ip_output.c revision 122922
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 122922 2003-11-20 20:07:39Z andre $ 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 * In the IP forwarding case, the packet will arrive with options already 130 * inserted, so must have a NULL opt pointer. 131 */ 132int 133ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, 134 int flags, struct ip_moptions *imo, struct inpcb *inp) 135{ 136 struct ip *ip; 137 struct ifnet *ifp = NULL; /* keep compiler happy */ 138 struct mbuf *m; 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#ifdef IPSEC 147 struct socket *so; 148 struct secpolicy *sp = NULL; 149#endif 150#ifdef FAST_IPSEC 151 struct m_tag *mtag; 152 struct secpolicy *sp = NULL; 153 struct tdb_ident *tdbi; 154 int s; 155#endif /* FAST_IPSEC */ 156 struct ip_fw_args args; 157 int src_was_INADDR_ANY = 0; /* as the name says... */ 158 159 args.eh = NULL; 160 args.rule = NULL; 161 args.next_hop = NULL; 162 args.divert_rule = 0; /* divert cookie */ 163 164 /* Grab info from MT_TAG mbufs prepended to the chain. */ 165 for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) { 166 switch(m0->_m_tag_id) { 167 default: 168 printf("ip_output: unrecognised MT_TAG tag %d\n", 169 m0->_m_tag_id); 170 break; 171 172 case PACKET_TAG_DUMMYNET: 173 /* 174 * the packet was already tagged, so part of the 175 * processing was already done, and we need to go down. 176 * Get parameters from the header. 177 */ 178 args.rule = ((struct dn_pkt *)m0)->rule; 179 opt = NULL ; 180 ro = & ( ((struct dn_pkt *)m0)->ro ) ; 181 imo = NULL ; 182 dst = ((struct dn_pkt *)m0)->dn_dst ; 183 ifp = ((struct dn_pkt *)m0)->ifp ; 184 flags = ((struct dn_pkt *)m0)->flags ; 185 break; 186 187 case PACKET_TAG_DIVERT: 188 args.divert_rule = (intptr_t)m0->m_data & 0xffff; 189 break; 190 191 case PACKET_TAG_IPFORWARD: 192 args.next_hop = (struct sockaddr_in *)m0->m_data; 193 break; 194 } 195 } 196 m = m0; 197 198#ifdef IPSEC 199 so = ipsec_getsocket(m); 200 (void)ipsec_setsocket(m, NULL); 201#endif /*IPSEC*/ 202 203 M_ASSERTPKTHDR(m); 204 205 if (ro == NULL) { 206 ro = &iproute; 207 bzero(ro, sizeof (*ro)); 208 } 209 210 if (inp != NULL) 211 INP_LOCK_ASSERT(inp); 212 213 if (args.rule != NULL) { /* dummynet already saw us */ 214 ip = mtod(m, struct ip *); 215 hlen = ip->ip_hl << 2 ; 216 if (ro->ro_rt) 217 ia = ifatoia(ro->ro_rt->rt_ifa); 218 goto sendit; 219 } 220 221 if (opt) { 222 len = 0; 223 m = ip_insertoptions(m, opt, &len); 224 if (len != 0) 225 hlen = len; 226 } 227 ip = mtod(m, struct ip *); 228 pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 229 230 /* 231 * Fill in IP header. If we are not allowing fragmentation, 232 * then the ip_id field is meaningless, so send it as zero 233 * to reduce information leakage. Otherwise, if we are not 234 * randomizing ip_id, then don't bother to convert it to network 235 * byte order -- it's just a nonce. Note that a 16-bit counter 236 * will wrap around in less than 10 seconds at 100 Mbit/s on a 237 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 238 * for Counting NATted Hosts", Proc. IMW'02, available at 239 * <http://www.research.att.com/~smb/papers/fnat.pdf>. 240 */ 241 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 242 ip->ip_v = IPVERSION; 243 ip->ip_hl = hlen >> 2; 244 if ((ip->ip_off & IP_DF) == 0) { 245 ip->ip_off = 0; 246#ifdef RANDOM_IP_ID 247 ip->ip_id = ip_randomid(); 248#else 249 ip->ip_id = ip_id++; 250#endif 251 } else { 252 ip->ip_off = IP_DF; 253 ip->ip_id = 0; 254 } 255 ipstat.ips_localout++; 256 } else { 257 hlen = ip->ip_hl << 2; 258 } 259 260 dst = (struct sockaddr_in *)&ro->ro_dst; 261 /* 262 * If there is a cached route, 263 * check that it is to the same destination 264 * and is still up. If not, free it and try again. 265 * The address family should also be checked in case of sharing the 266 * cache with IPv6. 267 */ 268 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 269 dst->sin_family != AF_INET || 270 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 271 RTFREE(ro->ro_rt); 272 ro->ro_rt = (struct rtentry *)0; 273 } 274 if (ro->ro_rt == 0) { 275 bzero(dst, sizeof(*dst)); 276 dst->sin_family = AF_INET; 277 dst->sin_len = sizeof(*dst); 278 dst->sin_addr = pkt_dst; 279 } 280 /* 281 * If routing to interface only, 282 * short circuit routing lookup. 283 */ 284 if (flags & IP_ROUTETOIF) { 285 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 286 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 287 ipstat.ips_noroute++; 288 error = ENETUNREACH; 289 goto bad; 290 } 291 ifp = ia->ia_ifp; 292 ip->ip_ttl = 1; 293 isbroadcast = in_broadcast(dst->sin_addr, ifp); 294 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 295 imo != NULL && imo->imo_multicast_ifp != NULL) { 296 /* 297 * Bypass the normal routing lookup for multicast 298 * packets if the interface is specified. 299 */ 300 ifp = imo->imo_multicast_ifp; 301 IFP_TO_IA(ifp, ia); 302 isbroadcast = 0; /* fool gcc */ 303 } else { 304 /* 305 * We want to do any cloning requested by the link layer, 306 * as this is probably required in all cases for correct 307 * operation (as it is for ARP). 308 */ 309 if (ro->ro_rt == 0) 310 rtalloc(ro); 311 if (ro->ro_rt == 0) { 312 ipstat.ips_noroute++; 313 error = EHOSTUNREACH; 314 goto bad; 315 } 316 ia = ifatoia(ro->ro_rt->rt_ifa); 317 ifp = ro->ro_rt->rt_ifp; 318 ro->ro_rt->rt_rmx.rmx_pksent++; 319 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 320 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 321 if (ro->ro_rt->rt_flags & RTF_HOST) 322 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 323 else 324 isbroadcast = in_broadcast(dst->sin_addr, ifp); 325 } 326 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 327 struct in_multi *inm; 328 329 m->m_flags |= M_MCAST; 330 /* 331 * IP destination address is multicast. Make sure "dst" 332 * still points to the address in "ro". (It may have been 333 * changed to point to a gateway address, above.) 334 */ 335 dst = (struct sockaddr_in *)&ro->ro_dst; 336 /* 337 * See if the caller provided any multicast options 338 */ 339 if (imo != NULL) { 340 ip->ip_ttl = imo->imo_multicast_ttl; 341 if (imo->imo_multicast_vif != -1) 342 ip->ip_src.s_addr = 343 ip_mcast_src ? 344 ip_mcast_src(imo->imo_multicast_vif) : 345 INADDR_ANY; 346 } else 347 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 348 /* 349 * Confirm that the outgoing interface supports multicast. 350 */ 351 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 352 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 353 ipstat.ips_noroute++; 354 error = ENETUNREACH; 355 goto bad; 356 } 357 } 358 /* 359 * If source address not specified yet, use address 360 * of outgoing interface. 361 */ 362 if (ip->ip_src.s_addr == INADDR_ANY) { 363 /* Interface may have no addresses. */ 364 if (ia != NULL) 365 ip->ip_src = IA_SIN(ia)->sin_addr; 366 } 367 368 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 369 /* 370 * XXX 371 * delayed checksums are not currently 372 * compatible with IP multicast routing 373 */ 374 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 375 in_delayed_cksum(m); 376 m->m_pkthdr.csum_flags &= 377 ~CSUM_DELAY_DATA; 378 } 379 } 380 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 381 if (inm != NULL && 382 (imo == NULL || imo->imo_multicast_loop)) { 383 /* 384 * If we belong to the destination multicast group 385 * on the outgoing interface, and the caller did not 386 * forbid loopback, loop back a copy. 387 */ 388 ip_mloopback(ifp, m, dst, hlen); 389 } 390 else { 391 /* 392 * If we are acting as a multicast router, perform 393 * multicast forwarding as if the packet had just 394 * arrived on the interface to which we are about 395 * to send. The multicast forwarding function 396 * recursively calls this function, using the 397 * IP_FORWARDING flag to prevent infinite recursion. 398 * 399 * Multicasts that are looped back by ip_mloopback(), 400 * above, will be forwarded by the ip_input() routine, 401 * if necessary. 402 */ 403 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 404 /* 405 * If rsvp daemon is not running, do not 406 * set ip_moptions. This ensures that the packet 407 * is multicast and not just sent down one link 408 * as prescribed by rsvpd. 409 */ 410 if (!rsvp_on) 411 imo = NULL; 412 if (ip_mforward && 413 ip_mforward(ip, ifp, m, imo) != 0) { 414 m_freem(m); 415 goto done; 416 } 417 } 418 } 419 420 /* 421 * Multicasts with a time-to-live of zero may be looped- 422 * back, above, but must not be transmitted on a network. 423 * Also, multicasts addressed to the loopback interface 424 * are not sent -- the above call to ip_mloopback() will 425 * loop back a copy if this host actually belongs to the 426 * destination group on the loopback interface. 427 */ 428 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 429 m_freem(m); 430 goto done; 431 } 432 433 goto sendit; 434 } 435#ifndef notdef 436 /* 437 * If the source address is not specified yet, use the address 438 * of the outoing interface. In case, keep note we did that, so 439 * if the the firewall changes the next-hop causing the output 440 * interface to change, we can fix that. 441 */ 442 if (ip->ip_src.s_addr == INADDR_ANY) { 443 /* Interface may have no addresses. */ 444 if (ia != NULL) { 445 ip->ip_src = IA_SIN(ia)->sin_addr; 446 src_was_INADDR_ANY = 1; 447 } 448 } 449#endif /* notdef */ 450 /* 451 * Verify that we have any chance at all of being able to queue 452 * the packet or packet fragments 453 */ 454 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 455 ifp->if_snd.ifq_maxlen) { 456 error = ENOBUFS; 457 ipstat.ips_odropped++; 458 goto bad; 459 } 460 461 /* 462 * Look for broadcast address and 463 * verify user is allowed to send 464 * such a packet. 465 */ 466 if (isbroadcast) { 467 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 468 error = EADDRNOTAVAIL; 469 goto bad; 470 } 471 if ((flags & IP_ALLOWBROADCAST) == 0) { 472 error = EACCES; 473 goto bad; 474 } 475 /* don't allow broadcast messages to be fragmented */ 476 if (ip->ip_len > ifp->if_mtu) { 477 error = EMSGSIZE; 478 goto bad; 479 } 480 if (flags & IP_SENDONES) 481 ip->ip_dst.s_addr = INADDR_BROADCAST; 482 m->m_flags |= M_BCAST; 483 } else { 484 m->m_flags &= ~M_BCAST; 485 } 486 487sendit: 488#ifdef IPSEC 489 /* get SP for this packet */ 490 if (so == NULL) 491 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 492 flags, &error); 493 else 494 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 495 496 if (sp == NULL) { 497 ipsecstat.out_inval++; 498 goto bad; 499 } 500 501 error = 0; 502 503 /* check policy */ 504 switch (sp->policy) { 505 case IPSEC_POLICY_DISCARD: 506 /* 507 * This packet is just discarded. 508 */ 509 ipsecstat.out_polvio++; 510 goto bad; 511 512 case IPSEC_POLICY_BYPASS: 513 case IPSEC_POLICY_NONE: 514 /* no need to do IPsec. */ 515 goto skip_ipsec; 516 517 case IPSEC_POLICY_IPSEC: 518 if (sp->req == NULL) { 519 /* acquire a policy */ 520 error = key_spdacquire(sp); 521 goto bad; 522 } 523 break; 524 525 case IPSEC_POLICY_ENTRUST: 526 default: 527 printf("ip_output: Invalid policy found. %d\n", sp->policy); 528 } 529 { 530 struct ipsec_output_state state; 531 bzero(&state, sizeof(state)); 532 state.m = m; 533 if (flags & IP_ROUTETOIF) { 534 state.ro = &iproute; 535 bzero(&iproute, sizeof(iproute)); 536 } else 537 state.ro = ro; 538 state.dst = (struct sockaddr *)dst; 539 540 ip->ip_sum = 0; 541 542 /* 543 * XXX 544 * delayed checksums are not currently compatible with IPsec 545 */ 546 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 547 in_delayed_cksum(m); 548 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 549 } 550 551 ip->ip_len = htons(ip->ip_len); 552 ip->ip_off = htons(ip->ip_off); 553 554 error = ipsec4_output(&state, sp, flags); 555 556 m = state.m; 557 if (flags & IP_ROUTETOIF) { 558 /* 559 * if we have tunnel mode SA, we may need to ignore 560 * IP_ROUTETOIF. 561 */ 562 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 563 flags &= ~IP_ROUTETOIF; 564 ro = state.ro; 565 } 566 } else 567 ro = state.ro; 568 dst = (struct sockaddr_in *)state.dst; 569 if (error) { 570 /* mbuf is already reclaimed in ipsec4_output. */ 571 m0 = NULL; 572 switch (error) { 573 case EHOSTUNREACH: 574 case ENETUNREACH: 575 case EMSGSIZE: 576 case ENOBUFS: 577 case ENOMEM: 578 break; 579 default: 580 printf("ip4_output (ipsec): error code %d\n", error); 581 /*fall through*/ 582 case ENOENT: 583 /* don't show these error codes to the user */ 584 error = 0; 585 break; 586 } 587 goto bad; 588 } 589 } 590 591 /* be sure to update variables that are affected by ipsec4_output() */ 592 ip = mtod(m, struct ip *); 593 hlen = ip->ip_hl << 2; 594 if (ro->ro_rt == NULL) { 595 if ((flags & IP_ROUTETOIF) == 0) { 596 printf("ip_output: " 597 "can't update route after IPsec processing\n"); 598 error = EHOSTUNREACH; /*XXX*/ 599 goto bad; 600 } 601 } else { 602 ia = ifatoia(ro->ro_rt->rt_ifa); 603 ifp = ro->ro_rt->rt_ifp; 604 } 605 606 /* make it flipped, again. */ 607 ip->ip_len = ntohs(ip->ip_len); 608 ip->ip_off = ntohs(ip->ip_off); 609skip_ipsec: 610#endif /*IPSEC*/ 611#ifdef FAST_IPSEC 612 /* 613 * Check the security policy (SP) for the packet and, if 614 * required, do IPsec-related processing. There are two 615 * cases here; the first time a packet is sent through 616 * it will be untagged and handled by ipsec4_checkpolicy. 617 * If the packet is resubmitted to ip_output (e.g. after 618 * AH, ESP, etc. processing), there will be a tag to bypass 619 * the lookup and related policy checking. 620 */ 621 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 622 s = splnet(); 623 if (mtag != NULL) { 624 tdbi = (struct tdb_ident *)(mtag + 1); 625 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 626 if (sp == NULL) 627 error = -EINVAL; /* force silent drop */ 628 m_tag_delete(m, mtag); 629 } else { 630 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 631 &error, inp); 632 } 633 /* 634 * There are four return cases: 635 * sp != NULL apply IPsec policy 636 * sp == NULL, error == 0 no IPsec handling needed 637 * sp == NULL, error == -EINVAL discard packet w/o error 638 * sp == NULL, error != 0 discard packet, report error 639 */ 640 if (sp != NULL) { 641 /* Loop detection, check if ipsec processing already done */ 642 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 643 for (mtag = m_tag_first(m); mtag != NULL; 644 mtag = m_tag_next(m, mtag)) { 645 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 646 continue; 647 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 648 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 649 continue; 650 /* 651 * Check if policy has an SA associated with it. 652 * This can happen when an SP has yet to acquire 653 * an SA; e.g. on first reference. If it occurs, 654 * then we let ipsec4_process_packet do its thing. 655 */ 656 if (sp->req->sav == NULL) 657 break; 658 tdbi = (struct tdb_ident *)(mtag + 1); 659 if (tdbi->spi == sp->req->sav->spi && 660 tdbi->proto == sp->req->sav->sah->saidx.proto && 661 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 662 sizeof (union sockaddr_union)) == 0) { 663 /* 664 * No IPsec processing is needed, free 665 * reference to SP. 666 * 667 * NB: null pointer to avoid free at 668 * done: below. 669 */ 670 KEY_FREESP(&sp), sp = NULL; 671 splx(s); 672 goto spd_done; 673 } 674 } 675 676 /* 677 * Do delayed checksums now because we send before 678 * this is done in the normal processing path. 679 */ 680 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 681 in_delayed_cksum(m); 682 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 683 } 684 685 ip->ip_len = htons(ip->ip_len); 686 ip->ip_off = htons(ip->ip_off); 687 688 /* NB: callee frees mbuf */ 689 error = ipsec4_process_packet(m, sp->req, flags, 0); 690 /* 691 * Preserve KAME behaviour: ENOENT can be returned 692 * when an SA acquire is in progress. Don't propagate 693 * this to user-level; it confuses applications. 694 * 695 * XXX this will go away when the SADB is redone. 696 */ 697 if (error == ENOENT) 698 error = 0; 699 splx(s); 700 goto done; 701 } else { 702 splx(s); 703 704 if (error != 0) { 705 /* 706 * Hack: -EINVAL is used to signal that a packet 707 * should be silently discarded. This is typically 708 * because we asked key management for an SA and 709 * it was delayed (e.g. kicked up to IKE). 710 */ 711 if (error == -EINVAL) 712 error = 0; 713 goto bad; 714 } else { 715 /* No IPsec processing for this packet. */ 716 } 717#ifdef notyet 718 /* 719 * If deferred crypto processing is needed, check that 720 * the interface supports it. 721 */ 722 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 723 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 724 /* notify IPsec to do its own crypto */ 725 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 726 error = EHOSTUNREACH; 727 goto bad; 728 } 729#endif 730 } 731spd_done: 732#endif /* FAST_IPSEC */ 733 734 /* 735 * IpHack's section. 736 * - Xlate: translate packet's addr/port (NAT). 737 * - Firewall: deny/allow/etc. 738 * - Wrap: fake packet's addr/port <unimpl.> 739 * - Encapsulate: put it in another IP and send out. <unimp.> 740 */ 741#ifdef PFIL_HOOKS 742 /* 743 * Run through list of hooks for output packets. 744 */ 745 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT); 746 if (error != 0 || m == NULL) 747 goto done; 748 ip = mtod(m, struct ip *); 749#endif /* PFIL_HOOKS */ 750 751 /* 752 * Check with the firewall... 753 * but not if we are already being fwd'd from a firewall. 754 */ 755 if (fw_enable && IPFW_LOADED && !args.next_hop) { 756 struct sockaddr_in *old = dst; 757 758 args.m = m; 759 args.next_hop = dst; 760 args.oif = ifp; 761 off = ip_fw_chk_ptr(&args); 762 m = args.m; 763 dst = args.next_hop; 764 765 /* 766 * On return we must do the following: 767 * m == NULL -> drop the pkt (old interface, deprecated) 768 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 769 * 1<=off<= 0xffff -> DIVERT 770 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 771 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 772 * dst != old -> IPFIREWALL_FORWARD 773 * off==0, dst==old -> accept 774 * If some of the above modules are not compiled in, then 775 * we should't have to check the corresponding condition 776 * (because the ipfw control socket should not accept 777 * unsupported rules), but better play safe and drop 778 * packets in case of doubt. 779 */ 780 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 781 if (m) 782 m_freem(m); 783 error = EACCES; 784 goto done; 785 } 786 ip = mtod(m, struct ip *); 787 if (off == 0 && dst == old) /* common case */ 788 goto pass; 789 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 790 /* 791 * pass the pkt to dummynet. Need to include 792 * pipe number, m, ifp, ro, dst because these are 793 * not recomputed in the next pass. 794 * All other parameters have been already used and 795 * so they are not needed anymore. 796 * XXX note: if the ifp or ro entry are deleted 797 * while a pkt is in dummynet, we are in trouble! 798 */ 799 args.ro = ro; 800 args.dst = dst; 801 args.flags = flags; 802 803 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 804 &args); 805 goto done; 806 } 807#ifdef IPDIVERT 808 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 809 struct mbuf *clone = NULL; 810 811 /* Clone packet if we're doing a 'tee' */ 812 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 813 clone = m_dup(m, M_DONTWAIT); 814 815 /* 816 * XXX 817 * delayed checksums are not currently compatible 818 * with divert sockets. 819 */ 820 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 821 in_delayed_cksum(m); 822 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 823 } 824 825 /* Restore packet header fields to original values */ 826 ip->ip_len = htons(ip->ip_len); 827 ip->ip_off = htons(ip->ip_off); 828 829 /* Deliver packet to divert input routine */ 830 divert_packet(m, 0, off & 0xffff, args.divert_rule); 831 832 /* If 'tee', continue with original packet */ 833 if (clone != NULL) { 834 m = clone; 835 ip = mtod(m, struct ip *); 836 goto pass; 837 } 838 goto done; 839 } 840#endif 841 842 /* IPFIREWALL_FORWARD */ 843 /* 844 * Check dst to make sure it is directly reachable on the 845 * interface we previously thought it was. 846 * If it isn't (which may be likely in some situations) we have 847 * to re-route it (ie, find a route for the next-hop and the 848 * associated interface) and set them here. This is nested 849 * forwarding which in most cases is undesirable, except where 850 * such control is nigh impossible. So we do it here. 851 * And I'm babbling. 852 */ 853 if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 854#if 0 855 /* 856 * XXX To improve readability, this block should be 857 * changed into a function call as below: 858 */ 859 error = ip_ipforward(&m, &dst, &ifp); 860 if (error) 861 goto bad; 862 if (m == NULL) /* ip_input consumed the mbuf */ 863 goto done; 864#else 865 struct in_ifaddr *ia; 866 867 /* 868 * XXX sro_fwd below is static, and a pointer 869 * to it gets passed to routines downstream. 870 * This could have surprisingly bad results in 871 * practice, because its content is overwritten 872 * by subsequent packets. 873 */ 874 /* There must be a better way to do this next line... */ 875 static struct route sro_fwd; 876 struct route *ro_fwd = &sro_fwd; 877 878#if 0 879 print_ip("IPFIREWALL_FORWARD: New dst ip: ", 880 dst->sin_addr, "\n"); 881#endif 882 883 /* 884 * We need to figure out if we have been forwarded 885 * to a local socket. If so, then we should somehow 886 * "loop back" to ip_input, and get directed to the 887 * PCB as if we had received this packet. This is 888 * because it may be dificult to identify the packets 889 * you want to forward until they are being output 890 * and have selected an interface. (e.g. locally 891 * initiated packets) If we used the loopback inteface, 892 * we would not be able to control what happens 893 * as the packet runs through ip_input() as 894 * it is done through an ISR. 895 */ 896 LIST_FOREACH(ia, 897 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 898 /* 899 * If the addr to forward to is one 900 * of ours, we pretend to 901 * be the destination for this packet. 902 */ 903 if (IA_SIN(ia)->sin_addr.s_addr == 904 dst->sin_addr.s_addr) 905 break; 906 } 907 if (ia) { /* tell ip_input "dont filter" */ 908 struct m_hdr tag; 909 910 tag.mh_type = MT_TAG; 911 tag.mh_flags = PACKET_TAG_IPFORWARD; 912 tag.mh_data = (caddr_t)args.next_hop; 913 tag.mh_next = m; 914 tag.mh_nextpkt = NULL; 915 916 if (m->m_pkthdr.rcvif == NULL) 917 m->m_pkthdr.rcvif = ifunit("lo0"); 918 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 919 m->m_pkthdr.csum_flags |= 920 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 921 m0->m_pkthdr.csum_data = 0xffff; 922 } 923 m->m_pkthdr.csum_flags |= 924 CSUM_IP_CHECKED | CSUM_IP_VALID; 925 ip->ip_len = htons(ip->ip_len); 926 ip->ip_off = htons(ip->ip_off); 927 ip_input((struct mbuf *)&tag); 928 goto done; 929 } 930 /* 931 * Some of the logic for this was 932 * nicked from above. 933 */ 934 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 935 936 ro_fwd->ro_rt = 0; 937 rtalloc_ign(ro_fwd, RTF_CLONING); 938 939 if (ro_fwd->ro_rt == 0) { 940 ipstat.ips_noroute++; 941 error = EHOSTUNREACH; 942 goto bad; 943 } 944 945 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 946 ifp = ro_fwd->ro_rt->rt_ifp; 947 ro_fwd->ro_rt->rt_rmx.rmx_pksent++; 948 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 949 dst = (struct sockaddr_in *) 950 ro_fwd->ro_rt->rt_gateway; 951 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 952 isbroadcast = 953 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 954 else 955 isbroadcast = in_broadcast(dst->sin_addr, ifp); 956 if (ro->ro_rt) 957 RTFREE(ro->ro_rt); 958 ro->ro_rt = ro_fwd->ro_rt; 959 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 960 961#endif /* ... block to be put into a function */ 962 /* 963 * If we added a default src ip earlier, 964 * which would have been gotten from the-then 965 * interface, do it again, from the new one. 966 */ 967 if (src_was_INADDR_ANY) 968 ip->ip_src = IA_SIN(ia)->sin_addr; 969 goto pass ; 970 } 971 972 /* 973 * if we get here, none of the above matches, and 974 * we have to drop the pkt 975 */ 976 m_freem(m); 977 error = EACCES; /* not sure this is the right error msg */ 978 goto done; 979 } 980 981pass: 982 /* 127/8 must not appear on wire - RFC1122. */ 983 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 984 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 985 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 986 ipstat.ips_badaddr++; 987 error = EADDRNOTAVAIL; 988 goto bad; 989 } 990 } 991 992 m->m_pkthdr.csum_flags |= CSUM_IP; 993 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 994 if (sw_csum & CSUM_DELAY_DATA) { 995 in_delayed_cksum(m); 996 sw_csum &= ~CSUM_DELAY_DATA; 997 } 998 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 999 1000 /* 1001 * If small enough for interface, or the interface will take 1002 * care of the fragmentation for us, can just send directly. 1003 */ 1004 if (ip->ip_len <= ifp->if_mtu || (ifp->if_hwassist & CSUM_FRAGMENT && 1005 ((ip->ip_off & IP_DF) == 0))) { 1006 ip->ip_len = htons(ip->ip_len); 1007 ip->ip_off = htons(ip->ip_off); 1008 ip->ip_sum = 0; 1009 if (sw_csum & CSUM_DELAY_IP) 1010 ip->ip_sum = in_cksum(m, hlen); 1011 1012 /* Record statistics for this interface address. */ 1013 if (!(flags & IP_FORWARDING) && ia) { 1014 ia->ia_ifa.if_opackets++; 1015 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1016 } 1017 1018#ifdef IPSEC 1019 /* clean ipsec history once it goes out of the node */ 1020 ipsec_delaux(m); 1021#endif 1022 1023#ifdef MBUF_STRESS_TEST 1024 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 1025 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 1026#endif 1027 error = (*ifp->if_output)(ifp, m, 1028 (struct sockaddr *)dst, ro->ro_rt); 1029 goto done; 1030 } 1031 1032 if (ip->ip_off & IP_DF) { 1033 error = EMSGSIZE; 1034 /* 1035 * This case can happen if the user changed the MTU 1036 * of an interface after enabling IP on it. Because 1037 * most netifs don't keep track of routes pointing to 1038 * them, there is no way for one to update all its 1039 * routes when the MTU is changed. 1040 */ 1041 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 1042 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 1043 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 1044 } 1045 ipstat.ips_cantfrag++; 1046 goto bad; 1047 } 1048 1049 /* 1050 * Too large for interface; fragment if possible. If successful, 1051 * on return, m will point to a list of packets to be sent. 1052 */ 1053 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 1054 if (error) 1055 goto bad; 1056 for (; m; m = m0) { 1057 m0 = m->m_nextpkt; 1058 m->m_nextpkt = 0; 1059#ifdef IPSEC 1060 /* clean ipsec history once it goes out of the node */ 1061 ipsec_delaux(m); 1062#endif 1063 if (error == 0) { 1064 /* Record statistics for this interface address. */ 1065 if (ia != NULL) { 1066 ia->ia_ifa.if_opackets++; 1067 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1068 } 1069 1070 error = (*ifp->if_output)(ifp, m, 1071 (struct sockaddr *)dst, ro->ro_rt); 1072 } else 1073 m_freem(m); 1074 } 1075 1076 if (error == 0) 1077 ipstat.ips_fragmented++; 1078 1079done: 1080 if (ro == &iproute && ro->ro_rt) { 1081 RTFREE(ro->ro_rt); 1082 ro->ro_rt = NULL; 1083 } 1084#ifdef IPSEC 1085 if (sp != NULL) { 1086 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1087 printf("DP ip_output call free SP:%p\n", sp)); 1088 key_freesp(sp); 1089 } 1090#endif 1091#ifdef FAST_IPSEC 1092 if (sp != NULL) 1093 KEY_FREESP(&sp); 1094#endif 1095 return (error); 1096bad: 1097 m_freem(m); 1098 goto done; 1099} 1100 1101/* 1102 * Create a chain of fragments which fit the given mtu. m_frag points to the 1103 * mbuf to be fragmented; on return it points to the chain with the fragments. 1104 * Return 0 if no error. If error, m_frag may contain a partially built 1105 * chain of fragments that should be freed by the caller. 1106 * 1107 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 1108 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 1109 */ 1110int 1111ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 1112 u_long if_hwassist_flags, int sw_csum) 1113{ 1114 int error = 0; 1115 int hlen = ip->ip_hl << 2; 1116 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 1117 int off; 1118 struct mbuf *m0 = *m_frag; /* the original packet */ 1119 int firstlen; 1120 struct mbuf **mnext; 1121 int nfrags; 1122 1123 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 1124 ipstat.ips_cantfrag++; 1125 return EMSGSIZE; 1126 } 1127 1128 /* 1129 * Must be able to put at least 8 bytes per fragment. 1130 */ 1131 if (len < 8) 1132 return EMSGSIZE; 1133 1134 /* 1135 * If the interface will not calculate checksums on 1136 * fragmented packets, then do it here. 1137 */ 1138 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 1139 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 1140 in_delayed_cksum(m0); 1141 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1142 } 1143 1144 if (len > PAGE_SIZE) { 1145 /* 1146 * Fragment large datagrams such that each segment 1147 * contains a multiple of PAGE_SIZE amount of data, 1148 * plus headers. This enables a receiver to perform 1149 * page-flipping zero-copy optimizations. 1150 * 1151 * XXX When does this help given that sender and receiver 1152 * could have different page sizes, and also mtu could 1153 * be less than the receiver's page size ? 1154 */ 1155 int newlen; 1156 struct mbuf *m; 1157 1158 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 1159 off += m->m_len; 1160 1161 /* 1162 * firstlen (off - hlen) must be aligned on an 1163 * 8-byte boundary 1164 */ 1165 if (off < hlen) 1166 goto smart_frag_failure; 1167 off = ((off - hlen) & ~7) + hlen; 1168 newlen = (~PAGE_MASK) & mtu; 1169 if ((newlen + sizeof (struct ip)) > mtu) { 1170 /* we failed, go back the default */ 1171smart_frag_failure: 1172 newlen = len; 1173 off = hlen + len; 1174 } 1175 len = newlen; 1176 1177 } else { 1178 off = hlen + len; 1179 } 1180 1181 firstlen = off - hlen; 1182 mnext = &m0->m_nextpkt; /* pointer to next packet */ 1183 1184 /* 1185 * Loop through length of segment after first fragment, 1186 * make new header and copy data of each part and link onto chain. 1187 * Here, m0 is the original packet, m is the fragment being created. 1188 * The fragments are linked off the m_nextpkt of the original 1189 * packet, which after processing serves as the first fragment. 1190 */ 1191 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 1192 struct ip *mhip; /* ip header on the fragment */ 1193 struct mbuf *m; 1194 int mhlen = sizeof (struct ip); 1195 1196 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1197 if (m == 0) { 1198 error = ENOBUFS; 1199 ipstat.ips_odropped++; 1200 goto done; 1201 } 1202 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1203 /* 1204 * In the first mbuf, leave room for the link header, then 1205 * copy the original IP header including options. The payload 1206 * goes into an additional mbuf chain returned by m_copy(). 1207 */ 1208 m->m_data += max_linkhdr; 1209 mhip = mtod(m, struct ip *); 1210 *mhip = *ip; 1211 if (hlen > sizeof (struct ip)) { 1212 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1213 mhip->ip_v = IPVERSION; 1214 mhip->ip_hl = mhlen >> 2; 1215 } 1216 m->m_len = mhlen; 1217 /* XXX do we need to add ip->ip_off below ? */ 1218 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1219 if (off + len >= ip->ip_len) { /* last fragment */ 1220 len = ip->ip_len - off; 1221 m->m_flags |= M_LASTFRAG; 1222 } else 1223 mhip->ip_off |= IP_MF; 1224 mhip->ip_len = htons((u_short)(len + mhlen)); 1225 m->m_next = m_copy(m0, off, len); 1226 if (m->m_next == 0) { /* copy failed */ 1227 m_free(m); 1228 error = ENOBUFS; /* ??? */ 1229 ipstat.ips_odropped++; 1230 goto done; 1231 } 1232 m->m_pkthdr.len = mhlen + len; 1233 m->m_pkthdr.rcvif = (struct ifnet *)0; 1234#ifdef MAC 1235 mac_create_fragment(m0, m); 1236#endif 1237 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1238 mhip->ip_off = htons(mhip->ip_off); 1239 mhip->ip_sum = 0; 1240 if (sw_csum & CSUM_DELAY_IP) 1241 mhip->ip_sum = in_cksum(m, mhlen); 1242 *mnext = m; 1243 mnext = &m->m_nextpkt; 1244 } 1245 ipstat.ips_ofragments += nfrags; 1246 1247 /* set first marker for fragment chain */ 1248 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1249 m0->m_pkthdr.csum_data = nfrags; 1250 1251 /* 1252 * Update first fragment by trimming what's been copied out 1253 * and updating header. 1254 */ 1255 m_adj(m0, hlen + firstlen - ip->ip_len); 1256 m0->m_pkthdr.len = hlen + firstlen; 1257 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 1258 ip->ip_off |= IP_MF; 1259 ip->ip_off = htons(ip->ip_off); 1260 ip->ip_sum = 0; 1261 if (sw_csum & CSUM_DELAY_IP) 1262 ip->ip_sum = in_cksum(m0, hlen); 1263 1264done: 1265 *m_frag = m0; 1266 return error; 1267} 1268 1269void 1270in_delayed_cksum(struct mbuf *m) 1271{ 1272 struct ip *ip; 1273 u_short csum, offset; 1274 1275 ip = mtod(m, struct ip *); 1276 offset = ip->ip_hl << 2 ; 1277 csum = in_cksum_skip(m, ip->ip_len, offset); 1278 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1279 csum = 0xffff; 1280 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1281 1282 if (offset + sizeof(u_short) > m->m_len) { 1283 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1284 m->m_len, offset, ip->ip_p); 1285 /* 1286 * XXX 1287 * this shouldn't happen, but if it does, the 1288 * correct behavior may be to insert the checksum 1289 * in the existing chain instead of rearranging it. 1290 */ 1291 m = m_pullup(m, offset + sizeof(u_short)); 1292 } 1293 *(u_short *)(m->m_data + offset) = csum; 1294} 1295 1296/* 1297 * Insert IP options into preformed packet. 1298 * Adjust IP destination as required for IP source routing, 1299 * as indicated by a non-zero in_addr at the start of the options. 1300 * 1301 * XXX This routine assumes that the packet has no options in place. 1302 */ 1303static struct mbuf * 1304ip_insertoptions(m, opt, phlen) 1305 register struct mbuf *m; 1306 struct mbuf *opt; 1307 int *phlen; 1308{ 1309 register struct ipoption *p = mtod(opt, struct ipoption *); 1310 struct mbuf *n; 1311 register struct ip *ip = mtod(m, struct ip *); 1312 unsigned optlen; 1313 1314 optlen = opt->m_len - sizeof(p->ipopt_dst); 1315 if (optlen + ip->ip_len > IP_MAXPACKET) { 1316 *phlen = 0; 1317 return (m); /* XXX should fail */ 1318 } 1319 if (p->ipopt_dst.s_addr) 1320 ip->ip_dst = p->ipopt_dst; 1321 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1322 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1323 if (n == 0) { 1324 *phlen = 0; 1325 return (m); 1326 } 1327 n->m_pkthdr.rcvif = (struct ifnet *)0; 1328#ifdef MAC 1329 mac_create_mbuf_from_mbuf(m, n); 1330#endif 1331 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1332 m->m_len -= sizeof(struct ip); 1333 m->m_data += sizeof(struct ip); 1334 n->m_next = m; 1335 m = n; 1336 m->m_len = optlen + sizeof(struct ip); 1337 m->m_data += max_linkhdr; 1338 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1339 } else { 1340 m->m_data -= optlen; 1341 m->m_len += optlen; 1342 m->m_pkthdr.len += optlen; 1343 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1344 } 1345 ip = mtod(m, struct ip *); 1346 bcopy(p->ipopt_list, ip + 1, optlen); 1347 *phlen = sizeof(struct ip) + optlen; 1348 ip->ip_v = IPVERSION; 1349 ip->ip_hl = *phlen >> 2; 1350 ip->ip_len += optlen; 1351 return (m); 1352} 1353 1354/* 1355 * Copy options from ip to jp, 1356 * omitting those not copied during fragmentation. 1357 */ 1358int 1359ip_optcopy(ip, jp) 1360 struct ip *ip, *jp; 1361{ 1362 register u_char *cp, *dp; 1363 int opt, optlen, cnt; 1364 1365 cp = (u_char *)(ip + 1); 1366 dp = (u_char *)(jp + 1); 1367 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1368 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1369 opt = cp[0]; 1370 if (opt == IPOPT_EOL) 1371 break; 1372 if (opt == IPOPT_NOP) { 1373 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1374 *dp++ = IPOPT_NOP; 1375 optlen = 1; 1376 continue; 1377 } 1378 1379 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1380 ("ip_optcopy: malformed ipv4 option")); 1381 optlen = cp[IPOPT_OLEN]; 1382 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1383 ("ip_optcopy: malformed ipv4 option")); 1384 1385 /* bogus lengths should have been caught by ip_dooptions */ 1386 if (optlen > cnt) 1387 optlen = cnt; 1388 if (IPOPT_COPIED(opt)) { 1389 bcopy(cp, dp, optlen); 1390 dp += optlen; 1391 } 1392 } 1393 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1394 *dp++ = IPOPT_EOL; 1395 return (optlen); 1396} 1397 1398/* 1399 * IP socket option processing. 1400 */ 1401int 1402ip_ctloutput(so, sopt) 1403 struct socket *so; 1404 struct sockopt *sopt; 1405{ 1406 struct inpcb *inp = sotoinpcb(so); 1407 int error, optval; 1408 1409 error = optval = 0; 1410 if (sopt->sopt_level != IPPROTO_IP) { 1411 return (EINVAL); 1412 } 1413 1414 switch (sopt->sopt_dir) { 1415 case SOPT_SET: 1416 switch (sopt->sopt_name) { 1417 case IP_OPTIONS: 1418#ifdef notyet 1419 case IP_RETOPTS: 1420#endif 1421 { 1422 struct mbuf *m; 1423 if (sopt->sopt_valsize > MLEN) { 1424 error = EMSGSIZE; 1425 break; 1426 } 1427 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1428 if (m == 0) { 1429 error = ENOBUFS; 1430 break; 1431 } 1432 m->m_len = sopt->sopt_valsize; 1433 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1434 m->m_len); 1435 1436 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1437 m)); 1438 } 1439 1440 case IP_TOS: 1441 case IP_TTL: 1442 case IP_RECVOPTS: 1443 case IP_RECVRETOPTS: 1444 case IP_RECVDSTADDR: 1445 case IP_RECVTTL: 1446 case IP_RECVIF: 1447 case IP_FAITH: 1448 case IP_ONESBCAST: 1449 error = sooptcopyin(sopt, &optval, sizeof optval, 1450 sizeof optval); 1451 if (error) 1452 break; 1453 1454 switch (sopt->sopt_name) { 1455 case IP_TOS: 1456 inp->inp_ip_tos = optval; 1457 break; 1458 1459 case IP_TTL: 1460 inp->inp_ip_ttl = optval; 1461 break; 1462#define OPTSET(bit) \ 1463 if (optval) \ 1464 inp->inp_flags |= bit; \ 1465 else \ 1466 inp->inp_flags &= ~bit; 1467 1468 case IP_RECVOPTS: 1469 OPTSET(INP_RECVOPTS); 1470 break; 1471 1472 case IP_RECVRETOPTS: 1473 OPTSET(INP_RECVRETOPTS); 1474 break; 1475 1476 case IP_RECVDSTADDR: 1477 OPTSET(INP_RECVDSTADDR); 1478 break; 1479 1480 case IP_RECVTTL: 1481 OPTSET(INP_RECVTTL); 1482 break; 1483 1484 case IP_RECVIF: 1485 OPTSET(INP_RECVIF); 1486 break; 1487 1488 case IP_FAITH: 1489 OPTSET(INP_FAITH); 1490 break; 1491 1492 case IP_ONESBCAST: 1493 OPTSET(INP_ONESBCAST); 1494 break; 1495 } 1496 break; 1497#undef OPTSET 1498 1499 case IP_MULTICAST_IF: 1500 case IP_MULTICAST_VIF: 1501 case IP_MULTICAST_TTL: 1502 case IP_MULTICAST_LOOP: 1503 case IP_ADD_MEMBERSHIP: 1504 case IP_DROP_MEMBERSHIP: 1505 error = ip_setmoptions(sopt, &inp->inp_moptions); 1506 break; 1507 1508 case IP_PORTRANGE: 1509 error = sooptcopyin(sopt, &optval, sizeof optval, 1510 sizeof optval); 1511 if (error) 1512 break; 1513 1514 switch (optval) { 1515 case IP_PORTRANGE_DEFAULT: 1516 inp->inp_flags &= ~(INP_LOWPORT); 1517 inp->inp_flags &= ~(INP_HIGHPORT); 1518 break; 1519 1520 case IP_PORTRANGE_HIGH: 1521 inp->inp_flags &= ~(INP_LOWPORT); 1522 inp->inp_flags |= INP_HIGHPORT; 1523 break; 1524 1525 case IP_PORTRANGE_LOW: 1526 inp->inp_flags &= ~(INP_HIGHPORT); 1527 inp->inp_flags |= INP_LOWPORT; 1528 break; 1529 1530 default: 1531 error = EINVAL; 1532 break; 1533 } 1534 break; 1535 1536#if defined(IPSEC) || defined(FAST_IPSEC) 1537 case IP_IPSEC_POLICY: 1538 { 1539 caddr_t req; 1540 size_t len = 0; 1541 int priv; 1542 struct mbuf *m; 1543 int optname; 1544 1545 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1546 break; 1547 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1548 break; 1549 priv = (sopt->sopt_td != NULL && 1550 suser(sopt->sopt_td) != 0) ? 0 : 1; 1551 req = mtod(m, caddr_t); 1552 len = m->m_len; 1553 optname = sopt->sopt_name; 1554 error = ipsec4_set_policy(inp, optname, req, len, priv); 1555 m_freem(m); 1556 break; 1557 } 1558#endif /*IPSEC*/ 1559 1560 default: 1561 error = ENOPROTOOPT; 1562 break; 1563 } 1564 break; 1565 1566 case SOPT_GET: 1567 switch (sopt->sopt_name) { 1568 case IP_OPTIONS: 1569 case IP_RETOPTS: 1570 if (inp->inp_options) 1571 error = sooptcopyout(sopt, 1572 mtod(inp->inp_options, 1573 char *), 1574 inp->inp_options->m_len); 1575 else 1576 sopt->sopt_valsize = 0; 1577 break; 1578 1579 case IP_TOS: 1580 case IP_TTL: 1581 case IP_RECVOPTS: 1582 case IP_RECVRETOPTS: 1583 case IP_RECVDSTADDR: 1584 case IP_RECVTTL: 1585 case IP_RECVIF: 1586 case IP_PORTRANGE: 1587 case IP_FAITH: 1588 case IP_ONESBCAST: 1589 switch (sopt->sopt_name) { 1590 1591 case IP_TOS: 1592 optval = inp->inp_ip_tos; 1593 break; 1594 1595 case IP_TTL: 1596 optval = inp->inp_ip_ttl; 1597 break; 1598 1599#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1600 1601 case IP_RECVOPTS: 1602 optval = OPTBIT(INP_RECVOPTS); 1603 break; 1604 1605 case IP_RECVRETOPTS: 1606 optval = OPTBIT(INP_RECVRETOPTS); 1607 break; 1608 1609 case IP_RECVDSTADDR: 1610 optval = OPTBIT(INP_RECVDSTADDR); 1611 break; 1612 1613 case IP_RECVTTL: 1614 optval = OPTBIT(INP_RECVTTL); 1615 break; 1616 1617 case IP_RECVIF: 1618 optval = OPTBIT(INP_RECVIF); 1619 break; 1620 1621 case IP_PORTRANGE: 1622 if (inp->inp_flags & INP_HIGHPORT) 1623 optval = IP_PORTRANGE_HIGH; 1624 else if (inp->inp_flags & INP_LOWPORT) 1625 optval = IP_PORTRANGE_LOW; 1626 else 1627 optval = 0; 1628 break; 1629 1630 case IP_FAITH: 1631 optval = OPTBIT(INP_FAITH); 1632 break; 1633 1634 case IP_ONESBCAST: 1635 optval = OPTBIT(INP_ONESBCAST); 1636 break; 1637 } 1638 error = sooptcopyout(sopt, &optval, sizeof optval); 1639 break; 1640 1641 case IP_MULTICAST_IF: 1642 case IP_MULTICAST_VIF: 1643 case IP_MULTICAST_TTL: 1644 case IP_MULTICAST_LOOP: 1645 case IP_ADD_MEMBERSHIP: 1646 case IP_DROP_MEMBERSHIP: 1647 error = ip_getmoptions(sopt, inp->inp_moptions); 1648 break; 1649 1650#if defined(IPSEC) || defined(FAST_IPSEC) 1651 case IP_IPSEC_POLICY: 1652 { 1653 struct mbuf *m = NULL; 1654 caddr_t req = NULL; 1655 size_t len = 0; 1656 1657 if (m != 0) { 1658 req = mtod(m, caddr_t); 1659 len = m->m_len; 1660 } 1661 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1662 if (error == 0) 1663 error = soopt_mcopyout(sopt, m); /* XXX */ 1664 if (error == 0) 1665 m_freem(m); 1666 break; 1667 } 1668#endif /*IPSEC*/ 1669 1670 default: 1671 error = ENOPROTOOPT; 1672 break; 1673 } 1674 break; 1675 } 1676 return (error); 1677} 1678 1679/* 1680 * Set up IP options in pcb for insertion in output packets. 1681 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1682 * with destination address if source routed. 1683 */ 1684static int 1685ip_pcbopts(optname, pcbopt, m) 1686 int optname; 1687 struct mbuf **pcbopt; 1688 register struct mbuf *m; 1689{ 1690 register int cnt, optlen; 1691 register u_char *cp; 1692 u_char opt; 1693 1694 /* turn off any old options */ 1695 if (*pcbopt) 1696 (void)m_free(*pcbopt); 1697 *pcbopt = 0; 1698 if (m == (struct mbuf *)0 || m->m_len == 0) { 1699 /* 1700 * Only turning off any previous options. 1701 */ 1702 if (m) 1703 (void)m_free(m); 1704 return (0); 1705 } 1706 1707 if (m->m_len % sizeof(int32_t)) 1708 goto bad; 1709 /* 1710 * IP first-hop destination address will be stored before 1711 * actual options; move other options back 1712 * and clear it when none present. 1713 */ 1714 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1715 goto bad; 1716 cnt = m->m_len; 1717 m->m_len += sizeof(struct in_addr); 1718 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1719 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1720 bzero(mtod(m, void *), sizeof(struct in_addr)); 1721 1722 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1723 opt = cp[IPOPT_OPTVAL]; 1724 if (opt == IPOPT_EOL) 1725 break; 1726 if (opt == IPOPT_NOP) 1727 optlen = 1; 1728 else { 1729 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1730 goto bad; 1731 optlen = cp[IPOPT_OLEN]; 1732 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1733 goto bad; 1734 } 1735 switch (opt) { 1736 1737 default: 1738 break; 1739 1740 case IPOPT_LSRR: 1741 case IPOPT_SSRR: 1742 /* 1743 * user process specifies route as: 1744 * ->A->B->C->D 1745 * D must be our final destination (but we can't 1746 * check that since we may not have connected yet). 1747 * A is first hop destination, which doesn't appear in 1748 * actual IP option, but is stored before the options. 1749 */ 1750 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1751 goto bad; 1752 m->m_len -= sizeof(struct in_addr); 1753 cnt -= sizeof(struct in_addr); 1754 optlen -= sizeof(struct in_addr); 1755 cp[IPOPT_OLEN] = optlen; 1756 /* 1757 * Move first hop before start of options. 1758 */ 1759 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1760 sizeof(struct in_addr)); 1761 /* 1762 * Then copy rest of options back 1763 * to close up the deleted entry. 1764 */ 1765 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1766 &cp[IPOPT_OFFSET+1], 1767 (unsigned)cnt + sizeof(struct in_addr)); 1768 break; 1769 } 1770 } 1771 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1772 goto bad; 1773 *pcbopt = m; 1774 return (0); 1775 1776bad: 1777 (void)m_free(m); 1778 return (EINVAL); 1779} 1780 1781/* 1782 * XXX 1783 * The whole multicast option thing needs to be re-thought. 1784 * Several of these options are equally applicable to non-multicast 1785 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1786 * standard option (IP_TTL). 1787 */ 1788 1789/* 1790 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1791 */ 1792static struct ifnet * 1793ip_multicast_if(a, ifindexp) 1794 struct in_addr *a; 1795 int *ifindexp; 1796{ 1797 int ifindex; 1798 struct ifnet *ifp; 1799 1800 if (ifindexp) 1801 *ifindexp = 0; 1802 if (ntohl(a->s_addr) >> 24 == 0) { 1803 ifindex = ntohl(a->s_addr) & 0xffffff; 1804 if (ifindex < 0 || if_index < ifindex) 1805 return NULL; 1806 ifp = ifnet_byindex(ifindex); 1807 if (ifindexp) 1808 *ifindexp = ifindex; 1809 } else { 1810 INADDR_TO_IFP(*a, ifp); 1811 } 1812 return ifp; 1813} 1814 1815/* 1816 * Set the IP multicast options in response to user setsockopt(). 1817 */ 1818static int 1819ip_setmoptions(sopt, imop) 1820 struct sockopt *sopt; 1821 struct ip_moptions **imop; 1822{ 1823 int error = 0; 1824 int i; 1825 struct in_addr addr; 1826 struct ip_mreq mreq; 1827 struct ifnet *ifp; 1828 struct ip_moptions *imo = *imop; 1829 struct route ro; 1830 struct sockaddr_in *dst; 1831 int ifindex; 1832 int s; 1833 1834 if (imo == NULL) { 1835 /* 1836 * No multicast option buffer attached to the pcb; 1837 * allocate one and initialize to default values. 1838 */ 1839 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1840 M_WAITOK); 1841 1842 if (imo == NULL) 1843 return (ENOBUFS); 1844 *imop = imo; 1845 imo->imo_multicast_ifp = NULL; 1846 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1847 imo->imo_multicast_vif = -1; 1848 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1849 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1850 imo->imo_num_memberships = 0; 1851 } 1852 1853 switch (sopt->sopt_name) { 1854 /* store an index number for the vif you wanna use in the send */ 1855 case IP_MULTICAST_VIF: 1856 if (legal_vif_num == 0) { 1857 error = EOPNOTSUPP; 1858 break; 1859 } 1860 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1861 if (error) 1862 break; 1863 if (!legal_vif_num(i) && (i != -1)) { 1864 error = EINVAL; 1865 break; 1866 } 1867 imo->imo_multicast_vif = i; 1868 break; 1869 1870 case IP_MULTICAST_IF: 1871 /* 1872 * Select the interface for outgoing multicast packets. 1873 */ 1874 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1875 if (error) 1876 break; 1877 /* 1878 * INADDR_ANY is used to remove a previous selection. 1879 * When no interface is selected, a default one is 1880 * chosen every time a multicast packet is sent. 1881 */ 1882 if (addr.s_addr == INADDR_ANY) { 1883 imo->imo_multicast_ifp = NULL; 1884 break; 1885 } 1886 /* 1887 * The selected interface is identified by its local 1888 * IP address. Find the interface and confirm that 1889 * it supports multicasting. 1890 */ 1891 s = splimp(); 1892 ifp = ip_multicast_if(&addr, &ifindex); 1893 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1894 splx(s); 1895 error = EADDRNOTAVAIL; 1896 break; 1897 } 1898 imo->imo_multicast_ifp = ifp; 1899 if (ifindex) 1900 imo->imo_multicast_addr = addr; 1901 else 1902 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1903 splx(s); 1904 break; 1905 1906 case IP_MULTICAST_TTL: 1907 /* 1908 * Set the IP time-to-live for outgoing multicast packets. 1909 * The original multicast API required a char argument, 1910 * which is inconsistent with the rest of the socket API. 1911 * We allow either a char or an int. 1912 */ 1913 if (sopt->sopt_valsize == 1) { 1914 u_char ttl; 1915 error = sooptcopyin(sopt, &ttl, 1, 1); 1916 if (error) 1917 break; 1918 imo->imo_multicast_ttl = ttl; 1919 } else { 1920 u_int ttl; 1921 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1922 sizeof ttl); 1923 if (error) 1924 break; 1925 if (ttl > 255) 1926 error = EINVAL; 1927 else 1928 imo->imo_multicast_ttl = ttl; 1929 } 1930 break; 1931 1932 case IP_MULTICAST_LOOP: 1933 /* 1934 * Set the loopback flag for outgoing multicast packets. 1935 * Must be zero or one. The original multicast API required a 1936 * char argument, which is inconsistent with the rest 1937 * of the socket API. We allow either a char or an int. 1938 */ 1939 if (sopt->sopt_valsize == 1) { 1940 u_char loop; 1941 error = sooptcopyin(sopt, &loop, 1, 1); 1942 if (error) 1943 break; 1944 imo->imo_multicast_loop = !!loop; 1945 } else { 1946 u_int loop; 1947 error = sooptcopyin(sopt, &loop, sizeof loop, 1948 sizeof loop); 1949 if (error) 1950 break; 1951 imo->imo_multicast_loop = !!loop; 1952 } 1953 break; 1954 1955 case IP_ADD_MEMBERSHIP: 1956 /* 1957 * Add a multicast group membership. 1958 * Group must be a valid IP multicast address. 1959 */ 1960 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1961 if (error) 1962 break; 1963 1964 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1965 error = EINVAL; 1966 break; 1967 } 1968 s = splimp(); 1969 /* 1970 * If no interface address was provided, use the interface of 1971 * the route to the given multicast address. 1972 */ 1973 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1974 bzero((caddr_t)&ro, sizeof(ro)); 1975 dst = (struct sockaddr_in *)&ro.ro_dst; 1976 dst->sin_len = sizeof(*dst); 1977 dst->sin_family = AF_INET; 1978 dst->sin_addr = mreq.imr_multiaddr; 1979 rtalloc_ign(&ro, RTF_CLONING); 1980 if (ro.ro_rt == NULL) { 1981 error = EADDRNOTAVAIL; 1982 splx(s); 1983 break; 1984 } 1985 ifp = ro.ro_rt->rt_ifp; 1986 RTFREE(ro.ro_rt); 1987 } 1988 else { 1989 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1990 } 1991 1992 /* 1993 * See if we found an interface, and confirm that it 1994 * supports multicast. 1995 */ 1996 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1997 error = EADDRNOTAVAIL; 1998 splx(s); 1999 break; 2000 } 2001 /* 2002 * See if the membership already exists or if all the 2003 * membership slots are full. 2004 */ 2005 for (i = 0; i < imo->imo_num_memberships; ++i) { 2006 if (imo->imo_membership[i]->inm_ifp == ifp && 2007 imo->imo_membership[i]->inm_addr.s_addr 2008 == mreq.imr_multiaddr.s_addr) 2009 break; 2010 } 2011 if (i < imo->imo_num_memberships) { 2012 error = EADDRINUSE; 2013 splx(s); 2014 break; 2015 } 2016 if (i == IP_MAX_MEMBERSHIPS) { 2017 error = ETOOMANYREFS; 2018 splx(s); 2019 break; 2020 } 2021 /* 2022 * Everything looks good; add a new record to the multicast 2023 * address list for the given interface. 2024 */ 2025 if ((imo->imo_membership[i] = 2026 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 2027 error = ENOBUFS; 2028 splx(s); 2029 break; 2030 } 2031 ++imo->imo_num_memberships; 2032 splx(s); 2033 break; 2034 2035 case IP_DROP_MEMBERSHIP: 2036 /* 2037 * Drop a multicast group membership. 2038 * Group must be a valid IP multicast address. 2039 */ 2040 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 2041 if (error) 2042 break; 2043 2044 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 2045 error = EINVAL; 2046 break; 2047 } 2048 2049 s = splimp(); 2050 /* 2051 * If an interface address was specified, get a pointer 2052 * to its ifnet structure. 2053 */ 2054 if (mreq.imr_interface.s_addr == INADDR_ANY) 2055 ifp = NULL; 2056 else { 2057 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 2058 if (ifp == NULL) { 2059 error = EADDRNOTAVAIL; 2060 splx(s); 2061 break; 2062 } 2063 } 2064 /* 2065 * Find the membership in the membership array. 2066 */ 2067 for (i = 0; i < imo->imo_num_memberships; ++i) { 2068 if ((ifp == NULL || 2069 imo->imo_membership[i]->inm_ifp == ifp) && 2070 imo->imo_membership[i]->inm_addr.s_addr == 2071 mreq.imr_multiaddr.s_addr) 2072 break; 2073 } 2074 if (i == imo->imo_num_memberships) { 2075 error = EADDRNOTAVAIL; 2076 splx(s); 2077 break; 2078 } 2079 /* 2080 * Give up the multicast address record to which the 2081 * membership points. 2082 */ 2083 in_delmulti(imo->imo_membership[i]); 2084 /* 2085 * Remove the gap in the membership array. 2086 */ 2087 for (++i; i < imo->imo_num_memberships; ++i) 2088 imo->imo_membership[i-1] = imo->imo_membership[i]; 2089 --imo->imo_num_memberships; 2090 splx(s); 2091 break; 2092 2093 default: 2094 error = EOPNOTSUPP; 2095 break; 2096 } 2097 2098 /* 2099 * If all options have default values, no need to keep the mbuf. 2100 */ 2101 if (imo->imo_multicast_ifp == NULL && 2102 imo->imo_multicast_vif == -1 && 2103 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2104 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2105 imo->imo_num_memberships == 0) { 2106 free(*imop, M_IPMOPTS); 2107 *imop = NULL; 2108 } 2109 2110 return (error); 2111} 2112 2113/* 2114 * Return the IP multicast options in response to user getsockopt(). 2115 */ 2116static int 2117ip_getmoptions(sopt, imo) 2118 struct sockopt *sopt; 2119 register struct ip_moptions *imo; 2120{ 2121 struct in_addr addr; 2122 struct in_ifaddr *ia; 2123 int error, optval; 2124 u_char coptval; 2125 2126 error = 0; 2127 switch (sopt->sopt_name) { 2128 case IP_MULTICAST_VIF: 2129 if (imo != NULL) 2130 optval = imo->imo_multicast_vif; 2131 else 2132 optval = -1; 2133 error = sooptcopyout(sopt, &optval, sizeof optval); 2134 break; 2135 2136 case IP_MULTICAST_IF: 2137 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2138 addr.s_addr = INADDR_ANY; 2139 else if (imo->imo_multicast_addr.s_addr) { 2140 /* return the value user has set */ 2141 addr = imo->imo_multicast_addr; 2142 } else { 2143 IFP_TO_IA(imo->imo_multicast_ifp, ia); 2144 addr.s_addr = (ia == NULL) ? INADDR_ANY 2145 : IA_SIN(ia)->sin_addr.s_addr; 2146 } 2147 error = sooptcopyout(sopt, &addr, sizeof addr); 2148 break; 2149 2150 case IP_MULTICAST_TTL: 2151 if (imo == 0) 2152 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2153 else 2154 optval = coptval = imo->imo_multicast_ttl; 2155 if (sopt->sopt_valsize == 1) 2156 error = sooptcopyout(sopt, &coptval, 1); 2157 else 2158 error = sooptcopyout(sopt, &optval, sizeof optval); 2159 break; 2160 2161 case IP_MULTICAST_LOOP: 2162 if (imo == 0) 2163 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2164 else 2165 optval = coptval = imo->imo_multicast_loop; 2166 if (sopt->sopt_valsize == 1) 2167 error = sooptcopyout(sopt, &coptval, 1); 2168 else 2169 error = sooptcopyout(sopt, &optval, sizeof optval); 2170 break; 2171 2172 default: 2173 error = ENOPROTOOPT; 2174 break; 2175 } 2176 return (error); 2177} 2178 2179/* 2180 * Discard the IP multicast options. 2181 */ 2182void 2183ip_freemoptions(imo) 2184 register struct ip_moptions *imo; 2185{ 2186 register int i; 2187 2188 if (imo != NULL) { 2189 for (i = 0; i < imo->imo_num_memberships; ++i) 2190 in_delmulti(imo->imo_membership[i]); 2191 free(imo, M_IPMOPTS); 2192 } 2193} 2194 2195/* 2196 * Routine called from ip_output() to loop back a copy of an IP multicast 2197 * packet to the input queue of a specified interface. Note that this 2198 * calls the output routine of the loopback "driver", but with an interface 2199 * pointer that might NOT be a loopback interface -- evil, but easier than 2200 * replicating that code here. 2201 */ 2202static void 2203ip_mloopback(ifp, m, dst, hlen) 2204 struct ifnet *ifp; 2205 register struct mbuf *m; 2206 register struct sockaddr_in *dst; 2207 int hlen; 2208{ 2209 register struct ip *ip; 2210 struct mbuf *copym; 2211 2212 copym = m_copy(m, 0, M_COPYALL); 2213 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2214 copym = m_pullup(copym, hlen); 2215 if (copym != NULL) { 2216 /* 2217 * We don't bother to fragment if the IP length is greater 2218 * than the interface's MTU. Can this possibly matter? 2219 */ 2220 ip = mtod(copym, struct ip *); 2221 ip->ip_len = htons(ip->ip_len); 2222 ip->ip_off = htons(ip->ip_off); 2223 ip->ip_sum = 0; 2224 ip->ip_sum = in_cksum(copym, hlen); 2225 /* 2226 * NB: 2227 * It's not clear whether there are any lingering 2228 * reentrancy problems in other areas which might 2229 * be exposed by using ip_input directly (in 2230 * particular, everything which modifies the packet 2231 * in-place). Yet another option is using the 2232 * protosw directly to deliver the looped back 2233 * packet. For the moment, we'll err on the side 2234 * of safety by using if_simloop(). 2235 */ 2236#if 1 /* XXX */ 2237 if (dst->sin_family != AF_INET) { 2238 printf("ip_mloopback: bad address family %d\n", 2239 dst->sin_family); 2240 dst->sin_family = AF_INET; 2241 } 2242#endif 2243 2244#ifdef notdef 2245 copym->m_pkthdr.rcvif = ifp; 2246 ip_input(copym); 2247#else 2248 /* if the checksum hasn't been computed, mark it as valid */ 2249 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2250 copym->m_pkthdr.csum_flags |= 2251 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2252 copym->m_pkthdr.csum_data = 0xffff; 2253 } 2254 if_simloop(ifp, copym, dst->sin_family, 0); 2255#endif 2256 } 2257} 2258