ip_output.c revision 122921
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 122921 2003-11-20 19:47:31Z 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 * If this is the case, we probably don't want to allocate 306 * a protocol-cloned route since we didn't get one from the 307 * ULP. This lets TCP do its thing, while not burdening 308 * forwarding or ICMP with the overhead of cloning a route. 309 * Of course, we still want to do any cloning requested by 310 * the link layer, as this is probably required in all cases 311 * for correct operation (as it is for ARP). 312 */ 313 if (ro->ro_rt == 0) 314 rtalloc(ro); 315 if (ro->ro_rt == 0) { 316 ipstat.ips_noroute++; 317 error = EHOSTUNREACH; 318 goto bad; 319 } 320 ia = ifatoia(ro->ro_rt->rt_ifa); 321 ifp = ro->ro_rt->rt_ifp; 322 ro->ro_rt->rt_use++; 323 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 324 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 325 if (ro->ro_rt->rt_flags & RTF_HOST) 326 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 327 else 328 isbroadcast = in_broadcast(dst->sin_addr, ifp); 329 } 330 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 331 struct in_multi *inm; 332 333 m->m_flags |= M_MCAST; 334 /* 335 * IP destination address is multicast. Make sure "dst" 336 * still points to the address in "ro". (It may have been 337 * changed to point to a gateway address, above.) 338 */ 339 dst = (struct sockaddr_in *)&ro->ro_dst; 340 /* 341 * See if the caller provided any multicast options 342 */ 343 if (imo != NULL) { 344 ip->ip_ttl = imo->imo_multicast_ttl; 345 if (imo->imo_multicast_vif != -1) 346 ip->ip_src.s_addr = 347 ip_mcast_src ? 348 ip_mcast_src(imo->imo_multicast_vif) : 349 INADDR_ANY; 350 } else 351 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 352 /* 353 * Confirm that the outgoing interface supports multicast. 354 */ 355 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 356 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 357 ipstat.ips_noroute++; 358 error = ENETUNREACH; 359 goto bad; 360 } 361 } 362 /* 363 * If source address not specified yet, use address 364 * of outgoing interface. 365 */ 366 if (ip->ip_src.s_addr == INADDR_ANY) { 367 /* Interface may have no addresses. */ 368 if (ia != NULL) 369 ip->ip_src = IA_SIN(ia)->sin_addr; 370 } 371 372 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 373 /* 374 * XXX 375 * delayed checksums are not currently 376 * compatible with IP multicast routing 377 */ 378 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 379 in_delayed_cksum(m); 380 m->m_pkthdr.csum_flags &= 381 ~CSUM_DELAY_DATA; 382 } 383 } 384 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 385 if (inm != NULL && 386 (imo == NULL || imo->imo_multicast_loop)) { 387 /* 388 * If we belong to the destination multicast group 389 * on the outgoing interface, and the caller did not 390 * forbid loopback, loop back a copy. 391 */ 392 ip_mloopback(ifp, m, dst, hlen); 393 } 394 else { 395 /* 396 * If we are acting as a multicast router, perform 397 * multicast forwarding as if the packet had just 398 * arrived on the interface to which we are about 399 * to send. The multicast forwarding function 400 * recursively calls this function, using the 401 * IP_FORWARDING flag to prevent infinite recursion. 402 * 403 * Multicasts that are looped back by ip_mloopback(), 404 * above, will be forwarded by the ip_input() routine, 405 * if necessary. 406 */ 407 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 408 /* 409 * If rsvp daemon is not running, do not 410 * set ip_moptions. This ensures that the packet 411 * is multicast and not just sent down one link 412 * as prescribed by rsvpd. 413 */ 414 if (!rsvp_on) 415 imo = NULL; 416 if (ip_mforward && 417 ip_mforward(ip, ifp, m, imo) != 0) { 418 m_freem(m); 419 goto done; 420 } 421 } 422 } 423 424 /* 425 * Multicasts with a time-to-live of zero may be looped- 426 * back, above, but must not be transmitted on a network. 427 * Also, multicasts addressed to the loopback interface 428 * are not sent -- the above call to ip_mloopback() will 429 * loop back a copy if this host actually belongs to the 430 * destination group on the loopback interface. 431 */ 432 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 433 m_freem(m); 434 goto done; 435 } 436 437 goto sendit; 438 } 439#ifndef notdef 440 /* 441 * If the source address is not specified yet, use the address 442 * of the outoing interface. In case, keep note we did that, so 443 * if the the firewall changes the next-hop causing the output 444 * interface to change, we can fix that. 445 */ 446 if (ip->ip_src.s_addr == INADDR_ANY) { 447 /* Interface may have no addresses. */ 448 if (ia != NULL) { 449 ip->ip_src = IA_SIN(ia)->sin_addr; 450 src_was_INADDR_ANY = 1; 451 } 452 } 453#endif /* notdef */ 454 /* 455 * Verify that we have any chance at all of being able to queue 456 * the packet or packet fragments 457 */ 458 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 459 ifp->if_snd.ifq_maxlen) { 460 error = ENOBUFS; 461 ipstat.ips_odropped++; 462 goto bad; 463 } 464 465 /* 466 * Look for broadcast address and 467 * verify user is allowed to send 468 * such a packet. 469 */ 470 if (isbroadcast) { 471 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 472 error = EADDRNOTAVAIL; 473 goto bad; 474 } 475 if ((flags & IP_ALLOWBROADCAST) == 0) { 476 error = EACCES; 477 goto bad; 478 } 479 /* don't allow broadcast messages to be fragmented */ 480 if (ip->ip_len > ifp->if_mtu) { 481 error = EMSGSIZE; 482 goto bad; 483 } 484 if (flags & IP_SENDONES) 485 ip->ip_dst.s_addr = INADDR_BROADCAST; 486 m->m_flags |= M_BCAST; 487 } else { 488 m->m_flags &= ~M_BCAST; 489 } 490 491sendit: 492#ifdef IPSEC 493 /* get SP for this packet */ 494 if (so == NULL) 495 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 496 flags, &error); 497 else 498 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 499 500 if (sp == NULL) { 501 ipsecstat.out_inval++; 502 goto bad; 503 } 504 505 error = 0; 506 507 /* check policy */ 508 switch (sp->policy) { 509 case IPSEC_POLICY_DISCARD: 510 /* 511 * This packet is just discarded. 512 */ 513 ipsecstat.out_polvio++; 514 goto bad; 515 516 case IPSEC_POLICY_BYPASS: 517 case IPSEC_POLICY_NONE: 518 /* no need to do IPsec. */ 519 goto skip_ipsec; 520 521 case IPSEC_POLICY_IPSEC: 522 if (sp->req == NULL) { 523 /* acquire a policy */ 524 error = key_spdacquire(sp); 525 goto bad; 526 } 527 break; 528 529 case IPSEC_POLICY_ENTRUST: 530 default: 531 printf("ip_output: Invalid policy found. %d\n", sp->policy); 532 } 533 { 534 struct ipsec_output_state state; 535 bzero(&state, sizeof(state)); 536 state.m = m; 537 if (flags & IP_ROUTETOIF) { 538 state.ro = &iproute; 539 bzero(&iproute, sizeof(iproute)); 540 } else 541 state.ro = ro; 542 state.dst = (struct sockaddr *)dst; 543 544 ip->ip_sum = 0; 545 546 /* 547 * XXX 548 * delayed checksums are not currently compatible with IPsec 549 */ 550 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 551 in_delayed_cksum(m); 552 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 553 } 554 555 ip->ip_len = htons(ip->ip_len); 556 ip->ip_off = htons(ip->ip_off); 557 558 error = ipsec4_output(&state, sp, flags); 559 560 m = state.m; 561 if (flags & IP_ROUTETOIF) { 562 /* 563 * if we have tunnel mode SA, we may need to ignore 564 * IP_ROUTETOIF. 565 */ 566 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 567 flags &= ~IP_ROUTETOIF; 568 ro = state.ro; 569 } 570 } else 571 ro = state.ro; 572 dst = (struct sockaddr_in *)state.dst; 573 if (error) { 574 /* mbuf is already reclaimed in ipsec4_output. */ 575 m0 = NULL; 576 switch (error) { 577 case EHOSTUNREACH: 578 case ENETUNREACH: 579 case EMSGSIZE: 580 case ENOBUFS: 581 case ENOMEM: 582 break; 583 default: 584 printf("ip4_output (ipsec): error code %d\n", error); 585 /*fall through*/ 586 case ENOENT: 587 /* don't show these error codes to the user */ 588 error = 0; 589 break; 590 } 591 goto bad; 592 } 593 } 594 595 /* be sure to update variables that are affected by ipsec4_output() */ 596 ip = mtod(m, struct ip *); 597 hlen = ip->ip_hl << 2; 598 if (ro->ro_rt == NULL) { 599 if ((flags & IP_ROUTETOIF) == 0) { 600 printf("ip_output: " 601 "can't update route after IPsec processing\n"); 602 error = EHOSTUNREACH; /*XXX*/ 603 goto bad; 604 } 605 } else { 606 ia = ifatoia(ro->ro_rt->rt_ifa); 607 ifp = ro->ro_rt->rt_ifp; 608 } 609 610 /* make it flipped, again. */ 611 ip->ip_len = ntohs(ip->ip_len); 612 ip->ip_off = ntohs(ip->ip_off); 613skip_ipsec: 614#endif /*IPSEC*/ 615#ifdef FAST_IPSEC 616 /* 617 * Check the security policy (SP) for the packet and, if 618 * required, do IPsec-related processing. There are two 619 * cases here; the first time a packet is sent through 620 * it will be untagged and handled by ipsec4_checkpolicy. 621 * If the packet is resubmitted to ip_output (e.g. after 622 * AH, ESP, etc. processing), there will be a tag to bypass 623 * the lookup and related policy checking. 624 */ 625 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 626 s = splnet(); 627 if (mtag != NULL) { 628 tdbi = (struct tdb_ident *)(mtag + 1); 629 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 630 if (sp == NULL) 631 error = -EINVAL; /* force silent drop */ 632 m_tag_delete(m, mtag); 633 } else { 634 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 635 &error, inp); 636 } 637 /* 638 * There are four return cases: 639 * sp != NULL apply IPsec policy 640 * sp == NULL, error == 0 no IPsec handling needed 641 * sp == NULL, error == -EINVAL discard packet w/o error 642 * sp == NULL, error != 0 discard packet, report error 643 */ 644 if (sp != NULL) { 645 /* Loop detection, check if ipsec processing already done */ 646 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 647 for (mtag = m_tag_first(m); mtag != NULL; 648 mtag = m_tag_next(m, mtag)) { 649 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 650 continue; 651 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 652 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 653 continue; 654 /* 655 * Check if policy has an SA associated with it. 656 * This can happen when an SP has yet to acquire 657 * an SA; e.g. on first reference. If it occurs, 658 * then we let ipsec4_process_packet do its thing. 659 */ 660 if (sp->req->sav == NULL) 661 break; 662 tdbi = (struct tdb_ident *)(mtag + 1); 663 if (tdbi->spi == sp->req->sav->spi && 664 tdbi->proto == sp->req->sav->sah->saidx.proto && 665 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 666 sizeof (union sockaddr_union)) == 0) { 667 /* 668 * No IPsec processing is needed, free 669 * reference to SP. 670 * 671 * NB: null pointer to avoid free at 672 * done: below. 673 */ 674 KEY_FREESP(&sp), sp = NULL; 675 splx(s); 676 goto spd_done; 677 } 678 } 679 680 /* 681 * Do delayed checksums now because we send before 682 * this is done in the normal processing path. 683 */ 684 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 685 in_delayed_cksum(m); 686 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 687 } 688 689 ip->ip_len = htons(ip->ip_len); 690 ip->ip_off = htons(ip->ip_off); 691 692 /* NB: callee frees mbuf */ 693 error = ipsec4_process_packet(m, sp->req, flags, 0); 694 /* 695 * Preserve KAME behaviour: ENOENT can be returned 696 * when an SA acquire is in progress. Don't propagate 697 * this to user-level; it confuses applications. 698 * 699 * XXX this will go away when the SADB is redone. 700 */ 701 if (error == ENOENT) 702 error = 0; 703 splx(s); 704 goto done; 705 } else { 706 splx(s); 707 708 if (error != 0) { 709 /* 710 * Hack: -EINVAL is used to signal that a packet 711 * should be silently discarded. This is typically 712 * because we asked key management for an SA and 713 * it was delayed (e.g. kicked up to IKE). 714 */ 715 if (error == -EINVAL) 716 error = 0; 717 goto bad; 718 } else { 719 /* No IPsec processing for this packet. */ 720 } 721#ifdef notyet 722 /* 723 * If deferred crypto processing is needed, check that 724 * the interface supports it. 725 */ 726 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 727 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 728 /* notify IPsec to do its own crypto */ 729 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 730 error = EHOSTUNREACH; 731 goto bad; 732 } 733#endif 734 } 735spd_done: 736#endif /* FAST_IPSEC */ 737 738 /* 739 * IpHack's section. 740 * - Xlate: translate packet's addr/port (NAT). 741 * - Firewall: deny/allow/etc. 742 * - Wrap: fake packet's addr/port <unimpl.> 743 * - Encapsulate: put it in another IP and send out. <unimp.> 744 */ 745#ifdef PFIL_HOOKS 746 /* 747 * Run through list of hooks for output packets. 748 */ 749 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT); 750 if (error != 0 || m == NULL) 751 goto done; 752 ip = mtod(m, struct ip *); 753#endif /* PFIL_HOOKS */ 754 755 /* 756 * Check with the firewall... 757 * but not if we are already being fwd'd from a firewall. 758 */ 759 if (fw_enable && IPFW_LOADED && !args.next_hop) { 760 struct sockaddr_in *old = dst; 761 762 args.m = m; 763 args.next_hop = dst; 764 args.oif = ifp; 765 off = ip_fw_chk_ptr(&args); 766 m = args.m; 767 dst = args.next_hop; 768 769 /* 770 * On return we must do the following: 771 * m == NULL -> drop the pkt (old interface, deprecated) 772 * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 773 * 1<=off<= 0xffff -> DIVERT 774 * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 775 * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 776 * dst != old -> IPFIREWALL_FORWARD 777 * off==0, dst==old -> accept 778 * If some of the above modules are not compiled in, then 779 * we should't have to check the corresponding condition 780 * (because the ipfw control socket should not accept 781 * unsupported rules), but better play safe and drop 782 * packets in case of doubt. 783 */ 784 if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 785 if (m) 786 m_freem(m); 787 error = EACCES; 788 goto done; 789 } 790 ip = mtod(m, struct ip *); 791 if (off == 0 && dst == old) /* common case */ 792 goto pass; 793 if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 794 /* 795 * pass the pkt to dummynet. Need to include 796 * pipe number, m, ifp, ro, dst because these are 797 * not recomputed in the next pass. 798 * All other parameters have been already used and 799 * so they are not needed anymore. 800 * XXX note: if the ifp or ro entry are deleted 801 * while a pkt is in dummynet, we are in trouble! 802 */ 803 args.ro = ro; 804 args.dst = dst; 805 args.flags = flags; 806 807 error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 808 &args); 809 goto done; 810 } 811#ifdef IPDIVERT 812 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 813 struct mbuf *clone = NULL; 814 815 /* Clone packet if we're doing a 'tee' */ 816 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 817 clone = m_dup(m, M_DONTWAIT); 818 819 /* 820 * XXX 821 * delayed checksums are not currently compatible 822 * with divert sockets. 823 */ 824 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 825 in_delayed_cksum(m); 826 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 827 } 828 829 /* Restore packet header fields to original values */ 830 ip->ip_len = htons(ip->ip_len); 831 ip->ip_off = htons(ip->ip_off); 832 833 /* Deliver packet to divert input routine */ 834 divert_packet(m, 0, off & 0xffff, args.divert_rule); 835 836 /* If 'tee', continue with original packet */ 837 if (clone != NULL) { 838 m = clone; 839 ip = mtod(m, struct ip *); 840 goto pass; 841 } 842 goto done; 843 } 844#endif 845 846 /* IPFIREWALL_FORWARD */ 847 /* 848 * Check dst to make sure it is directly reachable on the 849 * interface we previously thought it was. 850 * If it isn't (which may be likely in some situations) we have 851 * to re-route it (ie, find a route for the next-hop and the 852 * associated interface) and set them here. This is nested 853 * forwarding which in most cases is undesirable, except where 854 * such control is nigh impossible. So we do it here. 855 * And I'm babbling. 856 */ 857 if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 858#if 0 859 /* 860 * XXX To improve readability, this block should be 861 * changed into a function call as below: 862 */ 863 error = ip_ipforward(&m, &dst, &ifp); 864 if (error) 865 goto bad; 866 if (m == NULL) /* ip_input consumed the mbuf */ 867 goto done; 868#else 869 struct in_ifaddr *ia; 870 871 /* 872 * XXX sro_fwd below is static, and a pointer 873 * to it gets passed to routines downstream. 874 * This could have surprisingly bad results in 875 * practice, because its content is overwritten 876 * by subsequent packets. 877 */ 878 /* There must be a better way to do this next line... */ 879 static struct route sro_fwd; 880 struct route *ro_fwd = &sro_fwd; 881 882#if 0 883 print_ip("IPFIREWALL_FORWARD: New dst ip: ", 884 dst->sin_addr, "\n"); 885#endif 886 887 /* 888 * We need to figure out if we have been forwarded 889 * to a local socket. If so, then we should somehow 890 * "loop back" to ip_input, and get directed to the 891 * PCB as if we had received this packet. This is 892 * because it may be dificult to identify the packets 893 * you want to forward until they are being output 894 * and have selected an interface. (e.g. locally 895 * initiated packets) If we used the loopback inteface, 896 * we would not be able to control what happens 897 * as the packet runs through ip_input() as 898 * it is done through an ISR. 899 */ 900 LIST_FOREACH(ia, 901 INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 902 /* 903 * If the addr to forward to is one 904 * of ours, we pretend to 905 * be the destination for this packet. 906 */ 907 if (IA_SIN(ia)->sin_addr.s_addr == 908 dst->sin_addr.s_addr) 909 break; 910 } 911 if (ia) { /* tell ip_input "dont filter" */ 912 struct m_hdr tag; 913 914 tag.mh_type = MT_TAG; 915 tag.mh_flags = PACKET_TAG_IPFORWARD; 916 tag.mh_data = (caddr_t)args.next_hop; 917 tag.mh_next = m; 918 tag.mh_nextpkt = NULL; 919 920 if (m->m_pkthdr.rcvif == NULL) 921 m->m_pkthdr.rcvif = ifunit("lo0"); 922 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 923 m->m_pkthdr.csum_flags |= 924 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 925 m0->m_pkthdr.csum_data = 0xffff; 926 } 927 m->m_pkthdr.csum_flags |= 928 CSUM_IP_CHECKED | CSUM_IP_VALID; 929 ip->ip_len = htons(ip->ip_len); 930 ip->ip_off = htons(ip->ip_off); 931 ip_input((struct mbuf *)&tag); 932 goto done; 933 } 934 /* Some of the logic for this was 935 * nicked from above. 936 * 937 * This rewrites the cached route in a local PCB. 938 * Is this what we want to do? 939 */ 940 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 941 942 ro_fwd->ro_rt = 0; 943 rtalloc(ro_fwd); 944 945 if (ro_fwd->ro_rt == 0) { 946 ipstat.ips_noroute++; 947 error = EHOSTUNREACH; 948 goto bad; 949 } 950 951 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 952 ifp = ro_fwd->ro_rt->rt_ifp; 953 ro_fwd->ro_rt->rt_use++; 954 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 955 dst = (struct sockaddr_in *) 956 ro_fwd->ro_rt->rt_gateway; 957 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 958 isbroadcast = 959 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 960 else 961 isbroadcast = in_broadcast(dst->sin_addr, ifp); 962 if (ro->ro_rt) 963 RTFREE(ro->ro_rt); 964 ro->ro_rt = ro_fwd->ro_rt; 965 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 966 967#endif /* ... block to be put into a function */ 968 /* 969 * If we added a default src ip earlier, 970 * which would have been gotten from the-then 971 * interface, do it again, from the new one. 972 */ 973 if (src_was_INADDR_ANY) 974 ip->ip_src = IA_SIN(ia)->sin_addr; 975 goto pass ; 976 } 977 978 /* 979 * if we get here, none of the above matches, and 980 * we have to drop the pkt 981 */ 982 m_freem(m); 983 error = EACCES; /* not sure this is the right error msg */ 984 goto done; 985 } 986 987pass: 988 /* 127/8 must not appear on wire - RFC1122. */ 989 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 990 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 991 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 992 ipstat.ips_badaddr++; 993 error = EADDRNOTAVAIL; 994 goto bad; 995 } 996 } 997 998 m->m_pkthdr.csum_flags |= CSUM_IP; 999 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 1000 if (sw_csum & CSUM_DELAY_DATA) { 1001 in_delayed_cksum(m); 1002 sw_csum &= ~CSUM_DELAY_DATA; 1003 } 1004 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 1005 1006 /* 1007 * If small enough for interface, or the interface will take 1008 * care of the fragmentation for us, can just send directly. 1009 */ 1010 if (ip->ip_len <= ifp->if_mtu || (ifp->if_hwassist & CSUM_FRAGMENT && 1011 ((ip->ip_off & IP_DF) == 0))) { 1012 ip->ip_len = htons(ip->ip_len); 1013 ip->ip_off = htons(ip->ip_off); 1014 ip->ip_sum = 0; 1015 if (sw_csum & CSUM_DELAY_IP) 1016 ip->ip_sum = in_cksum(m, hlen); 1017 1018 /* Record statistics for this interface address. */ 1019 if (!(flags & IP_FORWARDING) && ia) { 1020 ia->ia_ifa.if_opackets++; 1021 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1022 } 1023 1024#ifdef IPSEC 1025 /* clean ipsec history once it goes out of the node */ 1026 ipsec_delaux(m); 1027#endif 1028 1029#ifdef MBUF_STRESS_TEST 1030 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 1031 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 1032#endif 1033 error = (*ifp->if_output)(ifp, m, 1034 (struct sockaddr *)dst, ro->ro_rt); 1035 goto done; 1036 } 1037 1038 if (ip->ip_off & IP_DF) { 1039 error = EMSGSIZE; 1040 /* 1041 * This case can happen if the user changed the MTU 1042 * of an interface after enabling IP on it. Because 1043 * most netifs don't keep track of routes pointing to 1044 * them, there is no way for one to update all its 1045 * routes when the MTU is changed. 1046 */ 1047 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 1048 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) && 1049 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 1050 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 1051 } 1052 ipstat.ips_cantfrag++; 1053 goto bad; 1054 } 1055 1056 /* 1057 * Too large for interface; fragment if possible. If successful, 1058 * on return, m will point to a list of packets to be sent. 1059 */ 1060 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 1061 if (error) 1062 goto bad; 1063 for (; m; m = m0) { 1064 m0 = m->m_nextpkt; 1065 m->m_nextpkt = 0; 1066#ifdef IPSEC 1067 /* clean ipsec history once it goes out of the node */ 1068 ipsec_delaux(m); 1069#endif 1070 if (error == 0) { 1071 /* Record statistics for this interface address. */ 1072 if (ia != NULL) { 1073 ia->ia_ifa.if_opackets++; 1074 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1075 } 1076 1077 error = (*ifp->if_output)(ifp, m, 1078 (struct sockaddr *)dst, ro->ro_rt); 1079 } else 1080 m_freem(m); 1081 } 1082 1083 if (error == 0) 1084 ipstat.ips_fragmented++; 1085 1086done: 1087 if (ro == &iproute && ro->ro_rt) { 1088 RTFREE(ro->ro_rt); 1089 ro->ro_rt = NULL; 1090 } 1091#ifdef IPSEC 1092 if (sp != NULL) { 1093 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1094 printf("DP ip_output call free SP:%p\n", sp)); 1095 key_freesp(sp); 1096 } 1097#endif 1098#ifdef FAST_IPSEC 1099 if (sp != NULL) 1100 KEY_FREESP(&sp); 1101#endif 1102 return (error); 1103bad: 1104 m_freem(m); 1105 goto done; 1106} 1107 1108/* 1109 * Create a chain of fragments which fit the given mtu. m_frag points to the 1110 * mbuf to be fragmented; on return it points to the chain with the fragments. 1111 * Return 0 if no error. If error, m_frag may contain a partially built 1112 * chain of fragments that should be freed by the caller. 1113 * 1114 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 1115 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 1116 */ 1117int 1118ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 1119 u_long if_hwassist_flags, int sw_csum) 1120{ 1121 int error = 0; 1122 int hlen = ip->ip_hl << 2; 1123 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 1124 int off; 1125 struct mbuf *m0 = *m_frag; /* the original packet */ 1126 int firstlen; 1127 struct mbuf **mnext; 1128 int nfrags; 1129 1130 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 1131 ipstat.ips_cantfrag++; 1132 return EMSGSIZE; 1133 } 1134 1135 /* 1136 * Must be able to put at least 8 bytes per fragment. 1137 */ 1138 if (len < 8) 1139 return EMSGSIZE; 1140 1141 /* 1142 * If the interface will not calculate checksums on 1143 * fragmented packets, then do it here. 1144 */ 1145 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 1146 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 1147 in_delayed_cksum(m0); 1148 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1149 } 1150 1151 if (len > PAGE_SIZE) { 1152 /* 1153 * Fragment large datagrams such that each segment 1154 * contains a multiple of PAGE_SIZE amount of data, 1155 * plus headers. This enables a receiver to perform 1156 * page-flipping zero-copy optimizations. 1157 * 1158 * XXX When does this help given that sender and receiver 1159 * could have different page sizes, and also mtu could 1160 * be less than the receiver's page size ? 1161 */ 1162 int newlen; 1163 struct mbuf *m; 1164 1165 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 1166 off += m->m_len; 1167 1168 /* 1169 * firstlen (off - hlen) must be aligned on an 1170 * 8-byte boundary 1171 */ 1172 if (off < hlen) 1173 goto smart_frag_failure; 1174 off = ((off - hlen) & ~7) + hlen; 1175 newlen = (~PAGE_MASK) & mtu; 1176 if ((newlen + sizeof (struct ip)) > mtu) { 1177 /* we failed, go back the default */ 1178smart_frag_failure: 1179 newlen = len; 1180 off = hlen + len; 1181 } 1182 len = newlen; 1183 1184 } else { 1185 off = hlen + len; 1186 } 1187 1188 firstlen = off - hlen; 1189 mnext = &m0->m_nextpkt; /* pointer to next packet */ 1190 1191 /* 1192 * Loop through length of segment after first fragment, 1193 * make new header and copy data of each part and link onto chain. 1194 * Here, m0 is the original packet, m is the fragment being created. 1195 * The fragments are linked off the m_nextpkt of the original 1196 * packet, which after processing serves as the first fragment. 1197 */ 1198 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 1199 struct ip *mhip; /* ip header on the fragment */ 1200 struct mbuf *m; 1201 int mhlen = sizeof (struct ip); 1202 1203 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1204 if (m == 0) { 1205 error = ENOBUFS; 1206 ipstat.ips_odropped++; 1207 goto done; 1208 } 1209 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1210 /* 1211 * In the first mbuf, leave room for the link header, then 1212 * copy the original IP header including options. The payload 1213 * goes into an additional mbuf chain returned by m_copy(). 1214 */ 1215 m->m_data += max_linkhdr; 1216 mhip = mtod(m, struct ip *); 1217 *mhip = *ip; 1218 if (hlen > sizeof (struct ip)) { 1219 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1220 mhip->ip_v = IPVERSION; 1221 mhip->ip_hl = mhlen >> 2; 1222 } 1223 m->m_len = mhlen; 1224 /* XXX do we need to add ip->ip_off below ? */ 1225 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1226 if (off + len >= ip->ip_len) { /* last fragment */ 1227 len = ip->ip_len - off; 1228 m->m_flags |= M_LASTFRAG; 1229 } else 1230 mhip->ip_off |= IP_MF; 1231 mhip->ip_len = htons((u_short)(len + mhlen)); 1232 m->m_next = m_copy(m0, off, len); 1233 if (m->m_next == 0) { /* copy failed */ 1234 m_free(m); 1235 error = ENOBUFS; /* ??? */ 1236 ipstat.ips_odropped++; 1237 goto done; 1238 } 1239 m->m_pkthdr.len = mhlen + len; 1240 m->m_pkthdr.rcvif = (struct ifnet *)0; 1241#ifdef MAC 1242 mac_create_fragment(m0, m); 1243#endif 1244 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1245 mhip->ip_off = htons(mhip->ip_off); 1246 mhip->ip_sum = 0; 1247 if (sw_csum & CSUM_DELAY_IP) 1248 mhip->ip_sum = in_cksum(m, mhlen); 1249 *mnext = m; 1250 mnext = &m->m_nextpkt; 1251 } 1252 ipstat.ips_ofragments += nfrags; 1253 1254 /* set first marker for fragment chain */ 1255 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1256 m0->m_pkthdr.csum_data = nfrags; 1257 1258 /* 1259 * Update first fragment by trimming what's been copied out 1260 * and updating header. 1261 */ 1262 m_adj(m0, hlen + firstlen - ip->ip_len); 1263 m0->m_pkthdr.len = hlen + firstlen; 1264 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 1265 ip->ip_off |= IP_MF; 1266 ip->ip_off = htons(ip->ip_off); 1267 ip->ip_sum = 0; 1268 if (sw_csum & CSUM_DELAY_IP) 1269 ip->ip_sum = in_cksum(m0, hlen); 1270 1271done: 1272 *m_frag = m0; 1273 return error; 1274} 1275 1276void 1277in_delayed_cksum(struct mbuf *m) 1278{ 1279 struct ip *ip; 1280 u_short csum, offset; 1281 1282 ip = mtod(m, struct ip *); 1283 offset = ip->ip_hl << 2 ; 1284 csum = in_cksum_skip(m, ip->ip_len, offset); 1285 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1286 csum = 0xffff; 1287 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1288 1289 if (offset + sizeof(u_short) > m->m_len) { 1290 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1291 m->m_len, offset, ip->ip_p); 1292 /* 1293 * XXX 1294 * this shouldn't happen, but if it does, the 1295 * correct behavior may be to insert the checksum 1296 * in the existing chain instead of rearranging it. 1297 */ 1298 m = m_pullup(m, offset + sizeof(u_short)); 1299 } 1300 *(u_short *)(m->m_data + offset) = csum; 1301} 1302 1303/* 1304 * Insert IP options into preformed packet. 1305 * Adjust IP destination as required for IP source routing, 1306 * as indicated by a non-zero in_addr at the start of the options. 1307 * 1308 * XXX This routine assumes that the packet has no options in place. 1309 */ 1310static struct mbuf * 1311ip_insertoptions(m, opt, phlen) 1312 register struct mbuf *m; 1313 struct mbuf *opt; 1314 int *phlen; 1315{ 1316 register struct ipoption *p = mtod(opt, struct ipoption *); 1317 struct mbuf *n; 1318 register struct ip *ip = mtod(m, struct ip *); 1319 unsigned optlen; 1320 1321 optlen = opt->m_len - sizeof(p->ipopt_dst); 1322 if (optlen + ip->ip_len > IP_MAXPACKET) { 1323 *phlen = 0; 1324 return (m); /* XXX should fail */ 1325 } 1326 if (p->ipopt_dst.s_addr) 1327 ip->ip_dst = p->ipopt_dst; 1328 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1329 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1330 if (n == 0) { 1331 *phlen = 0; 1332 return (m); 1333 } 1334 n->m_pkthdr.rcvif = (struct ifnet *)0; 1335#ifdef MAC 1336 mac_create_mbuf_from_mbuf(m, n); 1337#endif 1338 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1339 m->m_len -= sizeof(struct ip); 1340 m->m_data += sizeof(struct ip); 1341 n->m_next = m; 1342 m = n; 1343 m->m_len = optlen + sizeof(struct ip); 1344 m->m_data += max_linkhdr; 1345 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1346 } else { 1347 m->m_data -= optlen; 1348 m->m_len += optlen; 1349 m->m_pkthdr.len += optlen; 1350 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1351 } 1352 ip = mtod(m, struct ip *); 1353 bcopy(p->ipopt_list, ip + 1, optlen); 1354 *phlen = sizeof(struct ip) + optlen; 1355 ip->ip_v = IPVERSION; 1356 ip->ip_hl = *phlen >> 2; 1357 ip->ip_len += optlen; 1358 return (m); 1359} 1360 1361/* 1362 * Copy options from ip to jp, 1363 * omitting those not copied during fragmentation. 1364 */ 1365int 1366ip_optcopy(ip, jp) 1367 struct ip *ip, *jp; 1368{ 1369 register u_char *cp, *dp; 1370 int opt, optlen, cnt; 1371 1372 cp = (u_char *)(ip + 1); 1373 dp = (u_char *)(jp + 1); 1374 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1375 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1376 opt = cp[0]; 1377 if (opt == IPOPT_EOL) 1378 break; 1379 if (opt == IPOPT_NOP) { 1380 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1381 *dp++ = IPOPT_NOP; 1382 optlen = 1; 1383 continue; 1384 } 1385 1386 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1387 ("ip_optcopy: malformed ipv4 option")); 1388 optlen = cp[IPOPT_OLEN]; 1389 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1390 ("ip_optcopy: malformed ipv4 option")); 1391 1392 /* bogus lengths should have been caught by ip_dooptions */ 1393 if (optlen > cnt) 1394 optlen = cnt; 1395 if (IPOPT_COPIED(opt)) { 1396 bcopy(cp, dp, optlen); 1397 dp += optlen; 1398 } 1399 } 1400 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1401 *dp++ = IPOPT_EOL; 1402 return (optlen); 1403} 1404 1405/* 1406 * IP socket option processing. 1407 */ 1408int 1409ip_ctloutput(so, sopt) 1410 struct socket *so; 1411 struct sockopt *sopt; 1412{ 1413 struct inpcb *inp = sotoinpcb(so); 1414 int error, optval; 1415 1416 error = optval = 0; 1417 if (sopt->sopt_level != IPPROTO_IP) { 1418 return (EINVAL); 1419 } 1420 1421 switch (sopt->sopt_dir) { 1422 case SOPT_SET: 1423 switch (sopt->sopt_name) { 1424 case IP_OPTIONS: 1425#ifdef notyet 1426 case IP_RETOPTS: 1427#endif 1428 { 1429 struct mbuf *m; 1430 if (sopt->sopt_valsize > MLEN) { 1431 error = EMSGSIZE; 1432 break; 1433 } 1434 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1435 if (m == 0) { 1436 error = ENOBUFS; 1437 break; 1438 } 1439 m->m_len = sopt->sopt_valsize; 1440 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1441 m->m_len); 1442 1443 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1444 m)); 1445 } 1446 1447 case IP_TOS: 1448 case IP_TTL: 1449 case IP_RECVOPTS: 1450 case IP_RECVRETOPTS: 1451 case IP_RECVDSTADDR: 1452 case IP_RECVTTL: 1453 case IP_RECVIF: 1454 case IP_FAITH: 1455 case IP_ONESBCAST: 1456 error = sooptcopyin(sopt, &optval, sizeof optval, 1457 sizeof optval); 1458 if (error) 1459 break; 1460 1461 switch (sopt->sopt_name) { 1462 case IP_TOS: 1463 inp->inp_ip_tos = optval; 1464 break; 1465 1466 case IP_TTL: 1467 inp->inp_ip_ttl = optval; 1468 break; 1469#define OPTSET(bit) \ 1470 if (optval) \ 1471 inp->inp_flags |= bit; \ 1472 else \ 1473 inp->inp_flags &= ~bit; 1474 1475 case IP_RECVOPTS: 1476 OPTSET(INP_RECVOPTS); 1477 break; 1478 1479 case IP_RECVRETOPTS: 1480 OPTSET(INP_RECVRETOPTS); 1481 break; 1482 1483 case IP_RECVDSTADDR: 1484 OPTSET(INP_RECVDSTADDR); 1485 break; 1486 1487 case IP_RECVTTL: 1488 OPTSET(INP_RECVTTL); 1489 break; 1490 1491 case IP_RECVIF: 1492 OPTSET(INP_RECVIF); 1493 break; 1494 1495 case IP_FAITH: 1496 OPTSET(INP_FAITH); 1497 break; 1498 1499 case IP_ONESBCAST: 1500 OPTSET(INP_ONESBCAST); 1501 break; 1502 } 1503 break; 1504#undef OPTSET 1505 1506 case IP_MULTICAST_IF: 1507 case IP_MULTICAST_VIF: 1508 case IP_MULTICAST_TTL: 1509 case IP_MULTICAST_LOOP: 1510 case IP_ADD_MEMBERSHIP: 1511 case IP_DROP_MEMBERSHIP: 1512 error = ip_setmoptions(sopt, &inp->inp_moptions); 1513 break; 1514 1515 case IP_PORTRANGE: 1516 error = sooptcopyin(sopt, &optval, sizeof optval, 1517 sizeof optval); 1518 if (error) 1519 break; 1520 1521 switch (optval) { 1522 case IP_PORTRANGE_DEFAULT: 1523 inp->inp_flags &= ~(INP_LOWPORT); 1524 inp->inp_flags &= ~(INP_HIGHPORT); 1525 break; 1526 1527 case IP_PORTRANGE_HIGH: 1528 inp->inp_flags &= ~(INP_LOWPORT); 1529 inp->inp_flags |= INP_HIGHPORT; 1530 break; 1531 1532 case IP_PORTRANGE_LOW: 1533 inp->inp_flags &= ~(INP_HIGHPORT); 1534 inp->inp_flags |= INP_LOWPORT; 1535 break; 1536 1537 default: 1538 error = EINVAL; 1539 break; 1540 } 1541 break; 1542 1543#if defined(IPSEC) || defined(FAST_IPSEC) 1544 case IP_IPSEC_POLICY: 1545 { 1546 caddr_t req; 1547 size_t len = 0; 1548 int priv; 1549 struct mbuf *m; 1550 int optname; 1551 1552 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1553 break; 1554 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1555 break; 1556 priv = (sopt->sopt_td != NULL && 1557 suser(sopt->sopt_td) != 0) ? 0 : 1; 1558 req = mtod(m, caddr_t); 1559 len = m->m_len; 1560 optname = sopt->sopt_name; 1561 error = ipsec4_set_policy(inp, optname, req, len, priv); 1562 m_freem(m); 1563 break; 1564 } 1565#endif /*IPSEC*/ 1566 1567 default: 1568 error = ENOPROTOOPT; 1569 break; 1570 } 1571 break; 1572 1573 case SOPT_GET: 1574 switch (sopt->sopt_name) { 1575 case IP_OPTIONS: 1576 case IP_RETOPTS: 1577 if (inp->inp_options) 1578 error = sooptcopyout(sopt, 1579 mtod(inp->inp_options, 1580 char *), 1581 inp->inp_options->m_len); 1582 else 1583 sopt->sopt_valsize = 0; 1584 break; 1585 1586 case IP_TOS: 1587 case IP_TTL: 1588 case IP_RECVOPTS: 1589 case IP_RECVRETOPTS: 1590 case IP_RECVDSTADDR: 1591 case IP_RECVTTL: 1592 case IP_RECVIF: 1593 case IP_PORTRANGE: 1594 case IP_FAITH: 1595 case IP_ONESBCAST: 1596 switch (sopt->sopt_name) { 1597 1598 case IP_TOS: 1599 optval = inp->inp_ip_tos; 1600 break; 1601 1602 case IP_TTL: 1603 optval = inp->inp_ip_ttl; 1604 break; 1605 1606#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1607 1608 case IP_RECVOPTS: 1609 optval = OPTBIT(INP_RECVOPTS); 1610 break; 1611 1612 case IP_RECVRETOPTS: 1613 optval = OPTBIT(INP_RECVRETOPTS); 1614 break; 1615 1616 case IP_RECVDSTADDR: 1617 optval = OPTBIT(INP_RECVDSTADDR); 1618 break; 1619 1620 case IP_RECVTTL: 1621 optval = OPTBIT(INP_RECVTTL); 1622 break; 1623 1624 case IP_RECVIF: 1625 optval = OPTBIT(INP_RECVIF); 1626 break; 1627 1628 case IP_PORTRANGE: 1629 if (inp->inp_flags & INP_HIGHPORT) 1630 optval = IP_PORTRANGE_HIGH; 1631 else if (inp->inp_flags & INP_LOWPORT) 1632 optval = IP_PORTRANGE_LOW; 1633 else 1634 optval = 0; 1635 break; 1636 1637 case IP_FAITH: 1638 optval = OPTBIT(INP_FAITH); 1639 break; 1640 1641 case IP_ONESBCAST: 1642 optval = OPTBIT(INP_ONESBCAST); 1643 break; 1644 } 1645 error = sooptcopyout(sopt, &optval, sizeof optval); 1646 break; 1647 1648 case IP_MULTICAST_IF: 1649 case IP_MULTICAST_VIF: 1650 case IP_MULTICAST_TTL: 1651 case IP_MULTICAST_LOOP: 1652 case IP_ADD_MEMBERSHIP: 1653 case IP_DROP_MEMBERSHIP: 1654 error = ip_getmoptions(sopt, inp->inp_moptions); 1655 break; 1656 1657#if defined(IPSEC) || defined(FAST_IPSEC) 1658 case IP_IPSEC_POLICY: 1659 { 1660 struct mbuf *m = NULL; 1661 caddr_t req = NULL; 1662 size_t len = 0; 1663 1664 if (m != 0) { 1665 req = mtod(m, caddr_t); 1666 len = m->m_len; 1667 } 1668 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1669 if (error == 0) 1670 error = soopt_mcopyout(sopt, m); /* XXX */ 1671 if (error == 0) 1672 m_freem(m); 1673 break; 1674 } 1675#endif /*IPSEC*/ 1676 1677 default: 1678 error = ENOPROTOOPT; 1679 break; 1680 } 1681 break; 1682 } 1683 return (error); 1684} 1685 1686/* 1687 * Set up IP options in pcb for insertion in output packets. 1688 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1689 * with destination address if source routed. 1690 */ 1691static int 1692ip_pcbopts(optname, pcbopt, m) 1693 int optname; 1694 struct mbuf **pcbopt; 1695 register struct mbuf *m; 1696{ 1697 register int cnt, optlen; 1698 register u_char *cp; 1699 u_char opt; 1700 1701 /* turn off any old options */ 1702 if (*pcbopt) 1703 (void)m_free(*pcbopt); 1704 *pcbopt = 0; 1705 if (m == (struct mbuf *)0 || m->m_len == 0) { 1706 /* 1707 * Only turning off any previous options. 1708 */ 1709 if (m) 1710 (void)m_free(m); 1711 return (0); 1712 } 1713 1714 if (m->m_len % sizeof(int32_t)) 1715 goto bad; 1716 /* 1717 * IP first-hop destination address will be stored before 1718 * actual options; move other options back 1719 * and clear it when none present. 1720 */ 1721 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1722 goto bad; 1723 cnt = m->m_len; 1724 m->m_len += sizeof(struct in_addr); 1725 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1726 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1727 bzero(mtod(m, void *), sizeof(struct in_addr)); 1728 1729 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1730 opt = cp[IPOPT_OPTVAL]; 1731 if (opt == IPOPT_EOL) 1732 break; 1733 if (opt == IPOPT_NOP) 1734 optlen = 1; 1735 else { 1736 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1737 goto bad; 1738 optlen = cp[IPOPT_OLEN]; 1739 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1740 goto bad; 1741 } 1742 switch (opt) { 1743 1744 default: 1745 break; 1746 1747 case IPOPT_LSRR: 1748 case IPOPT_SSRR: 1749 /* 1750 * user process specifies route as: 1751 * ->A->B->C->D 1752 * D must be our final destination (but we can't 1753 * check that since we may not have connected yet). 1754 * A is first hop destination, which doesn't appear in 1755 * actual IP option, but is stored before the options. 1756 */ 1757 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1758 goto bad; 1759 m->m_len -= sizeof(struct in_addr); 1760 cnt -= sizeof(struct in_addr); 1761 optlen -= sizeof(struct in_addr); 1762 cp[IPOPT_OLEN] = optlen; 1763 /* 1764 * Move first hop before start of options. 1765 */ 1766 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1767 sizeof(struct in_addr)); 1768 /* 1769 * Then copy rest of options back 1770 * to close up the deleted entry. 1771 */ 1772 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1773 &cp[IPOPT_OFFSET+1], 1774 (unsigned)cnt + sizeof(struct in_addr)); 1775 break; 1776 } 1777 } 1778 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1779 goto bad; 1780 *pcbopt = m; 1781 return (0); 1782 1783bad: 1784 (void)m_free(m); 1785 return (EINVAL); 1786} 1787 1788/* 1789 * XXX 1790 * The whole multicast option thing needs to be re-thought. 1791 * Several of these options are equally applicable to non-multicast 1792 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1793 * standard option (IP_TTL). 1794 */ 1795 1796/* 1797 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1798 */ 1799static struct ifnet * 1800ip_multicast_if(a, ifindexp) 1801 struct in_addr *a; 1802 int *ifindexp; 1803{ 1804 int ifindex; 1805 struct ifnet *ifp; 1806 1807 if (ifindexp) 1808 *ifindexp = 0; 1809 if (ntohl(a->s_addr) >> 24 == 0) { 1810 ifindex = ntohl(a->s_addr) & 0xffffff; 1811 if (ifindex < 0 || if_index < ifindex) 1812 return NULL; 1813 ifp = ifnet_byindex(ifindex); 1814 if (ifindexp) 1815 *ifindexp = ifindex; 1816 } else { 1817 INADDR_TO_IFP(*a, ifp); 1818 } 1819 return ifp; 1820} 1821 1822/* 1823 * Set the IP multicast options in response to user setsockopt(). 1824 */ 1825static int 1826ip_setmoptions(sopt, imop) 1827 struct sockopt *sopt; 1828 struct ip_moptions **imop; 1829{ 1830 int error = 0; 1831 int i; 1832 struct in_addr addr; 1833 struct ip_mreq mreq; 1834 struct ifnet *ifp; 1835 struct ip_moptions *imo = *imop; 1836 struct route ro; 1837 struct sockaddr_in *dst; 1838 int ifindex; 1839 int s; 1840 1841 if (imo == NULL) { 1842 /* 1843 * No multicast option buffer attached to the pcb; 1844 * allocate one and initialize to default values. 1845 */ 1846 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1847 M_WAITOK); 1848 1849 if (imo == NULL) 1850 return (ENOBUFS); 1851 *imop = imo; 1852 imo->imo_multicast_ifp = NULL; 1853 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1854 imo->imo_multicast_vif = -1; 1855 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1856 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1857 imo->imo_num_memberships = 0; 1858 } 1859 1860 switch (sopt->sopt_name) { 1861 /* store an index number for the vif you wanna use in the send */ 1862 case IP_MULTICAST_VIF: 1863 if (legal_vif_num == 0) { 1864 error = EOPNOTSUPP; 1865 break; 1866 } 1867 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1868 if (error) 1869 break; 1870 if (!legal_vif_num(i) && (i != -1)) { 1871 error = EINVAL; 1872 break; 1873 } 1874 imo->imo_multicast_vif = i; 1875 break; 1876 1877 case IP_MULTICAST_IF: 1878 /* 1879 * Select the interface for outgoing multicast packets. 1880 */ 1881 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1882 if (error) 1883 break; 1884 /* 1885 * INADDR_ANY is used to remove a previous selection. 1886 * When no interface is selected, a default one is 1887 * chosen every time a multicast packet is sent. 1888 */ 1889 if (addr.s_addr == INADDR_ANY) { 1890 imo->imo_multicast_ifp = NULL; 1891 break; 1892 } 1893 /* 1894 * The selected interface is identified by its local 1895 * IP address. Find the interface and confirm that 1896 * it supports multicasting. 1897 */ 1898 s = splimp(); 1899 ifp = ip_multicast_if(&addr, &ifindex); 1900 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1901 splx(s); 1902 error = EADDRNOTAVAIL; 1903 break; 1904 } 1905 imo->imo_multicast_ifp = ifp; 1906 if (ifindex) 1907 imo->imo_multicast_addr = addr; 1908 else 1909 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1910 splx(s); 1911 break; 1912 1913 case IP_MULTICAST_TTL: 1914 /* 1915 * Set the IP time-to-live for outgoing multicast packets. 1916 * The original multicast API required a char argument, 1917 * which is inconsistent with the rest of the socket API. 1918 * We allow either a char or an int. 1919 */ 1920 if (sopt->sopt_valsize == 1) { 1921 u_char ttl; 1922 error = sooptcopyin(sopt, &ttl, 1, 1); 1923 if (error) 1924 break; 1925 imo->imo_multicast_ttl = ttl; 1926 } else { 1927 u_int ttl; 1928 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1929 sizeof ttl); 1930 if (error) 1931 break; 1932 if (ttl > 255) 1933 error = EINVAL; 1934 else 1935 imo->imo_multicast_ttl = ttl; 1936 } 1937 break; 1938 1939 case IP_MULTICAST_LOOP: 1940 /* 1941 * Set the loopback flag for outgoing multicast packets. 1942 * Must be zero or one. The original multicast API required a 1943 * char argument, which is inconsistent with the rest 1944 * of the socket API. We allow either a char or an int. 1945 */ 1946 if (sopt->sopt_valsize == 1) { 1947 u_char loop; 1948 error = sooptcopyin(sopt, &loop, 1, 1); 1949 if (error) 1950 break; 1951 imo->imo_multicast_loop = !!loop; 1952 } else { 1953 u_int loop; 1954 error = sooptcopyin(sopt, &loop, sizeof loop, 1955 sizeof loop); 1956 if (error) 1957 break; 1958 imo->imo_multicast_loop = !!loop; 1959 } 1960 break; 1961 1962 case IP_ADD_MEMBERSHIP: 1963 /* 1964 * Add a multicast group membership. 1965 * Group must be a valid IP multicast address. 1966 */ 1967 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1968 if (error) 1969 break; 1970 1971 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1972 error = EINVAL; 1973 break; 1974 } 1975 s = splimp(); 1976 /* 1977 * If no interface address was provided, use the interface of 1978 * the route to the given multicast address. 1979 */ 1980 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1981 bzero((caddr_t)&ro, sizeof(ro)); 1982 dst = (struct sockaddr_in *)&ro.ro_dst; 1983 dst->sin_len = sizeof(*dst); 1984 dst->sin_family = AF_INET; 1985 dst->sin_addr = mreq.imr_multiaddr; 1986 rtalloc(&ro); 1987 if (ro.ro_rt == NULL) { 1988 error = EADDRNOTAVAIL; 1989 splx(s); 1990 break; 1991 } 1992 ifp = ro.ro_rt->rt_ifp; 1993 RTFREE(ro.ro_rt); 1994 } 1995 else { 1996 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1997 } 1998 1999 /* 2000 * See if we found an interface, and confirm that it 2001 * supports multicast. 2002 */ 2003 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 2004 error = EADDRNOTAVAIL; 2005 splx(s); 2006 break; 2007 } 2008 /* 2009 * See if the membership already exists or if all the 2010 * membership slots are full. 2011 */ 2012 for (i = 0; i < imo->imo_num_memberships; ++i) { 2013 if (imo->imo_membership[i]->inm_ifp == ifp && 2014 imo->imo_membership[i]->inm_addr.s_addr 2015 == mreq.imr_multiaddr.s_addr) 2016 break; 2017 } 2018 if (i < imo->imo_num_memberships) { 2019 error = EADDRINUSE; 2020 splx(s); 2021 break; 2022 } 2023 if (i == IP_MAX_MEMBERSHIPS) { 2024 error = ETOOMANYREFS; 2025 splx(s); 2026 break; 2027 } 2028 /* 2029 * Everything looks good; add a new record to the multicast 2030 * address list for the given interface. 2031 */ 2032 if ((imo->imo_membership[i] = 2033 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 2034 error = ENOBUFS; 2035 splx(s); 2036 break; 2037 } 2038 ++imo->imo_num_memberships; 2039 splx(s); 2040 break; 2041 2042 case IP_DROP_MEMBERSHIP: 2043 /* 2044 * Drop a multicast group membership. 2045 * Group must be a valid IP multicast address. 2046 */ 2047 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 2048 if (error) 2049 break; 2050 2051 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 2052 error = EINVAL; 2053 break; 2054 } 2055 2056 s = splimp(); 2057 /* 2058 * If an interface address was specified, get a pointer 2059 * to its ifnet structure. 2060 */ 2061 if (mreq.imr_interface.s_addr == INADDR_ANY) 2062 ifp = NULL; 2063 else { 2064 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 2065 if (ifp == NULL) { 2066 error = EADDRNOTAVAIL; 2067 splx(s); 2068 break; 2069 } 2070 } 2071 /* 2072 * Find the membership in the membership array. 2073 */ 2074 for (i = 0; i < imo->imo_num_memberships; ++i) { 2075 if ((ifp == NULL || 2076 imo->imo_membership[i]->inm_ifp == ifp) && 2077 imo->imo_membership[i]->inm_addr.s_addr == 2078 mreq.imr_multiaddr.s_addr) 2079 break; 2080 } 2081 if (i == imo->imo_num_memberships) { 2082 error = EADDRNOTAVAIL; 2083 splx(s); 2084 break; 2085 } 2086 /* 2087 * Give up the multicast address record to which the 2088 * membership points. 2089 */ 2090 in_delmulti(imo->imo_membership[i]); 2091 /* 2092 * Remove the gap in the membership array. 2093 */ 2094 for (++i; i < imo->imo_num_memberships; ++i) 2095 imo->imo_membership[i-1] = imo->imo_membership[i]; 2096 --imo->imo_num_memberships; 2097 splx(s); 2098 break; 2099 2100 default: 2101 error = EOPNOTSUPP; 2102 break; 2103 } 2104 2105 /* 2106 * If all options have default values, no need to keep the mbuf. 2107 */ 2108 if (imo->imo_multicast_ifp == NULL && 2109 imo->imo_multicast_vif == -1 && 2110 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2111 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2112 imo->imo_num_memberships == 0) { 2113 free(*imop, M_IPMOPTS); 2114 *imop = NULL; 2115 } 2116 2117 return (error); 2118} 2119 2120/* 2121 * Return the IP multicast options in response to user getsockopt(). 2122 */ 2123static int 2124ip_getmoptions(sopt, imo) 2125 struct sockopt *sopt; 2126 register struct ip_moptions *imo; 2127{ 2128 struct in_addr addr; 2129 struct in_ifaddr *ia; 2130 int error, optval; 2131 u_char coptval; 2132 2133 error = 0; 2134 switch (sopt->sopt_name) { 2135 case IP_MULTICAST_VIF: 2136 if (imo != NULL) 2137 optval = imo->imo_multicast_vif; 2138 else 2139 optval = -1; 2140 error = sooptcopyout(sopt, &optval, sizeof optval); 2141 break; 2142 2143 case IP_MULTICAST_IF: 2144 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2145 addr.s_addr = INADDR_ANY; 2146 else if (imo->imo_multicast_addr.s_addr) { 2147 /* return the value user has set */ 2148 addr = imo->imo_multicast_addr; 2149 } else { 2150 IFP_TO_IA(imo->imo_multicast_ifp, ia); 2151 addr.s_addr = (ia == NULL) ? INADDR_ANY 2152 : IA_SIN(ia)->sin_addr.s_addr; 2153 } 2154 error = sooptcopyout(sopt, &addr, sizeof addr); 2155 break; 2156 2157 case IP_MULTICAST_TTL: 2158 if (imo == 0) 2159 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2160 else 2161 optval = coptval = imo->imo_multicast_ttl; 2162 if (sopt->sopt_valsize == 1) 2163 error = sooptcopyout(sopt, &coptval, 1); 2164 else 2165 error = sooptcopyout(sopt, &optval, sizeof optval); 2166 break; 2167 2168 case IP_MULTICAST_LOOP: 2169 if (imo == 0) 2170 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2171 else 2172 optval = coptval = imo->imo_multicast_loop; 2173 if (sopt->sopt_valsize == 1) 2174 error = sooptcopyout(sopt, &coptval, 1); 2175 else 2176 error = sooptcopyout(sopt, &optval, sizeof optval); 2177 break; 2178 2179 default: 2180 error = ENOPROTOOPT; 2181 break; 2182 } 2183 return (error); 2184} 2185 2186/* 2187 * Discard the IP multicast options. 2188 */ 2189void 2190ip_freemoptions(imo) 2191 register struct ip_moptions *imo; 2192{ 2193 register int i; 2194 2195 if (imo != NULL) { 2196 for (i = 0; i < imo->imo_num_memberships; ++i) 2197 in_delmulti(imo->imo_membership[i]); 2198 free(imo, M_IPMOPTS); 2199 } 2200} 2201 2202/* 2203 * Routine called from ip_output() to loop back a copy of an IP multicast 2204 * packet to the input queue of a specified interface. Note that this 2205 * calls the output routine of the loopback "driver", but with an interface 2206 * pointer that might NOT be a loopback interface -- evil, but easier than 2207 * replicating that code here. 2208 */ 2209static void 2210ip_mloopback(ifp, m, dst, hlen) 2211 struct ifnet *ifp; 2212 register struct mbuf *m; 2213 register struct sockaddr_in *dst; 2214 int hlen; 2215{ 2216 register struct ip *ip; 2217 struct mbuf *copym; 2218 2219 copym = m_copy(m, 0, M_COPYALL); 2220 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2221 copym = m_pullup(copym, hlen); 2222 if (copym != NULL) { 2223 /* 2224 * We don't bother to fragment if the IP length is greater 2225 * than the interface's MTU. Can this possibly matter? 2226 */ 2227 ip = mtod(copym, struct ip *); 2228 ip->ip_len = htons(ip->ip_len); 2229 ip->ip_off = htons(ip->ip_off); 2230 ip->ip_sum = 0; 2231 ip->ip_sum = in_cksum(copym, hlen); 2232 /* 2233 * NB: 2234 * It's not clear whether there are any lingering 2235 * reentrancy problems in other areas which might 2236 * be exposed by using ip_input directly (in 2237 * particular, everything which modifies the packet 2238 * in-place). Yet another option is using the 2239 * protosw directly to deliver the looped back 2240 * packet. For the moment, we'll err on the side 2241 * of safety by using if_simloop(). 2242 */ 2243#if 1 /* XXX */ 2244 if (dst->sin_family != AF_INET) { 2245 printf("ip_mloopback: bad address family %d\n", 2246 dst->sin_family); 2247 dst->sin_family = AF_INET; 2248 } 2249#endif 2250 2251#ifdef notdef 2252 copym->m_pkthdr.rcvif = ifp; 2253 ip_input(copym); 2254#else 2255 /* if the checksum hasn't been computed, mark it as valid */ 2256 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2257 copym->m_pkthdr.csum_flags |= 2258 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2259 copym->m_pkthdr.csum_data = 0xffff; 2260 } 2261 if_simloop(ifp, copym, dst->sin_family, 0); 2262#endif 2263 } 2264} 2265