 770 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 771#endif 772 error = (*ifp->if_output)(ifp, m, 773 (struct sockaddr *)dst, ro->ro_rt); 774 goto done; 775 } 776 777 if (ip->ip_off & IP_DF) { 778 error = EMSGSIZE; 779 /* 780 * This case can happen if the user changed the MTU 781 * of an interface after enabling IP on it. Because 782 * most netifs don't keep track of routes pointing to 783 * them, there is no way for one to update all its 784 * routes when the MTU is changed. 785 */ 786 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 787 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 788 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 789 } 790 ipstat.ips_cantfrag++; 791 goto bad; 792 } 793 794 /* 795 * Too large for interface; fragment if possible. If successful, 796 * on return, m will point to a list of packets to be sent. 797 */ 798 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 799 if (error) 800 goto bad; 801 for (; m; m = m0) { 802 m0 = m->m_nextpkt; 803 m->m_nextpkt = 0; 804#ifdef IPSEC 805 /* clean ipsec history once it goes out of the node */ 806 ipsec_delaux(m); 807#endif 808 if (error == 0) { 809 /* Record statistics for this interface address. */ 810 if (ia != NULL) { 811 ia->ia_ifa.if_opackets++; 812 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 813 } 814 815 error = (*ifp->if_output)(ifp, m, 816 (struct sockaddr *)dst, ro->ro_rt); 817 } else 818 m_freem(m); 819 } 820 821 if (error == 0) 822 ipstat.ips_fragmented++; 823 824done: 825 if (ro == &iproute && ro->ro_rt) { 826 RTFREE(ro->ro_rt); 827 } 828#ifdef IPSEC 829 if (sp != NULL) { 830 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 831 printf("DP ip_output call free SP:%p\n", sp)); 832 key_freesp(sp); 833 } 834#endif 835#ifdef FAST_IPSEC 836 if (sp != NULL) 837 KEY_FREESP(&sp); 838#endif 839 return (error); 840bad: 841 m_freem(m); 842 goto done; 843} 844 845/* 846 * Create a chain of fragments which fit the given mtu. m_frag points to the 847 * mbuf to be fragmented; on return it points to the chain with the fragments. 848 * Return 0 if no error. If error, m_frag may contain a partially built 849 * chain of fragments that should be freed by the caller. 850 * 851 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 852 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 853 */ 854int 855ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 856 u_long if_hwassist_flags, int sw_csum) 857{ 858 int error = 0; 859 int hlen = ip->ip_hl << 2; 860 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 861 int off; 862 struct mbuf *m0 = *m_frag; /* the original packet */ 863 int firstlen; 864 struct mbuf **mnext; 865 int nfrags; 866 867 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 868 ipstat.ips_cantfrag++; 869 return EMSGSIZE; 870 } 871 872 /* 873 * Must be able to put at least 8 bytes per fragment. 874 */ 875 if (len < 8) 876 return EMSGSIZE; 877 878 /* 879 * If the interface will not calculate checksums on 880 * fragmented packets, then do it here. 881 */ 882 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 883 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 884 in_delayed_cksum(m0); 885 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 886 } 887 888 if (len > PAGE_SIZE) { 889 /* 890 * Fragment large datagrams such that each segment 891 * contains a multiple of PAGE_SIZE amount of data, 892 * plus headers. This enables a receiver to perform 893 * page-flipping zero-copy optimizations. 894 * 895 * XXX When does this help given that sender and receiver 896 * could have different page sizes, and also mtu could 897 * be less than the receiver's page size ? 898 */ 899 int newlen; 900 struct mbuf *m; 901 902 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 903 off += m->m_len; 904 905 /* 906 * firstlen (off - hlen) must be aligned on an 907 * 8-byte boundary 908 */ 909 if (off < hlen) 910 goto smart_frag_failure; 911 off = ((off - hlen) & ~7) + hlen; 912 newlen = (~PAGE_MASK) & mtu; 913 if ((newlen + sizeof (struct ip)) > mtu) { 914 /* we failed, go back the default */ 915smart_frag_failure: 916 newlen = len; 917 off = hlen + len; 918 } 919 len = newlen; 920 921 } else { 922 off = hlen + len; 923 } 924 925 firstlen = off - hlen; 926 mnext = &m0->m_nextpkt; /* pointer to next packet */ 927 928 /* 929 * Loop through length of segment after first fragment, 930 * make new header and copy data of each part and link onto chain. 931 * Here, m0 is the original packet, m is the fragment being created. 932 * The fragments are linked off the m_nextpkt of the original 933 * packet, which after processing serves as the first fragment. 934 */ 935 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 936 struct ip *mhip; /* ip header on the fragment */ 937 struct mbuf *m; 938 int mhlen = sizeof (struct ip); 939 940 MGETHDR(m, M_DONTWAIT, MT_HEADER); 941 if (m == NULL) { 942 error = ENOBUFS; 943 ipstat.ips_odropped++; 944 goto done; 945 } 946 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 947 /* 948 * In the first mbuf, leave room for the link header, then 949 * copy the original IP header including options. The payload 950 * goes into an additional mbuf chain returned by m_copy(). 951 */ 952 m->m_data += max_linkhdr; 953 mhip = mtod(m, struct ip *); 954 *mhip = *ip; 955 if (hlen > sizeof (struct ip)) { 956 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 957 mhip->ip_v = IPVERSION; 958 mhip->ip_hl = mhlen >> 2; 959 } 960 m->m_len = mhlen; 961 /* XXX do we need to add ip->ip_off below ? */ 962 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 963 if (off + len >= ip->ip_len) { /* last fragment */ 964 len = ip->ip_len - off; 965 m->m_flags |= M_LASTFRAG; 966 } else 967 mhip->ip_off |= IP_MF; 968 mhip->ip_len = htons((u_short)(len + mhlen)); 969 m->m_next = m_copy(m0, off, len); 970 if (m->m_next == NULL) { /* copy failed */ 971 m_free(m); 972 error = ENOBUFS; /* ??? */ 973 ipstat.ips_odropped++; 974 goto done; 975 } 976 m->m_pkthdr.len = mhlen + len; 977 m->m_pkthdr.rcvif = (struct ifnet *)0; 978#ifdef MAC 979 mac_create_fragment(m0, m); 980#endif 981 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 982 mhip->ip_off = htons(mhip->ip_off); 983 mhip->ip_sum = 0; 984 if (sw_csum & CSUM_DELAY_IP) 985 mhip->ip_sum = in_cksum(m, mhlen); 986 *mnext = m; 987 mnext = &m->m_nextpkt; 988 } 989 ipstat.ips_ofragments += nfrags; 990 991 /* set first marker for fragment chain */ 992 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 993 m0->m_pkthdr.csum_data = nfrags; 994 995 /* 996 * Update first fragment by trimming what's been copied out 997 * and updating header. 998 */ 999 m_adj(m0, hlen + firstlen - ip->ip_len); 1000 m0->m_pkthdr.len = hlen + firstlen; 1001 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 1002 ip->ip_off |= IP_MF; 1003 ip->ip_off = htons(ip->ip_off); 1004 ip->ip_sum = 0; 1005 if (sw_csum & CSUM_DELAY_IP) 1006 ip->ip_sum = in_cksum(m0, hlen); 1007 1008done: 1009 *m_frag = m0; 1010 return error; 1011} 1012 1013void 1014in_delayed_cksum(struct mbuf *m) 1015{ 1016 struct ip *ip; 1017 u_short csum, offset; 1018 1019 ip = mtod(m, struct ip *); 1020 offset = ip->ip_hl << 2 ; 1021 csum = in_cksum_skip(m, ip->ip_len, offset); 1022 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1023 csum = 0xffff; 1024 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1025 1026 if (offset + sizeof(u_short) > m->m_len) { 1027 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1028 m->m_len, offset, ip->ip_p); 1029 /* 1030 * XXX 1031 * this shouldn't happen, but if it does, the 1032 * correct behavior may be to insert the checksum 1033 * in the existing chain instead of rearranging it. 1034 */ 1035 m = m_pullup(m, offset + sizeof(u_short)); 1036 } 1037 *(u_short *)(m->m_data + offset) = csum; 1038} 1039 1040/* 1041 * Insert IP options into preformed packet. 1042 * Adjust IP destination as required for IP source routing, 1043 * as indicated by a non-zero in_addr at the start of the options. 1044 * 1045 * XXX This routine assumes that the packet has no options in place. 1046 */ 1047static struct mbuf * 1048ip_insertoptions(m, opt, phlen) 1049 register struct mbuf *m; 1050 struct mbuf *opt; 1051 int *phlen; 1052{ 1053 register struct ipoption *p = mtod(opt, struct ipoption *); 1054 struct mbuf *n; 1055 register struct ip *ip = mtod(m, struct ip *); 1056 unsigned optlen; 1057 1058 optlen = opt->m_len - sizeof(p->ipopt_dst); 1059 if (optlen + ip->ip_len > IP_MAXPACKET) { 1060 *phlen = 0; 1061 return (m); /* XXX should fail */ 1062 } 1063 if (p->ipopt_dst.s_addr) 1064 ip->ip_dst = p->ipopt_dst; 1065 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1066 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1067 if (n == NULL) { 1068 *phlen = 0; 1069 return (m); 1070 } 1071 M_MOVE_PKTHDR(n, m); 1072 n->m_pkthdr.rcvif = (struct ifnet *)0; 1073#ifdef MAC 1074 mac_create_mbuf_from_mbuf(m, n); 1075#endif 1076 n->m_pkthdr.len += optlen; 1077 m->m_len -= sizeof(struct ip); 1078 m->m_data += sizeof(struct ip); 1079 n->m_next = m; 1080 m = n; 1081 m->m_len = optlen + sizeof(struct ip); 1082 m->m_data += max_linkhdr; 1083 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1084 } else { 1085 m->m_data -= optlen; 1086 m->m_len += optlen; 1087 m->m_pkthdr.len += optlen; 1088 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1089 } 1090 ip = mtod(m, struct ip *); 1091 bcopy(p->ipopt_list, ip + 1, optlen); 1092 *phlen = sizeof(struct ip) + optlen; 1093 ip->ip_v = IPVERSION; 1094 ip->ip_hl = *phlen >> 2; 1095 ip->ip_len += optlen; 1096 return (m); 1097} 1098 1099/* 1100 * Copy options from ip to jp, 1101 * omitting those not copied during fragmentation. 1102 */ 1103int 1104ip_optcopy(ip, jp) 1105 struct ip *ip, *jp; 1106{ 1107 register u_char *cp, *dp; 1108 int opt, optlen, cnt; 1109 1110 cp = (u_char *)(ip + 1); 1111 dp = (u_char *)(jp + 1); 1112 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1113 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1114 opt = cp[0]; 1115 if (opt == IPOPT_EOL) 1116 break; 1117 if (opt == IPOPT_NOP) { 1118 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1119 *dp++ = IPOPT_NOP; 1120 optlen = 1; 1121 continue; 1122 } 1123 1124 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1125 ("ip_optcopy: malformed ipv4 option")); 1126 optlen = cp[IPOPT_OLEN]; 1127 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1128 ("ip_optcopy: malformed ipv4 option")); 1129 1130 /* bogus lengths should have been caught by ip_dooptions */ 1131 if (optlen > cnt) 1132 optlen = cnt; 1133 if (IPOPT_COPIED(opt)) { 1134 bcopy(cp, dp, optlen); 1135 dp += optlen; 1136 } 1137 } 1138 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1139 *dp++ = IPOPT_EOL; 1140 return (optlen); 1141} 1142 1143/* 1144 * IP socket option processing. 1145 */ 1146int 1147ip_ctloutput(so, sopt) 1148 struct socket *so; 1149 struct sockopt *sopt; 1150{ 1151 struct inpcb *inp = sotoinpcb(so); 1152 int error, optval; 1153 1154 error = optval = 0; 1155 if (sopt->sopt_level != IPPROTO_IP) { 1156 return (EINVAL); 1157 } 1158 1159 switch (sopt->sopt_dir) { 1160 case SOPT_SET: 1161 switch (sopt->sopt_name) { 1162 case IP_OPTIONS: 1163#ifdef notyet 1164 case IP_RETOPTS: 1165#endif 1166 { 1167 struct mbuf *m; 1168 if (sopt->sopt_valsize > MLEN) { 1169 error = EMSGSIZE; 1170 break; 1171 } 1172 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1173 if (m == NULL) { 1174 error = ENOBUFS; 1175 break; 1176 } 1177 m->m_len = sopt->sopt_valsize; 1178 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1179 m->m_len); 1180 INP_LOCK(inp); 1181 error = ip_pcbopts(inp, sopt->sopt_name, m); 1182 INP_UNLOCK(inp); 1183 return (error); 1184 } 1185 1186 case IP_TOS: 1187 case IP_TTL: 1188 case IP_RECVOPTS: 1189 case IP_RECVRETOPTS: 1190 case IP_RECVDSTADDR: 1191 case IP_RECVTTL: 1192 case IP_RECVIF: 1193 case IP_FAITH: 1194 case IP_ONESBCAST: 1195 error = sooptcopyin(sopt, &optval, sizeof optval, 1196 sizeof optval); 1197 if (error) 1198 break; 1199 1200 switch (sopt->sopt_name) { 1201 case IP_TOS: 1202 inp->inp_ip_tos = optval; 1203 break; 1204 1205 case IP_TTL: 1206 inp->inp_ip_ttl = optval; 1207 break; 1208#define OPTSET(bit) do { \ 1209 INP_LOCK(inp); \ 1210 if (optval) \ 1211 inp->inp_flags |= bit; \ 1212 else \ 1213 inp->inp_flags &= ~bit; \ 1214 INP_UNLOCK(inp); \ 1215} while (0) 1216 1217 case IP_RECVOPTS: 1218 OPTSET(INP_RECVOPTS); 1219 break; 1220 1221 case IP_RECVRETOPTS: 1222 OPTSET(INP_RECVRETOPTS); 1223 break; 1224 1225 case IP_RECVDSTADDR: 1226 OPTSET(INP_RECVDSTADDR); 1227 break; 1228 1229 case IP_RECVTTL: 1230 OPTSET(INP_RECVTTL); 1231 break; 1232 1233 case IP_RECVIF: 1234 OPTSET(INP_RECVIF); 1235 break; 1236 1237 case IP_FAITH: 1238 OPTSET(INP_FAITH); 1239 break; 1240 1241 case IP_ONESBCAST: 1242 OPTSET(INP_ONESBCAST); 1243 break; 1244 } 1245 break; 1246#undef OPTSET 1247 1248 case IP_MULTICAST_IF: 1249 case IP_MULTICAST_VIF: 1250 case IP_MULTICAST_TTL: 1251 case IP_MULTICAST_LOOP: 1252 case IP_ADD_MEMBERSHIP: 1253 case IP_DROP_MEMBERSHIP: 1254 error = ip_setmoptions(inp, sopt); 1255 break; 1256 1257 case IP_PORTRANGE: 1258 error = sooptcopyin(sopt, &optval, sizeof optval, 1259 sizeof optval); 1260 if (error) 1261 break; 1262 1263 INP_LOCK(inp); 1264 switch (optval) { 1265 case IP_PORTRANGE_DEFAULT: 1266 inp->inp_flags &= ~(INP_LOWPORT); 1267 inp->inp_flags &= ~(INP_HIGHPORT); 1268 break; 1269 1270 case IP_PORTRANGE_HIGH: 1271 inp->inp_flags &= ~(INP_LOWPORT); 1272 inp->inp_flags |= INP_HIGHPORT; 1273 break; 1274 1275 case IP_PORTRANGE_LOW: 1276 inp->inp_flags &= ~(INP_HIGHPORT); 1277 inp->inp_flags |= INP_LOWPORT; 1278 break; 1279 1280 default: 1281 error = EINVAL; 1282 break; 1283 } 1284 INP_UNLOCK(inp); 1285 break; 1286 1287#if defined(IPSEC) || defined(FAST_IPSEC) 1288 case IP_IPSEC_POLICY: 1289 { 1290 caddr_t req; 1291 size_t len = 0; 1292 int priv; 1293 struct mbuf *m; 1294 int optname; 1295 1296 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1297 break; 1298 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1299 break; 1300 priv = (sopt->sopt_td != NULL && 1301 suser(sopt->sopt_td) != 0) ? 0 : 1; 1302 req = mtod(m, caddr_t); 1303 len = m->m_len; 1304 optname = sopt->sopt_name; 1305 error = ipsec4_set_policy(inp, optname, req, len, priv); 1306 m_freem(m); 1307 break; 1308 } 1309#endif /*IPSEC*/ 1310 1311 default: 1312 error = ENOPROTOOPT; 1313 break; 1314 } 1315 break; 1316 1317 case SOPT_GET: 1318 switch (sopt->sopt_name) { 1319 case IP_OPTIONS: 1320 case IP_RETOPTS: 1321 if (inp->inp_options) 1322 error = sooptcopyout(sopt, 1323 mtod(inp->inp_options, 1324 char *), 1325 inp->inp_options->m_len); 1326 else 1327 sopt->sopt_valsize = 0; 1328 break; 1329 1330 case IP_TOS: 1331 case IP_TTL: 1332 case IP_RECVOPTS: 1333 case IP_RECVRETOPTS: 1334 case IP_RECVDSTADDR: 1335 case IP_RECVTTL: 1336 case IP_RECVIF: 1337 case IP_PORTRANGE: 1338 case IP_FAITH: 1339 case IP_ONESBCAST: 1340 switch (sopt->sopt_name) { 1341 1342 case IP_TOS: 1343 optval = inp->inp_ip_tos; 1344 break; 1345 1346 case IP_TTL: 1347 optval = inp->inp_ip_ttl; 1348 break; 1349 1350#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1351 1352 case IP_RECVOPTS: 1353 optval = OPTBIT(INP_RECVOPTS); 1354 break; 1355 1356 case IP_RECVRETOPTS: 1357 optval = OPTBIT(INP_RECVRETOPTS); 1358 break; 1359 1360 case IP_RECVDSTADDR: 1361 optval = OPTBIT(INP_RECVDSTADDR); 1362 break; 1363 1364 case IP_RECVTTL: 1365 optval = OPTBIT(INP_RECVTTL); 1366 break; 1367 1368 case IP_RECVIF: 1369 optval = OPTBIT(INP_RECVIF); 1370 break; 1371 1372 case IP_PORTRANGE: 1373 if (inp->inp_flags & INP_HIGHPORT) 1374 optval = IP_PORTRANGE_HIGH; 1375 else if (inp->inp_flags & INP_LOWPORT) 1376 optval = IP_PORTRANGE_LOW; 1377 else 1378 optval = 0; 1379 break; 1380 1381 case IP_FAITH: 1382 optval = OPTBIT(INP_FAITH); 1383 break; 1384 1385 case IP_ONESBCAST: 1386 optval = OPTBIT(INP_ONESBCAST); 1387 break; 1388 } 1389 error = sooptcopyout(sopt, &optval, sizeof optval); 1390 break; 1391 1392 case IP_MULTICAST_IF: 1393 case IP_MULTICAST_VIF: 1394 case IP_MULTICAST_TTL: 1395 case IP_MULTICAST_LOOP: 1396 case IP_ADD_MEMBERSHIP: 1397 case IP_DROP_MEMBERSHIP: 1398 error = ip_getmoptions(inp, sopt); 1399 break; 1400 1401#if defined(IPSEC) || defined(FAST_IPSEC) 1402 case IP_IPSEC_POLICY: 1403 { 1404 struct mbuf *m = NULL; 1405 caddr_t req = NULL; 1406 size_t len = 0; 1407 1408 if (m != 0) { 1409 req = mtod(m, caddr_t); 1410 len = m->m_len; 1411 } 1412 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1413 if (error == 0) 1414 error = soopt_mcopyout(sopt, m); /* XXX */ 1415 if (error == 0) 1416 m_freem(m); 1417 break; 1418 } 1419#endif /*IPSEC*/ 1420 1421 default: 1422 error = ENOPROTOOPT; 1423 break; 1424 } 1425 break; 1426 } 1427 return (error); 1428} 1429 1430/* 1431 * Set up IP options in pcb for insertion in output packets. 1432 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1433 * with destination address if source routed. 1434 */ 1435static int 1436ip_pcbopts(struct inpcb *inp, int optname, struct mbuf *m) 1437{ 1438 register int cnt, optlen; 1439 register u_char *cp; 1440 struct mbuf **pcbopt; 1441 u_char opt; 1442 1443 INP_LOCK_ASSERT(inp); 1444 1445 pcbopt = &inp->inp_options; 1446 1447 /* turn off any old options */ 1448 if (*pcbopt) 1449 (void)m_free(*pcbopt); 1450 *pcbopt = 0; 1451 if (m == NULL || m->m_len == 0) { 1452 /* 1453 * Only turning off any previous options. 1454 */ 1455 if (m != NULL) 1456 (void)m_free(m); 1457 return (0); 1458 } 1459 1460 if (m->m_len % sizeof(int32_t)) 1461 goto bad; 1462 /* 1463 * IP first-hop destination address will be stored before 1464 * actual options; move other options back 1465 * and clear it when none present. 1466 */ 1467 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1468 goto bad; 1469 cnt = m->m_len; 1470 m->m_len += sizeof(struct in_addr); 1471 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1472 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1473 bzero(mtod(m, void *), sizeof(struct in_addr)); 1474 1475 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1476 opt = cp[IPOPT_OPTVAL]; 1477 if (opt == IPOPT_EOL) 1478 break; 1479 if (opt == IPOPT_NOP) 1480 optlen = 1; 1481 else { 1482 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1483 goto bad; 1484 optlen = cp[IPOPT_OLEN]; 1485 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1486 goto bad; 1487 } 1488 switch (opt) { 1489 1490 default: 1491 break; 1492 1493 case IPOPT_LSRR: 1494 case IPOPT_SSRR: 1495 /* 1496 * user process specifies route as: 1497 * ->A->B->C->D 1498 * D must be our final destination (but we can't 1499 * check that since we may not have connected yet). 1500 * A is first hop destination, which doesn't appear in 1501 * actual IP option, but is stored before the options. 1502 */ 1503 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1504 goto bad; 1505 m->m_len -= sizeof(struct in_addr); 1506 cnt -= sizeof(struct in_addr); 1507 optlen -= sizeof(struct in_addr); 1508 cp[IPOPT_OLEN] = optlen; 1509 /* 1510 * Move first hop before start of options. 1511 */ 1512 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1513 sizeof(struct in_addr)); 1514 /* 1515 * Then copy rest of options back 1516 * to close up the deleted entry. 1517 */ 1518 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1519 &cp[IPOPT_OFFSET+1], 1520 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1521 break; 1522 } 1523 } 1524 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1525 goto bad; 1526 *pcbopt = m; 1527 return (0); 1528 1529bad: 1530 (void)m_free(m); 1531 return (EINVAL); 1532} 1533 1534/* 1535 * XXX 1536 * The whole multicast option thing needs to be re-thought. 1537 * Several of these options are equally applicable to non-multicast 1538 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1539 * standard option (IP_TTL). 1540 */ 1541 1542/* 1543 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1544 */ 1545static struct ifnet * 1546ip_multicast_if(a, ifindexp) 1547 struct in_addr *a; 1548 int *ifindexp; 1549{ 1550 int ifindex; 1551 struct ifnet *ifp; 1552 1553 if (ifindexp) 1554 *ifindexp = 0; 1555 if (ntohl(a->s_addr) >> 24 == 0) { 1556 ifindex = ntohl(a->s_addr) & 0xffffff; 1557 if (ifindex < 0 || if_index < ifindex) 1558 return NULL; 1559 ifp = ifnet_byindex(ifindex); 1560 if (ifindexp) 1561 *ifindexp = ifindex; 1562 } else { 1563 INADDR_TO_IFP(*a, ifp); 1564 } 1565 return ifp; 1566} 1567 1568/* 1569 * Set the IP multicast options in response to user setsockopt(). 1570 */ 1571static int 1572ip_setmoptions(struct inpcb *inp, struct sockopt *sopt) 1573{ 1574 int error = 0; 1575 int i; 1576 struct in_addr addr; 1577 struct ip_mreq mreq; 1578 struct ifnet *ifp; 1579 struct ip_moptions *imo; 1580 struct route ro; 1581 struct sockaddr_in *dst; 1582 int ifindex; 1583 int s; 1584 1585 imo = inp->inp_moptions; 1586 if (imo == NULL) { 1587 /* 1588 * No multicast option buffer attached to the pcb; 1589 * allocate one and initialize to default values. 1590 */ 1591 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1592 M_WAITOK); 1593 1594 if (imo == NULL) 1595 return (ENOBUFS); 1596 inp->inp_moptions = imo; 1597 imo->imo_multicast_ifp = NULL; 1598 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1599 imo->imo_multicast_vif = -1; 1600 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1601 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1602 imo->imo_num_memberships = 0; 1603 } 1604 1605 switch (sopt->sopt_name) { 1606 /* store an index number for the vif you wanna use in the send */ 1607 case IP_MULTICAST_VIF: 1608 if (legal_vif_num == 0) { 1609 error = EOPNOTSUPP; 1610 break; 1611 } 1612 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1613 if (error) 1614 break; 1615 if (!legal_vif_num(i) && (i != -1)) { 1616 error = EINVAL; 1617 break; 1618 } 1619 imo->imo_multicast_vif = i; 1620 break; 1621 1622 case IP_MULTICAST_IF: 1623 /* 1624 * Select the interface for outgoing multicast packets. 1625 */ 1626 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1627 if (error) 1628 break; 1629 /* 1630 * INADDR_ANY is used to remove a previous selection. 1631 * When no interface is selected, a default one is 1632 * chosen every time a multicast packet is sent. 1633 */ 1634 if (addr.s_addr == INADDR_ANY) { 1635 imo->imo_multicast_ifp = NULL; 1636 break; 1637 } 1638 /* 1639 * The selected interface is identified by its local 1640 * IP address. Find the interface and confirm that 1641 * it supports multicasting. 1642 */ 1643 s = splimp(); 1644 ifp = ip_multicast_if(&addr, &ifindex); 1645 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1646 splx(s); 1647 error = EADDRNOTAVAIL; 1648 break; 1649 } 1650 imo->imo_multicast_ifp = ifp; 1651 if (ifindex) 1652 imo->imo_multicast_addr = addr; 1653 else 1654 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1655 splx(s); 1656 break; 1657 1658 case IP_MULTICAST_TTL: 1659 /* 1660 * Set the IP time-to-live for outgoing multicast packets. 1661 * The original multicast API required a char argument, 1662 * which is inconsistent with the rest of the socket API. 1663 * We allow either a char or an int. 1664 */ 1665 if (sopt->sopt_valsize == 1) { 1666 u_char ttl; 1667 error = sooptcopyin(sopt, &ttl, 1, 1); 1668 if (error) 1669 break; 1670 imo->imo_multicast_ttl = ttl; 1671 } else { 1672 u_int ttl; 1673 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1674 sizeof ttl); 1675 if (error) 1676 break; 1677 if (ttl > 255) 1678 error = EINVAL; 1679 else 1680 imo->imo_multicast_ttl = ttl; 1681 } 1682 break; 1683 1684 case IP_MULTICAST_LOOP: 1685 /* 1686 * Set the loopback flag for outgoing multicast packets. 1687 * Must be zero or one. The original multicast API required a 1688 * char argument, which is inconsistent with the rest 1689 * of the socket API. We allow either a char or an int. 1690 */ 1691 if (sopt->sopt_valsize == 1) { 1692 u_char loop; 1693 error = sooptcopyin(sopt, &loop, 1, 1); 1694 if (error) 1695 break; 1696 imo->imo_multicast_loop = !!loop; 1697 } else { 1698 u_int loop; 1699 error = sooptcopyin(sopt, &loop, sizeof loop, 1700 sizeof loop); 1701 if (error) 1702 break; 1703 imo->imo_multicast_loop = !!loop; 1704 } 1705 break; 1706 1707 case IP_ADD_MEMBERSHIP: 1708 /* 1709 * Add a multicast group membership. 1710 * Group must be a valid IP multicast address. 1711 */ 1712 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1713 if (error) 1714 break; 1715 1716 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1717 error = EINVAL; 1718 break; 1719 } 1720 s = splimp(); 1721 /* 1722 * If no interface address was provided, use the interface of 1723 * the route to the given multicast address. 1724 */ 1725 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1726 bzero((caddr_t)&ro, sizeof(ro)); 1727 dst = (struct sockaddr_in *)&ro.ro_dst; 1728 dst->sin_len = sizeof(*dst); 1729 dst->sin_family = AF_INET; 1730 dst->sin_addr = mreq.imr_multiaddr; 1731 rtalloc_ign(&ro, RTF_CLONING); 1732 if (ro.ro_rt == NULL) { 1733 error = EADDRNOTAVAIL; 1734 splx(s); 1735 break; 1736 } 1737 ifp = ro.ro_rt->rt_ifp; 1738 RTFREE(ro.ro_rt); 1739 } 1740 else { 1741 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1742 } 1743 1744 /* 1745 * See if we found an interface, and confirm that it 1746 * supports multicast. 1747 */ 1748 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1749 error = EADDRNOTAVAIL; 1750 splx(s); 1751 break; 1752 } 1753 /* 1754 * See if the membership already exists or if all the 1755 * membership slots are full. 1756 */ 1757 for (i = 0; i < imo->imo_num_memberships; ++i) { 1758 if (imo->imo_membership[i]->inm_ifp == ifp && 1759 imo->imo_membership[i]->inm_addr.s_addr 1760 == mreq.imr_multiaddr.s_addr) 1761 break; 1762 } 1763 if (i < imo->imo_num_memberships) { 1764 error = EADDRINUSE; 1765 splx(s); 1766 break; 1767 } 1768 if (i == IP_MAX_MEMBERSHIPS) { 1769 error = ETOOMANYREFS; 1770 splx(s); 1771 break; 1772 } 1773 /* 1774 * Everything looks good; add a new record to the multicast 1775 * address list for the given interface. 1776 */ 1777 if ((imo->imo_membership[i] = 1778 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1779 error = ENOBUFS; 1780 splx(s); 1781 break; 1782 } 1783 ++imo->imo_num_memberships; 1784 splx(s); 1785 break; 1786 1787 case IP_DROP_MEMBERSHIP: 1788 /* 1789 * Drop a multicast group membership. 1790 * Group must be a valid IP multicast address. 1791 */ 1792 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1793 if (error) 1794 break; 1795 1796 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1797 error = EINVAL; 1798 break; 1799 } 1800 1801 s = splimp(); 1802 /* 1803 * If an interface address was specified, get a pointer 1804 * to its ifnet structure. 1805 */ 1806 if (mreq.imr_interface.s_addr == INADDR_ANY) 1807 ifp = NULL; 1808 else { 1809 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1810 if (ifp == NULL) { 1811 error = EADDRNOTAVAIL; 1812 splx(s); 1813 break; 1814 } 1815 } 1816 /* 1817 * Find the membership in the membership array. 1818 */ 1819 for (i = 0; i < imo->imo_num_memberships; ++i) { 1820 if ((ifp == NULL || 1821 imo->imo_membership[i]->inm_ifp == ifp) && 1822 imo->imo_membership[i]->inm_addr.s_addr == 1823 mreq.imr_multiaddr.s_addr) 1824 break; 1825 } 1826 if (i == imo->imo_num_memberships) { 1827 error = EADDRNOTAVAIL; 1828 splx(s); 1829 break; 1830 } 1831 /* 1832 * Give up the multicast address record to which the 1833 * membership points. 1834 */ 1835 in_delmulti(imo->imo_membership[i]); 1836 /* 1837 * Remove the gap in the membership array. 1838 */ 1839 for (++i; i < imo->imo_num_memberships; ++i) 1840 imo->imo_membership[i-1] = imo->imo_membership[i]; 1841 --imo->imo_num_memberships; 1842 splx(s); 1843 break; 1844 1845 default: 1846 error = EOPNOTSUPP; 1847 break; 1848 } 1849 1850 /* 1851 * If all options have default values, no need to keep the mbuf. 1852 */ 1853 if (imo->imo_multicast_ifp == NULL && 1854 imo->imo_multicast_vif == -1 && 1855 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1856 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1857 imo->imo_num_memberships == 0) { 1858 free(inp->inp_moptions, M_IPMOPTS); 1859 inp->inp_moptions = NULL; 1860 } 1861 1862 return (error); 1863} 1864 1865/* 1866 * Return the IP multicast options in response to user getsockopt(). 1867 */ 1868static int 1869ip_getmoptions(struct inpcb *inp, struct sockopt *sopt) 1870{ 1871 struct ip_moptions *imo; 1872 struct in_addr addr; 1873 struct in_ifaddr *ia; 1874 int error, optval; 1875 u_char coptval; 1876 1877 INP_LOCK(inp); 1878 imo = inp->inp_moptions; 1879 1880 error = 0; 1881 switch (sopt->sopt_name) { 1882 case IP_MULTICAST_VIF: 1883 if (imo != NULL) 1884 optval = imo->imo_multicast_vif; 1885 else 1886 optval = -1; 1887 INP_UNLOCK(inp); 1888 error = sooptcopyout(sopt, &optval, sizeof optval); 1889 break; 1890 1891 case IP_MULTICAST_IF: 1892 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1893 addr.s_addr = INADDR_ANY; 1894 else if (imo->imo_multicast_addr.s_addr) { 1895 /* return the value user has set */ 1896 addr = imo->imo_multicast_addr; 1897 } else { 1898 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1899 addr.s_addr = (ia == NULL) ? INADDR_ANY 1900 : IA_SIN(ia)->sin_addr.s_addr; 1901 } 1902 INP_UNLOCK(inp); 1903 error = sooptcopyout(sopt, &addr, sizeof addr); 1904 break; 1905 1906 case IP_MULTICAST_TTL: 1907 if (imo == 0) 1908 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1909 else 1910 optval = coptval = imo->imo_multicast_ttl; 1911 INP_UNLOCK(inp); 1912 if (sopt->sopt_valsize == 1) 1913 error = sooptcopyout(sopt, &coptval, 1); 1914 else 1915 error = sooptcopyout(sopt, &optval, sizeof optval); 1916 break; 1917 1918 case IP_MULTICAST_LOOP: 1919 if (imo == 0) 1920 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1921 else 1922 optval = coptval = imo->imo_multicast_loop; 1923 INP_UNLOCK(inp); 1924 if (sopt->sopt_valsize == 1) 1925 error = sooptcopyout(sopt, &coptval, 1); 1926 else 1927 error = sooptcopyout(sopt, &optval, sizeof optval); 1928 break; 1929 1930 default: 1931 INP_UNLOCK(inp); 1932 error = ENOPROTOOPT; 1933 break; 1934 } 1935 INP_UNLOCK_ASSERT(inp); 1936 1937 return (error); 1938} 1939 1940/* 1941 * Discard the IP multicast options. 1942 */ 1943void 1944ip_freemoptions(imo) 1945 register struct ip_moptions *imo; 1946{ 1947 register int i; 1948 1949 if (imo != NULL) { 1950 for (i = 0; i < imo->imo_num_memberships; ++i) 1951 in_delmulti(imo->imo_membership[i]); 1952 free(imo, M_IPMOPTS); 1953 } 1954} 1955 1956/* 1957 * Routine called from ip_output() to loop back a copy of an IP multicast 1958 * packet to the input queue of a specified interface. Note that this 1959 * calls the output routine of the loopback "driver", but with an interface 1960 * pointer that might NOT be a loopback interface -- evil, but easier than 1961 * replicating that code here. 1962 */ 1963static void 1964ip_mloopback(ifp, m, dst, hlen) 1965 struct ifnet *ifp; 1966 register struct mbuf *m; 1967 register struct sockaddr_in *dst; 1968 int hlen; 1969{ 1970 register struct ip *ip; 1971 struct mbuf *copym; 1972 1973 copym = m_copy(m, 0, M_COPYALL); 1974 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1975 copym = m_pullup(copym, hlen); 1976 if (copym != NULL) { 1977 /* If needed, compute the checksum and mark it as valid. */ 1978 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1979 in_delayed_cksum(copym); 1980 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1981 copym->m_pkthdr.csum_flags |= 1982 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1983 copym->m_pkthdr.csum_data = 0xffff; 1984 } 1985 /* 1986 * We don't bother to fragment if the IP length is greater 1987 * than the interface's MTU. Can this possibly matter? 1988 */ 1989 ip = mtod(copym, struct ip *); 1990 ip->ip_len = htons(ip->ip_len); 1991 ip->ip_off = htons(ip->ip_off); 1992 ip->ip_sum = 0; 1993 ip->ip_sum = in_cksum(copym, hlen); 1994 /* 1995 * NB: 1996 * It's not clear whether there are any lingering 1997 * reentrancy problems in other areas which might 1998 * be exposed by using ip_input directly (in 1999 * particular, everything which modifies the packet 2000 * in-place). Yet another option is using the 2001 * protosw directly to deliver the looped back 2002 * packet. For the moment, we'll err on the side 2003 * of safety by using if_simloop(). 2004 */ 2005#if 1 /* XXX */ 2006 if (dst->sin_family != AF_INET) { 2007 printf("ip_mloopback: bad address family %d\n", 2008 dst->sin_family); 2009 dst->sin_family = AF_INET; 2010 } 2011#endif 2012 2013#ifdef notdef 2014 copym->m_pkthdr.rcvif = ifp; 2015 ip_input(copym); 2016#else 2017 if_simloop(ifp, copym, dst->sin_family, 0); 2018#endif 2019 } 2020} 2021