ip_output.c revision 126239
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 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 34 * $FreeBSD: head/sys/netinet/ip_output.c 126239 2004-02-25 19:55:29Z mlaier $ 35 */ 36 37#include "opt_ipfw.h" 38#include "opt_ipdn.h" 39#include "opt_ipdivert.h" 40#include "opt_ipfilter.h" 41#include "opt_ipsec.h" 42#include "opt_mac.h" 43#include "opt_pfil_hooks.h" 44#include "opt_random_ip_id.h" 45#include "opt_mbuf_stress_test.h" 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/kernel.h> 50#include <sys/mac.h> 51#include <sys/malloc.h> 52#include <sys/mbuf.h> 53#include <sys/protosw.h> 54#include <sys/socket.h> 55#include <sys/socketvar.h> 56#include <sys/sysctl.h> 57 58#include <net/if.h> 59#include <net/route.h> 60 61#include <netinet/in.h> 62#include <netinet/in_systm.h> 63#include <netinet/ip.h> 64#include <netinet/in_pcb.h> 65#include <netinet/in_var.h> 66#include <netinet/ip_var.h> 67 68#ifdef PFIL_HOOKS 69#include <net/pfil.h> 70#endif 71 72#include <machine/in_cksum.h> 73 74static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 75 76#ifdef IPSEC 77#include <netinet6/ipsec.h> 78#include <netkey/key.h> 79#ifdef IPSEC_DEBUG 80#include <netkey/key_debug.h> 81#else 82#define KEYDEBUG(lev,arg) 83#endif 84#endif /*IPSEC*/ 85 86#ifdef FAST_IPSEC 87#include <netipsec/ipsec.h> 88#include <netipsec/xform.h> 89#include <netipsec/key.h> 90#endif /*FAST_IPSEC*/ 91 92#include <netinet/ip_fw.h> 93#include <netinet/ip_divert.h> 94#include <netinet/ip_dummynet.h> 95 96#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 97 x, (ntohl(a.s_addr)>>24)&0xFF,\ 98 (ntohl(a.s_addr)>>16)&0xFF,\ 99 (ntohl(a.s_addr)>>8)&0xFF,\ 100 (ntohl(a.s_addr))&0xFF, y); 101 102u_short ip_id; 103 104#ifdef MBUF_STRESS_TEST 105int mbuf_frag_size = 0; 106SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 107 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 108#endif 109 110static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 111static struct ifnet *ip_multicast_if(struct in_addr *, int *); 112static void ip_mloopback 113 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 114static int ip_getmoptions 115 (struct sockopt *, struct ip_moptions *); 116static int ip_pcbopts(int, struct mbuf **, struct mbuf *); 117static int ip_setmoptions 118 (struct sockopt *, struct ip_moptions **); 119 120int ip_optcopy(struct ip *, struct ip *); 121 122 123extern struct protosw inetsw[]; 124 125/* 126 * IP output. The packet in mbuf chain m contains a skeletal IP 127 * header (with len, off, ttl, proto, tos, src, dst). 128 * The mbuf chain containing the packet will be freed. 129 * The mbuf opt, if present, will not be freed. 130 * In the IP forwarding case, the packet will arrive with options already 131 * inserted, so must have a NULL opt pointer. 132 */ 133int 134ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, 135 int flags, struct ip_moptions *imo, struct inpcb *inp) 136{ 137 struct ip *ip; 138 struct ifnet *ifp = NULL; /* keep compiler happy */ 139 struct mbuf *m0; 140 int hlen = sizeof (struct ip); 141 int len, off, error = 0; 142 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 143 struct in_ifaddr *ia = NULL; 144 int isbroadcast, sw_csum; 145 struct in_addr pkt_dst; 146 struct route iproute; 147 struct m_tag *dummytag; 148#ifdef IPSEC 149 struct secpolicy *sp = NULL; 150#endif 151#ifdef FAST_IPSEC 152 struct m_tag *mtag; 153 struct secpolicy *sp = NULL; 154 struct tdb_ident *tdbi; 155 int s; 156#endif /* FAST_IPSEC */ 157 struct ip_fw_args args; 158 int src_was_INADDR_ANY = 0; /* as the name says... */ 159 160 args.eh = NULL; 161 args.rule = NULL; 162 163 M_ASSERTPKTHDR(m); 164 165 args.next_hop = ip_claim_next_hop(m); 166 dummytag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 167 if (dummytag != NULL) { 168 struct dn_pkt_tag *dt = (struct dn_pkt_tag *)(dummytag+1); 169 /* 170 * Prevent lower layers from finding the tag 171 * Cleanup and free is done below 172 */ 173 m_tag_unlink(m, dummytag); 174 /* 175 * the packet was already tagged, so part of the 176 * processing was already done, and we need to go down. 177 * Get parameters from the header. 178 */ 179 args.rule = dt->rule; 180 ro = &(dt->ro); 181 dst = dt->dn_dst; 182 ifp = dt->ifp; 183 } 184 185 if (ro == NULL) { 186 ro = &iproute; 187 bzero(ro, sizeof (*ro)); 188 } 189 190 if (inp != NULL) 191 INP_LOCK_ASSERT(inp); 192 193 if (args.rule != NULL) { /* dummynet already saw us */ 194 ip = mtod(m, struct ip *); 195 hlen = ip->ip_hl << 2 ; 196 if (ro->ro_rt) 197 ia = ifatoia(ro->ro_rt->rt_ifa); 198 goto sendit; 199 } 200 201 if (opt) { 202 len = 0; 203 m = ip_insertoptions(m, opt, &len); 204 if (len != 0) 205 hlen = len; 206 } 207 ip = mtod(m, struct ip *); 208 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 209 210 /* 211 * Fill in IP header. If we are not allowing fragmentation, 212 * then the ip_id field is meaningless, but we don't set it 213 * to zero. Doing so causes various problems when devices along 214 * the path (routers, load balancers, firewalls, etc.) illegally 215 * disable DF on our packet. Note that a 16-bit counter 216 * will wrap around in less than 10 seconds at 100 Mbit/s on a 217 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 218 * for Counting NATted Hosts", Proc. IMW'02, available at 219 * <http://www.research.att.com/~smb/papers/fnat.pdf>. 220 */ 221 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 222 ip->ip_v = IPVERSION; 223 ip->ip_hl = hlen >> 2; 224#ifdef RANDOM_IP_ID 225 ip->ip_id = ip_randomid(); 226#else 227 ip->ip_id = htons(ip_id++); 228#endif 229 ipstat.ips_localout++; 230 } else { 231 hlen = ip->ip_hl << 2; 232 } 233 234 dst = (struct sockaddr_in *)&ro->ro_dst; 235 /* 236 * If there is a cached route, 237 * check that it is to the same destination 238 * and is still up. If not, free it and try again. 239 * The address family should also be checked in case of sharing the 240 * cache with IPv6. 241 */ 242 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 243 dst->sin_family != AF_INET || 244 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 245 RTFREE(ro->ro_rt); 246 ro->ro_rt = (struct rtentry *)0; 247 } 248 if (ro->ro_rt == 0) { 249 bzero(dst, sizeof(*dst)); 250 dst->sin_family = AF_INET; 251 dst->sin_len = sizeof(*dst); 252 dst->sin_addr = pkt_dst; 253 } 254 /* 255 * If routing to interface only, 256 * short circuit routing lookup. 257 */ 258 if (flags & IP_ROUTETOIF) { 259 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 260 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 261 ipstat.ips_noroute++; 262 error = ENETUNREACH; 263 goto bad; 264 } 265 ifp = ia->ia_ifp; 266 ip->ip_ttl = 1; 267 isbroadcast = in_broadcast(dst->sin_addr, ifp); 268 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 269 imo != NULL && imo->imo_multicast_ifp != NULL) { 270 /* 271 * Bypass the normal routing lookup for multicast 272 * packets if the interface is specified. 273 */ 274 ifp = imo->imo_multicast_ifp; 275 IFP_TO_IA(ifp, ia); 276 isbroadcast = 0; /* fool gcc */ 277 } else { 278 /* 279 * We want to do any cloning requested by the link layer, 280 * as this is probably required in all cases for correct 281 * operation (as it is for ARP). 282 */ 283 if (ro->ro_rt == 0) 284 rtalloc(ro); 285 if (ro->ro_rt == 0) { 286 ipstat.ips_noroute++; 287 error = EHOSTUNREACH; 288 goto bad; 289 } 290 ia = ifatoia(ro->ro_rt->rt_ifa); 291 ifp = ro->ro_rt->rt_ifp; 292 ro->ro_rt->rt_rmx.rmx_pksent++; 293 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 294 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 295 if (ro->ro_rt->rt_flags & RTF_HOST) 296 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 297 else 298 isbroadcast = in_broadcast(dst->sin_addr, ifp); 299 } 300 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 301 struct in_multi *inm; 302 303 m->m_flags |= M_MCAST; 304 /* 305 * IP destination address is multicast. Make sure "dst" 306 * still points to the address in "ro". (It may have been 307 * changed to point to a gateway address, above.) 308 */ 309 dst = (struct sockaddr_in *)&ro->ro_dst; 310 /* 311 * See if the caller provided any multicast options 312 */ 313 if (imo != NULL) { 314 ip->ip_ttl = imo->imo_multicast_ttl; 315 if (imo->imo_multicast_vif != -1) 316 ip->ip_src.s_addr = 317 ip_mcast_src ? 318 ip_mcast_src(imo->imo_multicast_vif) : 319 INADDR_ANY; 320 } else 321 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 322 /* 323 * Confirm that the outgoing interface supports multicast. 324 */ 325 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 326 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 327 ipstat.ips_noroute++; 328 error = ENETUNREACH; 329 goto bad; 330 } 331 } 332 /* 333 * If source address not specified yet, use address 334 * of outgoing interface. 335 */ 336 if (ip->ip_src.s_addr == INADDR_ANY) { 337 /* Interface may have no addresses. */ 338 if (ia != NULL) 339 ip->ip_src = IA_SIN(ia)->sin_addr; 340 } 341 342 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 343 /* 344 * XXX 345 * delayed checksums are not currently 346 * compatible with IP multicast routing 347 */ 348 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 349 in_delayed_cksum(m); 350 m->m_pkthdr.csum_flags &= 351 ~CSUM_DELAY_DATA; 352 } 353 } 354 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 355 if (inm != NULL && 356 (imo == NULL || imo->imo_multicast_loop)) { 357 /* 358 * If we belong to the destination multicast group 359 * on the outgoing interface, and the caller did not 360 * forbid loopback, loop back a copy. 361 */ 362 ip_mloopback(ifp, m, dst, hlen); 363 } 364 else { 365 /* 366 * If we are acting as a multicast router, perform 367 * multicast forwarding as if the packet had just 368 * arrived on the interface to which we are about 369 * to send. The multicast forwarding function 370 * recursively calls this function, using the 371 * IP_FORWARDING flag to prevent infinite recursion. 372 * 373 * Multicasts that are looped back by ip_mloopback(), 374 * above, will be forwarded by the ip_input() routine, 375 * if necessary. 376 */ 377 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 378 /* 379 * If rsvp daemon is not running, do not 380 * set ip_moptions. This ensures that the packet 381 * is multicast and not just sent down one link 382 * as prescribed by rsvpd. 383 */ 384 if (!rsvp_on) 385 imo = NULL; 386 if (ip_mforward && 387 ip_mforward(ip, ifp, m, imo) != 0) { 388 m_freem(m); 389 goto done; 390 } 391 } 392 } 393 394 /* 395 * Multicasts with a time-to-live of zero may be looped- 396 * back, above, but must not be transmitted on a network. 397 * Also, multicasts addressed to the loopback interface 398 * are not sent -- the above call to ip_mloopback() will 399 * loop back a copy if this host actually belongs to the 400 * destination group on the loopback interface. 401 */ 402 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 403 m_freem(m); 404 goto done; 405 } 406 407 goto sendit; 408 } 409#ifndef notdef 410 /* 411 * If the source address is not specified yet, use the address 412 * of the outoing interface. In case, keep note we did that, so 413 * if the the firewall changes the next-hop causing the output 414 * interface to change, we can fix that. 415 */ 416 if (ip->ip_src.s_addr == INADDR_ANY) { 417 /* Interface may have no addresses. */ 418 if (ia != NULL) { 419 ip->ip_src = IA_SIN(ia)->sin_addr; 420 src_was_INADDR_ANY = 1; 421 } 422 } 423#endif /* notdef */ 424 /* 425 * Verify that we have any chance at all of being able to queue 426 * the packet or packet fragments 427 */ 428 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 429 ifp->if_snd.ifq_maxlen) { 430 error = ENOBUFS; 431 ipstat.ips_odropped++; 432 goto bad; 433 } 434 435 /* 436 * Look for broadcast address and 437 * verify user is allowed to send 438 * such a packet. 439 */ 440 if (isbroadcast) { 441 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 442 error = EADDRNOTAVAIL; 443 goto bad; 444 } 445 if ((flags & IP_ALLOWBROADCAST) == 0) { 446 error = EACCES; 447 goto bad; 448 } 449 /* don't allow broadcast messages to be fragmented */ 450 if (ip->ip_len > ifp->if_mtu) { 451 error = EMSGSIZE; 452 goto bad; 453 } 454 if (flags & IP_SENDONES) 455 ip->ip_dst.s_addr = INADDR_BROADCAST; 456 m->m_flags |= M_BCAST; 457 } else { 458 m->m_flags &= ~M_BCAST; 459 } 460 461sendit: 462#ifdef IPSEC 463 /* get SP for this packet */ 464 if (inp == NULL) 465 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 466 flags, &error); 467 else 468 sp = ipsec4_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error); 469 470 if (sp == NULL) { 471 ipsecstat.out_inval++; 472 goto bad; 473 } 474 475 error = 0; 476 477 /* check policy */ 478 switch (sp->policy) { 479 case IPSEC_POLICY_DISCARD: 480 /* 481 * This packet is just discarded. 482 */ 483 ipsecstat.out_polvio++; 484 goto bad; 485 486 case IPSEC_POLICY_BYPASS: 487 case IPSEC_POLICY_NONE: 488 case IPSEC_POLICY_TCP: 489 /* no need to do IPsec. */ 490 goto skip_ipsec; 491 492 case IPSEC_POLICY_IPSEC: 493 if (sp->req == NULL) { 494 /* acquire a policy */ 495 error = key_spdacquire(sp); 496 goto bad; 497 } 498 break; 499 500 case IPSEC_POLICY_ENTRUST: 501 default: 502 printf("ip_output: Invalid policy found. %d\n", sp->policy); 503 } 504 { 505 struct ipsec_output_state state; 506 bzero(&state, sizeof(state)); 507 state.m = m; 508 if (flags & IP_ROUTETOIF) { 509 state.ro = &iproute; 510 bzero(&iproute, sizeof(iproute)); 511 } else 512 state.ro = ro; 513 state.dst = (struct sockaddr *)dst; 514 515 ip->ip_sum = 0; 516 517 /* 518 * XXX 519 * delayed checksums are not currently compatible with IPsec 520 */ 521 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 522 in_delayed_cksum(m); 523 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 524 } 525 526 ip->ip_len = htons(ip->ip_len); 527 ip->ip_off = htons(ip->ip_off); 528 529 error = ipsec4_output(&state, sp, flags); 530 531 m = state.m; 532 if (flags & IP_ROUTETOIF) { 533 /* 534 * if we have tunnel mode SA, we may need to ignore 535 * IP_ROUTETOIF. 536 */ 537 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 538 flags &= ~IP_ROUTETOIF; 539 ro = state.ro; 540 } 541 } else 542 ro = state.ro; 543 dst = (struct sockaddr_in *)state.dst; 544 if (error) { 545 /* mbuf is already reclaimed in ipsec4_output. */ 546 m = NULL; 547 switch (error) { 548 case EHOSTUNREACH: 549 case ENETUNREACH: 550 case EMSGSIZE: 551 case ENOBUFS: 552 case ENOMEM: 553 break; 554 default: 555 printf("ip4_output (ipsec): error code %d\n", error); 556 /*fall through*/ 557 case ENOENT: 558 /* don't show these error codes to the user */ 559 error = 0; 560 break; 561 } 562 goto bad; 563 } 564 565 /* be sure to update variables that are affected by ipsec4_output() */ 566 ip = mtod(m, struct ip *); 567 hlen = ip->ip_hl << 2; 568 if (ro->ro_rt == NULL) { 569 if ((flags & IP_ROUTETOIF) == 0) { 570 printf("ip_output: " 571 "can't update route after IPsec processing\n"); 572 error = EHOSTUNREACH; /*XXX*/ 573 goto bad; 574 } 575 } else { 576 if (state.encap) { 577 ia = ifatoia(ro->ro_rt->rt_ifa); 578 ifp = ro->ro_rt->rt_ifp; 579 } 580 } 581 } 582 583 /* make it flipped, again. */ 584 ip->ip_len = ntohs(ip->ip_len); 585 ip->ip_off = ntohs(ip->ip_off); 586skip_ipsec: 587#endif /*IPSEC*/ 588#ifdef FAST_IPSEC 589 /* 590 * Check the security policy (SP) for the packet and, if 591 * required, do IPsec-related processing. There are two 592 * cases here; the first time a packet is sent through 593 * it will be untagged and handled by ipsec4_checkpolicy. 594 * If the packet is resubmitted to ip_output (e.g. after 595 * AH, ESP, etc. processing), there will be a tag to bypass 596 * the lookup and related policy checking. 597 */ 598 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 599 s = splnet(); 600 if (mtag != NULL) { 601 tdbi = (struct tdb_ident *)(mtag + 1); 602 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 603 if (sp == NULL) 604 error = -EINVAL; /* force silent drop */ 605 m_tag_delete(m, mtag); 606 } else { 607 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 608 &error, inp); 609 } 610 /* 611 * There are four return cases: 612 * sp != NULL apply IPsec policy 613 * sp == NULL, error == 0 no IPsec handling needed 614 * sp == NULL, error == -EINVAL discard packet w/o error 615 * sp == NULL, error != 0 discard packet, report error 616 */ 617 if (sp != NULL) { 618 /* Loop detection, check if ipsec processing already done */ 619 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 620 for (mtag = m_tag_first(m); mtag != NULL; 621 mtag = m_tag_next(m, mtag)) { 622 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 623 continue; 624 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 625 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 626 continue; 627 /* 628 * Check if policy has an SA associated with it. 629 * This can happen when an SP has yet to acquire 630 * an SA; e.g. on first reference. If it occurs, 631 * then we let ipsec4_process_packet do its thing. 632 */ 633 if (sp->req->sav == NULL) 634 break; 635 tdbi = (struct tdb_ident *)(mtag + 1); 636 if (tdbi->spi == sp->req->sav->spi && 637 tdbi->proto == sp->req->sav->sah->saidx.proto && 638 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 639 sizeof (union sockaddr_union)) == 0) { 640 /* 641 * No IPsec processing is needed, free 642 * reference to SP. 643 * 644 * NB: null pointer to avoid free at 645 * done: below. 646 */ 647 KEY_FREESP(&sp), sp = NULL; 648 splx(s); 649 goto spd_done; 650 } 651 } 652 653 /* 654 * Do delayed checksums now because we send before 655 * this is done in the normal processing path. 656 */ 657 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 658 in_delayed_cksum(m); 659 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 660 } 661 662 ip->ip_len = htons(ip->ip_len); 663 ip->ip_off = htons(ip->ip_off); 664 665 /* NB: callee frees mbuf */ 666 error = ipsec4_process_packet(m, sp->req, flags, 0); 667 /* 668 * Preserve KAME behaviour: ENOENT can be returned 669 * when an SA acquire is in progress. Don't propagate 670 * this to user-level; it confuses applications. 671 * 672 * XXX this will go away when the SADB is redone. 673 */ 674 if (error == ENOENT) 675 error = 0; 676 splx(s); 677 goto done; 678 } else { 679 splx(s); 680 681 if (error != 0) { 682 /* 683 * Hack: -EINVAL is used to signal that a packet 684 * should be silently discarded. This is typically 685 * because we asked key management for an SA and 686 * it was delayed (e.g. kicked up to IKE). 687 */ 688 if (error == -EINVAL) 689 error = 0; 690 goto bad; 691 } else { 692 /* No IPsec processing for this packet. */ 693 } 694#ifdef notyet 695 /* 696 * If deferred crypto processing is needed, check that 697 * the interface supports it. 698 */ 699 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 700 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 701 /* notify IPsec to do its own crypto */ 702 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 703 error = EHOSTUNREACH; 704 goto bad; 705 } 706#endif 707 } 708spd_done: 709#endif /* FAST_IPSEC */ 710 711 /* 712 * IpHack's section. 713 * - Xlate: translate packet's addr/port (NAT). 714 * - Firewall: deny/allow/etc. 715 * - Wrap: fake packet's addr/port <unimpl.> 716 * - Encapsulate: put it in another IP and send out. <unimp.> 717 */ 718#ifdef PFIL_HOOKS 719 /* 720 * Run through list of hooks for output packets. 721 */ 722 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT); 723 if (error != 0 || m == NULL) 724 goto done; 725 ip = mtod(m, struct ip *); 726#endif /* PFIL_HOOKS */ 727 728 /* 729 * Check with the firewall... 730 * but not if we are already being fwd'd from a firewall. 731 */ 732 if (fw_enable && IPFW_LOADED && !args.next_hop) { 733 struct sockaddr_in *old = dst; 734 735 args.m = m; 736 args.next_hop = dst; 737 args.oif = ifp; 738 off = ip_fw_chk_ptr(&args); 739 m = args.m; 740 dst = args.next_hop; 741 742 /* 743 * On return we must do the following: 744 * m == NULL -> drop the pkt (old interface, deprecated) 745 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 746 * 1<=off<= 0xffff -> DIVERT 747 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 748 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 749 * dst != old -> IPFIREWALL_FORWARD 750 * off==0, dst==old -> accept 751 * If some of the above modules are not compiled in, then 752 * we should't have to check the corresponding condition 753 * (because the ipfw control socket should not accept 754 * unsupported rules), but better play safe and drop 755 * packets in case of doubt. 756 */ 757 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 758 if (m) 759 m_freem(m); 760 error = EACCES; 761 goto done; 762 } 763 ip = mtod(m, struct ip *); 764 if (off == 0 && dst == old) /* common case */ 765 goto pass; 766 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 767 /* 768 * pass the pkt to dummynet. Need to include 769 * pipe number, m, ifp, ro, dst because these are 770 * not recomputed in the next pass. 771 * All other parameters have been already used and 772 * so they are not needed anymore. 773 * XXX note: if the ifp or ro entry are deleted 774 * while a pkt is in dummynet, we are in trouble! 775 */ 776 args.ro = ro; 777 args.dst = dst; 778 args.flags = flags; 779 780 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 781 &args); 782 goto done; 783 } 784#ifdef IPDIVERT 785 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 786 struct mbuf *clone; 787 788 /* Clone packet if we're doing a 'tee' */ 789 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 790 clone = divert_clone(m); 791 else 792 clone = NULL; 793 794 /* 795 * XXX 796 * delayed checksums are not currently compatible 797 * with divert sockets. 798 */ 799 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 800 in_delayed_cksum(m); 801 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 802 } 803 804 /* Restore packet header fields to original values */ 805 ip->ip_len = htons(ip->ip_len); 806 ip->ip_off = htons(ip->ip_off); 807 808 /* Deliver packet to divert input routine */ 809 divert_packet(m, 0); 810 811 /* If 'tee', continue with original packet */ 812 if (clone != NULL) { 813 m = clone; 814 ip = mtod(m, struct ip *); 815 goto pass; 816 } 817 goto done; 818 } 819#endif 820 821 /* IPFIREWALL_FORWARD */ 822 /* 823 * Check dst to make sure it is directly reachable on the 824 * interface we previously thought it was. 825 * If it isn't (which may be likely in some situations) we have 826 * to re-route it (ie, find a route for the next-hop and the 827 * associated interface) and set them here. This is nested 828 * forwarding which in most cases is undesirable, except where 829 * such control is nigh impossible. So we do it here. 830 * And I'm babbling. 831 */ 832 if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 833#if 0 834 /* 835 * XXX To improve readability, this block should be 836 * changed into a function call as below: 837 */ 838 error = ip_ipforward(&m, &dst, &ifp); 839 if (error) 840 goto bad; 841 if (m == NULL) /* ip_input consumed the mbuf */ 842 goto done; 843#else 844 struct in_ifaddr *ia; 845 846 /* 847 * XXX sro_fwd below is static, and a pointer 848 * to it gets passed to routines downstream. 849 * This could have surprisingly bad results in 850 * practice, because its content is overwritten 851 * by subsequent packets. 852 */ 853 /* There must be a better way to do this next line... */ 854 static struct route sro_fwd; 855 struct route *ro_fwd = &sro_fwd; 856 857#if 0 858 print_ip("IPFIREWALL_FORWARD: New dst ip: ", 859 dst->sin_addr, "\n"); 860#endif 861 862 /* 863 * We need to figure out if we have been forwarded 864 * to a local socket. If so, then we should somehow 865 * "loop back" to ip_input, and get directed to the 866 * PCB as if we had received this packet. This is 867 * because it may be dificult to identify the packets 868 * you want to forward until they are being output 869 * and have selected an interface. (e.g. locally 870 * initiated packets) If we used the loopback inteface, 871 * we would not be able to control what happens 872 * as the packet runs through ip_input() as 873 * it is done through an ISR. 874 */ 875 LIST_FOREACH(ia, 876 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 877 /* 878 * If the addr to forward to is one 879 * of ours, we pretend to 880 * be the destination for this packet. 881 */ 882 if (IA_SIN(ia)->sin_addr.s_addr == 883 dst->sin_addr.s_addr) 884 break; 885 } 886 if (ia) { /* tell ip_input "dont filter" */ 887 struct m_tag *mtag = m_tag_get( 888 PACKET_TAG_IPFORWARD, 889 sizeof(struct sockaddr_in *), M_NOWAIT); 890 if (mtag == NULL) { 891 error = ENOBUFS; 892 goto bad; 893 } 894 *(struct sockaddr_in **)(mtag+1) = 895 args.next_hop; 896 m_tag_prepend(m, mtag); 897 898 if (m->m_pkthdr.rcvif == NULL) 899 m->m_pkthdr.rcvif = ifunit("lo0"); 900 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 901 m->m_pkthdr.csum_flags |= 902 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 903 m->m_pkthdr.csum_data = 0xffff; 904 } 905 m->m_pkthdr.csum_flags |= 906 CSUM_IP_CHECKED | CSUM_IP_VALID; 907 ip->ip_len = htons(ip->ip_len); 908 ip->ip_off = htons(ip->ip_off); 909 ip_input(m); 910 goto done; 911 } 912 /* 913 * Some of the logic for this was 914 * nicked from above. 915 */ 916 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 917 918 ro_fwd->ro_rt = 0; 919 rtalloc_ign(ro_fwd, RTF_CLONING); 920 921 if (ro_fwd->ro_rt == 0) { 922 ipstat.ips_noroute++; 923 error = EHOSTUNREACH; 924 goto bad; 925 } 926 927 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 928 ifp = ro_fwd->ro_rt->rt_ifp; 929 ro_fwd->ro_rt->rt_rmx.rmx_pksent++; 930 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 931 dst = (struct sockaddr_in *) 932 ro_fwd->ro_rt->rt_gateway; 933 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 934 isbroadcast = 935 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 936 else 937 isbroadcast = in_broadcast(dst->sin_addr, ifp); 938 if (ro->ro_rt) 939 RTFREE(ro->ro_rt); 940 ro->ro_rt = ro_fwd->ro_rt; 941 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 942 943#endif /* ... block to be put into a function */ 944 /* 945 * If we added a default src ip earlier, 946 * which would have been gotten from the-then 947 * interface, do it again, from the new one. 948 */ 949 if (src_was_INADDR_ANY) 950 ip->ip_src = IA_SIN(ia)->sin_addr; 951 goto pass ; 952 } 953 954 /* 955 * if we get here, none of the above matches, and 956 * we have to drop the pkt 957 */ 958 m_freem(m); 959 error = EACCES; /* not sure this is the right error msg */ 960 goto done; 961 } 962 963pass: 964 /* 127/8 must not appear on wire - RFC1122. */ 965 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 966 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 967 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 968 ipstat.ips_badaddr++; 969 error = EADDRNOTAVAIL; 970 goto bad; 971 } 972 } 973 974 m->m_pkthdr.csum_flags |= CSUM_IP; 975 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 976 if (sw_csum & CSUM_DELAY_DATA) { 977 in_delayed_cksum(m); 978 sw_csum &= ~CSUM_DELAY_DATA; 979 } 980 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 981 982 /* 983 * If small enough for interface, or the interface will take 984 * care of the fragmentation for us, can just send directly. 985 */ 986 if (ip->ip_len <= ifp->if_mtu || (ifp->if_hwassist & CSUM_FRAGMENT && 987 ((ip->ip_off & IP_DF) == 0))) { 988 ip->ip_len = htons(ip->ip_len); 989 ip->ip_off = htons(ip->ip_off); 990 ip->ip_sum = 0; 991 if (sw_csum & CSUM_DELAY_IP) 992 ip->ip_sum = in_cksum(m, hlen); 993 994 /* Record statistics for this interface address. */ 995 if (!(flags & IP_FORWARDING) && ia) { 996 ia->ia_ifa.if_opackets++; 997 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 998 } 999 1000#ifdef IPSEC 1001 /* clean ipsec history once it goes out of the node */ 1002 ipsec_delaux(m); 1003#endif 1004 1005#ifdef MBUF_STRESS_TEST 1006 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 1007 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 1008#endif 1009 error = (*ifp->if_output)(ifp, m, 1010 (struct sockaddr *)dst, ro->ro_rt); 1011 goto done; 1012 } 1013 1014 if (ip->ip_off & IP_DF) { 1015 error = EMSGSIZE; 1016 /* 1017 * This case can happen if the user changed the MTU 1018 * of an interface after enabling IP on it. Because 1019 * most netifs don't keep track of routes pointing to 1020 * them, there is no way for one to update all its 1021 * routes when the MTU is changed. 1022 */ 1023 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 1024 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 1025 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 1026 } 1027 ipstat.ips_cantfrag++; 1028 goto bad; 1029 } 1030 1031 /* 1032 * Too large for interface; fragment if possible. If successful, 1033 * on return, m will point to a list of packets to be sent. 1034 */ 1035 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 1036 if (error) 1037 goto bad; 1038 for (; m; m = m0) { 1039 m0 = m->m_nextpkt; 1040 m->m_nextpkt = 0; 1041#ifdef IPSEC 1042 /* clean ipsec history once it goes out of the node */ 1043 ipsec_delaux(m); 1044#endif 1045 if (error == 0) { 1046 /* Record statistics for this interface address. */ 1047 if (ia != NULL) { 1048 ia->ia_ifa.if_opackets++; 1049 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1050 } 1051 1052 error = (*ifp->if_output)(ifp, m, 1053 (struct sockaddr *)dst, ro->ro_rt); 1054 } else 1055 m_freem(m); 1056 } 1057 1058 if (error == 0) 1059 ipstat.ips_fragmented++; 1060 1061done: 1062 if (ro == &iproute && ro->ro_rt) { 1063 RTFREE(ro->ro_rt); 1064 ro->ro_rt = NULL; 1065 } 1066 if (dummytag) { 1067 struct dn_pkt_tag *dt = (struct dn_pkt_tag *)(dummytag+1); 1068 if (dt->ro.ro_rt) 1069 RTFREE(dt->ro.ro_rt); 1070 m_tag_free(dummytag); 1071 } 1072#ifdef IPSEC 1073 if (sp != NULL) { 1074 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1075 printf("DP ip_output call free SP:%p\n", sp)); 1076 key_freesp(sp); 1077 } 1078#endif 1079#ifdef FAST_IPSEC 1080 if (sp != NULL) 1081 KEY_FREESP(&sp); 1082#endif 1083 return (error); 1084bad: 1085 m_freem(m); 1086 goto done; 1087} 1088 1089/* 1090 * Create a chain of fragments which fit the given mtu. m_frag points to the 1091 * mbuf to be fragmented; on return it points to the chain with the fragments. 1092 * Return 0 if no error. If error, m_frag may contain a partially built 1093 * chain of fragments that should be freed by the caller. 1094 * 1095 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 1096 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 1097 */ 1098int 1099ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 1100 u_long if_hwassist_flags, int sw_csum) 1101{ 1102 int error = 0; 1103 int hlen = ip->ip_hl << 2; 1104 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 1105 int off; 1106 struct mbuf *m0 = *m_frag; /* the original packet */ 1107 int firstlen; 1108 struct mbuf **mnext; 1109 int nfrags; 1110 1111 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 1112 ipstat.ips_cantfrag++; 1113 return EMSGSIZE; 1114 } 1115 1116 /* 1117 * Must be able to put at least 8 bytes per fragment. 1118 */ 1119 if (len < 8) 1120 return EMSGSIZE; 1121 1122 /* 1123 * If the interface will not calculate checksums on 1124 * fragmented packets, then do it here. 1125 */ 1126 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 1127 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 1128 in_delayed_cksum(m0); 1129 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1130 } 1131 1132 if (len > PAGE_SIZE) { 1133 /* 1134 * Fragment large datagrams such that each segment 1135 * contains a multiple of PAGE_SIZE amount of data, 1136 * plus headers. This enables a receiver to perform 1137 * page-flipping zero-copy optimizations. 1138 * 1139 * XXX When does this help given that sender and receiver 1140 * could have different page sizes, and also mtu could 1141 * be less than the receiver's page size ? 1142 */ 1143 int newlen; 1144 struct mbuf *m; 1145 1146 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 1147 off += m->m_len; 1148 1149 /* 1150 * firstlen (off - hlen) must be aligned on an 1151 * 8-byte boundary 1152 */ 1153 if (off < hlen) 1154 goto smart_frag_failure; 1155 off = ((off - hlen) & ~7) + hlen; 1156 newlen = (~PAGE_MASK) & mtu; 1157 if ((newlen + sizeof (struct ip)) > mtu) { 1158 /* we failed, go back the default */ 1159smart_frag_failure: 1160 newlen = len; 1161 off = hlen + len; 1162 } 1163 len = newlen; 1164 1165 } else { 1166 off = hlen + len; 1167 } 1168 1169 firstlen = off - hlen; 1170 mnext = &m0->m_nextpkt; /* pointer to next packet */ 1171 1172 /* 1173 * Loop through length of segment after first fragment, 1174 * make new header and copy data of each part and link onto chain. 1175 * Here, m0 is the original packet, m is the fragment being created. 1176 * The fragments are linked off the m_nextpkt of the original 1177 * packet, which after processing serves as the first fragment. 1178 */ 1179 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 1180 struct ip *mhip; /* ip header on the fragment */ 1181 struct mbuf *m; 1182 int mhlen = sizeof (struct ip); 1183 1184 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1185 if (m == 0) { 1186 error = ENOBUFS; 1187 ipstat.ips_odropped++; 1188 goto done; 1189 } 1190 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1191 /* 1192 * In the first mbuf, leave room for the link header, then 1193 * copy the original IP header including options. The payload 1194 * goes into an additional mbuf chain returned by m_copy(). 1195 */ 1196 m->m_data += max_linkhdr; 1197 mhip = mtod(m, struct ip *); 1198 *mhip = *ip; 1199 if (hlen > sizeof (struct ip)) { 1200 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1201 mhip->ip_v = IPVERSION; 1202 mhip->ip_hl = mhlen >> 2; 1203 } 1204 m->m_len = mhlen; 1205 /* XXX do we need to add ip->ip_off below ? */ 1206 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1207 if (off + len >= ip->ip_len) { /* last fragment */ 1208 len = ip->ip_len - off; 1209 m->m_flags |= M_LASTFRAG; 1210 } else 1211 mhip->ip_off |= IP_MF; 1212 mhip->ip_len = htons((u_short)(len + mhlen)); 1213 m->m_next = m_copy(m0, off, len); 1214 if (m->m_next == 0) { /* copy failed */ 1215 m_free(m); 1216 error = ENOBUFS; /* ??? */ 1217 ipstat.ips_odropped++; 1218 goto done; 1219 } 1220 m->m_pkthdr.len = mhlen + len; 1221 m->m_pkthdr.rcvif = (struct ifnet *)0; 1222#ifdef MAC 1223 mac_create_fragment(m0, m); 1224#endif 1225 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1226 mhip->ip_off = htons(mhip->ip_off); 1227 mhip->ip_sum = 0; 1228 if (sw_csum & CSUM_DELAY_IP) 1229 mhip->ip_sum = in_cksum(m, mhlen); 1230 *mnext = m; 1231 mnext = &m->m_nextpkt; 1232 } 1233 ipstat.ips_ofragments += nfrags; 1234 1235 /* set first marker for fragment chain */ 1236 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1237 m0->m_pkthdr.csum_data = nfrags; 1238 1239 /* 1240 * Update first fragment by trimming what's been copied out 1241 * and updating header. 1242 */ 1243 m_adj(m0, hlen + firstlen - ip->ip_len); 1244 m0->m_pkthdr.len = hlen + firstlen; 1245 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 1246 ip->ip_off |= IP_MF; 1247 ip->ip_off = htons(ip->ip_off); 1248 ip->ip_sum = 0; 1249 if (sw_csum & CSUM_DELAY_IP) 1250 ip->ip_sum = in_cksum(m0, hlen); 1251 1252done: 1253 *m_frag = m0; 1254 return error; 1255} 1256 1257void 1258in_delayed_cksum(struct mbuf *m) 1259{ 1260 struct ip *ip; 1261 u_short csum, offset; 1262 1263 ip = mtod(m, struct ip *); 1264 offset = ip->ip_hl << 2 ; 1265 csum = in_cksum_skip(m, ip->ip_len, offset); 1266 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1267 csum = 0xffff; 1268 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1269 1270 if (offset + sizeof(u_short) > m->m_len) { 1271 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1272 m->m_len, offset, ip->ip_p); 1273 /* 1274 * XXX 1275 * this shouldn't happen, but if it does, the 1276 * correct behavior may be to insert the checksum 1277 * in the existing chain instead of rearranging it. 1278 */ 1279 m = m_pullup(m, offset + sizeof(u_short)); 1280 } 1281 *(u_short *)(m->m_data + offset) = csum; 1282} 1283 1284/* 1285 * Insert IP options into preformed packet. 1286 * Adjust IP destination as required for IP source routing, 1287 * as indicated by a non-zero in_addr at the start of the options. 1288 * 1289 * XXX This routine assumes that the packet has no options in place. 1290 */ 1291static struct mbuf * 1292ip_insertoptions(m, opt, phlen) 1293 register struct mbuf *m; 1294 struct mbuf *opt; 1295 int *phlen; 1296{ 1297 register struct ipoption *p = mtod(opt, struct ipoption *); 1298 struct mbuf *n; 1299 register struct ip *ip = mtod(m, struct ip *); 1300 unsigned optlen; 1301 1302 optlen = opt->m_len - sizeof(p->ipopt_dst); 1303 if (optlen + ip->ip_len > IP_MAXPACKET) { 1304 *phlen = 0; 1305 return (m); /* XXX should fail */ 1306 } 1307 if (p->ipopt_dst.s_addr) 1308 ip->ip_dst = p->ipopt_dst; 1309 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1310 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1311 if (n == 0) { 1312 *phlen = 0; 1313 return (m); 1314 } 1315 n->m_pkthdr.rcvif = (struct ifnet *)0; 1316#ifdef MAC 1317 mac_create_mbuf_from_mbuf(m, n); 1318#endif 1319 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1320 m->m_len -= sizeof(struct ip); 1321 m->m_data += sizeof(struct ip); 1322 n->m_next = m; 1323 m = n; 1324 m->m_len = optlen + sizeof(struct ip); 1325 m->m_data += max_linkhdr; 1326 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1327 } else { 1328 m->m_data -= optlen; 1329 m->m_len += optlen; 1330 m->m_pkthdr.len += optlen; 1331 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1332 } 1333 ip = mtod(m, struct ip *); 1334 bcopy(p->ipopt_list, ip + 1, optlen); 1335 *phlen = sizeof(struct ip) + optlen; 1336 ip->ip_v = IPVERSION; 1337 ip->ip_hl = *phlen >> 2; 1338 ip->ip_len += optlen; 1339 return (m); 1340} 1341 1342/* 1343 * Copy options from ip to jp, 1344 * omitting those not copied during fragmentation. 1345 */ 1346int 1347ip_optcopy(ip, jp) 1348 struct ip *ip, *jp; 1349{ 1350 register u_char *cp, *dp; 1351 int opt, optlen, cnt; 1352 1353 cp = (u_char *)(ip + 1); 1354 dp = (u_char *)(jp + 1); 1355 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1356 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1357 opt = cp[0]; 1358 if (opt == IPOPT_EOL) 1359 break; 1360 if (opt == IPOPT_NOP) { 1361 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1362 *dp++ = IPOPT_NOP; 1363 optlen = 1; 1364 continue; 1365 } 1366 1367 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1368 ("ip_optcopy: malformed ipv4 option")); 1369 optlen = cp[IPOPT_OLEN]; 1370 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1371 ("ip_optcopy: malformed ipv4 option")); 1372 1373 /* bogus lengths should have been caught by ip_dooptions */ 1374 if (optlen > cnt) 1375 optlen = cnt; 1376 if (IPOPT_COPIED(opt)) { 1377 bcopy(cp, dp, optlen); 1378 dp += optlen; 1379 } 1380 } 1381 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1382 *dp++ = IPOPT_EOL; 1383 return (optlen); 1384} 1385 1386/* 1387 * IP socket option processing. 1388 */ 1389int 1390ip_ctloutput(so, sopt) 1391 struct socket *so; 1392 struct sockopt *sopt; 1393{ 1394 struct inpcb *inp = sotoinpcb(so); 1395 int error, optval; 1396 1397 error = optval = 0; 1398 if (sopt->sopt_level != IPPROTO_IP) { 1399 return (EINVAL); 1400 } 1401 1402 switch (sopt->sopt_dir) { 1403 case SOPT_SET: 1404 switch (sopt->sopt_name) { 1405 case IP_OPTIONS: 1406#ifdef notyet 1407 case IP_RETOPTS: 1408#endif 1409 { 1410 struct mbuf *m; 1411 if (sopt->sopt_valsize > MLEN) { 1412 error = EMSGSIZE; 1413 break; 1414 } 1415 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1416 if (m == 0) { 1417 error = ENOBUFS; 1418 break; 1419 } 1420 m->m_len = sopt->sopt_valsize; 1421 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1422 m->m_len); 1423 1424 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1425 m)); 1426 } 1427 1428 case IP_TOS: 1429 case IP_TTL: 1430 case IP_RECVOPTS: 1431 case IP_RECVRETOPTS: 1432 case IP_RECVDSTADDR: 1433 case IP_RECVTTL: 1434 case IP_RECVIF: 1435 case IP_FAITH: 1436 case IP_ONESBCAST: 1437 error = sooptcopyin(sopt, &optval, sizeof optval, 1438 sizeof optval); 1439 if (error) 1440 break; 1441 1442 switch (sopt->sopt_name) { 1443 case IP_TOS: 1444 inp->inp_ip_tos = optval; 1445 break; 1446 1447 case IP_TTL: 1448 inp->inp_ip_ttl = optval; 1449 break; 1450#define OPTSET(bit) \ 1451 if (optval) \ 1452 inp->inp_flags |= bit; \ 1453 else \ 1454 inp->inp_flags &= ~bit; 1455 1456 case IP_RECVOPTS: 1457 OPTSET(INP_RECVOPTS); 1458 break; 1459 1460 case IP_RECVRETOPTS: 1461 OPTSET(INP_RECVRETOPTS); 1462 break; 1463 1464 case IP_RECVDSTADDR: 1465 OPTSET(INP_RECVDSTADDR); 1466 break; 1467 1468 case IP_RECVTTL: 1469 OPTSET(INP_RECVTTL); 1470 break; 1471 1472 case IP_RECVIF: 1473 OPTSET(INP_RECVIF); 1474 break; 1475 1476 case IP_FAITH: 1477 OPTSET(INP_FAITH); 1478 break; 1479 1480 case IP_ONESBCAST: 1481 OPTSET(INP_ONESBCAST); 1482 break; 1483 } 1484 break; 1485#undef OPTSET 1486 1487 case IP_MULTICAST_IF: 1488 case IP_MULTICAST_VIF: 1489 case IP_MULTICAST_TTL: 1490 case IP_MULTICAST_LOOP: 1491 case IP_ADD_MEMBERSHIP: 1492 case IP_DROP_MEMBERSHIP: 1493 error = ip_setmoptions(sopt, &inp->inp_moptions); 1494 break; 1495 1496 case IP_PORTRANGE: 1497 error = sooptcopyin(sopt, &optval, sizeof optval, 1498 sizeof optval); 1499 if (error) 1500 break; 1501 1502 switch (optval) { 1503 case IP_PORTRANGE_DEFAULT: 1504 inp->inp_flags &= ~(INP_LOWPORT); 1505 inp->inp_flags &= ~(INP_HIGHPORT); 1506 break; 1507 1508 case IP_PORTRANGE_HIGH: 1509 inp->inp_flags &= ~(INP_LOWPORT); 1510 inp->inp_flags |= INP_HIGHPORT; 1511 break; 1512 1513 case IP_PORTRANGE_LOW: 1514 inp->inp_flags &= ~(INP_HIGHPORT); 1515 inp->inp_flags |= INP_LOWPORT; 1516 break; 1517 1518 default: 1519 error = EINVAL; 1520 break; 1521 } 1522 break; 1523 1524#if defined(IPSEC) || defined(FAST_IPSEC) 1525 case IP_IPSEC_POLICY: 1526 { 1527 caddr_t req; 1528 size_t len = 0; 1529 int priv; 1530 struct mbuf *m; 1531 int optname; 1532 1533 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1534 break; 1535 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1536 break; 1537 priv = (sopt->sopt_td != NULL && 1538 suser(sopt->sopt_td) != 0) ? 0 : 1; 1539 req = mtod(m, caddr_t); 1540 len = m->m_len; 1541 optname = sopt->sopt_name; 1542 error = ipsec4_set_policy(inp, optname, req, len, priv); 1543 m_freem(m); 1544 break; 1545 } 1546#endif /*IPSEC*/ 1547 1548 default: 1549 error = ENOPROTOOPT; 1550 break; 1551 } 1552 break; 1553 1554 case SOPT_GET: 1555 switch (sopt->sopt_name) { 1556 case IP_OPTIONS: 1557 case IP_RETOPTS: 1558 if (inp->inp_options) 1559 error = sooptcopyout(sopt, 1560 mtod(inp->inp_options, 1561 char *), 1562 inp->inp_options->m_len); 1563 else 1564 sopt->sopt_valsize = 0; 1565 break; 1566 1567 case IP_TOS: 1568 case IP_TTL: 1569 case IP_RECVOPTS: 1570 case IP_RECVRETOPTS: 1571 case IP_RECVDSTADDR: 1572 case IP_RECVTTL: 1573 case IP_RECVIF: 1574 case IP_PORTRANGE: 1575 case IP_FAITH: 1576 case IP_ONESBCAST: 1577 switch (sopt->sopt_name) { 1578 1579 case IP_TOS: 1580 optval = inp->inp_ip_tos; 1581 break; 1582 1583 case IP_TTL: 1584 optval = inp->inp_ip_ttl; 1585 break; 1586 1587#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1588 1589 case IP_RECVOPTS: 1590 optval = OPTBIT(INP_RECVOPTS); 1591 break; 1592 1593 case IP_RECVRETOPTS: 1594 optval = OPTBIT(INP_RECVRETOPTS); 1595 break; 1596 1597 case IP_RECVDSTADDR: 1598 optval = OPTBIT(INP_RECVDSTADDR); 1599 break; 1600 1601 case IP_RECVTTL: 1602 optval = OPTBIT(INP_RECVTTL); 1603 break; 1604 1605 case IP_RECVIF: 1606 optval = OPTBIT(INP_RECVIF); 1607 break; 1608 1609 case IP_PORTRANGE: 1610 if (inp->inp_flags & INP_HIGHPORT) 1611 optval = IP_PORTRANGE_HIGH; 1612 else if (inp->inp_flags & INP_LOWPORT) 1613 optval = IP_PORTRANGE_LOW; 1614 else 1615 optval = 0; 1616 break; 1617 1618 case IP_FAITH: 1619 optval = OPTBIT(INP_FAITH); 1620 break; 1621 1622 case IP_ONESBCAST: 1623 optval = OPTBIT(INP_ONESBCAST); 1624 break; 1625 } 1626 error = sooptcopyout(sopt, &optval, sizeof optval); 1627 break; 1628 1629 case IP_MULTICAST_IF: 1630 case IP_MULTICAST_VIF: 1631 case IP_MULTICAST_TTL: 1632 case IP_MULTICAST_LOOP: 1633 case IP_ADD_MEMBERSHIP: 1634 case IP_DROP_MEMBERSHIP: 1635 error = ip_getmoptions(sopt, inp->inp_moptions); 1636 break; 1637 1638#if defined(IPSEC) || defined(FAST_IPSEC) 1639 case IP_IPSEC_POLICY: 1640 { 1641 struct mbuf *m = NULL; 1642 caddr_t req = NULL; 1643 size_t len = 0; 1644 1645 if (m != 0) { 1646 req = mtod(m, caddr_t); 1647 len = m->m_len; 1648 } 1649 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1650 if (error == 0) 1651 error = soopt_mcopyout(sopt, m); /* XXX */ 1652 if (error == 0) 1653 m_freem(m); 1654 break; 1655 } 1656#endif /*IPSEC*/ 1657 1658 default: 1659 error = ENOPROTOOPT; 1660 break; 1661 } 1662 break; 1663 } 1664 return (error); 1665} 1666 1667/* 1668 * Set up IP options in pcb for insertion in output packets. 1669 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1670 * with destination address if source routed. 1671 */ 1672static int 1673ip_pcbopts(optname, pcbopt, m) 1674 int optname; 1675 struct mbuf **pcbopt; 1676 register struct mbuf *m; 1677{ 1678 register int cnt, optlen; 1679 register u_char *cp; 1680 u_char opt; 1681 1682 /* turn off any old options */ 1683 if (*pcbopt) 1684 (void)m_free(*pcbopt); 1685 *pcbopt = 0; 1686 if (m == (struct mbuf *)0 || m->m_len == 0) { 1687 /* 1688 * Only turning off any previous options. 1689 */ 1690 if (m) 1691 (void)m_free(m); 1692 return (0); 1693 } 1694 1695 if (m->m_len % sizeof(int32_t)) 1696 goto bad; 1697 /* 1698 * IP first-hop destination address will be stored before 1699 * actual options; move other options back 1700 * and clear it when none present. 1701 */ 1702 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1703 goto bad; 1704 cnt = m->m_len; 1705 m->m_len += sizeof(struct in_addr); 1706 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1707 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1708 bzero(mtod(m, void *), sizeof(struct in_addr)); 1709 1710 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1711 opt = cp[IPOPT_OPTVAL]; 1712 if (opt == IPOPT_EOL) 1713 break; 1714 if (opt == IPOPT_NOP) 1715 optlen = 1; 1716 else { 1717 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1718 goto bad; 1719 optlen = cp[IPOPT_OLEN]; 1720 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1721 goto bad; 1722 } 1723 switch (opt) { 1724 1725 default: 1726 break; 1727 1728 case IPOPT_LSRR: 1729 case IPOPT_SSRR: 1730 /* 1731 * user process specifies route as: 1732 * ->A->B->C->D 1733 * D must be our final destination (but we can't 1734 * check that since we may not have connected yet). 1735 * A is first hop destination, which doesn't appear in 1736 * actual IP option, but is stored before the options. 1737 */ 1738 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1739 goto bad; 1740 m->m_len -= sizeof(struct in_addr); 1741 cnt -= sizeof(struct in_addr); 1742 optlen -= sizeof(struct in_addr); 1743 cp[IPOPT_OLEN] = optlen; 1744 /* 1745 * Move first hop before start of options. 1746 */ 1747 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1748 sizeof(struct in_addr)); 1749 /* 1750 * Then copy rest of options back 1751 * to close up the deleted entry. 1752 */ 1753 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1754 &cp[IPOPT_OFFSET+1], 1755 (unsigned)cnt + sizeof(struct in_addr)); 1756 break; 1757 } 1758 } 1759 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1760 goto bad; 1761 *pcbopt = m; 1762 return (0); 1763 1764bad: 1765 (void)m_free(m); 1766 return (EINVAL); 1767} 1768 1769/* 1770 * XXX 1771 * The whole multicast option thing needs to be re-thought. 1772 * Several of these options are equally applicable to non-multicast 1773 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1774 * standard option (IP_TTL). 1775 */ 1776 1777/* 1778 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1779 */ 1780static struct ifnet * 1781ip_multicast_if(a, ifindexp) 1782 struct in_addr *a; 1783 int *ifindexp; 1784{ 1785 int ifindex; 1786 struct ifnet *ifp; 1787 1788 if (ifindexp) 1789 *ifindexp = 0; 1790 if (ntohl(a->s_addr) >> 24 == 0) { 1791 ifindex = ntohl(a->s_addr) & 0xffffff; 1792 if (ifindex < 0 || if_index < ifindex) 1793 return NULL; 1794 ifp = ifnet_byindex(ifindex); 1795 if (ifindexp) 1796 *ifindexp = ifindex; 1797 } else { 1798 INADDR_TO_IFP(*a, ifp); 1799 } 1800 return ifp; 1801} 1802 1803/* 1804 * Set the IP multicast options in response to user setsockopt(). 1805 */ 1806static int 1807ip_setmoptions(sopt, imop) 1808 struct sockopt *sopt; 1809 struct ip_moptions **imop; 1810{ 1811 int error = 0; 1812 int i; 1813 struct in_addr addr; 1814 struct ip_mreq mreq; 1815 struct ifnet *ifp; 1816 struct ip_moptions *imo = *imop; 1817 struct route ro; 1818 struct sockaddr_in *dst; 1819 int ifindex; 1820 int s; 1821 1822 if (imo == NULL) { 1823 /* 1824 * No multicast option buffer attached to the pcb; 1825 * allocate one and initialize to default values. 1826 */ 1827 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1828 M_WAITOK); 1829 1830 if (imo == NULL) 1831 return (ENOBUFS); 1832 *imop = imo; 1833 imo->imo_multicast_ifp = NULL; 1834 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1835 imo->imo_multicast_vif = -1; 1836 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1837 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1838 imo->imo_num_memberships = 0; 1839 } 1840 1841 switch (sopt->sopt_name) { 1842 /* store an index number for the vif you wanna use in the send */ 1843 case IP_MULTICAST_VIF: 1844 if (legal_vif_num == 0) { 1845 error = EOPNOTSUPP; 1846 break; 1847 } 1848 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1849 if (error) 1850 break; 1851 if (!legal_vif_num(i) && (i != -1)) { 1852 error = EINVAL; 1853 break; 1854 } 1855 imo->imo_multicast_vif = i; 1856 break; 1857 1858 case IP_MULTICAST_IF: 1859 /* 1860 * Select the interface for outgoing multicast packets. 1861 */ 1862 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1863 if (error) 1864 break; 1865 /* 1866 * INADDR_ANY is used to remove a previous selection. 1867 * When no interface is selected, a default one is 1868 * chosen every time a multicast packet is sent. 1869 */ 1870 if (addr.s_addr == INADDR_ANY) { 1871 imo->imo_multicast_ifp = NULL; 1872 break; 1873 } 1874 /* 1875 * The selected interface is identified by its local 1876 * IP address. Find the interface and confirm that 1877 * it supports multicasting. 1878 */ 1879 s = splimp(); 1880 ifp = ip_multicast_if(&addr, &ifindex); 1881 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1882 splx(s); 1883 error = EADDRNOTAVAIL; 1884 break; 1885 } 1886 imo->imo_multicast_ifp = ifp; 1887 if (ifindex) 1888 imo->imo_multicast_addr = addr; 1889 else 1890 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1891 splx(s); 1892 break; 1893 1894 case IP_MULTICAST_TTL: 1895 /* 1896 * Set the IP time-to-live for outgoing multicast packets. 1897 * The original multicast API required a char argument, 1898 * which is inconsistent with the rest of the socket API. 1899 * We allow either a char or an int. 1900 */ 1901 if (sopt->sopt_valsize == 1) { 1902 u_char ttl; 1903 error = sooptcopyin(sopt, &ttl, 1, 1); 1904 if (error) 1905 break; 1906 imo->imo_multicast_ttl = ttl; 1907 } else { 1908 u_int ttl; 1909 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1910 sizeof ttl); 1911 if (error) 1912 break; 1913 if (ttl > 255) 1914 error = EINVAL; 1915 else 1916 imo->imo_multicast_ttl = ttl; 1917 } 1918 break; 1919 1920 case IP_MULTICAST_LOOP: 1921 /* 1922 * Set the loopback flag for outgoing multicast packets. 1923 * Must be zero or one. The original multicast API required a 1924 * char argument, which is inconsistent with the rest 1925 * of the socket API. We allow either a char or an int. 1926 */ 1927 if (sopt->sopt_valsize == 1) { 1928 u_char loop; 1929 error = sooptcopyin(sopt, &loop, 1, 1); 1930 if (error) 1931 break; 1932 imo->imo_multicast_loop = !!loop; 1933 } else { 1934 u_int loop; 1935 error = sooptcopyin(sopt, &loop, sizeof loop, 1936 sizeof loop); 1937 if (error) 1938 break; 1939 imo->imo_multicast_loop = !!loop; 1940 } 1941 break; 1942 1943 case IP_ADD_MEMBERSHIP: 1944 /* 1945 * Add a multicast group membership. 1946 * Group must be a valid IP multicast address. 1947 */ 1948 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1949 if (error) 1950 break; 1951 1952 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1953 error = EINVAL; 1954 break; 1955 } 1956 s = splimp(); 1957 /* 1958 * If no interface address was provided, use the interface of 1959 * the route to the given multicast address. 1960 */ 1961 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1962 bzero((caddr_t)&ro, sizeof(ro)); 1963 dst = (struct sockaddr_in *)&ro.ro_dst; 1964 dst->sin_len = sizeof(*dst); 1965 dst->sin_family = AF_INET; 1966 dst->sin_addr = mreq.imr_multiaddr; 1967 rtalloc_ign(&ro, RTF_CLONING); 1968 if (ro.ro_rt == NULL) { 1969 error = EADDRNOTAVAIL; 1970 splx(s); 1971 break; 1972 } 1973 ifp = ro.ro_rt->rt_ifp; 1974 RTFREE(ro.ro_rt); 1975 } 1976 else { 1977 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1978 } 1979 1980 /* 1981 * See if we found an interface, and confirm that it 1982 * supports multicast. 1983 */ 1984 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1985 error = EADDRNOTAVAIL; 1986 splx(s); 1987 break; 1988 } 1989 /* 1990 * See if the membership already exists or if all the 1991 * membership slots are full. 1992 */ 1993 for (i = 0; i < imo->imo_num_memberships; ++i) { 1994 if (imo->imo_membership[i]->inm_ifp == ifp && 1995 imo->imo_membership[i]->inm_addr.s_addr 1996 == mreq.imr_multiaddr.s_addr) 1997 break; 1998 } 1999 if (i < imo->imo_num_memberships) { 2000 error = EADDRINUSE; 2001 splx(s); 2002 break; 2003 } 2004 if (i == IP_MAX_MEMBERSHIPS) { 2005 error = ETOOMANYREFS; 2006 splx(s); 2007 break; 2008 } 2009 /* 2010 * Everything looks good; add a new record to the multicast 2011 * address list for the given interface. 2012 */ 2013 if ((imo->imo_membership[i] = 2014 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 2015 error = ENOBUFS; 2016 splx(s); 2017 break; 2018 } 2019 ++imo->imo_num_memberships; 2020 splx(s); 2021 break; 2022 2023 case IP_DROP_MEMBERSHIP: 2024 /* 2025 * Drop a multicast group membership. 2026 * Group must be a valid IP multicast address. 2027 */ 2028 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 2029 if (error) 2030 break; 2031 2032 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 2033 error = EINVAL; 2034 break; 2035 } 2036 2037 s = splimp(); 2038 /* 2039 * If an interface address was specified, get a pointer 2040 * to its ifnet structure. 2041 */ 2042 if (mreq.imr_interface.s_addr == INADDR_ANY) 2043 ifp = NULL; 2044 else { 2045 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 2046 if (ifp == NULL) { 2047 error = EADDRNOTAVAIL; 2048 splx(s); 2049 break; 2050 } 2051 } 2052 /* 2053 * Find the membership in the membership array. 2054 */ 2055 for (i = 0; i < imo->imo_num_memberships; ++i) { 2056 if ((ifp == NULL || 2057 imo->imo_membership[i]->inm_ifp == ifp) && 2058 imo->imo_membership[i]->inm_addr.s_addr == 2059 mreq.imr_multiaddr.s_addr) 2060 break; 2061 } 2062 if (i == imo->imo_num_memberships) { 2063 error = EADDRNOTAVAIL; 2064 splx(s); 2065 break; 2066 } 2067 /* 2068 * Give up the multicast address record to which the 2069 * membership points. 2070 */ 2071 in_delmulti(imo->imo_membership[i]); 2072 /* 2073 * Remove the gap in the membership array. 2074 */ 2075 for (++i; i < imo->imo_num_memberships; ++i) 2076 imo->imo_membership[i-1] = imo->imo_membership[i]; 2077 --imo->imo_num_memberships; 2078 splx(s); 2079 break; 2080 2081 default: 2082 error = EOPNOTSUPP; 2083 break; 2084 } 2085 2086 /* 2087 * If all options have default values, no need to keep the mbuf. 2088 */ 2089 if (imo->imo_multicast_ifp == NULL && 2090 imo->imo_multicast_vif == -1 && 2091 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2092 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2093 imo->imo_num_memberships == 0) { 2094 free(*imop, M_IPMOPTS); 2095 *imop = NULL; 2096 } 2097 2098 return (error); 2099} 2100 2101/* 2102 * Return the IP multicast options in response to user getsockopt(). 2103 */ 2104static int 2105ip_getmoptions(sopt, imo) 2106 struct sockopt *sopt; 2107 register struct ip_moptions *imo; 2108{ 2109 struct in_addr addr; 2110 struct in_ifaddr *ia; 2111 int error, optval; 2112 u_char coptval; 2113 2114 error = 0; 2115 switch (sopt->sopt_name) { 2116 case IP_MULTICAST_VIF: 2117 if (imo != NULL) 2118 optval = imo->imo_multicast_vif; 2119 else 2120 optval = -1; 2121 error = sooptcopyout(sopt, &optval, sizeof optval); 2122 break; 2123 2124 case IP_MULTICAST_IF: 2125 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2126 addr.s_addr = INADDR_ANY; 2127 else if (imo->imo_multicast_addr.s_addr) { 2128 /* return the value user has set */ 2129 addr = imo->imo_multicast_addr; 2130 } else { 2131 IFP_TO_IA(imo->imo_multicast_ifp, ia); 2132 addr.s_addr = (ia == NULL) ? INADDR_ANY 2133 : IA_SIN(ia)->sin_addr.s_addr; 2134 } 2135 error = sooptcopyout(sopt, &addr, sizeof addr); 2136 break; 2137 2138 case IP_MULTICAST_TTL: 2139 if (imo == 0) 2140 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2141 else 2142 optval = coptval = imo->imo_multicast_ttl; 2143 if (sopt->sopt_valsize == 1) 2144 error = sooptcopyout(sopt, &coptval, 1); 2145 else 2146 error = sooptcopyout(sopt, &optval, sizeof optval); 2147 break; 2148 2149 case IP_MULTICAST_LOOP: 2150 if (imo == 0) 2151 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2152 else 2153 optval = coptval = imo->imo_multicast_loop; 2154 if (sopt->sopt_valsize == 1) 2155 error = sooptcopyout(sopt, &coptval, 1); 2156 else 2157 error = sooptcopyout(sopt, &optval, sizeof optval); 2158 break; 2159 2160 default: 2161 error = ENOPROTOOPT; 2162 break; 2163 } 2164 return (error); 2165} 2166 2167/* 2168 * Discard the IP multicast options. 2169 */ 2170void 2171ip_freemoptions(imo) 2172 register struct ip_moptions *imo; 2173{ 2174 register int i; 2175 2176 if (imo != NULL) { 2177 for (i = 0; i < imo->imo_num_memberships; ++i) 2178 in_delmulti(imo->imo_membership[i]); 2179 free(imo, M_IPMOPTS); 2180 } 2181} 2182 2183/* 2184 * Routine called from ip_output() to loop back a copy of an IP multicast 2185 * packet to the input queue of a specified interface. Note that this 2186 * calls the output routine of the loopback "driver", but with an interface 2187 * pointer that might NOT be a loopback interface -- evil, but easier than 2188 * replicating that code here. 2189 */ 2190static void 2191ip_mloopback(ifp, m, dst, hlen) 2192 struct ifnet *ifp; 2193 register struct mbuf *m; 2194 register struct sockaddr_in *dst; 2195 int hlen; 2196{ 2197 register struct ip *ip; 2198 struct mbuf *copym; 2199 2200 copym = m_copy(m, 0, M_COPYALL); 2201 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2202 copym = m_pullup(copym, hlen); 2203 if (copym != NULL) { 2204 /* 2205 * We don't bother to fragment if the IP length is greater 2206 * than the interface's MTU. Can this possibly matter? 2207 */ 2208 ip = mtod(copym, struct ip *); 2209 ip->ip_len = htons(ip->ip_len); 2210 ip->ip_off = htons(ip->ip_off); 2211 ip->ip_sum = 0; 2212 ip->ip_sum = in_cksum(copym, hlen); 2213 /* 2214 * NB: 2215 * It's not clear whether there are any lingering 2216 * reentrancy problems in other areas which might 2217 * be exposed by using ip_input directly (in 2218 * particular, everything which modifies the packet 2219 * in-place). Yet another option is using the 2220 * protosw directly to deliver the looped back 2221 * packet. For the moment, we'll err on the side 2222 * of safety by using if_simloop(). 2223 */ 2224#if 1 /* XXX */ 2225 if (dst->sin_family != AF_INET) { 2226 printf("ip_mloopback: bad address family %d\n", 2227 dst->sin_family); 2228 dst->sin_family = AF_INET; 2229 } 2230#endif 2231 2232#ifdef notdef 2233 copym->m_pkthdr.rcvif = ifp; 2234 ip_input(copym); 2235#else 2236 /* if the checksum hasn't been computed, mark it as valid */ 2237 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2238 copym->m_pkthdr.csum_flags |= 2239 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2240 copym->m_pkthdr.csum_data = 0xffff; 2241 } 2242 if_simloop(ifp, copym, dst->sin_family, 0); 2243#endif 2244 } 2245} 2246