ip_output.c revision 139310
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 139310 2004-12-25 22:59:42Z 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#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 if (!in_localip(ip->ip_src) && !in_localaddr(ip->ip_dst)) { 710 dst = (struct sockaddr_in *)&ro->ro_dst; 711 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 712 m->m_flags |= M_SKIP_FIREWALL; 713 m_tag_delete(m, fwd_tag); 714 goto again; 715 } else { 716 m_tag_delete(m, fwd_tag); 717 /* Continue. */ 718 } 719 } 720#endif 721 722passout: 723 /* 127/8 must not appear on wire - RFC1122. */ 724 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 725 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 726 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 727 ipstat.ips_badaddr++; 728 error = EADDRNOTAVAIL; 729 goto bad; 730 } 731 } 732 733 m->m_pkthdr.csum_flags |= CSUM_IP; 734 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 735 if (sw_csum & CSUM_DELAY_DATA) { 736 in_delayed_cksum(m); 737 sw_csum &= ~CSUM_DELAY_DATA; 738 } 739 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 740 741 /* 742 * If small enough for interface, or the interface will take 743 * care of the fragmentation for us, can just send directly. 744 */ 745 if (ip->ip_len <= ifp->if_mtu || (ifp->if_hwassist & CSUM_FRAGMENT && 746 ((ip->ip_off & IP_DF) == 0))) { 747 ip->ip_len = htons(ip->ip_len); 748 ip->ip_off = htons(ip->ip_off); 749 ip->ip_sum = 0; 750 if (sw_csum & CSUM_DELAY_IP) 751 ip->ip_sum = in_cksum(m, hlen); 752 753 /* Record statistics for this interface address. */ 754 if (!(flags & IP_FORWARDING) && ia) { 755 ia->ia_ifa.if_opackets++; 756 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 757 } 758 759#ifdef IPSEC 760 /* clean ipsec history once it goes out of the node */ 761 ipsec_delaux(m); 762#endif 763 764#ifdef MBUF_STRESS_TEST 765 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 766 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 767#endif 768 error = (*ifp->if_output)(ifp, m, 769 (struct sockaddr *)dst, ro->ro_rt); 770 goto done; 771 } 772 773 if (ip->ip_off & IP_DF) { 774 error = EMSGSIZE; 775 /* 776 * This case can happen if the user changed the MTU 777 * of an interface after enabling IP on it. Because 778 * most netifs don't keep track of routes pointing to 779 * them, there is no way for one to update all its 780 * routes when the MTU is changed. 781 */ 782 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 783 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 784 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 785 } 786 ipstat.ips_cantfrag++; 787 goto bad; 788 } 789 790 /* 791 * Too large for interface; fragment if possible. If successful, 792 * on return, m will point to a list of packets to be sent. 793 */ 794 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 795 if (error) 796 goto bad; 797 for (; m; m = m0) { 798 m0 = m->m_nextpkt; 799 m->m_nextpkt = 0; 800#ifdef IPSEC 801 /* clean ipsec history once it goes out of the node */ 802 ipsec_delaux(m); 803#endif 804 if (error == 0) { 805 /* Record statistics for this interface address. */ 806 if (ia != NULL) { 807 ia->ia_ifa.if_opackets++; 808 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 809 } 810 811 error = (*ifp->if_output)(ifp, m, 812 (struct sockaddr *)dst, ro->ro_rt); 813 } else 814 m_freem(m); 815 } 816 817 if (error == 0) 818 ipstat.ips_fragmented++; 819 820done: 821 if (ro == &iproute && ro->ro_rt) { 822 RTFREE(ro->ro_rt); 823 } 824#ifdef IPSEC 825 if (sp != NULL) { 826 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 827 printf("DP ip_output call free SP:%p\n", sp)); 828 key_freesp(sp); 829 } 830#endif 831#ifdef FAST_IPSEC 832 if (sp != NULL) 833 KEY_FREESP(&sp); 834#endif 835 return (error); 836bad: 837 m_freem(m); 838 goto done; 839} 840 841/* 842 * Create a chain of fragments which fit the given mtu. m_frag points to the 843 * mbuf to be fragmented; on return it points to the chain with the fragments. 844 * Return 0 if no error. If error, m_frag may contain a partially built 845 * chain of fragments that should be freed by the caller. 846 * 847 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 848 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 849 */ 850int 851ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 852 u_long if_hwassist_flags, int sw_csum) 853{ 854 int error = 0; 855 int hlen = ip->ip_hl << 2; 856 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 857 int off; 858 struct mbuf *m0 = *m_frag; /* the original packet */ 859 int firstlen; 860 struct mbuf **mnext; 861 int nfrags; 862 863 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 864 ipstat.ips_cantfrag++; 865 return EMSGSIZE; 866 } 867 868 /* 869 * Must be able to put at least 8 bytes per fragment. 870 */ 871 if (len < 8) 872 return EMSGSIZE; 873 874 /* 875 * If the interface will not calculate checksums on 876 * fragmented packets, then do it here. 877 */ 878 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 879 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 880 in_delayed_cksum(m0); 881 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 882 } 883 884 if (len > PAGE_SIZE) { 885 /* 886 * Fragment large datagrams such that each segment 887 * contains a multiple of PAGE_SIZE amount of data, 888 * plus headers. This enables a receiver to perform 889 * page-flipping zero-copy optimizations. 890 * 891 * XXX When does this help given that sender and receiver 892 * could have different page sizes, and also mtu could 893 * be less than the receiver's page size ? 894 */ 895 int newlen; 896 struct mbuf *m; 897 898 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 899 off += m->m_len; 900 901 /* 902 * firstlen (off - hlen) must be aligned on an 903 * 8-byte boundary 904 */ 905 if (off < hlen) 906 goto smart_frag_failure; 907 off = ((off - hlen) & ~7) + hlen; 908 newlen = (~PAGE_MASK) & mtu; 909 if ((newlen + sizeof (struct ip)) > mtu) { 910 /* we failed, go back the default */ 911smart_frag_failure: 912 newlen = len; 913 off = hlen + len; 914 } 915 len = newlen; 916 917 } else { 918 off = hlen + len; 919 } 920 921 firstlen = off - hlen; 922 mnext = &m0->m_nextpkt; /* pointer to next packet */ 923 924 /* 925 * Loop through length of segment after first fragment, 926 * make new header and copy data of each part and link onto chain. 927 * Here, m0 is the original packet, m is the fragment being created. 928 * The fragments are linked off the m_nextpkt of the original 929 * packet, which after processing serves as the first fragment. 930 */ 931 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 932 struct ip *mhip; /* ip header on the fragment */ 933 struct mbuf *m; 934 int mhlen = sizeof (struct ip); 935 936 MGETHDR(m, M_DONTWAIT, MT_HEADER); 937 if (m == NULL) { 938 error = ENOBUFS; 939 ipstat.ips_odropped++; 940 goto done; 941 } 942 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 943 /* 944 * In the first mbuf, leave room for the link header, then 945 * copy the original IP header including options. The payload 946 * goes into an additional mbuf chain returned by m_copy(). 947 */ 948 m->m_data += max_linkhdr; 949 mhip = mtod(m, struct ip *); 950 *mhip = *ip; 951 if (hlen > sizeof (struct ip)) { 952 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 953 mhip->ip_v = IPVERSION; 954 mhip->ip_hl = mhlen >> 2; 955 } 956 m->m_len = mhlen; 957 /* XXX do we need to add ip->ip_off below ? */ 958 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 959 if (off + len >= ip->ip_len) { /* last fragment */ 960 len = ip->ip_len - off; 961 m->m_flags |= M_LASTFRAG; 962 } else 963 mhip->ip_off |= IP_MF; 964 mhip->ip_len = htons((u_short)(len + mhlen)); 965 m->m_next = m_copy(m0, off, len); 966 if (m->m_next == NULL) { /* copy failed */ 967 m_free(m); 968 error = ENOBUFS; /* ??? */ 969 ipstat.ips_odropped++; 970 goto done; 971 } 972 m->m_pkthdr.len = mhlen + len; 973 m->m_pkthdr.rcvif = (struct ifnet *)0; 974#ifdef MAC 975 mac_create_fragment(m0, m); 976#endif 977 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 978 mhip->ip_off = htons(mhip->ip_off); 979 mhip->ip_sum = 0; 980 if (sw_csum & CSUM_DELAY_IP) 981 mhip->ip_sum = in_cksum(m, mhlen); 982 *mnext = m; 983 mnext = &m->m_nextpkt; 984 } 985 ipstat.ips_ofragments += nfrags; 986 987 /* set first marker for fragment chain */ 988 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 989 m0->m_pkthdr.csum_data = nfrags; 990 991 /* 992 * Update first fragment by trimming what's been copied out 993 * and updating header. 994 */ 995 m_adj(m0, hlen + firstlen - ip->ip_len); 996 m0->m_pkthdr.len = hlen + firstlen; 997 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 998 ip->ip_off |= IP_MF; 999 ip->ip_off = htons(ip->ip_off); 1000 ip->ip_sum = 0; 1001 if (sw_csum & CSUM_DELAY_IP) 1002 ip->ip_sum = in_cksum(m0, hlen); 1003 1004done: 1005 *m_frag = m0; 1006 return error; 1007} 1008 1009void 1010in_delayed_cksum(struct mbuf *m) 1011{ 1012 struct ip *ip; 1013 u_short csum, offset; 1014 1015 ip = mtod(m, struct ip *); 1016 offset = ip->ip_hl << 2 ; 1017 csum = in_cksum_skip(m, ip->ip_len, offset); 1018 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1019 csum = 0xffff; 1020 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1021 1022 if (offset + sizeof(u_short) > m->m_len) { 1023 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1024 m->m_len, offset, ip->ip_p); 1025 /* 1026 * XXX 1027 * this shouldn't happen, but if it does, the 1028 * correct behavior may be to insert the checksum 1029 * in the existing chain instead of rearranging it. 1030 */ 1031 m = m_pullup(m, offset + sizeof(u_short)); 1032 } 1033 *(u_short *)(m->m_data + offset) = csum; 1034} 1035 1036/* 1037 * Insert IP options into preformed packet. 1038 * Adjust IP destination as required for IP source routing, 1039 * as indicated by a non-zero in_addr at the start of the options. 1040 * 1041 * XXX This routine assumes that the packet has no options in place. 1042 */ 1043static struct mbuf * 1044ip_insertoptions(m, opt, phlen) 1045 register struct mbuf *m; 1046 struct mbuf *opt; 1047 int *phlen; 1048{ 1049 register struct ipoption *p = mtod(opt, struct ipoption *); 1050 struct mbuf *n; 1051 register struct ip *ip = mtod(m, struct ip *); 1052 unsigned optlen; 1053 1054 optlen = opt->m_len - sizeof(p->ipopt_dst); 1055 if (optlen + ip->ip_len > IP_MAXPACKET) { 1056 *phlen = 0; 1057 return (m); /* XXX should fail */ 1058 } 1059 if (p->ipopt_dst.s_addr) 1060 ip->ip_dst = p->ipopt_dst; 1061 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1062 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1063 if (n == NULL) { 1064 *phlen = 0; 1065 return (m); 1066 } 1067 n->m_pkthdr.rcvif = (struct ifnet *)0; 1068#ifdef MAC 1069 mac_create_mbuf_from_mbuf(m, n); 1070#endif 1071 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1072 m->m_len -= sizeof(struct ip); 1073 m->m_data += sizeof(struct ip); 1074 n->m_next = m; 1075 m = n; 1076 m->m_len = optlen + sizeof(struct ip); 1077 m->m_data += max_linkhdr; 1078 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1079 } else { 1080 m->m_data -= optlen; 1081 m->m_len += optlen; 1082 m->m_pkthdr.len += optlen; 1083 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1084 } 1085 ip = mtod(m, struct ip *); 1086 bcopy(p->ipopt_list, ip + 1, optlen); 1087 *phlen = sizeof(struct ip) + optlen; 1088 ip->ip_v = IPVERSION; 1089 ip->ip_hl = *phlen >> 2; 1090 ip->ip_len += optlen; 1091 return (m); 1092} 1093 1094/* 1095 * Copy options from ip to jp, 1096 * omitting those not copied during fragmentation. 1097 */ 1098int 1099ip_optcopy(ip, jp) 1100 struct ip *ip, *jp; 1101{ 1102 register u_char *cp, *dp; 1103 int opt, optlen, cnt; 1104 1105 cp = (u_char *)(ip + 1); 1106 dp = (u_char *)(jp + 1); 1107 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1108 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1109 opt = cp[0]; 1110 if (opt == IPOPT_EOL) 1111 break; 1112 if (opt == IPOPT_NOP) { 1113 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1114 *dp++ = IPOPT_NOP; 1115 optlen = 1; 1116 continue; 1117 } 1118 1119 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1120 ("ip_optcopy: malformed ipv4 option")); 1121 optlen = cp[IPOPT_OLEN]; 1122 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1123 ("ip_optcopy: malformed ipv4 option")); 1124 1125 /* bogus lengths should have been caught by ip_dooptions */ 1126 if (optlen > cnt) 1127 optlen = cnt; 1128 if (IPOPT_COPIED(opt)) { 1129 bcopy(cp, dp, optlen); 1130 dp += optlen; 1131 } 1132 } 1133 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1134 *dp++ = IPOPT_EOL; 1135 return (optlen); 1136} 1137 1138/* 1139 * IP socket option processing. 1140 */ 1141int 1142ip_ctloutput(so, sopt) 1143 struct socket *so; 1144 struct sockopt *sopt; 1145{ 1146 struct inpcb *inp = sotoinpcb(so); 1147 int error, optval; 1148 1149 error = optval = 0; 1150 if (sopt->sopt_level != IPPROTO_IP) { 1151 return (EINVAL); 1152 } 1153 1154 switch (sopt->sopt_dir) { 1155 case SOPT_SET: 1156 switch (sopt->sopt_name) { 1157 case IP_OPTIONS: 1158#ifdef notyet 1159 case IP_RETOPTS: 1160#endif 1161 { 1162 struct mbuf *m; 1163 if (sopt->sopt_valsize > MLEN) { 1164 error = EMSGSIZE; 1165 break; 1166 } 1167 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1168 if (m == NULL) { 1169 error = ENOBUFS; 1170 break; 1171 } 1172 m->m_len = sopt->sopt_valsize; 1173 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1174 m->m_len); 1175 INP_LOCK(inp); 1176 error = ip_pcbopts(inp, sopt->sopt_name, m); 1177 INP_UNLOCK(inp); 1178 return (error); 1179 } 1180 1181 case IP_TOS: 1182 case IP_TTL: 1183 case IP_RECVOPTS: 1184 case IP_RECVRETOPTS: 1185 case IP_RECVDSTADDR: 1186 case IP_RECVTTL: 1187 case IP_RECVIF: 1188 case IP_FAITH: 1189 case IP_ONESBCAST: 1190 error = sooptcopyin(sopt, &optval, sizeof optval, 1191 sizeof optval); 1192 if (error) 1193 break; 1194 1195 switch (sopt->sopt_name) { 1196 case IP_TOS: 1197 inp->inp_ip_tos = optval; 1198 break; 1199 1200 case IP_TTL: 1201 inp->inp_ip_ttl = optval; 1202 break; 1203#define OPTSET(bit) do { \ 1204 INP_LOCK(inp); \ 1205 if (optval) \ 1206 inp->inp_flags |= bit; \ 1207 else \ 1208 inp->inp_flags &= ~bit; \ 1209 INP_UNLOCK(inp); \ 1210} while (0) 1211 1212 case IP_RECVOPTS: 1213 OPTSET(INP_RECVOPTS); 1214 break; 1215 1216 case IP_RECVRETOPTS: 1217 OPTSET(INP_RECVRETOPTS); 1218 break; 1219 1220 case IP_RECVDSTADDR: 1221 OPTSET(INP_RECVDSTADDR); 1222 break; 1223 1224 case IP_RECVTTL: 1225 OPTSET(INP_RECVTTL); 1226 break; 1227 1228 case IP_RECVIF: 1229 OPTSET(INP_RECVIF); 1230 break; 1231 1232 case IP_FAITH: 1233 OPTSET(INP_FAITH); 1234 break; 1235 1236 case IP_ONESBCAST: 1237 OPTSET(INP_ONESBCAST); 1238 break; 1239 } 1240 break; 1241#undef OPTSET 1242 1243 case IP_MULTICAST_IF: 1244 case IP_MULTICAST_VIF: 1245 case IP_MULTICAST_TTL: 1246 case IP_MULTICAST_LOOP: 1247 case IP_ADD_MEMBERSHIP: 1248 case IP_DROP_MEMBERSHIP: 1249 error = ip_setmoptions(inp, sopt); 1250 break; 1251 1252 case IP_PORTRANGE: 1253 error = sooptcopyin(sopt, &optval, sizeof optval, 1254 sizeof optval); 1255 if (error) 1256 break; 1257 1258 INP_LOCK(inp); 1259 switch (optval) { 1260 case IP_PORTRANGE_DEFAULT: 1261 inp->inp_flags &= ~(INP_LOWPORT); 1262 inp->inp_flags &= ~(INP_HIGHPORT); 1263 break; 1264 1265 case IP_PORTRANGE_HIGH: 1266 inp->inp_flags &= ~(INP_LOWPORT); 1267 inp->inp_flags |= INP_HIGHPORT; 1268 break; 1269 1270 case IP_PORTRANGE_LOW: 1271 inp->inp_flags &= ~(INP_HIGHPORT); 1272 inp->inp_flags |= INP_LOWPORT; 1273 break; 1274 1275 default: 1276 error = EINVAL; 1277 break; 1278 } 1279 INP_UNLOCK(inp); 1280 break; 1281 1282#if defined(IPSEC) || defined(FAST_IPSEC) 1283 case IP_IPSEC_POLICY: 1284 { 1285 caddr_t req; 1286 size_t len = 0; 1287 int priv; 1288 struct mbuf *m; 1289 int optname; 1290 1291 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1292 break; 1293 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1294 break; 1295 priv = (sopt->sopt_td != NULL && 1296 suser(sopt->sopt_td) != 0) ? 0 : 1; 1297 req = mtod(m, caddr_t); 1298 len = m->m_len; 1299 optname = sopt->sopt_name; 1300 error = ipsec4_set_policy(inp, optname, req, len, priv); 1301 m_freem(m); 1302 break; 1303 } 1304#endif /*IPSEC*/ 1305 1306 default: 1307 error = ENOPROTOOPT; 1308 break; 1309 } 1310 break; 1311 1312 case SOPT_GET: 1313 switch (sopt->sopt_name) { 1314 case IP_OPTIONS: 1315 case IP_RETOPTS: 1316 if (inp->inp_options) 1317 error = sooptcopyout(sopt, 1318 mtod(inp->inp_options, 1319 char *), 1320 inp->inp_options->m_len); 1321 else 1322 sopt->sopt_valsize = 0; 1323 break; 1324 1325 case IP_TOS: 1326 case IP_TTL: 1327 case IP_RECVOPTS: 1328 case IP_RECVRETOPTS: 1329 case IP_RECVDSTADDR: 1330 case IP_RECVTTL: 1331 case IP_RECVIF: 1332 case IP_PORTRANGE: 1333 case IP_FAITH: 1334 case IP_ONESBCAST: 1335 switch (sopt->sopt_name) { 1336 1337 case IP_TOS: 1338 optval = inp->inp_ip_tos; 1339 break; 1340 1341 case IP_TTL: 1342 optval = inp->inp_ip_ttl; 1343 break; 1344 1345#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1346 1347 case IP_RECVOPTS: 1348 optval = OPTBIT(INP_RECVOPTS); 1349 break; 1350 1351 case IP_RECVRETOPTS: 1352 optval = OPTBIT(INP_RECVRETOPTS); 1353 break; 1354 1355 case IP_RECVDSTADDR: 1356 optval = OPTBIT(INP_RECVDSTADDR); 1357 break; 1358 1359 case IP_RECVTTL: 1360 optval = OPTBIT(INP_RECVTTL); 1361 break; 1362 1363 case IP_RECVIF: 1364 optval = OPTBIT(INP_RECVIF); 1365 break; 1366 1367 case IP_PORTRANGE: 1368 if (inp->inp_flags & INP_HIGHPORT) 1369 optval = IP_PORTRANGE_HIGH; 1370 else if (inp->inp_flags & INP_LOWPORT) 1371 optval = IP_PORTRANGE_LOW; 1372 else 1373 optval = 0; 1374 break; 1375 1376 case IP_FAITH: 1377 optval = OPTBIT(INP_FAITH); 1378 break; 1379 1380 case IP_ONESBCAST: 1381 optval = OPTBIT(INP_ONESBCAST); 1382 break; 1383 } 1384 error = sooptcopyout(sopt, &optval, sizeof optval); 1385 break; 1386 1387 case IP_MULTICAST_IF: 1388 case IP_MULTICAST_VIF: 1389 case IP_MULTICAST_TTL: 1390 case IP_MULTICAST_LOOP: 1391 case IP_ADD_MEMBERSHIP: 1392 case IP_DROP_MEMBERSHIP: 1393 error = ip_getmoptions(inp, sopt); 1394 break; 1395 1396#if defined(IPSEC) || defined(FAST_IPSEC) 1397 case IP_IPSEC_POLICY: 1398 { 1399 struct mbuf *m = NULL; 1400 caddr_t req = NULL; 1401 size_t len = 0; 1402 1403 if (m != 0) { 1404 req = mtod(m, caddr_t); 1405 len = m->m_len; 1406 } 1407 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1408 if (error == 0) 1409 error = soopt_mcopyout(sopt, m); /* XXX */ 1410 if (error == 0) 1411 m_freem(m); 1412 break; 1413 } 1414#endif /*IPSEC*/ 1415 1416 default: 1417 error = ENOPROTOOPT; 1418 break; 1419 } 1420 break; 1421 } 1422 return (error); 1423} 1424 1425/* 1426 * Set up IP options in pcb for insertion in output packets. 1427 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1428 * with destination address if source routed. 1429 */ 1430static int 1431ip_pcbopts(struct inpcb *inp, int optname, struct mbuf *m) 1432{ 1433 register int cnt, optlen; 1434 register u_char *cp; 1435 struct mbuf **pcbopt; 1436 u_char opt; 1437 1438 INP_LOCK_ASSERT(inp); 1439 1440 pcbopt = &inp->inp_options; 1441 1442 /* turn off any old options */ 1443 if (*pcbopt) 1444 (void)m_free(*pcbopt); 1445 *pcbopt = 0; 1446 if (m == NULL || m->m_len == 0) { 1447 /* 1448 * Only turning off any previous options. 1449 */ 1450 if (m != NULL) 1451 (void)m_free(m); 1452 return (0); 1453 } 1454 1455 if (m->m_len % sizeof(int32_t)) 1456 goto bad; 1457 /* 1458 * IP first-hop destination address will be stored before 1459 * actual options; move other options back 1460 * and clear it when none present. 1461 */ 1462 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1463 goto bad; 1464 cnt = m->m_len; 1465 m->m_len += sizeof(struct in_addr); 1466 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1467 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1468 bzero(mtod(m, void *), sizeof(struct in_addr)); 1469 1470 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1471 opt = cp[IPOPT_OPTVAL]; 1472 if (opt == IPOPT_EOL) 1473 break; 1474 if (opt == IPOPT_NOP) 1475 optlen = 1; 1476 else { 1477 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1478 goto bad; 1479 optlen = cp[IPOPT_OLEN]; 1480 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1481 goto bad; 1482 } 1483 switch (opt) { 1484 1485 default: 1486 break; 1487 1488 case IPOPT_LSRR: 1489 case IPOPT_SSRR: 1490 /* 1491 * user process specifies route as: 1492 * ->A->B->C->D 1493 * D must be our final destination (but we can't 1494 * check that since we may not have connected yet). 1495 * A is first hop destination, which doesn't appear in 1496 * actual IP option, but is stored before the options. 1497 */ 1498 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1499 goto bad; 1500 m->m_len -= sizeof(struct in_addr); 1501 cnt -= sizeof(struct in_addr); 1502 optlen -= sizeof(struct in_addr); 1503 cp[IPOPT_OLEN] = optlen; 1504 /* 1505 * Move first hop before start of options. 1506 */ 1507 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1508 sizeof(struct in_addr)); 1509 /* 1510 * Then copy rest of options back 1511 * to close up the deleted entry. 1512 */ 1513 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1514 &cp[IPOPT_OFFSET+1], 1515 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1516 break; 1517 } 1518 } 1519 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1520 goto bad; 1521 *pcbopt = m; 1522 return (0); 1523 1524bad: 1525 (void)m_free(m); 1526 return (EINVAL); 1527} 1528 1529/* 1530 * XXX 1531 * The whole multicast option thing needs to be re-thought. 1532 * Several of these options are equally applicable to non-multicast 1533 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1534 * standard option (IP_TTL). 1535 */ 1536 1537/* 1538 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1539 */ 1540static struct ifnet * 1541ip_multicast_if(a, ifindexp) 1542 struct in_addr *a; 1543 int *ifindexp; 1544{ 1545 int ifindex; 1546 struct ifnet *ifp; 1547 1548 if (ifindexp) 1549 *ifindexp = 0; 1550 if (ntohl(a->s_addr) >> 24 == 0) { 1551 ifindex = ntohl(a->s_addr) & 0xffffff; 1552 if (ifindex < 0 || if_index < ifindex) 1553 return NULL; 1554 ifp = ifnet_byindex(ifindex); 1555 if (ifindexp) 1556 *ifindexp = ifindex; 1557 } else { 1558 INADDR_TO_IFP(*a, ifp); 1559 } 1560 return ifp; 1561} 1562 1563/* 1564 * Set the IP multicast options in response to user setsockopt(). 1565 */ 1566static int 1567ip_setmoptions(struct inpcb *inp, struct sockopt *sopt) 1568{ 1569 int error = 0; 1570 int i; 1571 struct in_addr addr; 1572 struct ip_mreq mreq; 1573 struct ifnet *ifp; 1574 struct ip_moptions *imo; 1575 struct route ro; 1576 struct sockaddr_in *dst; 1577 int ifindex; 1578 int s; 1579 1580 imo = inp->inp_moptions; 1581 if (imo == NULL) { 1582 /* 1583 * No multicast option buffer attached to the pcb; 1584 * allocate one and initialize to default values. 1585 */ 1586 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1587 M_WAITOK); 1588 1589 if (imo == NULL) 1590 return (ENOBUFS); 1591 inp->inp_moptions = imo; 1592 imo->imo_multicast_ifp = NULL; 1593 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1594 imo->imo_multicast_vif = -1; 1595 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1596 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1597 imo->imo_num_memberships = 0; 1598 } 1599 1600 switch (sopt->sopt_name) { 1601 /* store an index number for the vif you wanna use in the send */ 1602 case IP_MULTICAST_VIF: 1603 if (legal_vif_num == 0) { 1604 error = EOPNOTSUPP; 1605 break; 1606 } 1607 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1608 if (error) 1609 break; 1610 if (!legal_vif_num(i) && (i != -1)) { 1611 error = EINVAL; 1612 break; 1613 } 1614 imo->imo_multicast_vif = i; 1615 break; 1616 1617 case IP_MULTICAST_IF: 1618 /* 1619 * Select the interface for outgoing multicast packets. 1620 */ 1621 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1622 if (error) 1623 break; 1624 /* 1625 * INADDR_ANY is used to remove a previous selection. 1626 * When no interface is selected, a default one is 1627 * chosen every time a multicast packet is sent. 1628 */ 1629 if (addr.s_addr == INADDR_ANY) { 1630 imo->imo_multicast_ifp = NULL; 1631 break; 1632 } 1633 /* 1634 * The selected interface is identified by its local 1635 * IP address. Find the interface and confirm that 1636 * it supports multicasting. 1637 */ 1638 s = splimp(); 1639 ifp = ip_multicast_if(&addr, &ifindex); 1640 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1641 splx(s); 1642 error = EADDRNOTAVAIL; 1643 break; 1644 } 1645 imo->imo_multicast_ifp = ifp; 1646 if (ifindex) 1647 imo->imo_multicast_addr = addr; 1648 else 1649 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1650 splx(s); 1651 break; 1652 1653 case IP_MULTICAST_TTL: 1654 /* 1655 * Set the IP time-to-live for outgoing multicast packets. 1656 * The original multicast API required a char argument, 1657 * which is inconsistent with the rest of the socket API. 1658 * We allow either a char or an int. 1659 */ 1660 if (sopt->sopt_valsize == 1) { 1661 u_char ttl; 1662 error = sooptcopyin(sopt, &ttl, 1, 1); 1663 if (error) 1664 break; 1665 imo->imo_multicast_ttl = ttl; 1666 } else { 1667 u_int ttl; 1668 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1669 sizeof ttl); 1670 if (error) 1671 break; 1672 if (ttl > 255) 1673 error = EINVAL; 1674 else 1675 imo->imo_multicast_ttl = ttl; 1676 } 1677 break; 1678 1679 case IP_MULTICAST_LOOP: 1680 /* 1681 * Set the loopback flag for outgoing multicast packets. 1682 * Must be zero or one. The original multicast API required a 1683 * char argument, which is inconsistent with the rest 1684 * of the socket API. We allow either a char or an int. 1685 */ 1686 if (sopt->sopt_valsize == 1) { 1687 u_char loop; 1688 error = sooptcopyin(sopt, &loop, 1, 1); 1689 if (error) 1690 break; 1691 imo->imo_multicast_loop = !!loop; 1692 } else { 1693 u_int loop; 1694 error = sooptcopyin(sopt, &loop, sizeof loop, 1695 sizeof loop); 1696 if (error) 1697 break; 1698 imo->imo_multicast_loop = !!loop; 1699 } 1700 break; 1701 1702 case IP_ADD_MEMBERSHIP: 1703 /* 1704 * Add a multicast group membership. 1705 * Group must be a valid IP multicast address. 1706 */ 1707 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1708 if (error) 1709 break; 1710 1711 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1712 error = EINVAL; 1713 break; 1714 } 1715 s = splimp(); 1716 /* 1717 * If no interface address was provided, use the interface of 1718 * the route to the given multicast address. 1719 */ 1720 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1721 bzero((caddr_t)&ro, sizeof(ro)); 1722 dst = (struct sockaddr_in *)&ro.ro_dst; 1723 dst->sin_len = sizeof(*dst); 1724 dst->sin_family = AF_INET; 1725 dst->sin_addr = mreq.imr_multiaddr; 1726 rtalloc_ign(&ro, RTF_CLONING); 1727 if (ro.ro_rt == NULL) { 1728 error = EADDRNOTAVAIL; 1729 splx(s); 1730 break; 1731 } 1732 ifp = ro.ro_rt->rt_ifp; 1733 RTFREE(ro.ro_rt); 1734 } 1735 else { 1736 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1737 } 1738 1739 /* 1740 * See if we found an interface, and confirm that it 1741 * supports multicast. 1742 */ 1743 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1744 error = EADDRNOTAVAIL; 1745 splx(s); 1746 break; 1747 } 1748 /* 1749 * See if the membership already exists or if all the 1750 * membership slots are full. 1751 */ 1752 for (i = 0; i < imo->imo_num_memberships; ++i) { 1753 if (imo->imo_membership[i]->inm_ifp == ifp && 1754 imo->imo_membership[i]->inm_addr.s_addr 1755 == mreq.imr_multiaddr.s_addr) 1756 break; 1757 } 1758 if (i < imo->imo_num_memberships) { 1759 error = EADDRINUSE; 1760 splx(s); 1761 break; 1762 } 1763 if (i == IP_MAX_MEMBERSHIPS) { 1764 error = ETOOMANYREFS; 1765 splx(s); 1766 break; 1767 } 1768 /* 1769 * Everything looks good; add a new record to the multicast 1770 * address list for the given interface. 1771 */ 1772 if ((imo->imo_membership[i] = 1773 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1774 error = ENOBUFS; 1775 splx(s); 1776 break; 1777 } 1778 ++imo->imo_num_memberships; 1779 splx(s); 1780 break; 1781 1782 case IP_DROP_MEMBERSHIP: 1783 /* 1784 * Drop a multicast group membership. 1785 * Group must be a valid IP multicast address. 1786 */ 1787 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1788 if (error) 1789 break; 1790 1791 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1792 error = EINVAL; 1793 break; 1794 } 1795 1796 s = splimp(); 1797 /* 1798 * If an interface address was specified, get a pointer 1799 * to its ifnet structure. 1800 */ 1801 if (mreq.imr_interface.s_addr == INADDR_ANY) 1802 ifp = NULL; 1803 else { 1804 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1805 if (ifp == NULL) { 1806 error = EADDRNOTAVAIL; 1807 splx(s); 1808 break; 1809 } 1810 } 1811 /* 1812 * Find the membership in the membership array. 1813 */ 1814 for (i = 0; i < imo->imo_num_memberships; ++i) { 1815 if ((ifp == NULL || 1816 imo->imo_membership[i]->inm_ifp == ifp) && 1817 imo->imo_membership[i]->inm_addr.s_addr == 1818 mreq.imr_multiaddr.s_addr) 1819 break; 1820 } 1821 if (i == imo->imo_num_memberships) { 1822 error = EADDRNOTAVAIL; 1823 splx(s); 1824 break; 1825 } 1826 /* 1827 * Give up the multicast address record to which the 1828 * membership points. 1829 */ 1830 in_delmulti(imo->imo_membership[i]); 1831 /* 1832 * Remove the gap in the membership array. 1833 */ 1834 for (++i; i < imo->imo_num_memberships; ++i) 1835 imo->imo_membership[i-1] = imo->imo_membership[i]; 1836 --imo->imo_num_memberships; 1837 splx(s); 1838 break; 1839 1840 default: 1841 error = EOPNOTSUPP; 1842 break; 1843 } 1844 1845 /* 1846 * If all options have default values, no need to keep the mbuf. 1847 */ 1848 if (imo->imo_multicast_ifp == NULL && 1849 imo->imo_multicast_vif == -1 && 1850 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1851 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1852 imo->imo_num_memberships == 0) { 1853 free(inp->inp_moptions, M_IPMOPTS); 1854 inp->inp_moptions = NULL; 1855 } 1856 1857 return (error); 1858} 1859 1860/* 1861 * Return the IP multicast options in response to user getsockopt(). 1862 */ 1863static int 1864ip_getmoptions(struct inpcb *inp, struct sockopt *sopt) 1865{ 1866 struct ip_moptions *imo; 1867 struct in_addr addr; 1868 struct in_ifaddr *ia; 1869 int error, optval; 1870 u_char coptval; 1871 1872 INP_LOCK(inp); 1873 imo = inp->inp_moptions; 1874 1875 error = 0; 1876 switch (sopt->sopt_name) { 1877 case IP_MULTICAST_VIF: 1878 if (imo != NULL) 1879 optval = imo->imo_multicast_vif; 1880 else 1881 optval = -1; 1882 INP_UNLOCK(inp); 1883 error = sooptcopyout(sopt, &optval, sizeof optval); 1884 break; 1885 1886 case IP_MULTICAST_IF: 1887 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1888 addr.s_addr = INADDR_ANY; 1889 else if (imo->imo_multicast_addr.s_addr) { 1890 /* return the value user has set */ 1891 addr = imo->imo_multicast_addr; 1892 } else { 1893 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1894 addr.s_addr = (ia == NULL) ? INADDR_ANY 1895 : IA_SIN(ia)->sin_addr.s_addr; 1896 } 1897 INP_UNLOCK(inp); 1898 error = sooptcopyout(sopt, &addr, sizeof addr); 1899 break; 1900 1901 case IP_MULTICAST_TTL: 1902 if (imo == 0) 1903 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1904 else 1905 optval = coptval = imo->imo_multicast_ttl; 1906 INP_UNLOCK(inp); 1907 if (sopt->sopt_valsize == 1) 1908 error = sooptcopyout(sopt, &coptval, 1); 1909 else 1910 error = sooptcopyout(sopt, &optval, sizeof optval); 1911 break; 1912 1913 case IP_MULTICAST_LOOP: 1914 if (imo == 0) 1915 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1916 else 1917 optval = coptval = imo->imo_multicast_loop; 1918 INP_UNLOCK(inp); 1919 if (sopt->sopt_valsize == 1) 1920 error = sooptcopyout(sopt, &coptval, 1); 1921 else 1922 error = sooptcopyout(sopt, &optval, sizeof optval); 1923 break; 1924 1925 default: 1926 INP_UNLOCK(inp); 1927 error = ENOPROTOOPT; 1928 break; 1929 } 1930 INP_UNLOCK_ASSERT(inp); 1931 1932 return (error); 1933} 1934 1935/* 1936 * Discard the IP multicast options. 1937 */ 1938void 1939ip_freemoptions(imo) 1940 register struct ip_moptions *imo; 1941{ 1942 register int i; 1943 1944 if (imo != NULL) { 1945 for (i = 0; i < imo->imo_num_memberships; ++i) 1946 in_delmulti(imo->imo_membership[i]); 1947 free(imo, M_IPMOPTS); 1948 } 1949} 1950 1951/* 1952 * Routine called from ip_output() to loop back a copy of an IP multicast 1953 * packet to the input queue of a specified interface. Note that this 1954 * calls the output routine of the loopback "driver", but with an interface 1955 * pointer that might NOT be a loopback interface -- evil, but easier than 1956 * replicating that code here. 1957 */ 1958static void 1959ip_mloopback(ifp, m, dst, hlen) 1960 struct ifnet *ifp; 1961 register struct mbuf *m; 1962 register struct sockaddr_in *dst; 1963 int hlen; 1964{ 1965 register struct ip *ip; 1966 struct mbuf *copym; 1967 1968 copym = m_copy(m, 0, M_COPYALL); 1969 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1970 copym = m_pullup(copym, hlen); 1971 if (copym != NULL) { 1972 /* If needed, compute the checksum and mark it as valid. */ 1973 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1974 in_delayed_cksum(copym); 1975 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1976 copym->m_pkthdr.csum_flags |= 1977 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1978 copym->m_pkthdr.csum_data = 0xffff; 1979 } 1980 /* 1981 * We don't bother to fragment if the IP length is greater 1982 * than the interface's MTU. Can this possibly matter? 1983 */ 1984 ip = mtod(copym, struct ip *); 1985 ip->ip_len = htons(ip->ip_len); 1986 ip->ip_off = htons(ip->ip_off); 1987 ip->ip_sum = 0; 1988 ip->ip_sum = in_cksum(copym, hlen); 1989 /* 1990 * NB: 1991 * It's not clear whether there are any lingering 1992 * reentrancy problems in other areas which might 1993 * be exposed by using ip_input directly (in 1994 * particular, everything which modifies the packet 1995 * in-place). Yet another option is using the 1996 * protosw directly to deliver the looped back 1997 * packet. For the moment, we'll err on the side 1998 * of safety by using if_simloop(). 1999 */ 2000#if 1 /* XXX */ 2001 if (dst->sin_family != AF_INET) { 2002 printf("ip_mloopback: bad address family %d\n", 2003 dst->sin_family); 2004 dst->sin_family = AF_INET; 2005 } 2006#endif 2007 2008#ifdef notdef 2009 copym->m_pkthdr.rcvif = ifp; 2010 ip_input(copym); 2011#else 2012 if_simloop(ifp, copym, dst->sin_family, 0); 2013#endif 2014 } 2015} 2016