ip_output.c revision 120727
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 120727 2003-10-04 03:44:50Z sam $ 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_dummynet.h> 94 95#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 96 x, (ntohl(a.s_addr)>>24)&0xFF,\ 97 (ntohl(a.s_addr)>>16)&0xFF,\ 98 (ntohl(a.s_addr)>>8)&0xFF,\ 99 (ntohl(a.s_addr))&0xFF, y); 100 101u_short ip_id; 102 103#ifdef MBUF_STRESS_TEST 104int mbuf_frag_size = 0; 105SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 106 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 107#endif 108 109static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 110static struct ifnet *ip_multicast_if(struct in_addr *, int *); 111static void ip_mloopback 112 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 113static int ip_getmoptions 114 (struct sockopt *, struct ip_moptions *); 115static int ip_pcbopts(int, struct mbuf **, struct mbuf *); 116static int ip_setmoptions 117 (struct sockopt *, struct ip_moptions **); 118 119int ip_optcopy(struct ip *, struct ip *); 120 121 122extern struct protosw inetsw[]; 123 124/* 125 * IP output. The packet in mbuf chain m contains a skeletal IP 126 * header (with len, off, ttl, proto, tos, src, dst). 127 * The mbuf chain containing the packet will be freed. 128 * The mbuf opt, if present, will not be freed. 129 */ 130int 131ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, 132 int flags, struct ip_moptions *imo, struct inpcb *inp) 133{ 134 struct ip *ip; 135 struct ifnet *ifp = NULL; /* keep compiler happy */ 136 struct mbuf *m; 137 int hlen = sizeof (struct ip); 138 int len, off, error = 0; 139 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 140 struct in_ifaddr *ia = NULL; 141 int isbroadcast, sw_csum; 142 struct in_addr pkt_dst; 143#ifdef IPSEC 144 struct route iproute; 145 struct secpolicy *sp = NULL; 146#endif 147#ifdef FAST_IPSEC 148 struct route iproute; 149 struct m_tag *mtag; 150 struct secpolicy *sp = NULL; 151 struct tdb_ident *tdbi; 152 int s; 153#endif /* FAST_IPSEC */ 154 struct ip_fw_args args; 155 int src_was_INADDR_ANY = 0; /* as the name says... */ 156 157 args.eh = NULL; 158 args.rule = NULL; 159 args.next_hop = NULL; 160 args.divert_rule = 0; /* divert cookie */ 161 162 /* Grab info from MT_TAG mbufs prepended to the chain. */ 163 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) { 164 switch(m0->_m_tag_id) { 165 default: 166 printf("ip_output: unrecognised MT_TAG tag %d\n", 167 m0->_m_tag_id); 168 break; 169 170 case PACKET_TAG_DUMMYNET: 171 /* 172 * the packet was already tagged, so part of the 173 * processing was already done, and we need to go down. 174 * Get parameters from the header. 175 */ 176 args.rule = ((struct dn_pkt *)m0)->rule; 177 opt = NULL ; 178 ro = & ( ((struct dn_pkt *)m0)->ro ) ; 179 imo = NULL ; 180 dst = ((struct dn_pkt *)m0)->dn_dst ; 181 ifp = ((struct dn_pkt *)m0)->ifp ; 182 flags = ((struct dn_pkt *)m0)->flags ; 183 break; 184 185 case PACKET_TAG_DIVERT: 186 args.divert_rule = (intptr_t)m0->m_data & 0xffff; 187 break; 188 189 case PACKET_TAG_IPFORWARD: 190 args.next_hop = (struct sockaddr_in *)m0->m_data; 191 break; 192 } 193 } 194 m = m0; 195 196 M_ASSERTPKTHDR(m); 197#ifndef FAST_IPSEC 198 KASSERT(ro != NULL, ("ip_output: no route, proto %d", 199 mtod(m, struct ip *)->ip_p)); 200#endif 201 202 if (args.rule != NULL) { /* dummynet already saw us */ 203 ip = mtod(m, struct ip *); 204 hlen = ip->ip_hl << 2 ; 205 if (ro->ro_rt) 206 ia = ifatoia(ro->ro_rt->rt_ifa); 207 goto sendit; 208 } 209 210 if (opt) { 211 len = 0; 212 m = ip_insertoptions(m, opt, &len); 213 if (len != 0) 214 hlen = len; 215 } 216 ip = mtod(m, struct ip *); 217 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 218 219 /* 220 * Fill in IP header. If we are not allowing fragmentation, 221 * then the ip_id field is meaningless, so send it as zero 222 * to reduce information leakage. Otherwise, if we are not 223 * randomizing ip_id, then don't bother to convert it to network 224 * byte order -- it's just a nonce. Note that a 16-bit counter 225 * will wrap around in less than 10 seconds at 100 Mbit/s on a 226 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 227 * for Counting NATted Hosts", Proc. IMW'02, available at 228 * <http://www.research.att.com/~smb/papers/fnat.pdf>. 229 */ 230 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 231 ip->ip_v = IPVERSION; 232 ip->ip_hl = hlen >> 2; 233 if ((ip->ip_off & IP_DF) == 0) { 234 ip->ip_off = 0; 235#ifdef RANDOM_IP_ID 236 ip->ip_id = ip_randomid(); 237#else 238 ip->ip_id = ip_id++; 239#endif 240 } else { 241 ip->ip_off = IP_DF; 242 ip->ip_id = 0; 243 } 244 ipstat.ips_localout++; 245 } else { 246 hlen = ip->ip_hl << 2; 247 } 248 249#ifdef FAST_IPSEC 250 if (ro == NULL) { 251 ro = &iproute; 252 bzero(ro, sizeof (*ro)); 253 } 254#endif /* FAST_IPSEC */ 255 dst = (struct sockaddr_in *)&ro->ro_dst; 256 /* 257 * If there is a cached route, 258 * check that it is to the same destination 259 * and is still up. If not, free it and try again. 260 * The address family should also be checked in case of sharing the 261 * cache with IPv6. 262 */ 263 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 264 dst->sin_family != AF_INET || 265 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 266 RTFREE(ro->ro_rt); 267 ro->ro_rt = (struct rtentry *)0; 268 } 269 if (ro->ro_rt == 0) { 270 bzero(dst, sizeof(*dst)); 271 dst->sin_family = AF_INET; 272 dst->sin_len = sizeof(*dst); 273 dst->sin_addr = pkt_dst; 274 } 275 /* 276 * If routing to interface only, 277 * short circuit routing lookup. 278 */ 279 if (flags & IP_ROUTETOIF) { 280 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 281 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 282 ipstat.ips_noroute++; 283 error = ENETUNREACH; 284 goto bad; 285 } 286 ifp = ia->ia_ifp; 287 ip->ip_ttl = 1; 288 isbroadcast = in_broadcast(dst->sin_addr, ifp); 289 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 290 imo != NULL && imo->imo_multicast_ifp != NULL) { 291 /* 292 * Bypass the normal routing lookup for multicast 293 * packets if the interface is specified. 294 */ 295 ifp = imo->imo_multicast_ifp; 296 IFP_TO_IA(ifp, ia); 297 isbroadcast = 0; /* fool gcc */ 298 } else { 299 /* 300 * If this is the case, we probably don't want to allocate 301 * a protocol-cloned route since we didn't get one from the 302 * ULP. This lets TCP do its thing, while not burdening 303 * forwarding or ICMP with the overhead of cloning a route. 304 * Of course, we still want to do any cloning requested by 305 * the link layer, as this is probably required in all cases 306 * for correct operation (as it is for ARP). 307 */ 308 if (ro->ro_rt == 0) 309 rtalloc_ign(ro, RTF_PRCLONING); 310 if (ro->ro_rt == 0) { 311 ipstat.ips_noroute++; 312 error = EHOSTUNREACH; 313 goto bad; 314 } 315 ia = ifatoia(ro->ro_rt->rt_ifa); 316 ifp = ro->ro_rt->rt_ifp; 317 ro->ro_rt->rt_use++; 318 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 319 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 320 if (ro->ro_rt->rt_flags & RTF_HOST) 321 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 322 else 323 isbroadcast = in_broadcast(dst->sin_addr, ifp); 324 } 325 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 326 struct in_multi *inm; 327 328 m->m_flags |= M_MCAST; 329 /* 330 * IP destination address is multicast. Make sure "dst" 331 * still points to the address in "ro". (It may have been 332 * changed to point to a gateway address, above.) 333 */ 334 dst = (struct sockaddr_in *)&ro->ro_dst; 335 /* 336 * See if the caller provided any multicast options 337 */ 338 if (imo != NULL) { 339 ip->ip_ttl = imo->imo_multicast_ttl; 340 if (imo->imo_multicast_vif != -1) 341 ip->ip_src.s_addr = 342 ip_mcast_src ? 343 ip_mcast_src(imo->imo_multicast_vif) : 344 INADDR_ANY; 345 } else 346 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 347 /* 348 * Confirm that the outgoing interface supports multicast. 349 */ 350 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 351 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 352 ipstat.ips_noroute++; 353 error = ENETUNREACH; 354 goto bad; 355 } 356 } 357 /* 358 * If source address not specified yet, use address 359 * of outgoing interface. 360 */ 361 if (ip->ip_src.s_addr == INADDR_ANY) { 362 /* Interface may have no addresses. */ 363 if (ia != NULL) 364 ip->ip_src = IA_SIN(ia)->sin_addr; 365 } 366 367 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 368 /* 369 * XXX 370 * delayed checksums are not currently 371 * compatible with IP multicast routing 372 */ 373 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 374 in_delayed_cksum(m); 375 m->m_pkthdr.csum_flags &= 376 ~CSUM_DELAY_DATA; 377 } 378 } 379 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 380 if (inm != NULL && 381 (imo == NULL || imo->imo_multicast_loop)) { 382 /* 383 * If we belong to the destination multicast group 384 * on the outgoing interface, and the caller did not 385 * forbid loopback, loop back a copy. 386 */ 387 ip_mloopback(ifp, m, dst, hlen); 388 } 389 else { 390 /* 391 * If we are acting as a multicast router, perform 392 * multicast forwarding as if the packet had just 393 * arrived on the interface to which we are about 394 * to send. The multicast forwarding function 395 * recursively calls this function, using the 396 * IP_FORWARDING flag to prevent infinite recursion. 397 * 398 * Multicasts that are looped back by ip_mloopback(), 399 * above, will be forwarded by the ip_input() routine, 400 * if necessary. 401 */ 402 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 403 /* 404 * If rsvp daemon is not running, do not 405 * set ip_moptions. This ensures that the packet 406 * is multicast and not just sent down one link 407 * as prescribed by rsvpd. 408 */ 409 if (!rsvp_on) 410 imo = NULL; 411 if (ip_mforward && 412 ip_mforward(ip, ifp, m, imo) != 0) { 413 m_freem(m); 414 goto done; 415 } 416 } 417 } 418 419 /* 420 * Multicasts with a time-to-live of zero may be looped- 421 * back, above, but must not be transmitted on a network. 422 * Also, multicasts addressed to the loopback interface 423 * are not sent -- the above call to ip_mloopback() will 424 * loop back a copy if this host actually belongs to the 425 * destination group on the loopback interface. 426 */ 427 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 428 m_freem(m); 429 goto done; 430 } 431 432 goto sendit; 433 } 434#ifndef notdef 435 /* 436 * If the source address is not specified yet, use the address 437 * of the outoing interface. In case, keep note we did that, so 438 * if the the firewall changes the next-hop causing the output 439 * interface to change, we can fix that. 440 */ 441 if (ip->ip_src.s_addr == INADDR_ANY) { 442 /* Interface may have no addresses. */ 443 if (ia != NULL) { 444 ip->ip_src = IA_SIN(ia)->sin_addr; 445 src_was_INADDR_ANY = 1; 446 } 447 } 448#endif /* notdef */ 449 /* 450 * Verify that we have any chance at all of being able to queue 451 * the packet or packet fragments 452 */ 453 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 454 ifp->if_snd.ifq_maxlen) { 455 error = ENOBUFS; 456 ipstat.ips_odropped++; 457 goto bad; 458 } 459 460 /* 461 * Look for broadcast address and 462 * verify user is allowed to send 463 * such a packet. 464 */ 465 if (isbroadcast) { 466 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 467 error = EADDRNOTAVAIL; 468 goto bad; 469 } 470 if ((flags & IP_ALLOWBROADCAST) == 0) { 471 error = EACCES; 472 goto bad; 473 } 474 /* don't allow broadcast messages to be fragmented */ 475 if (ip->ip_len > ifp->if_mtu) { 476 error = EMSGSIZE; 477 goto bad; 478 } 479 if (flags & IP_SENDONES) 480 ip->ip_dst.s_addr = INADDR_BROADCAST; 481 m->m_flags |= M_BCAST; 482 } else { 483 m->m_flags &= ~M_BCAST; 484 } 485 486sendit: 487#ifdef IPSEC 488 /* get SP for this packet */ 489 if (inp == NULL) 490 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 491 else 492 sp = ipsec4_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error); 493 494 if (sp == NULL) { 495 ipsecstat.out_inval++; 496 goto bad; 497 } 498 499 error = 0; 500 501 /* check policy */ 502 switch (sp->policy) { 503 case IPSEC_POLICY_DISCARD: 504 /* 505 * This packet is just discarded. 506 */ 507 ipsecstat.out_polvio++; 508 goto bad; 509 510 case IPSEC_POLICY_BYPASS: 511 case IPSEC_POLICY_NONE: 512 /* no need to do IPsec. */ 513 goto skip_ipsec; 514 515 case IPSEC_POLICY_IPSEC: 516 if (sp->req == NULL) { 517 /* acquire a policy */ 518 error = key_spdacquire(sp); 519 goto bad; 520 } 521 break; 522 523 case IPSEC_POLICY_ENTRUST: 524 default: 525 printf("ip_output: Invalid policy found. %d\n", sp->policy); 526 } 527 { 528 struct ipsec_output_state state; 529 bzero(&state, sizeof(state)); 530 state.m = m; 531 if (flags & IP_ROUTETOIF) { 532 state.ro = &iproute; 533 bzero(&iproute, sizeof(iproute)); 534 } else 535 state.ro = ro; 536 state.dst = (struct sockaddr *)dst; 537 538 ip->ip_sum = 0; 539 540 /* 541 * XXX 542 * delayed checksums are not currently compatible with IPsec 543 */ 544 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 545 in_delayed_cksum(m); 546 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 547 } 548 549 ip->ip_len = htons(ip->ip_len); 550 ip->ip_off = htons(ip->ip_off); 551 552 error = ipsec4_output(&state, sp, flags); 553 554 m = state.m; 555 if (flags & IP_ROUTETOIF) { 556 /* 557 * if we have tunnel mode SA, we may need to ignore 558 * IP_ROUTETOIF. 559 */ 560 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 561 flags &= ~IP_ROUTETOIF; 562 ro = state.ro; 563 } 564 } else 565 ro = state.ro; 566 dst = (struct sockaddr_in *)state.dst; 567 if (error) { 568 /* mbuf is already reclaimed in ipsec4_output. */ 569 m0 = NULL; 570 switch (error) { 571 case EHOSTUNREACH: 572 case ENETUNREACH: 573 case EMSGSIZE: 574 case ENOBUFS: 575 case ENOMEM: 576 break; 577 default: 578 printf("ip4_output (ipsec): error code %d\n", error); 579 /*fall through*/ 580 case ENOENT: 581 /* don't show these error codes to the user */ 582 error = 0; 583 break; 584 } 585 goto bad; 586 } 587 } 588 589 /* be sure to update variables that are affected by ipsec4_output() */ 590 ip = mtod(m, struct ip *); 591 hlen = ip->ip_hl << 2; 592 if (ro->ro_rt == NULL) { 593 if ((flags & IP_ROUTETOIF) == 0) { 594 printf("ip_output: " 595 "can't update route after IPsec processing\n"); 596 error = EHOSTUNREACH; /*XXX*/ 597 goto bad; 598 } 599 } else { 600 ia = ifatoia(ro->ro_rt->rt_ifa); 601 ifp = ro->ro_rt->rt_ifp; 602 } 603 604 /* make it flipped, again. */ 605 ip->ip_len = ntohs(ip->ip_len); 606 ip->ip_off = ntohs(ip->ip_off); 607skip_ipsec: 608#endif /*IPSEC*/ 609#ifdef FAST_IPSEC 610 /* 611 * Check the security policy (SP) for the packet and, if 612 * required, do IPsec-related processing. There are two 613 * cases here; the first time a packet is sent through 614 * it will be untagged and handled by ipsec4_checkpolicy. 615 * If the packet is resubmitted to ip_output (e.g. after 616 * AH, ESP, etc. processing), there will be a tag to bypass 617 * the lookup and related policy checking. 618 */ 619 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 620 s = splnet(); 621 if (mtag != NULL) { 622 tdbi = (struct tdb_ident *)(mtag + 1); 623 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 624 if (sp == NULL) 625 error = -EINVAL; /* force silent drop */ 626 m_tag_delete(m, mtag); 627 } else { 628 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 629 &error, inp); 630 } 631 /* 632 * There are four return cases: 633 * sp != NULL apply IPsec policy 634 * sp == NULL, error == 0 no IPsec handling needed 635 * sp == NULL, error == -EINVAL discard packet w/o error 636 * sp == NULL, error != 0 discard packet, report error 637 */ 638 if (sp != NULL) { 639 /* Loop detection, check if ipsec processing already done */ 640 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 641 for (mtag = m_tag_first(m); mtag != NULL; 642 mtag = m_tag_next(m, mtag)) { 643 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 644 continue; 645 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 646 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 647 continue; 648 /* 649 * Check if policy has an SA associated with it. 650 * This can happen when an SP has yet to acquire 651 * an SA; e.g. on first reference. If it occurs, 652 * then we let ipsec4_process_packet do its thing. 653 */ 654 if (sp->req->sav == NULL) 655 break; 656 tdbi = (struct tdb_ident *)(mtag + 1); 657 if (tdbi->spi == sp->req->sav->spi && 658 tdbi->proto == sp->req->sav->sah->saidx.proto && 659 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 660 sizeof (union sockaddr_union)) == 0) { 661 /* 662 * No IPsec processing is needed, free 663 * reference to SP. 664 * 665 * NB: null pointer to avoid free at 666 * done: below. 667 */ 668 KEY_FREESP(&sp), sp = NULL; 669 splx(s); 670 goto spd_done; 671 } 672 } 673 674 /* 675 * Do delayed checksums now because we send before 676 * this is done in the normal processing path. 677 */ 678 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 679 in_delayed_cksum(m); 680 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 681 } 682 683 ip->ip_len = htons(ip->ip_len); 684 ip->ip_off = htons(ip->ip_off); 685 686 /* NB: callee frees mbuf */ 687 error = ipsec4_process_packet(m, sp->req, flags, 0); 688 /* 689 * Preserve KAME behaviour: ENOENT can be returned 690 * when an SA acquire is in progress. Don't propagate 691 * this to user-level; it confuses applications. 692 * 693 * XXX this will go away when the SADB is redone. 694 */ 695 if (error == ENOENT) 696 error = 0; 697 splx(s); 698 goto done; 699 } else { 700 splx(s); 701 702 if (error != 0) { 703 /* 704 * Hack: -EINVAL is used to signal that a packet 705 * should be silently discarded. This is typically 706 * because we asked key management for an SA and 707 * it was delayed (e.g. kicked up to IKE). 708 */ 709 if (error == -EINVAL) 710 error = 0; 711 goto bad; 712 } else { 713 /* No IPsec processing for this packet. */ 714 } 715#ifdef notyet 716 /* 717 * If deferred crypto processing is needed, check that 718 * the interface supports it. 719 */ 720 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 721 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 722 /* notify IPsec to do its own crypto */ 723 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 724 error = EHOSTUNREACH; 725 goto bad; 726 } 727#endif 728 } 729spd_done: 730#endif /* FAST_IPSEC */ 731 732 /* 733 * IpHack's section. 734 * - Xlate: translate packet's addr/port (NAT). 735 * - Firewall: deny/allow/etc. 736 * - Wrap: fake packet's addr/port <unimpl.> 737 * - Encapsulate: put it in another IP and send out. <unimp.> 738 */ 739#ifdef PFIL_HOOKS 740 /* 741 * Run through list of hooks for output packets. 742 */ 743 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT); 744 if (error != 0 || m == NULL) 745 goto done; 746 ip = mtod(m, struct ip *); 747#endif /* PFIL_HOOKS */ 748 749 /* 750 * Check with the firewall... 751 * but not if we are already being fwd'd from a firewall. 752 */ 753 if (fw_enable && IPFW_LOADED && !args.next_hop) { 754 struct sockaddr_in *old = dst; 755 756 args.m = m; 757 args.next_hop = dst; 758 args.oif = ifp; 759 off = ip_fw_chk_ptr(&args); 760 m = args.m; 761 dst = args.next_hop; 762 763 /* 764 * On return we must do the following: 765 * m == NULL -> drop the pkt (old interface, deprecated) 766 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 767 * 1<=off<= 0xffff -> DIVERT 768 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 769 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 770 * dst != old -> IPFIREWALL_FORWARD 771 * off==0, dst==old -> accept 772 * If some of the above modules are not compiled in, then 773 * we should't have to check the corresponding condition 774 * (because the ipfw control socket should not accept 775 * unsupported rules), but better play safe and drop 776 * packets in case of doubt. 777 */ 778 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 779 if (m) 780 m_freem(m); 781 error = EACCES; 782 goto done; 783 } 784 ip = mtod(m, struct ip *); 785 if (off == 0 && dst == old) /* common case */ 786 goto pass; 787 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 788 /* 789 * pass the pkt to dummynet. Need to include 790 * pipe number, m, ifp, ro, dst because these are 791 * not recomputed in the next pass. 792 * All other parameters have been already used and 793 * so they are not needed anymore. 794 * XXX note: if the ifp or ro entry are deleted 795 * while a pkt is in dummynet, we are in trouble! 796 */ 797 args.ro = ro; 798 args.dst = dst; 799 args.flags = flags; 800 801 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 802 &args); 803 goto done; 804 } 805#ifdef IPDIVERT 806 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 807 struct mbuf *clone = NULL; 808 809 /* Clone packet if we're doing a 'tee' */ 810 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 811 clone = m_dup(m, M_DONTWAIT); 812 813 /* 814 * XXX 815 * delayed checksums are not currently compatible 816 * with divert sockets. 817 */ 818 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 819 in_delayed_cksum(m); 820 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 821 } 822 823 /* Restore packet header fields to original values */ 824 ip->ip_len = htons(ip->ip_len); 825 ip->ip_off = htons(ip->ip_off); 826 827 /* Deliver packet to divert input routine */ 828 divert_packet(m, 0, off & 0xffff, args.divert_rule); 829 830 /* If 'tee', continue with original packet */ 831 if (clone != NULL) { 832 m = clone; 833 ip = mtod(m, struct ip *); 834 goto pass; 835 } 836 goto done; 837 } 838#endif 839 840 /* IPFIREWALL_FORWARD */ 841 /* 842 * Check dst to make sure it is directly reachable on the 843 * interface we previously thought it was. 844 * If it isn't (which may be likely in some situations) we have 845 * to re-route it (ie, find a route for the next-hop and the 846 * associated interface) and set them here. This is nested 847 * forwarding which in most cases is undesirable, except where 848 * such control is nigh impossible. So we do it here. 849 * And I'm babbling. 850 */ 851 if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 852#if 0 853 /* 854 * XXX To improve readability, this block should be 855 * changed into a function call as below: 856 */ 857 error = ip_ipforward(&m, &dst, &ifp); 858 if (error) 859 goto bad; 860 if (m == NULL) /* ip_input consumed the mbuf */ 861 goto done; 862#else 863 struct in_ifaddr *ia; 864 865 /* 866 * XXX sro_fwd below is static, and a pointer 867 * to it gets passed to routines downstream. 868 * This could have surprisingly bad results in 869 * practice, because its content is overwritten 870 * by subsequent packets. 871 */ 872 /* There must be a better way to do this next line... */ 873 static struct route sro_fwd; 874 struct route *ro_fwd = &sro_fwd; 875 876#if 0 877 print_ip("IPFIREWALL_FORWARD: New dst ip: ", 878 dst->sin_addr, "\n"); 879#endif 880 881 /* 882 * We need to figure out if we have been forwarded 883 * to a local socket. If so, then we should somehow 884 * "loop back" to ip_input, and get directed to the 885 * PCB as if we had received this packet. This is 886 * because it may be dificult to identify the packets 887 * you want to forward until they are being output 888 * and have selected an interface. (e.g. locally 889 * initiated packets) If we used the loopback inteface, 890 * we would not be able to control what happens 891 * as the packet runs through ip_input() as 892 * it is done through an ISR. 893 */ 894 LIST_FOREACH(ia, 895 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 896 /* 897 * If the addr to forward to is one 898 * of ours, we pretend to 899 * be the destination for this packet. 900 */ 901 if (IA_SIN(ia)->sin_addr.s_addr == 902 dst->sin_addr.s_addr) 903 break; 904 } 905 if (ia) { /* tell ip_input "dont filter" */ 906 struct m_hdr tag; 907 908 tag.mh_type = MT_TAG; 909 tag.mh_flags = PACKET_TAG_IPFORWARD; 910 tag.mh_data = (caddr_t)args.next_hop; 911 tag.mh_next = m; 912 913 if (m->m_pkthdr.rcvif == NULL) 914 m->m_pkthdr.rcvif = ifunit("lo0"); 915 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 916 m->m_pkthdr.csum_flags |= 917 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 918 m0->m_pkthdr.csum_data = 0xffff; 919 } 920 m->m_pkthdr.csum_flags |= 921 CSUM_IP_CHECKED | CSUM_IP_VALID; 922 ip->ip_len = htons(ip->ip_len); 923 ip->ip_off = htons(ip->ip_off); 924 ip_input((struct mbuf *)&tag); 925 goto done; 926 } 927 /* Some of the logic for this was 928 * nicked from above. 929 * 930 * This rewrites the cached route in a local PCB. 931 * Is this what we want to do? 932 */ 933 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 934 935 ro_fwd->ro_rt = 0; 936 rtalloc_ign(ro_fwd, RTF_PRCLONING); 937 938 if (ro_fwd->ro_rt == 0) { 939 ipstat.ips_noroute++; 940 error = EHOSTUNREACH; 941 goto bad; 942 } 943 944 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 945 ifp = ro_fwd->ro_rt->rt_ifp; 946 ro_fwd->ro_rt->rt_use++; 947 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 948 dst = (struct sockaddr_in *) 949 ro_fwd->ro_rt->rt_gateway; 950 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 951 isbroadcast = 952 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 953 else 954 isbroadcast = in_broadcast(dst->sin_addr, ifp); 955 if (ro->ro_rt) 956 RTFREE(ro->ro_rt); 957 ro->ro_rt = ro_fwd->ro_rt; 958 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 959 960#endif /* ... block to be put into a function */ 961 /* 962 * If we added a default src ip earlier, 963 * which would have been gotten from the-then 964 * interface, do it again, from the new one. 965 */ 966 if (src_was_INADDR_ANY) 967 ip->ip_src = IA_SIN(ia)->sin_addr; 968 goto pass ; 969 } 970 971 /* 972 * if we get here, none of the above matches, and 973 * we have to drop the pkt 974 */ 975 m_freem(m); 976 error = EACCES; /* not sure this is the right error msg */ 977 goto done; 978 } 979 980pass: 981 /* 127/8 must not appear on wire - RFC1122. */ 982 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 983 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 984 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 985 ipstat.ips_badaddr++; 986 error = EADDRNOTAVAIL; 987 goto bad; 988 } 989 } 990 991 m->m_pkthdr.csum_flags |= CSUM_IP; 992 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 993 if (sw_csum & CSUM_DELAY_DATA) { 994 in_delayed_cksum(m); 995 sw_csum &= ~CSUM_DELAY_DATA; 996 } 997 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 998 999 /* 1000 * If small enough for interface, or the interface will take 1001 * care of the fragmentation for us, can just send directly. 1002 */ 1003 if (ip->ip_len <= ifp->if_mtu || ifp->if_hwassist & CSUM_FRAGMENT) { 1004 ip->ip_len = htons(ip->ip_len); 1005 ip->ip_off = htons(ip->ip_off); 1006 ip->ip_sum = 0; 1007 if (sw_csum & CSUM_DELAY_IP) 1008 ip->ip_sum = in_cksum(m, hlen); 1009 1010 /* Record statistics for this interface address. */ 1011 if (!(flags & IP_FORWARDING) && ia) { 1012 ia->ia_ifa.if_opackets++; 1013 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1014 } 1015 1016#ifdef IPSEC 1017 /* clean ipsec history once it goes out of the node */ 1018 ipsec_delaux(m); 1019#endif 1020 1021#ifdef MBUF_STRESS_TEST 1022 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 1023 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 1024#endif 1025 error = (*ifp->if_output)(ifp, m, 1026 (struct sockaddr *)dst, ro->ro_rt); 1027 goto done; 1028 } 1029 1030 if (ip->ip_off & IP_DF) { 1031 error = EMSGSIZE; 1032 /* 1033 * This case can happen if the user changed the MTU 1034 * of an interface after enabling IP on it. Because 1035 * most netifs don't keep track of routes pointing to 1036 * them, there is no way for one to update all its 1037 * routes when the MTU is changed. 1038 */ 1039 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 1040 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) && 1041 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 1042 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 1043 } 1044 ipstat.ips_cantfrag++; 1045 goto bad; 1046 } 1047 1048 /* 1049 * Too large for interface; fragment if possible. If successful, 1050 * on return, m will point to a list of packets to be sent. 1051 */ 1052 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 1053 if (error) 1054 goto bad; 1055 for (; m; m = m0) { 1056 m0 = m->m_nextpkt; 1057 m->m_nextpkt = 0; 1058#ifdef IPSEC 1059 /* clean ipsec history once it goes out of the node */ 1060 ipsec_delaux(m); 1061#endif 1062 if (error == 0) { 1063 /* Record statistics for this interface address. */ 1064 if (ia != NULL) { 1065 ia->ia_ifa.if_opackets++; 1066 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1067 } 1068 1069 error = (*ifp->if_output)(ifp, m, 1070 (struct sockaddr *)dst, ro->ro_rt); 1071 } else 1072 m_freem(m); 1073 } 1074 1075 if (error == 0) 1076 ipstat.ips_fragmented++; 1077 1078done: 1079#ifdef IPSEC 1080 if (ro == &iproute && ro->ro_rt) { 1081 RTFREE(ro->ro_rt); 1082 ro->ro_rt = NULL; 1083 } 1084 if (sp != NULL) { 1085 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1086 printf("DP ip_output call free SP:%p\n", sp)); 1087 key_freesp(sp); 1088 } 1089#endif 1090#ifdef FAST_IPSEC 1091 if (ro == &iproute && ro->ro_rt) { 1092 RTFREE(ro->ro_rt); 1093 ro->ro_rt = NULL; 1094 } 1095 if (sp != NULL) 1096 KEY_FREESP(&sp); 1097#endif 1098 return (error); 1099bad: 1100 m_freem(m); 1101 goto done; 1102} 1103 1104/* 1105 * Create a chain of fragments which fit the given mtu. m_frag points to the 1106 * mbuf to be fragmented; on return it points to the chain with the fragments. 1107 * Return 0 if no error. If error, m_frag may contain a partially built 1108 * chain of fragments that should be freed by the caller. 1109 * 1110 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 1111 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 1112 */ 1113int 1114ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 1115 u_long if_hwassist_flags, int sw_csum) 1116{ 1117 int error = 0; 1118 int hlen = ip->ip_hl << 2; 1119 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 1120 int off; 1121 struct mbuf *m0 = *m_frag; /* the original packet */ 1122 int firstlen; 1123 struct mbuf **mnext; 1124 int nfrags; 1125 1126 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 1127 ipstat.ips_cantfrag++; 1128 return EMSGSIZE; 1129 } 1130 1131 /* 1132 * Must be able to put at least 8 bytes per fragment. 1133 */ 1134 if (len < 8) 1135 return EMSGSIZE; 1136 1137 /* 1138 * If the interface will not calculate checksums on 1139 * fragmented packets, then do it here. 1140 */ 1141 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 1142 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 1143 in_delayed_cksum(m0); 1144 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1145 } 1146 1147 if (len > PAGE_SIZE) { 1148 /* 1149 * Fragment large datagrams such that each segment 1150 * contains a multiple of PAGE_SIZE amount of data, 1151 * plus headers. This enables a receiver to perform 1152 * page-flipping zero-copy optimizations. 1153 * 1154 * XXX When does this help given that sender and receiver 1155 * could have different page sizes, and also mtu could 1156 * be less than the receiver's page size ? 1157 */ 1158 int newlen; 1159 struct mbuf *m; 1160 1161 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 1162 off += m->m_len; 1163 1164 /* 1165 * firstlen (off - hlen) must be aligned on an 1166 * 8-byte boundary 1167 */ 1168 if (off < hlen) 1169 goto smart_frag_failure; 1170 off = ((off - hlen) & ~7) + hlen; 1171 newlen = (~PAGE_MASK) & mtu; 1172 if ((newlen + sizeof (struct ip)) > mtu) { 1173 /* we failed, go back the default */ 1174smart_frag_failure: 1175 newlen = len; 1176 off = hlen + len; 1177 } 1178 len = newlen; 1179 1180 } else { 1181 off = hlen + len; 1182 } 1183 1184 firstlen = off - hlen; 1185 mnext = &m0->m_nextpkt; /* pointer to next packet */ 1186 1187 /* 1188 * Loop through length of segment after first fragment, 1189 * make new header and copy data of each part and link onto chain. 1190 * Here, m0 is the original packet, m is the fragment being created. 1191 * The fragments are linked off the m_nextpkt of the original 1192 * packet, which after processing serves as the first fragment. 1193 */ 1194 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 1195 struct ip *mhip; /* ip header on the fragment */ 1196 struct mbuf *m; 1197 int mhlen = sizeof (struct ip); 1198 1199 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1200 if (m == 0) { 1201 error = ENOBUFS; 1202 ipstat.ips_odropped++; 1203 goto done; 1204 } 1205 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1206 /* 1207 * In the first mbuf, leave room for the link header, then 1208 * copy the original IP header including options. The payload 1209 * goes into an additional mbuf chain returned by m_copy(). 1210 */ 1211 m->m_data += max_linkhdr; 1212 mhip = mtod(m, struct ip *); 1213 *mhip = *ip; 1214 if (hlen > sizeof (struct ip)) { 1215 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1216 mhip->ip_v = IPVERSION; 1217 mhip->ip_hl = mhlen >> 2; 1218 } 1219 m->m_len = mhlen; 1220 /* XXX do we need to add ip->ip_off below ? */ 1221 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1222 if (off + len >= ip->ip_len) { /* last fragment */ 1223 len = ip->ip_len - off; 1224 m->m_flags |= M_LASTFRAG; 1225 } else 1226 mhip->ip_off |= IP_MF; 1227 mhip->ip_len = htons((u_short)(len + mhlen)); 1228 m->m_next = m_copy(m0, off, len); 1229 if (m->m_next == 0) { /* copy failed */ 1230 m_free(m); 1231 error = ENOBUFS; /* ??? */ 1232 ipstat.ips_odropped++; 1233 goto done; 1234 } 1235 m->m_pkthdr.len = mhlen + len; 1236 m->m_pkthdr.rcvif = (struct ifnet *)0; 1237#ifdef MAC 1238 mac_create_fragment(m0, m); 1239#endif 1240 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1241 mhip->ip_off = htons(mhip->ip_off); 1242 mhip->ip_sum = 0; 1243 if (sw_csum & CSUM_DELAY_IP) 1244 mhip->ip_sum = in_cksum(m, mhlen); 1245 *mnext = m; 1246 mnext = &m->m_nextpkt; 1247 } 1248 ipstat.ips_ofragments += nfrags; 1249 1250 /* set first marker for fragment chain */ 1251 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1252 m0->m_pkthdr.csum_data = nfrags; 1253 1254 /* 1255 * Update first fragment by trimming what's been copied out 1256 * and updating header. 1257 */ 1258 m_adj(m0, hlen + firstlen - ip->ip_len); 1259 m0->m_pkthdr.len = hlen + firstlen; 1260 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 1261 ip->ip_off |= IP_MF; 1262 ip->ip_off = htons(ip->ip_off); 1263 ip->ip_sum = 0; 1264 if (sw_csum & CSUM_DELAY_IP) 1265 ip->ip_sum = in_cksum(m0, hlen); 1266 1267done: 1268 *m_frag = m0; 1269 return error; 1270} 1271 1272void 1273in_delayed_cksum(struct mbuf *m) 1274{ 1275 struct ip *ip; 1276 u_short csum, offset; 1277 1278 ip = mtod(m, struct ip *); 1279 offset = ip->ip_hl << 2 ; 1280 csum = in_cksum_skip(m, ip->ip_len, offset); 1281 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1282 csum = 0xffff; 1283 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1284 1285 if (offset + sizeof(u_short) > m->m_len) { 1286 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1287 m->m_len, offset, ip->ip_p); 1288 /* 1289 * XXX 1290 * this shouldn't happen, but if it does, the 1291 * correct behavior may be to insert the checksum 1292 * in the existing chain instead of rearranging it. 1293 */ 1294 m = m_pullup(m, offset + sizeof(u_short)); 1295 } 1296 *(u_short *)(m->m_data + offset) = csum; 1297} 1298 1299/* 1300 * Insert IP options into preformed packet. 1301 * Adjust IP destination as required for IP source routing, 1302 * as indicated by a non-zero in_addr at the start of the options. 1303 * 1304 * XXX This routine assumes that the packet has no options in place. 1305 */ 1306static struct mbuf * 1307ip_insertoptions(m, opt, phlen) 1308 register struct mbuf *m; 1309 struct mbuf *opt; 1310 int *phlen; 1311{ 1312 register struct ipoption *p = mtod(opt, struct ipoption *); 1313 struct mbuf *n; 1314 register struct ip *ip = mtod(m, struct ip *); 1315 unsigned optlen; 1316 1317 optlen = opt->m_len - sizeof(p->ipopt_dst); 1318 if (optlen + ip->ip_len > IP_MAXPACKET) { 1319 *phlen = 0; 1320 return (m); /* XXX should fail */ 1321 } 1322 if (p->ipopt_dst.s_addr) 1323 ip->ip_dst = p->ipopt_dst; 1324 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1325 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1326 if (n == 0) { 1327 *phlen = 0; 1328 return (m); 1329 } 1330 n->m_pkthdr.rcvif = (struct ifnet *)0; 1331#ifdef MAC 1332 mac_create_mbuf_from_mbuf(m, n); 1333#endif 1334 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1335 m->m_len -= sizeof(struct ip); 1336 m->m_data += sizeof(struct ip); 1337 n->m_next = m; 1338 m = n; 1339 m->m_len = optlen + sizeof(struct ip); 1340 m->m_data += max_linkhdr; 1341 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1342 } else { 1343 m->m_data -= optlen; 1344 m->m_len += optlen; 1345 m->m_pkthdr.len += optlen; 1346 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1347 } 1348 ip = mtod(m, struct ip *); 1349 bcopy(p->ipopt_list, ip + 1, optlen); 1350 *phlen = sizeof(struct ip) + optlen; 1351 ip->ip_v = IPVERSION; 1352 ip->ip_hl = *phlen >> 2; 1353 ip->ip_len += optlen; 1354 return (m); 1355} 1356 1357/* 1358 * Copy options from ip to jp, 1359 * omitting those not copied during fragmentation. 1360 */ 1361int 1362ip_optcopy(ip, jp) 1363 struct ip *ip, *jp; 1364{ 1365 register u_char *cp, *dp; 1366 int opt, optlen, cnt; 1367 1368 cp = (u_char *)(ip + 1); 1369 dp = (u_char *)(jp + 1); 1370 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1371 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1372 opt = cp[0]; 1373 if (opt == IPOPT_EOL) 1374 break; 1375 if (opt == IPOPT_NOP) { 1376 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1377 *dp++ = IPOPT_NOP; 1378 optlen = 1; 1379 continue; 1380 } 1381 1382 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1383 ("ip_optcopy: malformed ipv4 option")); 1384 optlen = cp[IPOPT_OLEN]; 1385 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1386 ("ip_optcopy: malformed ipv4 option")); 1387 1388 /* bogus lengths should have been caught by ip_dooptions */ 1389 if (optlen > cnt) 1390 optlen = cnt; 1391 if (IPOPT_COPIED(opt)) { 1392 bcopy(cp, dp, optlen); 1393 dp += optlen; 1394 } 1395 } 1396 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1397 *dp++ = IPOPT_EOL; 1398 return (optlen); 1399} 1400 1401/* 1402 * IP socket option processing. 1403 */ 1404int 1405ip_ctloutput(so, sopt) 1406 struct socket *so; 1407 struct sockopt *sopt; 1408{ 1409 struct inpcb *inp = sotoinpcb(so); 1410 int error, optval; 1411 1412 error = optval = 0; 1413 if (sopt->sopt_level != IPPROTO_IP) { 1414 return (EINVAL); 1415 } 1416 1417 switch (sopt->sopt_dir) { 1418 case SOPT_SET: 1419 switch (sopt->sopt_name) { 1420 case IP_OPTIONS: 1421#ifdef notyet 1422 case IP_RETOPTS: 1423#endif 1424 { 1425 struct mbuf *m; 1426 if (sopt->sopt_valsize > MLEN) { 1427 error = EMSGSIZE; 1428 break; 1429 } 1430 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1431 if (m == 0) { 1432 error = ENOBUFS; 1433 break; 1434 } 1435 m->m_len = sopt->sopt_valsize; 1436 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1437 m->m_len); 1438 1439 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1440 m)); 1441 } 1442 1443 case IP_TOS: 1444 case IP_TTL: 1445 case IP_RECVOPTS: 1446 case IP_RECVRETOPTS: 1447 case IP_RECVDSTADDR: 1448 case IP_RECVTTL: 1449 case IP_RECVIF: 1450 case IP_FAITH: 1451 case IP_ONESBCAST: 1452 error = sooptcopyin(sopt, &optval, sizeof optval, 1453 sizeof optval); 1454 if (error) 1455 break; 1456 1457 switch (sopt->sopt_name) { 1458 case IP_TOS: 1459 inp->inp_ip_tos = optval; 1460 break; 1461 1462 case IP_TTL: 1463 inp->inp_ip_ttl = optval; 1464 break; 1465#define OPTSET(bit) \ 1466 if (optval) \ 1467 inp->inp_flags |= bit; \ 1468 else \ 1469 inp->inp_flags &= ~bit; 1470 1471 case IP_RECVOPTS: 1472 OPTSET(INP_RECVOPTS); 1473 break; 1474 1475 case IP_RECVRETOPTS: 1476 OPTSET(INP_RECVRETOPTS); 1477 break; 1478 1479 case IP_RECVDSTADDR: 1480 OPTSET(INP_RECVDSTADDR); 1481 break; 1482 1483 case IP_RECVTTL: 1484 OPTSET(INP_RECVTTL); 1485 break; 1486 1487 case IP_RECVIF: 1488 OPTSET(INP_RECVIF); 1489 break; 1490 1491 case IP_FAITH: 1492 OPTSET(INP_FAITH); 1493 break; 1494 1495 case IP_ONESBCAST: 1496 OPTSET(INP_ONESBCAST); 1497 break; 1498 } 1499 break; 1500#undef OPTSET 1501 1502 case IP_MULTICAST_IF: 1503 case IP_MULTICAST_VIF: 1504 case IP_MULTICAST_TTL: 1505 case IP_MULTICAST_LOOP: 1506 case IP_ADD_MEMBERSHIP: 1507 case IP_DROP_MEMBERSHIP: 1508 error = ip_setmoptions(sopt, &inp->inp_moptions); 1509 break; 1510 1511 case IP_PORTRANGE: 1512 error = sooptcopyin(sopt, &optval, sizeof optval, 1513 sizeof optval); 1514 if (error) 1515 break; 1516 1517 switch (optval) { 1518 case IP_PORTRANGE_DEFAULT: 1519 inp->inp_flags &= ~(INP_LOWPORT); 1520 inp->inp_flags &= ~(INP_HIGHPORT); 1521 break; 1522 1523 case IP_PORTRANGE_HIGH: 1524 inp->inp_flags &= ~(INP_LOWPORT); 1525 inp->inp_flags |= INP_HIGHPORT; 1526 break; 1527 1528 case IP_PORTRANGE_LOW: 1529 inp->inp_flags &= ~(INP_HIGHPORT); 1530 inp->inp_flags |= INP_LOWPORT; 1531 break; 1532 1533 default: 1534 error = EINVAL; 1535 break; 1536 } 1537 break; 1538 1539#if defined(IPSEC) || defined(FAST_IPSEC) 1540 case IP_IPSEC_POLICY: 1541 { 1542 caddr_t req; 1543 size_t len = 0; 1544 int priv; 1545 struct mbuf *m; 1546 int optname; 1547 1548 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1549 break; 1550 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1551 break; 1552 priv = (sopt->sopt_td != NULL && 1553 suser(sopt->sopt_td) != 0) ? 0 : 1; 1554 req = mtod(m, caddr_t); 1555 len = m->m_len; 1556 optname = sopt->sopt_name; 1557 error = ipsec4_set_policy(inp, optname, req, len, priv); 1558 m_freem(m); 1559 break; 1560 } 1561#endif /*IPSEC*/ 1562 1563 default: 1564 error = ENOPROTOOPT; 1565 break; 1566 } 1567 break; 1568 1569 case SOPT_GET: 1570 switch (sopt->sopt_name) { 1571 case IP_OPTIONS: 1572 case IP_RETOPTS: 1573 if (inp->inp_options) 1574 error = sooptcopyout(sopt, 1575 mtod(inp->inp_options, 1576 char *), 1577 inp->inp_options->m_len); 1578 else 1579 sopt->sopt_valsize = 0; 1580 break; 1581 1582 case IP_TOS: 1583 case IP_TTL: 1584 case IP_RECVOPTS: 1585 case IP_RECVRETOPTS: 1586 case IP_RECVDSTADDR: 1587 case IP_RECVTTL: 1588 case IP_RECVIF: 1589 case IP_PORTRANGE: 1590 case IP_FAITH: 1591 case IP_ONESBCAST: 1592 switch (sopt->sopt_name) { 1593 1594 case IP_TOS: 1595 optval = inp->inp_ip_tos; 1596 break; 1597 1598 case IP_TTL: 1599 optval = inp->inp_ip_ttl; 1600 break; 1601 1602#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1603 1604 case IP_RECVOPTS: 1605 optval = OPTBIT(INP_RECVOPTS); 1606 break; 1607 1608 case IP_RECVRETOPTS: 1609 optval = OPTBIT(INP_RECVRETOPTS); 1610 break; 1611 1612 case IP_RECVDSTADDR: 1613 optval = OPTBIT(INP_RECVDSTADDR); 1614 break; 1615 1616 case IP_RECVTTL: 1617 optval = OPTBIT(INP_RECVTTL); 1618 break; 1619 1620 case IP_RECVIF: 1621 optval = OPTBIT(INP_RECVIF); 1622 break; 1623 1624 case IP_PORTRANGE: 1625 if (inp->inp_flags & INP_HIGHPORT) 1626 optval = IP_PORTRANGE_HIGH; 1627 else if (inp->inp_flags & INP_LOWPORT) 1628 optval = IP_PORTRANGE_LOW; 1629 else 1630 optval = 0; 1631 break; 1632 1633 case IP_FAITH: 1634 optval = OPTBIT(INP_FAITH); 1635 break; 1636 1637 case IP_ONESBCAST: 1638 optval = OPTBIT(INP_ONESBCAST); 1639 break; 1640 } 1641 error = sooptcopyout(sopt, &optval, sizeof optval); 1642 break; 1643 1644 case IP_MULTICAST_IF: 1645 case IP_MULTICAST_VIF: 1646 case IP_MULTICAST_TTL: 1647 case IP_MULTICAST_LOOP: 1648 case IP_ADD_MEMBERSHIP: 1649 case IP_DROP_MEMBERSHIP: 1650 error = ip_getmoptions(sopt, inp->inp_moptions); 1651 break; 1652 1653#if defined(IPSEC) || defined(FAST_IPSEC) 1654 case IP_IPSEC_POLICY: 1655 { 1656 struct mbuf *m = NULL; 1657 caddr_t req = NULL; 1658 size_t len = 0; 1659 1660 if (m != 0) { 1661 req = mtod(m, caddr_t); 1662 len = m->m_len; 1663 } 1664 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1665 if (error == 0) 1666 error = soopt_mcopyout(sopt, m); /* XXX */ 1667 if (error == 0) 1668 m_freem(m); 1669 break; 1670 } 1671#endif /*IPSEC*/ 1672 1673 default: 1674 error = ENOPROTOOPT; 1675 break; 1676 } 1677 break; 1678 } 1679 return (error); 1680} 1681 1682/* 1683 * Set up IP options in pcb for insertion in output packets. 1684 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1685 * with destination address if source routed. 1686 */ 1687static int 1688ip_pcbopts(optname, pcbopt, m) 1689 int optname; 1690 struct mbuf **pcbopt; 1691 register struct mbuf *m; 1692{ 1693 register int cnt, optlen; 1694 register u_char *cp; 1695 u_char opt; 1696 1697 /* turn off any old options */ 1698 if (*pcbopt) 1699 (void)m_free(*pcbopt); 1700 *pcbopt = 0; 1701 if (m == (struct mbuf *)0 || m->m_len == 0) { 1702 /* 1703 * Only turning off any previous options. 1704 */ 1705 if (m) 1706 (void)m_free(m); 1707 return (0); 1708 } 1709 1710 if (m->m_len % sizeof(int32_t)) 1711 goto bad; 1712 /* 1713 * IP first-hop destination address will be stored before 1714 * actual options; move other options back 1715 * and clear it when none present. 1716 */ 1717 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1718 goto bad; 1719 cnt = m->m_len; 1720 m->m_len += sizeof(struct in_addr); 1721 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1722 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1723 bzero(mtod(m, void *), sizeof(struct in_addr)); 1724 1725 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1726 opt = cp[IPOPT_OPTVAL]; 1727 if (opt == IPOPT_EOL) 1728 break; 1729 if (opt == IPOPT_NOP) 1730 optlen = 1; 1731 else { 1732 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1733 goto bad; 1734 optlen = cp[IPOPT_OLEN]; 1735 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1736 goto bad; 1737 } 1738 switch (opt) { 1739 1740 default: 1741 break; 1742 1743 case IPOPT_LSRR: 1744 case IPOPT_SSRR: 1745 /* 1746 * user process specifies route as: 1747 * ->A->B->C->D 1748 * D must be our final destination (but we can't 1749 * check that since we may not have connected yet). 1750 * A is first hop destination, which doesn't appear in 1751 * actual IP option, but is stored before the options. 1752 */ 1753 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1754 goto bad; 1755 m->m_len -= sizeof(struct in_addr); 1756 cnt -= sizeof(struct in_addr); 1757 optlen -= sizeof(struct in_addr); 1758 cp[IPOPT_OLEN] = optlen; 1759 /* 1760 * Move first hop before start of options. 1761 */ 1762 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1763 sizeof(struct in_addr)); 1764 /* 1765 * Then copy rest of options back 1766 * to close up the deleted entry. 1767 */ 1768 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1769 &cp[IPOPT_OFFSET+1], 1770 (unsigned)cnt + sizeof(struct in_addr)); 1771 break; 1772 } 1773 } 1774 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1775 goto bad; 1776 *pcbopt = m; 1777 return (0); 1778 1779bad: 1780 (void)m_free(m); 1781 return (EINVAL); 1782} 1783 1784/* 1785 * XXX 1786 * The whole multicast option thing needs to be re-thought. 1787 * Several of these options are equally applicable to non-multicast 1788 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1789 * standard option (IP_TTL). 1790 */ 1791 1792/* 1793 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1794 */ 1795static struct ifnet * 1796ip_multicast_if(a, ifindexp) 1797 struct in_addr *a; 1798 int *ifindexp; 1799{ 1800 int ifindex; 1801 struct ifnet *ifp; 1802 1803 if (ifindexp) 1804 *ifindexp = 0; 1805 if (ntohl(a->s_addr) >> 24 == 0) { 1806 ifindex = ntohl(a->s_addr) & 0xffffff; 1807 if (ifindex < 0 || if_index < ifindex) 1808 return NULL; 1809 ifp = ifnet_byindex(ifindex); 1810 if (ifindexp) 1811 *ifindexp = ifindex; 1812 } else { 1813 INADDR_TO_IFP(*a, ifp); 1814 } 1815 return ifp; 1816} 1817 1818/* 1819 * Set the IP multicast options in response to user setsockopt(). 1820 */ 1821static int 1822ip_setmoptions(sopt, imop) 1823 struct sockopt *sopt; 1824 struct ip_moptions **imop; 1825{ 1826 int error = 0; 1827 int i; 1828 struct in_addr addr; 1829 struct ip_mreq mreq; 1830 struct ifnet *ifp; 1831 struct ip_moptions *imo = *imop; 1832 struct route ro; 1833 struct sockaddr_in *dst; 1834 int ifindex; 1835 int s; 1836 1837 if (imo == NULL) { 1838 /* 1839 * No multicast option buffer attached to the pcb; 1840 * allocate one and initialize to default values. 1841 */ 1842 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1843 M_WAITOK); 1844 1845 if (imo == NULL) 1846 return (ENOBUFS); 1847 *imop = imo; 1848 imo->imo_multicast_ifp = NULL; 1849 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1850 imo->imo_multicast_vif = -1; 1851 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1852 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1853 imo->imo_num_memberships = 0; 1854 } 1855 1856 switch (sopt->sopt_name) { 1857 /* store an index number for the vif you wanna use in the send */ 1858 case IP_MULTICAST_VIF: 1859 if (legal_vif_num == 0) { 1860 error = EOPNOTSUPP; 1861 break; 1862 } 1863 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1864 if (error) 1865 break; 1866 if (!legal_vif_num(i) && (i != -1)) { 1867 error = EINVAL; 1868 break; 1869 } 1870 imo->imo_multicast_vif = i; 1871 break; 1872 1873 case IP_MULTICAST_IF: 1874 /* 1875 * Select the interface for outgoing multicast packets. 1876 */ 1877 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1878 if (error) 1879 break; 1880 /* 1881 * INADDR_ANY is used to remove a previous selection. 1882 * When no interface is selected, a default one is 1883 * chosen every time a multicast packet is sent. 1884 */ 1885 if (addr.s_addr == INADDR_ANY) { 1886 imo->imo_multicast_ifp = NULL; 1887 break; 1888 } 1889 /* 1890 * The selected interface is identified by its local 1891 * IP address. Find the interface and confirm that 1892 * it supports multicasting. 1893 */ 1894 s = splimp(); 1895 ifp = ip_multicast_if(&addr, &ifindex); 1896 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1897 splx(s); 1898 error = EADDRNOTAVAIL; 1899 break; 1900 } 1901 imo->imo_multicast_ifp = ifp; 1902 if (ifindex) 1903 imo->imo_multicast_addr = addr; 1904 else 1905 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1906 splx(s); 1907 break; 1908 1909 case IP_MULTICAST_TTL: 1910 /* 1911 * Set the IP time-to-live for outgoing multicast packets. 1912 * The original multicast API required a char argument, 1913 * which is inconsistent with the rest of the socket API. 1914 * We allow either a char or an int. 1915 */ 1916 if (sopt->sopt_valsize == 1) { 1917 u_char ttl; 1918 error = sooptcopyin(sopt, &ttl, 1, 1); 1919 if (error) 1920 break; 1921 imo->imo_multicast_ttl = ttl; 1922 } else { 1923 u_int ttl; 1924 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1925 sizeof ttl); 1926 if (error) 1927 break; 1928 if (ttl > 255) 1929 error = EINVAL; 1930 else 1931 imo->imo_multicast_ttl = ttl; 1932 } 1933 break; 1934 1935 case IP_MULTICAST_LOOP: 1936 /* 1937 * Set the loopback flag for outgoing multicast packets. 1938 * Must be zero or one. The original multicast API required a 1939 * char argument, which is inconsistent with the rest 1940 * of the socket API. We allow either a char or an int. 1941 */ 1942 if (sopt->sopt_valsize == 1) { 1943 u_char loop; 1944 error = sooptcopyin(sopt, &loop, 1, 1); 1945 if (error) 1946 break; 1947 imo->imo_multicast_loop = !!loop; 1948 } else { 1949 u_int loop; 1950 error = sooptcopyin(sopt, &loop, sizeof loop, 1951 sizeof loop); 1952 if (error) 1953 break; 1954 imo->imo_multicast_loop = !!loop; 1955 } 1956 break; 1957 1958 case IP_ADD_MEMBERSHIP: 1959 /* 1960 * Add a multicast group membership. 1961 * Group must be a valid IP multicast address. 1962 */ 1963 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1964 if (error) 1965 break; 1966 1967 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1968 error = EINVAL; 1969 break; 1970 } 1971 s = splimp(); 1972 /* 1973 * If no interface address was provided, use the interface of 1974 * the route to the given multicast address. 1975 */ 1976 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1977 bzero((caddr_t)&ro, sizeof(ro)); 1978 dst = (struct sockaddr_in *)&ro.ro_dst; 1979 dst->sin_len = sizeof(*dst); 1980 dst->sin_family = AF_INET; 1981 dst->sin_addr = mreq.imr_multiaddr; 1982 rtalloc(&ro); 1983 if (ro.ro_rt == NULL) { 1984 error = EADDRNOTAVAIL; 1985 splx(s); 1986 break; 1987 } 1988 ifp = ro.ro_rt->rt_ifp; 1989 RTFREE(ro.ro_rt); 1990 } 1991 else { 1992 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1993 } 1994 1995 /* 1996 * See if we found an interface, and confirm that it 1997 * supports multicast. 1998 */ 1999 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2000 error = EADDRNOTAVAIL; 2001 splx(s); 2002 break; 2003 } 2004 /* 2005 * See if the membership already exists or if all the 2006 * membership slots are full. 2007 */ 2008 for (i = 0; i < imo->imo_num_memberships; ++i) { 2009 if (imo->imo_membership[i]->inm_ifp == ifp && 2010 imo->imo_membership[i]->inm_addr.s_addr 2011 == mreq.imr_multiaddr.s_addr) 2012 break; 2013 } 2014 if (i < imo->imo_num_memberships) { 2015 error = EADDRINUSE; 2016 splx(s); 2017 break; 2018 } 2019 if (i == IP_MAX_MEMBERSHIPS) { 2020 error = ETOOMANYREFS; 2021 splx(s); 2022 break; 2023 } 2024 /* 2025 * Everything looks good; add a new record to the multicast 2026 * address list for the given interface. 2027 */ 2028 if ((imo->imo_membership[i] = 2029 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 2030 error = ENOBUFS; 2031 splx(s); 2032 break; 2033 } 2034 ++imo->imo_num_memberships; 2035 splx(s); 2036 break; 2037 2038 case IP_DROP_MEMBERSHIP: 2039 /* 2040 * Drop a multicast group membership. 2041 * Group must be a valid IP multicast address. 2042 */ 2043 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 2044 if (error) 2045 break; 2046 2047 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 2048 error = EINVAL; 2049 break; 2050 } 2051 2052 s = splimp(); 2053 /* 2054 * If an interface address was specified, get a pointer 2055 * to its ifnet structure. 2056 */ 2057 if (mreq.imr_interface.s_addr == INADDR_ANY) 2058 ifp = NULL; 2059 else { 2060 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 2061 if (ifp == NULL) { 2062 error = EADDRNOTAVAIL; 2063 splx(s); 2064 break; 2065 } 2066 } 2067 /* 2068 * Find the membership in the membership array. 2069 */ 2070 for (i = 0; i < imo->imo_num_memberships; ++i) { 2071 if ((ifp == NULL || 2072 imo->imo_membership[i]->inm_ifp == ifp) && 2073 imo->imo_membership[i]->inm_addr.s_addr == 2074 mreq.imr_multiaddr.s_addr) 2075 break; 2076 } 2077 if (i == imo->imo_num_memberships) { 2078 error = EADDRNOTAVAIL; 2079 splx(s); 2080 break; 2081 } 2082 /* 2083 * Give up the multicast address record to which the 2084 * membership points. 2085 */ 2086 in_delmulti(imo->imo_membership[i]); 2087 /* 2088 * Remove the gap in the membership array. 2089 */ 2090 for (++i; i < imo->imo_num_memberships; ++i) 2091 imo->imo_membership[i-1] = imo->imo_membership[i]; 2092 --imo->imo_num_memberships; 2093 splx(s); 2094 break; 2095 2096 default: 2097 error = EOPNOTSUPP; 2098 break; 2099 } 2100 2101 /* 2102 * If all options have default values, no need to keep the mbuf. 2103 */ 2104 if (imo->imo_multicast_ifp == NULL && 2105 imo->imo_multicast_vif == -1 && 2106 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2107 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2108 imo->imo_num_memberships == 0) { 2109 free(*imop, M_IPMOPTS); 2110 *imop = NULL; 2111 } 2112 2113 return (error); 2114} 2115 2116/* 2117 * Return the IP multicast options in response to user getsockopt(). 2118 */ 2119static int 2120ip_getmoptions(sopt, imo) 2121 struct sockopt *sopt; 2122 register struct ip_moptions *imo; 2123{ 2124 struct in_addr addr; 2125 struct in_ifaddr *ia; 2126 int error, optval; 2127 u_char coptval; 2128 2129 error = 0; 2130 switch (sopt->sopt_name) { 2131 case IP_MULTICAST_VIF: 2132 if (imo != NULL) 2133 optval = imo->imo_multicast_vif; 2134 else 2135 optval = -1; 2136 error = sooptcopyout(sopt, &optval, sizeof optval); 2137 break; 2138 2139 case IP_MULTICAST_IF: 2140 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2141 addr.s_addr = INADDR_ANY; 2142 else if (imo->imo_multicast_addr.s_addr) { 2143 /* return the value user has set */ 2144 addr = imo->imo_multicast_addr; 2145 } else { 2146 IFP_TO_IA(imo->imo_multicast_ifp, ia); 2147 addr.s_addr = (ia == NULL) ? INADDR_ANY 2148 : IA_SIN(ia)->sin_addr.s_addr; 2149 } 2150 error = sooptcopyout(sopt, &addr, sizeof addr); 2151 break; 2152 2153 case IP_MULTICAST_TTL: 2154 if (imo == 0) 2155 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2156 else 2157 optval = coptval = imo->imo_multicast_ttl; 2158 if (sopt->sopt_valsize == 1) 2159 error = sooptcopyout(sopt, &coptval, 1); 2160 else 2161 error = sooptcopyout(sopt, &optval, sizeof optval); 2162 break; 2163 2164 case IP_MULTICAST_LOOP: 2165 if (imo == 0) 2166 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2167 else 2168 optval = coptval = imo->imo_multicast_loop; 2169 if (sopt->sopt_valsize == 1) 2170 error = sooptcopyout(sopt, &coptval, 1); 2171 else 2172 error = sooptcopyout(sopt, &optval, sizeof optval); 2173 break; 2174 2175 default: 2176 error = ENOPROTOOPT; 2177 break; 2178 } 2179 return (error); 2180} 2181 2182/* 2183 * Discard the IP multicast options. 2184 */ 2185void 2186ip_freemoptions(imo) 2187 register struct ip_moptions *imo; 2188{ 2189 register int i; 2190 2191 if (imo != NULL) { 2192 for (i = 0; i < imo->imo_num_memberships; ++i) 2193 in_delmulti(imo->imo_membership[i]); 2194 free(imo, M_IPMOPTS); 2195 } 2196} 2197 2198/* 2199 * Routine called from ip_output() to loop back a copy of an IP multicast 2200 * packet to the input queue of a specified interface. Note that this 2201 * calls the output routine of the loopback "driver", but with an interface 2202 * pointer that might NOT be a loopback interface -- evil, but easier than 2203 * replicating that code here. 2204 */ 2205static void 2206ip_mloopback(ifp, m, dst, hlen) 2207 struct ifnet *ifp; 2208 register struct mbuf *m; 2209 register struct sockaddr_in *dst; 2210 int hlen; 2211{ 2212 register struct ip *ip; 2213 struct mbuf *copym; 2214 2215 copym = m_copy(m, 0, M_COPYALL); 2216 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2217 copym = m_pullup(copym, hlen); 2218 if (copym != NULL) { 2219 /* 2220 * We don't bother to fragment if the IP length is greater 2221 * than the interface's MTU. Can this possibly matter? 2222 */ 2223 ip = mtod(copym, struct ip *); 2224 ip->ip_len = htons(ip->ip_len); 2225 ip->ip_off = htons(ip->ip_off); 2226 ip->ip_sum = 0; 2227 ip->ip_sum = in_cksum(copym, hlen); 2228 /* 2229 * NB: 2230 * It's not clear whether there are any lingering 2231 * reentrancy problems in other areas which might 2232 * be exposed by using ip_input directly (in 2233 * particular, everything which modifies the packet 2234 * in-place). Yet another option is using the 2235 * protosw directly to deliver the looped back 2236 * packet. For the moment, we'll err on the side 2237 * of safety by using if_simloop(). 2238 */ 2239#if 1 /* XXX */ 2240 if (dst->sin_family != AF_INET) { 2241 printf("ip_mloopback: bad address family %d\n", 2242 dst->sin_family); 2243 dst->sin_family = AF_INET; 2244 } 2245#endif 2246 2247#ifdef notdef 2248 copym->m_pkthdr.rcvif = ifp; 2249 ip_input(copym); 2250#else 2251 /* if the checksum hasn't been computed, mark it as valid */ 2252 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2253 copym->m_pkthdr.csum_flags |= 2254 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2255 copym->m_pkthdr.csum_data = 0xffff; 2256 } 2257 if_simloop(ifp, copym, dst->sin_family, 0); 2258#endif 2259 } 2260} 2261