ip_output.c revision 125875
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 125875 2004-02-16 17:05:06Z ume $ 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 int hlen = sizeof (struct ip); 140 int len, off, error = 0; 141 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 142 struct in_ifaddr *ia = NULL; 143 int isbroadcast, sw_csum; 144 struct in_addr pkt_dst; 145 struct route iproute; 146 struct m_tag *dummytag; /* dummynet packet tag */ 147 struct m_tag *mtag; 148 struct mbuf *m0; /* XXX */ 149#ifdef IPSEC 150 struct secpolicy *sp = NULL; 151#endif 152#ifdef FAST_IPSEC 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 args.next_hop = ip_claim_next_hop(m); 163 164 M_ASSERTPKTHDR(m); 165 166 if (ro == NULL) { 167 ro = &iproute; 168 bzero(ro, sizeof (*ro)); 169 } 170 171 if (inp != NULL) 172 INP_LOCK_ASSERT(inp); 173 174 /* 175 * When packet comes from dummynet restore state from 176 * previous processing instead of the header. Yech! 177 * 178 * XXX add conditional compilation? 179 */ 180 dummytag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 181 if (dummytag != NULL) { 182 struct dn_pkt_tag *dt = (struct dn_pkt_tag *)(dummytag+1); 183 184 /* 185 * NB: the route in the tag is known to have a 186 * reference that must be free'd, but doing this 187 * before the storage is reclaimed is painful due 188 * to some of the contorted code in this routine. 189 * So instead unlink the tag from the mbuf so it 190 * doesn't get reclaimed and do the cleanup explicitly 191 * below. We should be able to do this automatically 192 * using a uma dtor method when m_tag's can be 193 * allocated from zones. 194 */ 195 m_tag_unlink(m, dummytag); 196 197 args.rule = dt->rule; 198 ro = &dt->ro; 199 dst = dt->dn_dst; 200 ifp = dt->ifp; 201 202 ip = mtod(m, struct ip *); 203 hlen = ip->ip_hl << 2 ; 204 if (ro->ro_rt) 205 ia = ifatoia(ro->ro_rt->rt_ifa); 206 goto sendit; 207 } 208 209 if (opt) { 210 len = 0; 211 m = ip_insertoptions(m, opt, &len); 212 if (len != 0) 213 hlen = len; 214 } 215 ip = mtod(m, struct ip *); 216 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 217 218 /* 219 * Fill in IP header. If we are not allowing fragmentation, 220 * then the ip_id field is meaningless, but we don't set it 221 * to zero. Doing so causes various problems when devices along 222 * the path (routers, load balancers, firewalls, etc.) illegally 223 * disable DF on our packet. Note that a 16-bit counter 224 * will wrap around in less than 10 seconds at 100 Mbit/s on a 225 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 226 * for Counting NATted Hosts", Proc. IMW'02, available at 227 * <http://www.research.att.com/~smb/papers/fnat.pdf>. 228 */ 229 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 230 ip->ip_v = IPVERSION; 231 ip->ip_hl = hlen >> 2; 232#ifdef RANDOM_IP_ID 233 ip->ip_id = ip_randomid(); 234#else 235 ip->ip_id = htons(ip_id++); 236#endif 237 ipstat.ips_localout++; 238 } else { 239 hlen = ip->ip_hl << 2; 240 } 241 242 dst = (struct sockaddr_in *)&ro->ro_dst; 243 /* 244 * If there is a cached route, 245 * check that it is to the same destination 246 * and is still up. If not, free it and try again. 247 * The address family should also be checked in case of sharing the 248 * cache with IPv6. 249 */ 250 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 251 dst->sin_family != AF_INET || 252 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 253 RTFREE(ro->ro_rt); 254 ro->ro_rt = (struct rtentry *)0; 255 } 256 if (ro->ro_rt == 0) { 257 bzero(dst, sizeof(*dst)); 258 dst->sin_family = AF_INET; 259 dst->sin_len = sizeof(*dst); 260 dst->sin_addr = pkt_dst; 261 } 262 /* 263 * If routing to interface only, 264 * short circuit routing lookup. 265 */ 266 if (flags & IP_ROUTETOIF) { 267 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 268 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 269 ipstat.ips_noroute++; 270 error = ENETUNREACH; 271 goto bad; 272 } 273 ifp = ia->ia_ifp; 274 ip->ip_ttl = 1; 275 isbroadcast = in_broadcast(dst->sin_addr, ifp); 276 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 277 imo != NULL && imo->imo_multicast_ifp != NULL) { 278 /* 279 * Bypass the normal routing lookup for multicast 280 * packets if the interface is specified. 281 */ 282 ifp = imo->imo_multicast_ifp; 283 IFP_TO_IA(ifp, ia); 284 isbroadcast = 0; /* fool gcc */ 285 } else { 286 /* 287 * We want to do any cloning requested by the link layer, 288 * as this is probably required in all cases for correct 289 * operation (as it is for ARP). 290 */ 291 if (ro->ro_rt == 0) 292 rtalloc(ro); 293 if (ro->ro_rt == 0) { 294 ipstat.ips_noroute++; 295 error = EHOSTUNREACH; 296 goto bad; 297 } 298 ia = ifatoia(ro->ro_rt->rt_ifa); 299 ifp = ro->ro_rt->rt_ifp; 300 ro->ro_rt->rt_rmx.rmx_pksent++; 301 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 302 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 303 if (ro->ro_rt->rt_flags & RTF_HOST) 304 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 305 else 306 isbroadcast = in_broadcast(dst->sin_addr, ifp); 307 } 308 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 309 struct in_multi *inm; 310 311 m->m_flags |= M_MCAST; 312 /* 313 * IP destination address is multicast. Make sure "dst" 314 * still points to the address in "ro". (It may have been 315 * changed to point to a gateway address, above.) 316 */ 317 dst = (struct sockaddr_in *)&ro->ro_dst; 318 /* 319 * See if the caller provided any multicast options 320 */ 321 if (imo != NULL) { 322 ip->ip_ttl = imo->imo_multicast_ttl; 323 if (imo->imo_multicast_vif != -1) 324 ip->ip_src.s_addr = 325 ip_mcast_src ? 326 ip_mcast_src(imo->imo_multicast_vif) : 327 INADDR_ANY; 328 } else 329 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 330 /* 331 * Confirm that the outgoing interface supports multicast. 332 */ 333 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 334 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 335 ipstat.ips_noroute++; 336 error = ENETUNREACH; 337 goto bad; 338 } 339 } 340 /* 341 * If source address not specified yet, use address 342 * of outgoing interface. 343 */ 344 if (ip->ip_src.s_addr == INADDR_ANY) { 345 /* Interface may have no addresses. */ 346 if (ia != NULL) 347 ip->ip_src = IA_SIN(ia)->sin_addr; 348 } 349 350 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 351 /* 352 * XXX 353 * delayed checksums are not currently 354 * compatible with IP multicast routing 355 */ 356 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 357 in_delayed_cksum(m); 358 m->m_pkthdr.csum_flags &= 359 ~CSUM_DELAY_DATA; 360 } 361 } 362 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 363 if (inm != NULL && 364 (imo == NULL || imo->imo_multicast_loop)) { 365 /* 366 * If we belong to the destination multicast group 367 * on the outgoing interface, and the caller did not 368 * forbid loopback, loop back a copy. 369 */ 370 ip_mloopback(ifp, m, dst, hlen); 371 } 372 else { 373 /* 374 * If we are acting as a multicast router, perform 375 * multicast forwarding as if the packet had just 376 * arrived on the interface to which we are about 377 * to send. The multicast forwarding function 378 * recursively calls this function, using the 379 * IP_FORWARDING flag to prevent infinite recursion. 380 * 381 * Multicasts that are looped back by ip_mloopback(), 382 * above, will be forwarded by the ip_input() routine, 383 * if necessary. 384 */ 385 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 386 /* 387 * If rsvp daemon is not running, do not 388 * set ip_moptions. This ensures that the packet 389 * is multicast and not just sent down one link 390 * as prescribed by rsvpd. 391 */ 392 if (!rsvp_on) 393 imo = NULL; 394 if (ip_mforward && 395 ip_mforward(ip, ifp, m, imo) != 0) { 396 m_freem(m); 397 goto done; 398 } 399 } 400 } 401 402 /* 403 * Multicasts with a time-to-live of zero may be looped- 404 * back, above, but must not be transmitted on a network. 405 * Also, multicasts addressed to the loopback interface 406 * are not sent -- the above call to ip_mloopback() will 407 * loop back a copy if this host actually belongs to the 408 * destination group on the loopback interface. 409 */ 410 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 411 m_freem(m); 412 goto done; 413 } 414 415 goto sendit; 416 } 417#ifndef notdef 418 /* 419 * If the source address is not specified yet, use the address 420 * of the outoing interface. In case, keep note we did that, so 421 * if the the firewall changes the next-hop causing the output 422 * interface to change, we can fix that. 423 */ 424 if (ip->ip_src.s_addr == INADDR_ANY) { 425 /* Interface may have no addresses. */ 426 if (ia != NULL) { 427 ip->ip_src = IA_SIN(ia)->sin_addr; 428 src_was_INADDR_ANY = 1; 429 } 430 } 431#endif /* notdef */ 432 /* 433 * Verify that we have any chance at all of being able to queue 434 * the packet or packet fragments 435 */ 436 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 437 ifp->if_snd.ifq_maxlen) { 438 error = ENOBUFS; 439 ipstat.ips_odropped++; 440 goto bad; 441 } 442 443 /* 444 * Look for broadcast address and 445 * verify user is allowed to send 446 * such a packet. 447 */ 448 if (isbroadcast) { 449 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 450 error = EADDRNOTAVAIL; 451 goto bad; 452 } 453 if ((flags & IP_ALLOWBROADCAST) == 0) { 454 error = EACCES; 455 goto bad; 456 } 457 /* don't allow broadcast messages to be fragmented */ 458 if (ip->ip_len > ifp->if_mtu) { 459 error = EMSGSIZE; 460 goto bad; 461 } 462 if (flags & IP_SENDONES) 463 ip->ip_dst.s_addr = INADDR_BROADCAST; 464 m->m_flags |= M_BCAST; 465 } else { 466 m->m_flags &= ~M_BCAST; 467 } 468 469sendit: 470#ifdef IPSEC 471 /* get SP for this packet */ 472 if (inp == NULL) 473 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 474 flags, &error); 475 else 476 sp = ipsec4_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error); 477 478 if (sp == NULL) { 479 ipsecstat.out_inval++; 480 goto bad; 481 } 482 483 error = 0; 484 485 /* check policy */ 486 switch (sp->policy) { 487 case IPSEC_POLICY_DISCARD: 488 /* 489 * This packet is just discarded. 490 */ 491 ipsecstat.out_polvio++; 492 goto bad; 493 494 case IPSEC_POLICY_BYPASS: 495 case IPSEC_POLICY_NONE: 496 case IPSEC_POLICY_TCP: 497 /* no need to do IPsec. */ 498 goto skip_ipsec; 499 500 case IPSEC_POLICY_IPSEC: 501 if (sp->req == NULL) { 502 /* acquire a policy */ 503 error = key_spdacquire(sp); 504 goto bad; 505 } 506 break; 507 508 case IPSEC_POLICY_ENTRUST: 509 default: 510 printf("ip_output: Invalid policy found. %d\n", sp->policy); 511 } 512 { 513 struct ipsec_output_state state; 514 bzero(&state, sizeof(state)); 515 state.m = m; 516 if (flags & IP_ROUTETOIF) { 517 state.ro = &iproute; 518 bzero(&iproute, sizeof(iproute)); 519 } else 520 state.ro = ro; 521 state.dst = (struct sockaddr *)dst; 522 523 ip->ip_sum = 0; 524 525 /* 526 * XXX 527 * delayed checksums are not currently compatible with IPsec 528 */ 529 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 530 in_delayed_cksum(m); 531 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 532 } 533 534 ip->ip_len = htons(ip->ip_len); 535 ip->ip_off = htons(ip->ip_off); 536 537 error = ipsec4_output(&state, sp, flags); 538 539 m = state.m; 540 if (flags & IP_ROUTETOIF) { 541 /* 542 * if we have tunnel mode SA, we may need to ignore 543 * IP_ROUTETOIF. 544 */ 545 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 546 flags &= ~IP_ROUTETOIF; 547 ro = state.ro; 548 } 549 } else 550 ro = state.ro; 551 dst = (struct sockaddr_in *)state.dst; 552 if (error) { 553 /* mbuf is already reclaimed in ipsec4_output. */ 554 m = NULL; 555 switch (error) { 556 case EHOSTUNREACH: 557 case ENETUNREACH: 558 case EMSGSIZE: 559 case ENOBUFS: 560 case ENOMEM: 561 break; 562 default: 563 printf("ip4_output (ipsec): error code %d\n", error); 564 /*fall through*/ 565 case ENOENT: 566 /* don't show these error codes to the user */ 567 error = 0; 568 break; 569 } 570 goto bad; 571 } 572 573 /* be sure to update variables that are affected by ipsec4_output() */ 574 ip = mtod(m, struct ip *); 575 hlen = ip->ip_hl << 2; 576 if (ro->ro_rt == NULL) { 577 if ((flags & IP_ROUTETOIF) == 0) { 578 printf("ip_output: " 579 "can't update route after IPsec processing\n"); 580 error = EHOSTUNREACH; /*XXX*/ 581 goto bad; 582 } 583 } else { 584 if (state.encap) { 585 ia = ifatoia(ro->ro_rt->rt_ifa); 586 ifp = ro->ro_rt->rt_ifp; 587 } 588 } 589 } 590 591 /* make it flipped, again. */ 592 ip->ip_len = ntohs(ip->ip_len); 593 ip->ip_off = ntohs(ip->ip_off); 594skip_ipsec: 595#endif /*IPSEC*/ 596#ifdef FAST_IPSEC 597 /* 598 * Check the security policy (SP) for the packet and, if 599 * required, do IPsec-related processing. There are two 600 * cases here; the first time a packet is sent through 601 * it will be untagged and handled by ipsec4_checkpolicy. 602 * If the packet is resubmitted to ip_output (e.g. after 603 * AH, ESP, etc. processing), there will be a tag to bypass 604 * the lookup and related policy checking. 605 */ 606 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 607 s = splnet(); 608 if (mtag != NULL) { 609 tdbi = (struct tdb_ident *)(mtag + 1); 610 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 611 if (sp == NULL) 612 error = -EINVAL; /* force silent drop */ 613 m_tag_delete(m, mtag); 614 } else { 615 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 616 &error, inp); 617 } 618 /* 619 * There are four return cases: 620 * sp != NULL apply IPsec policy 621 * sp == NULL, error == 0 no IPsec handling needed 622 * sp == NULL, error == -EINVAL discard packet w/o error 623 * sp == NULL, error != 0 discard packet, report error 624 */ 625 if (sp != NULL) { 626 /* Loop detection, check if ipsec processing already done */ 627 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 628 for (mtag = m_tag_first(m); mtag != NULL; 629 mtag = m_tag_next(m, mtag)) { 630 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 631 continue; 632 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 633 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 634 continue; 635 /* 636 * Check if policy has an SA associated with it. 637 * This can happen when an SP has yet to acquire 638 * an SA; e.g. on first reference. If it occurs, 639 * then we let ipsec4_process_packet do its thing. 640 */ 641 if (sp->req->sav == NULL) 642 break; 643 tdbi = (struct tdb_ident *)(mtag + 1); 644 if (tdbi->spi == sp->req->sav->spi && 645 tdbi->proto == sp->req->sav->sah->saidx.proto && 646 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 647 sizeof (union sockaddr_union)) == 0) { 648 /* 649 * No IPsec processing is needed, free 650 * reference to SP. 651 * 652 * NB: null pointer to avoid free at 653 * done: below. 654 */ 655 KEY_FREESP(&sp), sp = NULL; 656 splx(s); 657 goto spd_done; 658 } 659 } 660 661 /* 662 * Do delayed checksums now because we send before 663 * this is done in the normal processing path. 664 */ 665 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 666 in_delayed_cksum(m); 667 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 668 } 669 670 ip->ip_len = htons(ip->ip_len); 671 ip->ip_off = htons(ip->ip_off); 672 673 /* NB: callee frees mbuf */ 674 error = ipsec4_process_packet(m, sp->req, flags, 0); 675 /* 676 * Preserve KAME behaviour: ENOENT can be returned 677 * when an SA acquire is in progress. Don't propagate 678 * this to user-level; it confuses applications. 679 * 680 * XXX this will go away when the SADB is redone. 681 */ 682 if (error == ENOENT) 683 error = 0; 684 splx(s); 685 goto done; 686 } else { 687 splx(s); 688 689 if (error != 0) { 690 /* 691 * Hack: -EINVAL is used to signal that a packet 692 * should be silently discarded. This is typically 693 * because we asked key management for an SA and 694 * it was delayed (e.g. kicked up to IKE). 695 */ 696 if (error == -EINVAL) 697 error = 0; 698 goto bad; 699 } else { 700 /* No IPsec processing for this packet. */ 701 } 702#ifdef notyet 703 /* 704 * If deferred crypto processing is needed, check that 705 * the interface supports it. 706 */ 707 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 708 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 709 /* notify IPsec to do its own crypto */ 710 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 711 error = EHOSTUNREACH; 712 goto bad; 713 } 714#endif 715 } 716spd_done: 717#endif /* FAST_IPSEC */ 718 719 /* 720 * IpHack's section. 721 * - Xlate: translate packet's addr/port (NAT). 722 * - Firewall: deny/allow/etc. 723 * - Wrap: fake packet's addr/port <unimpl.> 724 * - Encapsulate: put it in another IP and send out. <unimp.> 725 */ 726#ifdef PFIL_HOOKS 727 /* 728 * Run through list of hooks for output packets. 729 */ 730 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT); 731 if (error != 0 || m == NULL) 732 goto done; 733 ip = mtod(m, struct ip *); 734#endif /* PFIL_HOOKS */ 735 736 /* 737 * Check with the firewall... 738 * but not if we are already being fwd'd from a firewall. 739 */ 740 if (fw_enable && IPFW_LOADED && !args.next_hop) { 741 struct sockaddr_in *old = dst; 742 743 args.m = m; 744 args.next_hop = dst; 745 args.oif = ifp; 746 off = ip_fw_chk_ptr(&args); 747 m = args.m; 748 dst = args.next_hop; 749 750 /* 751 * On return we must do the following: 752 * m == NULL -> drop the pkt (old interface, deprecated) 753 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 754 * 1<=off<= 0xffff -> DIVERT 755 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 756 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 757 * dst != old -> IPFIREWALL_FORWARD 758 * off==0, dst==old -> accept 759 * If some of the above modules are not compiled in, then 760 * we should't have to check the corresponding condition 761 * (because the ipfw control socket should not accept 762 * unsupported rules), but better play safe and drop 763 * packets in case of doubt. 764 */ 765 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 766 if (m) 767 m_freem(m); 768 error = EACCES; 769 goto done; 770 } 771 ip = mtod(m, struct ip *); 772 if (off == 0 && dst == old) /* common case */ 773 goto pass; 774 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 775 /* 776 * pass the pkt to dummynet. Need to include 777 * pipe number, m, ifp, ro, dst because these are 778 * not recomputed in the next pass. 779 * All other parameters have been already used and 780 * so they are not needed anymore. 781 * XXX note: if the ifp or ro entry are deleted 782 * while a pkt is in dummynet, we are in trouble! 783 */ 784 args.ro = ro; 785 args.dst = dst; 786 args.flags = flags; 787 788 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 789 &args); 790 goto done; 791 } 792#ifdef IPDIVERT 793 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 794 struct mbuf *clone; 795 796 /* Clone packet if we're doing a 'tee' */ 797 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 798 clone = divert_clone(m); 799 else 800 clone = NULL; 801 802 /* 803 * XXX 804 * delayed checksums are not currently compatible 805 * with divert sockets. 806 */ 807 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 808 in_delayed_cksum(m); 809 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 810 } 811 812 /* Restore packet header fields to original values */ 813 ip->ip_len = htons(ip->ip_len); 814 ip->ip_off = htons(ip->ip_off); 815 816 /* Deliver packet to divert input routine */ 817 divert_packet(m, 0); 818 819 /* If 'tee', continue with original packet */ 820 if (clone != NULL) { 821 m = clone; 822 ip = mtod(m, struct ip *); 823 goto pass; 824 } 825 goto done; 826 } 827#endif 828 829 /* IPFIREWALL_FORWARD */ 830 /* 831 * Check dst to make sure it is directly reachable on the 832 * interface we previously thought it was. 833 * If it isn't (which may be likely in some situations) we have 834 * to re-route it (ie, find a route for the next-hop and the 835 * associated interface) and set them here. This is nested 836 * forwarding which in most cases is undesirable, except where 837 * such control is nigh impossible. So we do it here. 838 * And I'm babbling. 839 */ 840 if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 841#if 0 842 /* 843 * XXX To improve readability, this block should be 844 * changed into a function call as below: 845 */ 846 error = ip_ipforward(&m, &dst, &ifp); 847 if (error) 848 goto bad; 849 if (m == NULL) /* ip_input consumed the mbuf */ 850 goto done; 851#else 852 struct in_ifaddr *ia; 853 854 /* 855 * XXX sro_fwd below is static, and a pointer 856 * to it gets passed to routines downstream. 857 * This could have surprisingly bad results in 858 * practice, because its content is overwritten 859 * by subsequent packets. 860 */ 861 /* There must be a better way to do this next line... */ 862 static struct route sro_fwd; 863 struct route *ro_fwd = &sro_fwd; 864 865#if 0 866 print_ip("IPFIREWALL_FORWARD: New dst ip: ", 867 dst->sin_addr, "\n"); 868#endif 869 870 /* 871 * We need to figure out if we have been forwarded 872 * to a local socket. If so, then we should somehow 873 * "loop back" to ip_input, and get directed to the 874 * PCB as if we had received this packet. This is 875 * because it may be dificult to identify the packets 876 * you want to forward until they are being output 877 * and have selected an interface. (e.g. locally 878 * initiated packets) If we used the loopback inteface, 879 * we would not be able to control what happens 880 * as the packet runs through ip_input() as 881 * it is done through an ISR. 882 */ 883 LIST_FOREACH(ia, 884 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 885 /* 886 * If the addr to forward to is one 887 * of ours, we pretend to 888 * be the destination for this packet. 889 */ 890 if (IA_SIN(ia)->sin_addr.s_addr == 891 dst->sin_addr.s_addr) 892 break; 893 } 894 if (ia) { /* tell ip_input "dont filter" */ 895 mtag = m_tag_get(PACKET_TAG_IPFORWARD, 896 sizeof(struct sockaddr_in *), 897 M_NOWAIT); 898 if (mtag == NULL) { 899 /* XXX statistic */ 900 error = ENOBUFS; /* XXX */ 901 goto bad; 902 } 903 *(struct sockaddr_in **)(mtag+1) = 904 args.next_hop; 905 m_tag_prepend(m, mtag); 906 907 if (m->m_pkthdr.rcvif == NULL) 908 m->m_pkthdr.rcvif = ifunit("lo0"); 909 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 910 m->m_pkthdr.csum_flags |= 911 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 912 m->m_pkthdr.csum_data = 0xffff; 913 } 914 m->m_pkthdr.csum_flags |= 915 CSUM_IP_CHECKED | CSUM_IP_VALID; 916 ip->ip_len = htons(ip->ip_len); 917 ip->ip_off = htons(ip->ip_off); 918 /* XXX netisr_queue(NETISR_IP, m); */ 919 ip_input(m); 920 goto done; 921 } 922 /* 923 * Some of the logic for this was 924 * nicked from above. 925 */ 926 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 927 928 ro_fwd->ro_rt = 0; 929 rtalloc_ign(ro_fwd, RTF_CLONING); 930 931 if (ro_fwd->ro_rt == 0) { 932 ipstat.ips_noroute++; 933 error = EHOSTUNREACH; 934 goto bad; 935 } 936 937 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 938 ifp = ro_fwd->ro_rt->rt_ifp; 939 ro_fwd->ro_rt->rt_rmx.rmx_pksent++; 940 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 941 dst = (struct sockaddr_in *) 942 ro_fwd->ro_rt->rt_gateway; 943 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 944 isbroadcast = 945 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 946 else 947 isbroadcast = in_broadcast(dst->sin_addr, ifp); 948 if (ro->ro_rt) 949 RTFREE(ro->ro_rt); 950 ro->ro_rt = ro_fwd->ro_rt; 951 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 952 953#endif /* ... block to be put into a function */ 954 /* 955 * If we added a default src ip earlier, 956 * which would have been gotten from the-then 957 * interface, do it again, from the new one. 958 */ 959 if (src_was_INADDR_ANY) 960 ip->ip_src = IA_SIN(ia)->sin_addr; 961 goto pass ; 962 } 963 964 /* 965 * if we get here, none of the above matches, and 966 * we have to drop the pkt 967 */ 968 m_freem(m); 969 error = EACCES; /* not sure this is the right error msg */ 970 goto done; 971 } 972 973pass: 974 /* 127/8 must not appear on wire - RFC1122. */ 975 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 976 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 977 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 978 ipstat.ips_badaddr++; 979 error = EADDRNOTAVAIL; 980 goto bad; 981 } 982 } 983 984 m->m_pkthdr.csum_flags |= CSUM_IP; 985 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 986 if (sw_csum & CSUM_DELAY_DATA) { 987 in_delayed_cksum(m); 988 sw_csum &= ~CSUM_DELAY_DATA; 989 } 990 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 991 992 /* 993 * If small enough for interface, or the interface will take 994 * care of the fragmentation for us, can just send directly. 995 */ 996 if (ip->ip_len <= ifp->if_mtu || (ifp->if_hwassist & CSUM_FRAGMENT && 997 ((ip->ip_off & IP_DF) == 0))) { 998 ip->ip_len = htons(ip->ip_len); 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(m, hlen); 1003 1004 /* Record statistics for this interface address. */ 1005 if (!(flags & IP_FORWARDING) && ia) { 1006 ia->ia_ifa.if_opackets++; 1007 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1008 } 1009 1010#ifdef IPSEC 1011 /* clean ipsec history once it goes out of the node */ 1012 ipsec_delaux(m); 1013#endif 1014 1015#ifdef MBUF_STRESS_TEST 1016 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 1017 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 1018#endif 1019 error = (*ifp->if_output)(ifp, m, 1020 (struct sockaddr *)dst, ro->ro_rt); 1021 goto done; 1022 } 1023 1024 if (ip->ip_off & IP_DF) { 1025 error = EMSGSIZE; 1026 /* 1027 * This case can happen if the user changed the MTU 1028 * of an interface after enabling IP on it. Because 1029 * most netifs don't keep track of routes pointing to 1030 * them, there is no way for one to update all its 1031 * routes when the MTU is changed. 1032 */ 1033 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 1034 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 1035 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 1036 } 1037 ipstat.ips_cantfrag++; 1038 goto bad; 1039 } 1040 1041 /* 1042 * Too large for interface; fragment if possible. If successful, 1043 * on return, m will point to a list of packets to be sent. 1044 */ 1045 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 1046 if (error) 1047 goto bad; 1048 for (; m; m = m0) { 1049 m0 = m->m_nextpkt; 1050 m->m_nextpkt = 0; 1051#ifdef IPSEC 1052 /* clean ipsec history once it goes out of the node */ 1053 ipsec_delaux(m); 1054#endif 1055 if (error == 0) { 1056 /* Record statistics for this interface address. */ 1057 if (ia != NULL) { 1058 ia->ia_ifa.if_opackets++; 1059 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1060 } 1061 1062 error = (*ifp->if_output)(ifp, m, 1063 (struct sockaddr *)dst, ro->ro_rt); 1064 } else 1065 m_freem(m); 1066 } 1067 1068 if (error == 0) 1069 ipstat.ips_fragmented++; 1070 1071done: 1072 if (ro == &iproute && ro->ro_rt) { 1073 RTFREE(ro->ro_rt); 1074 ro->ro_rt = NULL; 1075 } 1076 if (dummytag) { 1077 struct dn_pkt_tag *dt = (struct dn_pkt_tag *)(dummytag+1); 1078 if (dt->ro.ro_rt) 1079 RTFREE(dt->ro.ro_rt); 1080 m_tag_free(dummytag); 1081 } 1082#ifdef IPSEC 1083 if (sp != NULL) { 1084 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1085 printf("DP ip_output call free SP:%p\n", sp)); 1086 key_freesp(sp); 1087 } 1088#endif 1089#ifdef FAST_IPSEC 1090 if (sp != NULL) 1091 KEY_FREESP(&sp); 1092#endif 1093 return (error); 1094bad: 1095 m_freem(m); 1096 goto done; 1097} 1098 1099/* 1100 * Create a chain of fragments which fit the given mtu. m_frag points to the 1101 * mbuf to be fragmented; on return it points to the chain with the fragments. 1102 * Return 0 if no error. If error, m_frag may contain a partially built 1103 * chain of fragments that should be freed by the caller. 1104 * 1105 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 1106 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 1107 */ 1108int 1109ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 1110 u_long if_hwassist_flags, int sw_csum) 1111{ 1112 int error = 0; 1113 int hlen = ip->ip_hl << 2; 1114 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 1115 int off; 1116 struct mbuf *m0 = *m_frag; /* the original packet */ 1117 int firstlen; 1118 struct mbuf **mnext; 1119 int nfrags; 1120 1121 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 1122 ipstat.ips_cantfrag++; 1123 return EMSGSIZE; 1124 } 1125 1126 /* 1127 * Must be able to put at least 8 bytes per fragment. 1128 */ 1129 if (len < 8) 1130 return EMSGSIZE; 1131 1132 /* 1133 * If the interface will not calculate checksums on 1134 * fragmented packets, then do it here. 1135 */ 1136 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 1137 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 1138 in_delayed_cksum(m0); 1139 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1140 } 1141 1142 if (len > PAGE_SIZE) { 1143 /* 1144 * Fragment large datagrams such that each segment 1145 * contains a multiple of PAGE_SIZE amount of data, 1146 * plus headers. This enables a receiver to perform 1147 * page-flipping zero-copy optimizations. 1148 * 1149 * XXX When does this help given that sender and receiver 1150 * could have different page sizes, and also mtu could 1151 * be less than the receiver's page size ? 1152 */ 1153 int newlen; 1154 struct mbuf *m; 1155 1156 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 1157 off += m->m_len; 1158 1159 /* 1160 * firstlen (off - hlen) must be aligned on an 1161 * 8-byte boundary 1162 */ 1163 if (off < hlen) 1164 goto smart_frag_failure; 1165 off = ((off - hlen) & ~7) + hlen; 1166 newlen = (~PAGE_MASK) & mtu; 1167 if ((newlen + sizeof (struct ip)) > mtu) { 1168 /* we failed, go back the default */ 1169smart_frag_failure: 1170 newlen = len; 1171 off = hlen + len; 1172 } 1173 len = newlen; 1174 1175 } else { 1176 off = hlen + len; 1177 } 1178 1179 firstlen = off - hlen; 1180 mnext = &m0->m_nextpkt; /* pointer to next packet */ 1181 1182 /* 1183 * Loop through length of segment after first fragment, 1184 * make new header and copy data of each part and link onto chain. 1185 * Here, m0 is the original packet, m is the fragment being created. 1186 * The fragments are linked off the m_nextpkt of the original 1187 * packet, which after processing serves as the first fragment. 1188 */ 1189 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 1190 struct ip *mhip; /* ip header on the fragment */ 1191 struct mbuf *m; 1192 int mhlen = sizeof (struct ip); 1193 1194 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1195 if (m == 0) { 1196 error = ENOBUFS; 1197 ipstat.ips_odropped++; 1198 goto done; 1199 } 1200 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1201 /* 1202 * In the first mbuf, leave room for the link header, then 1203 * copy the original IP header including options. The payload 1204 * goes into an additional mbuf chain returned by m_copy(). 1205 */ 1206 m->m_data += max_linkhdr; 1207 mhip = mtod(m, struct ip *); 1208 *mhip = *ip; 1209 if (hlen > sizeof (struct ip)) { 1210 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1211 mhip->ip_v = IPVERSION; 1212 mhip->ip_hl = mhlen >> 2; 1213 } 1214 m->m_len = mhlen; 1215 /* XXX do we need to add ip->ip_off below ? */ 1216 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1217 if (off + len >= ip->ip_len) { /* last fragment */ 1218 len = ip->ip_len - off; 1219 m->m_flags |= M_LASTFRAG; 1220 } else 1221 mhip->ip_off |= IP_MF; 1222 mhip->ip_len = htons((u_short)(len + mhlen)); 1223 m->m_next = m_copy(m0, off, len); 1224 if (m->m_next == 0) { /* copy failed */ 1225 m_free(m); 1226 error = ENOBUFS; /* ??? */ 1227 ipstat.ips_odropped++; 1228 goto done; 1229 } 1230 m->m_pkthdr.len = mhlen + len; 1231 m->m_pkthdr.rcvif = (struct ifnet *)0; 1232#ifdef MAC 1233 mac_create_fragment(m0, m); 1234#endif 1235 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1236 mhip->ip_off = htons(mhip->ip_off); 1237 mhip->ip_sum = 0; 1238 if (sw_csum & CSUM_DELAY_IP) 1239 mhip->ip_sum = in_cksum(m, mhlen); 1240 *mnext = m; 1241 mnext = &m->m_nextpkt; 1242 } 1243 ipstat.ips_ofragments += nfrags; 1244 1245 /* set first marker for fragment chain */ 1246 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1247 m0->m_pkthdr.csum_data = nfrags; 1248 1249 /* 1250 * Update first fragment by trimming what's been copied out 1251 * and updating header. 1252 */ 1253 m_adj(m0, hlen + firstlen - ip->ip_len); 1254 m0->m_pkthdr.len = hlen + firstlen; 1255 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 1256 ip->ip_off |= IP_MF; 1257 ip->ip_off = htons(ip->ip_off); 1258 ip->ip_sum = 0; 1259 if (sw_csum & CSUM_DELAY_IP) 1260 ip->ip_sum = in_cksum(m0, hlen); 1261 1262done: 1263 *m_frag = m0; 1264 return error; 1265} 1266 1267void 1268in_delayed_cksum(struct mbuf *m) 1269{ 1270 struct ip *ip; 1271 u_short csum, offset; 1272 1273 ip = mtod(m, struct ip *); 1274 offset = ip->ip_hl << 2 ; 1275 csum = in_cksum_skip(m, ip->ip_len, offset); 1276 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1277 csum = 0xffff; 1278 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1279 1280 if (offset + sizeof(u_short) > m->m_len) { 1281 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1282 m->m_len, offset, ip->ip_p); 1283 /* 1284 * XXX 1285 * this shouldn't happen, but if it does, the 1286 * correct behavior may be to insert the checksum 1287 * in the existing chain instead of rearranging it. 1288 */ 1289 m = m_pullup(m, offset + sizeof(u_short)); 1290 } 1291 *(u_short *)(m->m_data + offset) = csum; 1292} 1293 1294/* 1295 * Insert IP options into preformed packet. 1296 * Adjust IP destination as required for IP source routing, 1297 * as indicated by a non-zero in_addr at the start of the options. 1298 * 1299 * XXX This routine assumes that the packet has no options in place. 1300 */ 1301static struct mbuf * 1302ip_insertoptions(m, opt, phlen) 1303 register struct mbuf *m; 1304 struct mbuf *opt; 1305 int *phlen; 1306{ 1307 register struct ipoption *p = mtod(opt, struct ipoption *); 1308 struct mbuf *n; 1309 register struct ip *ip = mtod(m, struct ip *); 1310 unsigned optlen; 1311 1312 optlen = opt->m_len - sizeof(p->ipopt_dst); 1313 if (optlen + ip->ip_len > IP_MAXPACKET) { 1314 *phlen = 0; 1315 return (m); /* XXX should fail */ 1316 } 1317 if (p->ipopt_dst.s_addr) 1318 ip->ip_dst = p->ipopt_dst; 1319 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1320 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1321 if (n == 0) { 1322 *phlen = 0; 1323 return (m); 1324 } 1325 n->m_pkthdr.rcvif = (struct ifnet *)0; 1326#ifdef MAC 1327 mac_create_mbuf_from_mbuf(m, n); 1328#endif 1329 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1330 m->m_len -= sizeof(struct ip); 1331 m->m_data += sizeof(struct ip); 1332 n->m_next = m; 1333 m = n; 1334 m->m_len = optlen + sizeof(struct ip); 1335 m->m_data += max_linkhdr; 1336 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1337 } else { 1338 m->m_data -= optlen; 1339 m->m_len += optlen; 1340 m->m_pkthdr.len += optlen; 1341 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1342 } 1343 ip = mtod(m, struct ip *); 1344 bcopy(p->ipopt_list, ip + 1, optlen); 1345 *phlen = sizeof(struct ip) + optlen; 1346 ip->ip_v = IPVERSION; 1347 ip->ip_hl = *phlen >> 2; 1348 ip->ip_len += optlen; 1349 return (m); 1350} 1351 1352/* 1353 * Copy options from ip to jp, 1354 * omitting those not copied during fragmentation. 1355 */ 1356int 1357ip_optcopy(ip, jp) 1358 struct ip *ip, *jp; 1359{ 1360 register u_char *cp, *dp; 1361 int opt, optlen, cnt; 1362 1363 cp = (u_char *)(ip + 1); 1364 dp = (u_char *)(jp + 1); 1365 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1366 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1367 opt = cp[0]; 1368 if (opt == IPOPT_EOL) 1369 break; 1370 if (opt == IPOPT_NOP) { 1371 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1372 *dp++ = IPOPT_NOP; 1373 optlen = 1; 1374 continue; 1375 } 1376 1377 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1378 ("ip_optcopy: malformed ipv4 option")); 1379 optlen = cp[IPOPT_OLEN]; 1380 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1381 ("ip_optcopy: malformed ipv4 option")); 1382 1383 /* bogus lengths should have been caught by ip_dooptions */ 1384 if (optlen > cnt) 1385 optlen = cnt; 1386 if (IPOPT_COPIED(opt)) { 1387 bcopy(cp, dp, optlen); 1388 dp += optlen; 1389 } 1390 } 1391 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1392 *dp++ = IPOPT_EOL; 1393 return (optlen); 1394} 1395 1396/* 1397 * IP socket option processing. 1398 */ 1399int 1400ip_ctloutput(so, sopt) 1401 struct socket *so; 1402 struct sockopt *sopt; 1403{ 1404 struct inpcb *inp = sotoinpcb(so); 1405 int error, optval; 1406 1407 error = optval = 0; 1408 if (sopt->sopt_level != IPPROTO_IP) { 1409 return (EINVAL); 1410 } 1411 1412 switch (sopt->sopt_dir) { 1413 case SOPT_SET: 1414 switch (sopt->sopt_name) { 1415 case IP_OPTIONS: 1416#ifdef notyet 1417 case IP_RETOPTS: 1418#endif 1419 { 1420 struct mbuf *m; 1421 if (sopt->sopt_valsize > MLEN) { 1422 error = EMSGSIZE; 1423 break; 1424 } 1425 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1426 if (m == 0) { 1427 error = ENOBUFS; 1428 break; 1429 } 1430 m->m_len = sopt->sopt_valsize; 1431 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1432 m->m_len); 1433 1434 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1435 m)); 1436 } 1437 1438 case IP_TOS: 1439 case IP_TTL: 1440 case IP_RECVOPTS: 1441 case IP_RECVRETOPTS: 1442 case IP_RECVDSTADDR: 1443 case IP_RECVTTL: 1444 case IP_RECVIF: 1445 case IP_FAITH: 1446 case IP_ONESBCAST: 1447 error = sooptcopyin(sopt, &optval, sizeof optval, 1448 sizeof optval); 1449 if (error) 1450 break; 1451 1452 switch (sopt->sopt_name) { 1453 case IP_TOS: 1454 inp->inp_ip_tos = optval; 1455 break; 1456 1457 case IP_TTL: 1458 inp->inp_ip_ttl = optval; 1459 break; 1460#define OPTSET(bit) \ 1461 if (optval) \ 1462 inp->inp_flags |= bit; \ 1463 else \ 1464 inp->inp_flags &= ~bit; 1465 1466 case IP_RECVOPTS: 1467 OPTSET(INP_RECVOPTS); 1468 break; 1469 1470 case IP_RECVRETOPTS: 1471 OPTSET(INP_RECVRETOPTS); 1472 break; 1473 1474 case IP_RECVDSTADDR: 1475 OPTSET(INP_RECVDSTADDR); 1476 break; 1477 1478 case IP_RECVTTL: 1479 OPTSET(INP_RECVTTL); 1480 break; 1481 1482 case IP_RECVIF: 1483 OPTSET(INP_RECVIF); 1484 break; 1485 1486 case IP_FAITH: 1487 OPTSET(INP_FAITH); 1488 break; 1489 1490 case IP_ONESBCAST: 1491 OPTSET(INP_ONESBCAST); 1492 break; 1493 } 1494 break; 1495#undef OPTSET 1496 1497 case IP_MULTICAST_IF: 1498 case IP_MULTICAST_VIF: 1499 case IP_MULTICAST_TTL: 1500 case IP_MULTICAST_LOOP: 1501 case IP_ADD_MEMBERSHIP: 1502 case IP_DROP_MEMBERSHIP: 1503 error = ip_setmoptions(sopt, &inp->inp_moptions); 1504 break; 1505 1506 case IP_PORTRANGE: 1507 error = sooptcopyin(sopt, &optval, sizeof optval, 1508 sizeof optval); 1509 if (error) 1510 break; 1511 1512 switch (optval) { 1513 case IP_PORTRANGE_DEFAULT: 1514 inp->inp_flags &= ~(INP_LOWPORT); 1515 inp->inp_flags &= ~(INP_HIGHPORT); 1516 break; 1517 1518 case IP_PORTRANGE_HIGH: 1519 inp->inp_flags &= ~(INP_LOWPORT); 1520 inp->inp_flags |= INP_HIGHPORT; 1521 break; 1522 1523 case IP_PORTRANGE_LOW: 1524 inp->inp_flags &= ~(INP_HIGHPORT); 1525 inp->inp_flags |= INP_LOWPORT; 1526 break; 1527 1528 default: 1529 error = EINVAL; 1530 break; 1531 } 1532 break; 1533 1534#if defined(IPSEC) || defined(FAST_IPSEC) 1535 case IP_IPSEC_POLICY: 1536 { 1537 caddr_t req; 1538 size_t len = 0; 1539 int priv; 1540 struct mbuf *m; 1541 int optname; 1542 1543 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1544 break; 1545 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1546 break; 1547 priv = (sopt->sopt_td != NULL && 1548 suser(sopt->sopt_td) != 0) ? 0 : 1; 1549 req = mtod(m, caddr_t); 1550 len = m->m_len; 1551 optname = sopt->sopt_name; 1552 error = ipsec4_set_policy(inp, optname, req, len, priv); 1553 m_freem(m); 1554 break; 1555 } 1556#endif /*IPSEC*/ 1557 1558 default: 1559 error = ENOPROTOOPT; 1560 break; 1561 } 1562 break; 1563 1564 case SOPT_GET: 1565 switch (sopt->sopt_name) { 1566 case IP_OPTIONS: 1567 case IP_RETOPTS: 1568 if (inp->inp_options) 1569 error = sooptcopyout(sopt, 1570 mtod(inp->inp_options, 1571 char *), 1572 inp->inp_options->m_len); 1573 else 1574 sopt->sopt_valsize = 0; 1575 break; 1576 1577 case IP_TOS: 1578 case IP_TTL: 1579 case IP_RECVOPTS: 1580 case IP_RECVRETOPTS: 1581 case IP_RECVDSTADDR: 1582 case IP_RECVTTL: 1583 case IP_RECVIF: 1584 case IP_PORTRANGE: 1585 case IP_FAITH: 1586 case IP_ONESBCAST: 1587 switch (sopt->sopt_name) { 1588 1589 case IP_TOS: 1590 optval = inp->inp_ip_tos; 1591 break; 1592 1593 case IP_TTL: 1594 optval = inp->inp_ip_ttl; 1595 break; 1596 1597#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1598 1599 case IP_RECVOPTS: 1600 optval = OPTBIT(INP_RECVOPTS); 1601 break; 1602 1603 case IP_RECVRETOPTS: 1604 optval = OPTBIT(INP_RECVRETOPTS); 1605 break; 1606 1607 case IP_RECVDSTADDR: 1608 optval = OPTBIT(INP_RECVDSTADDR); 1609 break; 1610 1611 case IP_RECVTTL: 1612 optval = OPTBIT(INP_RECVTTL); 1613 break; 1614 1615 case IP_RECVIF: 1616 optval = OPTBIT(INP_RECVIF); 1617 break; 1618 1619 case IP_PORTRANGE: 1620 if (inp->inp_flags & INP_HIGHPORT) 1621 optval = IP_PORTRANGE_HIGH; 1622 else if (inp->inp_flags & INP_LOWPORT) 1623 optval = IP_PORTRANGE_LOW; 1624 else 1625 optval = 0; 1626 break; 1627 1628 case IP_FAITH: 1629 optval = OPTBIT(INP_FAITH); 1630 break; 1631 1632 case IP_ONESBCAST: 1633 optval = OPTBIT(INP_ONESBCAST); 1634 break; 1635 } 1636 error = sooptcopyout(sopt, &optval, sizeof optval); 1637 break; 1638 1639 case IP_MULTICAST_IF: 1640 case IP_MULTICAST_VIF: 1641 case IP_MULTICAST_TTL: 1642 case IP_MULTICAST_LOOP: 1643 case IP_ADD_MEMBERSHIP: 1644 case IP_DROP_MEMBERSHIP: 1645 error = ip_getmoptions(sopt, inp->inp_moptions); 1646 break; 1647 1648#if defined(IPSEC) || defined(FAST_IPSEC) 1649 case IP_IPSEC_POLICY: 1650 { 1651 struct mbuf *m = NULL; 1652 caddr_t req = NULL; 1653 size_t len = 0; 1654 1655 if (m != 0) { 1656 req = mtod(m, caddr_t); 1657 len = m->m_len; 1658 } 1659 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1660 if (error == 0) 1661 error = soopt_mcopyout(sopt, m); /* XXX */ 1662 if (error == 0) 1663 m_freem(m); 1664 break; 1665 } 1666#endif /*IPSEC*/ 1667 1668 default: 1669 error = ENOPROTOOPT; 1670 break; 1671 } 1672 break; 1673 } 1674 return (error); 1675} 1676 1677/* 1678 * Set up IP options in pcb for insertion in output packets. 1679 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1680 * with destination address if source routed. 1681 */ 1682static int 1683ip_pcbopts(optname, pcbopt, m) 1684 int optname; 1685 struct mbuf **pcbopt; 1686 register struct mbuf *m; 1687{ 1688 register int cnt, optlen; 1689 register u_char *cp; 1690 u_char opt; 1691 1692 /* turn off any old options */ 1693 if (*pcbopt) 1694 (void)m_free(*pcbopt); 1695 *pcbopt = 0; 1696 if (m == (struct mbuf *)0 || m->m_len == 0) { 1697 /* 1698 * Only turning off any previous options. 1699 */ 1700 if (m) 1701 (void)m_free(m); 1702 return (0); 1703 } 1704 1705 if (m->m_len % sizeof(int32_t)) 1706 goto bad; 1707 /* 1708 * IP first-hop destination address will be stored before 1709 * actual options; move other options back 1710 * and clear it when none present. 1711 */ 1712 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1713 goto bad; 1714 cnt = m->m_len; 1715 m->m_len += sizeof(struct in_addr); 1716 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1717 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1718 bzero(mtod(m, void *), sizeof(struct in_addr)); 1719 1720 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1721 opt = cp[IPOPT_OPTVAL]; 1722 if (opt == IPOPT_EOL) 1723 break; 1724 if (opt == IPOPT_NOP) 1725 optlen = 1; 1726 else { 1727 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1728 goto bad; 1729 optlen = cp[IPOPT_OLEN]; 1730 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1731 goto bad; 1732 } 1733 switch (opt) { 1734 1735 default: 1736 break; 1737 1738 case IPOPT_LSRR: 1739 case IPOPT_SSRR: 1740 /* 1741 * user process specifies route as: 1742 * ->A->B->C->D 1743 * D must be our final destination (but we can't 1744 * check that since we may not have connected yet). 1745 * A is first hop destination, which doesn't appear in 1746 * actual IP option, but is stored before the options. 1747 */ 1748 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1749 goto bad; 1750 m->m_len -= sizeof(struct in_addr); 1751 cnt -= sizeof(struct in_addr); 1752 optlen -= sizeof(struct in_addr); 1753 cp[IPOPT_OLEN] = optlen; 1754 /* 1755 * Move first hop before start of options. 1756 */ 1757 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1758 sizeof(struct in_addr)); 1759 /* 1760 * Then copy rest of options back 1761 * to close up the deleted entry. 1762 */ 1763 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1764 &cp[IPOPT_OFFSET+1], 1765 (unsigned)cnt + sizeof(struct in_addr)); 1766 break; 1767 } 1768 } 1769 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1770 goto bad; 1771 *pcbopt = m; 1772 return (0); 1773 1774bad: 1775 (void)m_free(m); 1776 return (EINVAL); 1777} 1778 1779/* 1780 * XXX 1781 * The whole multicast option thing needs to be re-thought. 1782 * Several of these options are equally applicable to non-multicast 1783 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1784 * standard option (IP_TTL). 1785 */ 1786 1787/* 1788 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1789 */ 1790static struct ifnet * 1791ip_multicast_if(a, ifindexp) 1792 struct in_addr *a; 1793 int *ifindexp; 1794{ 1795 int ifindex; 1796 struct ifnet *ifp; 1797 1798 if (ifindexp) 1799 *ifindexp = 0; 1800 if (ntohl(a->s_addr) >> 24 == 0) { 1801 ifindex = ntohl(a->s_addr) & 0xffffff; 1802 if (ifindex < 0 || if_index < ifindex) 1803 return NULL; 1804 ifp = ifnet_byindex(ifindex); 1805 if (ifindexp) 1806 *ifindexp = ifindex; 1807 } else { 1808 INADDR_TO_IFP(*a, ifp); 1809 } 1810 return ifp; 1811} 1812 1813/* 1814 * Set the IP multicast options in response to user setsockopt(). 1815 */ 1816static int 1817ip_setmoptions(sopt, imop) 1818 struct sockopt *sopt; 1819 struct ip_moptions **imop; 1820{ 1821 int error = 0; 1822 int i; 1823 struct in_addr addr; 1824 struct ip_mreq mreq; 1825 struct ifnet *ifp; 1826 struct ip_moptions *imo = *imop; 1827 struct route ro; 1828 struct sockaddr_in *dst; 1829 int ifindex; 1830 int s; 1831 1832 if (imo == NULL) { 1833 /* 1834 * No multicast option buffer attached to the pcb; 1835 * allocate one and initialize to default values. 1836 */ 1837 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1838 M_WAITOK); 1839 1840 if (imo == NULL) 1841 return (ENOBUFS); 1842 *imop = imo; 1843 imo->imo_multicast_ifp = NULL; 1844 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1845 imo->imo_multicast_vif = -1; 1846 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1847 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1848 imo->imo_num_memberships = 0; 1849 } 1850 1851 switch (sopt->sopt_name) { 1852 /* store an index number for the vif you wanna use in the send */ 1853 case IP_MULTICAST_VIF: 1854 if (legal_vif_num == 0) { 1855 error = EOPNOTSUPP; 1856 break; 1857 } 1858 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1859 if (error) 1860 break; 1861 if (!legal_vif_num(i) && (i != -1)) { 1862 error = EINVAL; 1863 break; 1864 } 1865 imo->imo_multicast_vif = i; 1866 break; 1867 1868 case IP_MULTICAST_IF: 1869 /* 1870 * Select the interface for outgoing multicast packets. 1871 */ 1872 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1873 if (error) 1874 break; 1875 /* 1876 * INADDR_ANY is used to remove a previous selection. 1877 * When no interface is selected, a default one is 1878 * chosen every time a multicast packet is sent. 1879 */ 1880 if (addr.s_addr == INADDR_ANY) { 1881 imo->imo_multicast_ifp = NULL; 1882 break; 1883 } 1884 /* 1885 * The selected interface is identified by its local 1886 * IP address. Find the interface and confirm that 1887 * it supports multicasting. 1888 */ 1889 s = splimp(); 1890 ifp = ip_multicast_if(&addr, &ifindex); 1891 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1892 splx(s); 1893 error = EADDRNOTAVAIL; 1894 break; 1895 } 1896 imo->imo_multicast_ifp = ifp; 1897 if (ifindex) 1898 imo->imo_multicast_addr = addr; 1899 else 1900 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1901 splx(s); 1902 break; 1903 1904 case IP_MULTICAST_TTL: 1905 /* 1906 * Set the IP time-to-live for outgoing multicast packets. 1907 * The original multicast API required a char argument, 1908 * which is inconsistent with the rest of the socket API. 1909 * We allow either a char or an int. 1910 */ 1911 if (sopt->sopt_valsize == 1) { 1912 u_char ttl; 1913 error = sooptcopyin(sopt, &ttl, 1, 1); 1914 if (error) 1915 break; 1916 imo->imo_multicast_ttl = ttl; 1917 } else { 1918 u_int ttl; 1919 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1920 sizeof ttl); 1921 if (error) 1922 break; 1923 if (ttl > 255) 1924 error = EINVAL; 1925 else 1926 imo->imo_multicast_ttl = ttl; 1927 } 1928 break; 1929 1930 case IP_MULTICAST_LOOP: 1931 /* 1932 * Set the loopback flag for outgoing multicast packets. 1933 * Must be zero or one. The original multicast API required a 1934 * char argument, which is inconsistent with the rest 1935 * of the socket API. We allow either a char or an int. 1936 */ 1937 if (sopt->sopt_valsize == 1) { 1938 u_char loop; 1939 error = sooptcopyin(sopt, &loop, 1, 1); 1940 if (error) 1941 break; 1942 imo->imo_multicast_loop = !!loop; 1943 } else { 1944 u_int loop; 1945 error = sooptcopyin(sopt, &loop, sizeof loop, 1946 sizeof loop); 1947 if (error) 1948 break; 1949 imo->imo_multicast_loop = !!loop; 1950 } 1951 break; 1952 1953 case IP_ADD_MEMBERSHIP: 1954 /* 1955 * Add a multicast group membership. 1956 * Group must be a valid IP multicast address. 1957 */ 1958 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1959 if (error) 1960 break; 1961 1962 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1963 error = EINVAL; 1964 break; 1965 } 1966 s = splimp(); 1967 /* 1968 * If no interface address was provided, use the interface of 1969 * the route to the given multicast address. 1970 */ 1971 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1972 bzero((caddr_t)&ro, sizeof(ro)); 1973 dst = (struct sockaddr_in *)&ro.ro_dst; 1974 dst->sin_len = sizeof(*dst); 1975 dst->sin_family = AF_INET; 1976 dst->sin_addr = mreq.imr_multiaddr; 1977 rtalloc_ign(&ro, RTF_CLONING); 1978 if (ro.ro_rt == NULL) { 1979 error = EADDRNOTAVAIL; 1980 splx(s); 1981 break; 1982 } 1983 ifp = ro.ro_rt->rt_ifp; 1984 RTFREE(ro.ro_rt); 1985 } 1986 else { 1987 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1988 } 1989 1990 /* 1991 * See if we found an interface, and confirm that it 1992 * supports multicast. 1993 */ 1994 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1995 error = EADDRNOTAVAIL; 1996 splx(s); 1997 break; 1998 } 1999 /* 2000 * See if the membership already exists or if all the 2001 * membership slots are full. 2002 */ 2003 for (i = 0; i < imo->imo_num_memberships; ++i) { 2004 if (imo->imo_membership[i]->inm_ifp == ifp && 2005 imo->imo_membership[i]->inm_addr.s_addr 2006 == mreq.imr_multiaddr.s_addr) 2007 break; 2008 } 2009 if (i < imo->imo_num_memberships) { 2010 error = EADDRINUSE; 2011 splx(s); 2012 break; 2013 } 2014 if (i == IP_MAX_MEMBERSHIPS) { 2015 error = ETOOMANYREFS; 2016 splx(s); 2017 break; 2018 } 2019 /* 2020 * Everything looks good; add a new record to the multicast 2021 * address list for the given interface. 2022 */ 2023 if ((imo->imo_membership[i] = 2024 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 2025 error = ENOBUFS; 2026 splx(s); 2027 break; 2028 } 2029 ++imo->imo_num_memberships; 2030 splx(s); 2031 break; 2032 2033 case IP_DROP_MEMBERSHIP: 2034 /* 2035 * Drop a multicast group membership. 2036 * Group must be a valid IP multicast address. 2037 */ 2038 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 2039 if (error) 2040 break; 2041 2042 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 2043 error = EINVAL; 2044 break; 2045 } 2046 2047 s = splimp(); 2048 /* 2049 * If an interface address was specified, get a pointer 2050 * to its ifnet structure. 2051 */ 2052 if (mreq.imr_interface.s_addr == INADDR_ANY) 2053 ifp = NULL; 2054 else { 2055 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 2056 if (ifp == NULL) { 2057 error = EADDRNOTAVAIL; 2058 splx(s); 2059 break; 2060 } 2061 } 2062 /* 2063 * Find the membership in the membership array. 2064 */ 2065 for (i = 0; i < imo->imo_num_memberships; ++i) { 2066 if ((ifp == NULL || 2067 imo->imo_membership[i]->inm_ifp == ifp) && 2068 imo->imo_membership[i]->inm_addr.s_addr == 2069 mreq.imr_multiaddr.s_addr) 2070 break; 2071 } 2072 if (i == imo->imo_num_memberships) { 2073 error = EADDRNOTAVAIL; 2074 splx(s); 2075 break; 2076 } 2077 /* 2078 * Give up the multicast address record to which the 2079 * membership points. 2080 */ 2081 in_delmulti(imo->imo_membership[i]); 2082 /* 2083 * Remove the gap in the membership array. 2084 */ 2085 for (++i; i < imo->imo_num_memberships; ++i) 2086 imo->imo_membership[i-1] = imo->imo_membership[i]; 2087 --imo->imo_num_memberships; 2088 splx(s); 2089 break; 2090 2091 default: 2092 error = EOPNOTSUPP; 2093 break; 2094 } 2095 2096 /* 2097 * If all options have default values, no need to keep the mbuf. 2098 */ 2099 if (imo->imo_multicast_ifp == NULL && 2100 imo->imo_multicast_vif == -1 && 2101 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2102 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2103 imo->imo_num_memberships == 0) { 2104 free(*imop, M_IPMOPTS); 2105 *imop = NULL; 2106 } 2107 2108 return (error); 2109} 2110 2111/* 2112 * Return the IP multicast options in response to user getsockopt(). 2113 */ 2114static int 2115ip_getmoptions(sopt, imo) 2116 struct sockopt *sopt; 2117 register struct ip_moptions *imo; 2118{ 2119 struct in_addr addr; 2120 struct in_ifaddr *ia; 2121 int error, optval; 2122 u_char coptval; 2123 2124 error = 0; 2125 switch (sopt->sopt_name) { 2126 case IP_MULTICAST_VIF: 2127 if (imo != NULL) 2128 optval = imo->imo_multicast_vif; 2129 else 2130 optval = -1; 2131 error = sooptcopyout(sopt, &optval, sizeof optval); 2132 break; 2133 2134 case IP_MULTICAST_IF: 2135 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2136 addr.s_addr = INADDR_ANY; 2137 else if (imo->imo_multicast_addr.s_addr) { 2138 /* return the value user has set */ 2139 addr = imo->imo_multicast_addr; 2140 } else { 2141 IFP_TO_IA(imo->imo_multicast_ifp, ia); 2142 addr.s_addr = (ia == NULL) ? INADDR_ANY 2143 : IA_SIN(ia)->sin_addr.s_addr; 2144 } 2145 error = sooptcopyout(sopt, &addr, sizeof addr); 2146 break; 2147 2148 case IP_MULTICAST_TTL: 2149 if (imo == 0) 2150 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2151 else 2152 optval = coptval = imo->imo_multicast_ttl; 2153 if (sopt->sopt_valsize == 1) 2154 error = sooptcopyout(sopt, &coptval, 1); 2155 else 2156 error = sooptcopyout(sopt, &optval, sizeof optval); 2157 break; 2158 2159 case IP_MULTICAST_LOOP: 2160 if (imo == 0) 2161 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2162 else 2163 optval = coptval = imo->imo_multicast_loop; 2164 if (sopt->sopt_valsize == 1) 2165 error = sooptcopyout(sopt, &coptval, 1); 2166 else 2167 error = sooptcopyout(sopt, &optval, sizeof optval); 2168 break; 2169 2170 default: 2171 error = ENOPROTOOPT; 2172 break; 2173 } 2174 return (error); 2175} 2176 2177/* 2178 * Discard the IP multicast options. 2179 */ 2180void 2181ip_freemoptions(imo) 2182 register struct ip_moptions *imo; 2183{ 2184 register int i; 2185 2186 if (imo != NULL) { 2187 for (i = 0; i < imo->imo_num_memberships; ++i) 2188 in_delmulti(imo->imo_membership[i]); 2189 free(imo, M_IPMOPTS); 2190 } 2191} 2192 2193/* 2194 * Routine called from ip_output() to loop back a copy of an IP multicast 2195 * packet to the input queue of a specified interface. Note that this 2196 * calls the output routine of the loopback "driver", but with an interface 2197 * pointer that might NOT be a loopback interface -- evil, but easier than 2198 * replicating that code here. 2199 */ 2200static void 2201ip_mloopback(ifp, m, dst, hlen) 2202 struct ifnet *ifp; 2203 register struct mbuf *m; 2204 register struct sockaddr_in *dst; 2205 int hlen; 2206{ 2207 register struct ip *ip; 2208 struct mbuf *copym; 2209 2210 copym = m_copy(m, 0, M_COPYALL); 2211 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2212 copym = m_pullup(copym, hlen); 2213 if (copym != NULL) { 2214 /* 2215 * We don't bother to fragment if the IP length is greater 2216 * than the interface's MTU. Can this possibly matter? 2217 */ 2218 ip = mtod(copym, struct ip *); 2219 ip->ip_len = htons(ip->ip_len); 2220 ip->ip_off = htons(ip->ip_off); 2221 ip->ip_sum = 0; 2222 ip->ip_sum = in_cksum(copym, hlen); 2223 /* 2224 * NB: 2225 * It's not clear whether there are any lingering 2226 * reentrancy problems in other areas which might 2227 * be exposed by using ip_input directly (in 2228 * particular, everything which modifies the packet 2229 * in-place). Yet another option is using the 2230 * protosw directly to deliver the looped back 2231 * packet. For the moment, we'll err on the side 2232 * of safety by using if_simloop(). 2233 */ 2234#if 1 /* XXX */ 2235 if (dst->sin_family != AF_INET) { 2236 printf("ip_mloopback: bad address family %d\n", 2237 dst->sin_family); 2238 dst->sin_family = AF_INET; 2239 } 2240#endif 2241 2242#ifdef notdef 2243 copym->m_pkthdr.rcvif = ifp; 2244 ip_input(copym); 2245#else 2246 /* if the checksum hasn't been computed, mark it as valid */ 2247 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2248 copym->m_pkthdr.csum_flags |= 2249 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2250 copym->m_pkthdr.csum_data = 0xffff; 2251 } 2252 if_simloop(ifp, copym, dst->sin_family, 0); 2253#endif 2254 } 2255} 2256