ip_output.c revision 138404
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 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 30 * $FreeBSD: head/sys/netinet/ip_output.c 138404 2004-12-05 21:38:33Z rwatson $ 31 */ 32 33#include "opt_ipfw.h" 34#include "opt_ipsec.h" 35#include "opt_mac.h" 36#include "opt_mbuf_stress_test.h" 37 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/kernel.h> 41#include <sys/mac.h> 42#include <sys/malloc.h> 43#include <sys/mbuf.h> 44#include <sys/protosw.h> 45#include <sys/socket.h> 46#include <sys/socketvar.h> 47#include <sys/sysctl.h> 48 49#include <net/if.h> 50#include <net/netisr.h> 51#include <net/pfil.h> 52#include <net/route.h> 53 54#include <netinet/in.h> 55#include <netinet/in_systm.h> 56#include <netinet/ip.h> 57#include <netinet/in_pcb.h> 58#include <netinet/in_var.h> 59#include <netinet/ip_var.h> 60 61 62#include <machine/in_cksum.h> 63 64static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 65 66#ifdef IPSEC 67#include <netinet6/ipsec.h> 68#include <netkey/key.h> 69#ifdef IPSEC_DEBUG 70#include <netkey/key_debug.h> 71#else 72#define KEYDEBUG(lev,arg) 73#endif 74#endif /*IPSEC*/ 75 76#ifdef FAST_IPSEC 77#include <netipsec/ipsec.h> 78#include <netipsec/xform.h> 79#include <netipsec/key.h> 80#endif /*FAST_IPSEC*/ 81 82#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 83 x, (ntohl(a.s_addr)>>24)&0xFF,\ 84 (ntohl(a.s_addr)>>16)&0xFF,\ 85 (ntohl(a.s_addr)>>8)&0xFF,\ 86 (ntohl(a.s_addr))&0xFF, y); 87 88u_short ip_id; 89 90#ifdef MBUF_STRESS_TEST 91int mbuf_frag_size = 0; 92SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 93 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 94#endif 95 96static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 97static struct ifnet *ip_multicast_if(struct in_addr *, int *); 98static void ip_mloopback 99 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 100static int ip_getmoptions 101 (struct sockopt *, struct ip_moptions *); 102static int ip_pcbopts(struct inpcb *, int, struct mbuf *); 103static int ip_setmoptions(struct inpcb *, struct sockopt *); 104 105int ip_optcopy(struct ip *, struct ip *); 106 107 108extern struct protosw inetsw[]; 109 110/* 111 * IP output. The packet in mbuf chain m contains a skeletal IP 112 * header (with len, off, ttl, proto, tos, src, dst). 113 * The mbuf chain containing the packet will be freed. 114 * The mbuf opt, if present, will not be freed. 115 * In the IP forwarding case, the packet will arrive with options already 116 * inserted, so must have a NULL opt pointer. 117 */ 118int 119ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, 120 int flags, struct ip_moptions *imo, struct inpcb *inp) 121{ 122 struct ip *ip; 123 struct ifnet *ifp = NULL; /* keep compiler happy */ 124 struct mbuf *m0; 125 int hlen = sizeof (struct ip); 126 int len, error = 0; 127 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 128 struct in_ifaddr *ia = NULL; 129 int isbroadcast, sw_csum; 130 struct route iproute; 131 struct in_addr odst; 132#ifdef IPFIREWALL_FORWARD 133 struct m_tag *fwd_tag = NULL; 134#endif 135#ifdef IPSEC 136 struct secpolicy *sp = NULL; 137#endif 138#ifdef FAST_IPSEC 139 struct secpolicy *sp = NULL; 140 struct tdb_ident *tdbi; 141 struct m_tag *mtag; 142 int s; 143#endif /* FAST_IPSEC */ 144 145 M_ASSERTPKTHDR(m); 146 147 if (ro == NULL) { 148 ro = &iproute; 149 bzero(ro, sizeof (*ro)); 150 } 151 152 if (inp != NULL) 153 INP_LOCK_ASSERT(inp); 154 155 if (opt) { 156 len = 0; 157 m = ip_insertoptions(m, opt, &len); 158 if (len != 0) 159 hlen = len; 160 } 161 ip = mtod(m, struct ip *); 162 163 /* 164 * Fill in IP header. If we are not allowing fragmentation, 165 * then the ip_id field is meaningless, but we don't set it 166 * to zero. Doing so causes various problems when devices along 167 * the path (routers, load balancers, firewalls, etc.) illegally 168 * disable DF on our packet. Note that a 16-bit counter 169 * will wrap around in less than 10 seconds at 100 Mbit/s on a 170 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 171 * for Counting NATted Hosts", Proc. IMW'02, available at 172 * <http://www.research.att.com/~smb/papers/fnat.pdf>. 173 */ 174 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 175 ip->ip_v = IPVERSION; 176 ip->ip_hl = hlen >> 2; 177 ip->ip_id = ip_newid(); 178 ipstat.ips_localout++; 179 } else { 180 hlen = ip->ip_hl << 2; 181 } 182 183 dst = (struct sockaddr_in *)&ro->ro_dst; 184again: 185 /* 186 * If there is a cached route, 187 * check that it is to the same destination 188 * and is still up. If not, free it and try again. 189 * The address family should also be checked in case of sharing the 190 * cache with IPv6. 191 */ 192 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 193 dst->sin_family != AF_INET || 194 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 195 RTFREE(ro->ro_rt); 196 ro->ro_rt = (struct rtentry *)0; 197 } 198#ifdef IPFIREWALL_FORWARD 199 if (ro->ro_rt == NULL && fwd_tag == NULL) { 200#else 201 if (ro->ro_rt == NULL) { 202#endif 203 bzero(dst, sizeof(*dst)); 204 dst->sin_family = AF_INET; 205 dst->sin_len = sizeof(*dst); 206 dst->sin_addr = ip->ip_dst; 207 } 208 /* 209 * If routing to interface only, 210 * short circuit routing lookup. 211 */ 212 if (flags & IP_ROUTETOIF) { 213 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 214 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) { 215 ipstat.ips_noroute++; 216 error = ENETUNREACH; 217 goto bad; 218 } 219 ifp = ia->ia_ifp; 220 ip->ip_ttl = 1; 221 isbroadcast = in_broadcast(dst->sin_addr, ifp); 222 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 223 imo != NULL && imo->imo_multicast_ifp != NULL) { 224 /* 225 * Bypass the normal routing lookup for multicast 226 * packets if the interface is specified. 227 */ 228 ifp = imo->imo_multicast_ifp; 229 IFP_TO_IA(ifp, ia); 230 isbroadcast = 0; /* fool gcc */ 231 } else { 232 /* 233 * We want to do any cloning requested by the link layer, 234 * as this is probably required in all cases for correct 235 * operation (as it is for ARP). 236 */ 237 if (ro->ro_rt == NULL) 238 rtalloc_ign(ro, 0); 239 if (ro->ro_rt == NULL) { 240 ipstat.ips_noroute++; 241 error = EHOSTUNREACH; 242 goto bad; 243 } 244 ia = ifatoia(ro->ro_rt->rt_ifa); 245 ifp = ro->ro_rt->rt_ifp; 246 ro->ro_rt->rt_rmx.rmx_pksent++; 247 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 248 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 249 if (ro->ro_rt->rt_flags & RTF_HOST) 250 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 251 else 252 isbroadcast = in_broadcast(dst->sin_addr, ifp); 253 } 254 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 255 struct in_multi *inm; 256 257 m->m_flags |= M_MCAST; 258 /* 259 * IP destination address is multicast. Make sure "dst" 260 * still points to the address in "ro". (It may have been 261 * changed to point to a gateway address, above.) 262 */ 263 dst = (struct sockaddr_in *)&ro->ro_dst; 264 /* 265 * See if the caller provided any multicast options 266 */ 267 if (imo != NULL) { 268 ip->ip_ttl = imo->imo_multicast_ttl; 269 if (imo->imo_multicast_vif != -1) 270 ip->ip_src.s_addr = 271 ip_mcast_src ? 272 ip_mcast_src(imo->imo_multicast_vif) : 273 INADDR_ANY; 274 } else 275 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 276 /* 277 * Confirm that the outgoing interface supports multicast. 278 */ 279 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 280 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 281 ipstat.ips_noroute++; 282 error = ENETUNREACH; 283 goto bad; 284 } 285 } 286 /* 287 * If source address not specified yet, use address 288 * of outgoing interface. 289 */ 290 if (ip->ip_src.s_addr == INADDR_ANY) { 291 /* Interface may have no addresses. */ 292 if (ia != NULL) 293 ip->ip_src = IA_SIN(ia)->sin_addr; 294 } 295 296 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 297 if (inm != NULL && 298 (imo == NULL || imo->imo_multicast_loop)) { 299 /* 300 * If we belong to the destination multicast group 301 * on the outgoing interface, and the caller did not 302 * forbid loopback, loop back a copy. 303 */ 304 ip_mloopback(ifp, m, dst, hlen); 305 } 306 else { 307 /* 308 * If we are acting as a multicast router, perform 309 * multicast forwarding as if the packet had just 310 * arrived on the interface to which we are about 311 * to send. The multicast forwarding function 312 * recursively calls this function, using the 313 * IP_FORWARDING flag to prevent infinite recursion. 314 * 315 * Multicasts that are looped back by ip_mloopback(), 316 * above, will be forwarded by the ip_input() routine, 317 * if necessary. 318 */ 319 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 320 /* 321 * If rsvp daemon is not running, do not 322 * set ip_moptions. This ensures that the packet 323 * is multicast and not just sent down one link 324 * as prescribed by rsvpd. 325 */ 326 if (!rsvp_on) 327 imo = NULL; 328 if (ip_mforward && 329 ip_mforward(ip, ifp, m, imo) != 0) { 330 m_freem(m); 331 goto done; 332 } 333 } 334 } 335 336 /* 337 * Multicasts with a time-to-live of zero may be looped- 338 * back, above, but must not be transmitted on a network. 339 * Also, multicasts addressed to the loopback interface 340 * are not sent -- the above call to ip_mloopback() will 341 * loop back a copy if this host actually belongs to the 342 * destination group on the loopback interface. 343 */ 344 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 345 m_freem(m); 346 goto done; 347 } 348 349 goto sendit; 350 } 351#ifndef notdef 352 /* 353 * If the source address is not specified yet, use the address 354 * of the outoing interface. 355 */ 356 if (ip->ip_src.s_addr == INADDR_ANY) { 357 /* Interface may have no addresses. */ 358 if (ia != NULL) { 359 ip->ip_src = IA_SIN(ia)->sin_addr; 360 } 361 } 362#endif /* notdef */ 363 /* 364 * Verify that we have any chance at all of being able to queue the 365 * packet or packet fragments, unless ALTQ is enabled on the given 366 * interface in which case packetdrop should be done by queueing. 367 */ 368#ifdef ALTQ 369 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) && 370 ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 371 ifp->if_snd.ifq_maxlen)) 372#else 373 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 374 ifp->if_snd.ifq_maxlen) 375#endif /* ALTQ */ 376 { 377 error = ENOBUFS; 378 ipstat.ips_odropped++; 379 goto bad; 380 } 381 382 /* 383 * Look for broadcast address and 384 * verify user is allowed to send 385 * such a packet. 386 */ 387 if (isbroadcast) { 388 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 389 error = EADDRNOTAVAIL; 390 goto bad; 391 } 392 if ((flags & IP_ALLOWBROADCAST) == 0) { 393 error = EACCES; 394 goto bad; 395 } 396 /* don't allow broadcast messages to be fragmented */ 397 if (ip->ip_len > ifp->if_mtu) { 398 error = EMSGSIZE; 399 goto bad; 400 } 401 if (flags & IP_SENDONES) 402 ip->ip_dst.s_addr = INADDR_BROADCAST; 403 m->m_flags |= M_BCAST; 404 } else { 405 m->m_flags &= ~M_BCAST; 406 } 407 408sendit: 409#ifdef IPSEC 410 /* get SP for this packet */ 411 if (inp == NULL) 412 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 413 flags, &error); 414 else 415 sp = ipsec4_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error); 416 417 if (sp == NULL) { 418 ipsecstat.out_inval++; 419 goto bad; 420 } 421 422 error = 0; 423 424 /* check policy */ 425 switch (sp->policy) { 426 case IPSEC_POLICY_DISCARD: 427 /* 428 * This packet is just discarded. 429 */ 430 ipsecstat.out_polvio++; 431 goto bad; 432 433 case IPSEC_POLICY_BYPASS: 434 case IPSEC_POLICY_NONE: 435 case IPSEC_POLICY_TCP: 436 /* no need to do IPsec. */ 437 goto skip_ipsec; 438 439 case IPSEC_POLICY_IPSEC: 440 if (sp->req == NULL) { 441 /* acquire a policy */ 442 error = key_spdacquire(sp); 443 goto bad; 444 } 445 break; 446 447 case IPSEC_POLICY_ENTRUST: 448 default: 449 printf("ip_output: Invalid policy found. %d\n", sp->policy); 450 } 451 { 452 struct ipsec_output_state state; 453 bzero(&state, sizeof(state)); 454 state.m = m; 455 if (flags & IP_ROUTETOIF) { 456 state.ro = &iproute; 457 bzero(&iproute, sizeof(iproute)); 458 } else 459 state.ro = ro; 460 state.dst = (struct sockaddr *)dst; 461 462 ip->ip_sum = 0; 463 464 /* 465 * XXX 466 * delayed checksums are not currently compatible with IPsec 467 */ 468 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 469 in_delayed_cksum(m); 470 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 471 } 472 473 ip->ip_len = htons(ip->ip_len); 474 ip->ip_off = htons(ip->ip_off); 475 476 error = ipsec4_output(&state, sp, flags); 477 478 m = state.m; 479 if (flags & IP_ROUTETOIF) { 480 /* 481 * if we have tunnel mode SA, we may need to ignore 482 * IP_ROUTETOIF. 483 */ 484 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 485 flags &= ~IP_ROUTETOIF; 486 ro = state.ro; 487 } 488 } else 489 ro = state.ro; 490 dst = (struct sockaddr_in *)state.dst; 491 if (error) { 492 /* mbuf is already reclaimed in ipsec4_output. */ 493 m = NULL; 494 switch (error) { 495 case EHOSTUNREACH: 496 case ENETUNREACH: 497 case EMSGSIZE: 498 case ENOBUFS: 499 case ENOMEM: 500 break; 501 default: 502 printf("ip4_output (ipsec): error code %d\n", error); 503 /*fall through*/ 504 case ENOENT: 505 /* don't show these error codes to the user */ 506 error = 0; 507 break; 508 } 509 goto bad; 510 } 511 512 /* be sure to update variables that are affected by ipsec4_output() */ 513 ip = mtod(m, struct ip *); 514 hlen = ip->ip_hl << 2; 515 if (ro->ro_rt == NULL) { 516 if ((flags & IP_ROUTETOIF) == 0) { 517 printf("ip_output: " 518 "can't update route after IPsec processing\n"); 519 error = EHOSTUNREACH; /*XXX*/ 520 goto bad; 521 } 522 } else { 523 if (state.encap) { 524 ia = ifatoia(ro->ro_rt->rt_ifa); 525 ifp = ro->ro_rt->rt_ifp; 526 } 527 } 528 } 529 530 /* make it flipped, again. */ 531 ip->ip_len = ntohs(ip->ip_len); 532 ip->ip_off = ntohs(ip->ip_off); 533skip_ipsec: 534#endif /*IPSEC*/ 535#ifdef FAST_IPSEC 536 /* 537 * Check the security policy (SP) for the packet and, if 538 * required, do IPsec-related processing. There are two 539 * cases here; the first time a packet is sent through 540 * it will be untagged and handled by ipsec4_checkpolicy. 541 * If the packet is resubmitted to ip_output (e.g. after 542 * AH, ESP, etc. processing), there will be a tag to bypass 543 * the lookup and related policy checking. 544 */ 545 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 546 s = splnet(); 547 if (mtag != NULL) { 548 tdbi = (struct tdb_ident *)(mtag + 1); 549 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 550 if (sp == NULL) 551 error = -EINVAL; /* force silent drop */ 552 m_tag_delete(m, mtag); 553 } else { 554 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 555 &error, inp); 556 } 557 /* 558 * There are four return cases: 559 * sp != NULL apply IPsec policy 560 * sp == NULL, error == 0 no IPsec handling needed 561 * sp == NULL, error == -EINVAL discard packet w/o error 562 * sp == NULL, error != 0 discard packet, report error 563 */ 564 if (sp != NULL) { 565 /* Loop detection, check if ipsec processing already done */ 566 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 567 for (mtag = m_tag_first(m); mtag != NULL; 568 mtag = m_tag_next(m, mtag)) { 569 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 570 continue; 571 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 572 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 573 continue; 574 /* 575 * Check if policy has an SA associated with it. 576 * This can happen when an SP has yet to acquire 577 * an SA; e.g. on first reference. If it occurs, 578 * then we let ipsec4_process_packet do its thing. 579 */ 580 if (sp->req->sav == NULL) 581 break; 582 tdbi = (struct tdb_ident *)(mtag + 1); 583 if (tdbi->spi == sp->req->sav->spi && 584 tdbi->proto == sp->req->sav->sah->saidx.proto && 585 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 586 sizeof (union sockaddr_union)) == 0) { 587 /* 588 * No IPsec processing is needed, free 589 * reference to SP. 590 * 591 * NB: null pointer to avoid free at 592 * done: below. 593 */ 594 KEY_FREESP(&sp), sp = NULL; 595 splx(s); 596 goto spd_done; 597 } 598 } 599 600 /* 601 * Do delayed checksums now because we send before 602 * this is done in the normal processing path. 603 */ 604 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 605 in_delayed_cksum(m); 606 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 607 } 608 609 ip->ip_len = htons(ip->ip_len); 610 ip->ip_off = htons(ip->ip_off); 611 612 /* NB: callee frees mbuf */ 613 error = ipsec4_process_packet(m, sp->req, flags, 0); 614 /* 615 * Preserve KAME behaviour: ENOENT can be returned 616 * when an SA acquire is in progress. Don't propagate 617 * this to user-level; it confuses applications. 618 * 619 * XXX this will go away when the SADB is redone. 620 */ 621 if (error == ENOENT) 622 error = 0; 623 splx(s); 624 goto done; 625 } else { 626 splx(s); 627 628 if (error != 0) { 629 /* 630 * Hack: -EINVAL is used to signal that a packet 631 * should be silently discarded. This is typically 632 * because we asked key management for an SA and 633 * it was delayed (e.g. kicked up to IKE). 634 */ 635 if (error == -EINVAL) 636 error = 0; 637 goto bad; 638 } else { 639 /* No IPsec processing for this packet. */ 640 } 641#ifdef notyet 642 /* 643 * If deferred crypto processing is needed, check that 644 * the interface supports it. 645 */ 646 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 647 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 648 /* notify IPsec to do its own crypto */ 649 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 650 error = EHOSTUNREACH; 651 goto bad; 652 } 653#endif 654 } 655spd_done: 656#endif /* FAST_IPSEC */ 657 658 /* Jump over all PFIL processing if hooks are not active. */ 659 if (inet_pfil_hook.ph_busy_count == -1) 660 goto passout; 661 662 /* Run through list of hooks for output packets. */ 663 odst.s_addr = ip->ip_dst.s_addr; 664 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 665 if (error != 0 || m == NULL) 666 goto done; 667 668 ip = mtod(m, struct ip *); 669 670 /* See if destination IP address was changed by packet filter. */ 671 if (odst.s_addr != ip->ip_dst.s_addr) { 672 m->m_flags |= M_SKIP_FIREWALL; 673 /* If destination is now ourself drop to ip_input(). */ 674 if (in_localip(ip->ip_dst)) { 675 m->m_flags |= M_FASTFWD_OURS; 676 if (m->m_pkthdr.rcvif == NULL) 677 m->m_pkthdr.rcvif = loif; 678 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 679 m->m_pkthdr.csum_flags |= 680 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 681 m->m_pkthdr.csum_data = 0xffff; 682 } 683 m->m_pkthdr.csum_flags |= 684 CSUM_IP_CHECKED | CSUM_IP_VALID; 685 686 error = netisr_queue(NETISR_IP, m); 687 goto done; 688 } else 689 goto again; /* Redo the routing table lookup. */ 690 } 691 692#ifdef IPFIREWALL_FORWARD 693 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 694 if (m->m_flags & M_FASTFWD_OURS) { 695 if (m->m_pkthdr.rcvif == NULL) 696 m->m_pkthdr.rcvif = loif; 697 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 698 m->m_pkthdr.csum_flags |= 699 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 700 m->m_pkthdr.csum_data = 0xffff; 701 } 702 m->m_pkthdr.csum_flags |= 703 CSUM_IP_CHECKED | CSUM_IP_VALID; 704 705 error = netisr_queue(NETISR_IP, m); 706 goto done; 707 } 708 /* Or forward to some other address? */ 709 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 710 if (fwd_tag) { 711 if (!in_localip(ip->ip_src) && !in_localaddr(ip->ip_dst)) { 712 dst = (struct sockaddr_in *)&ro->ro_dst; 713 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 714 m->m_flags |= M_SKIP_FIREWALL; 715 m_tag_delete(m, fwd_tag); 716 goto again; 717 } else { 718 m_tag_delete(m, fwd_tag); 719 /* Continue. */ 720 } 721 } 722#endif 723 724passout: 725 /* 127/8 must not appear on wire - RFC1122. */ 726 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 727 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 728 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 729 ipstat.ips_badaddr++; 730 error = EADDRNOTAVAIL; 731 goto bad; 732 } 733 } 734 735 m->m_pkthdr.csum_flags |= CSUM_IP; 736 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 737 if (sw_csum & CSUM_DELAY_DATA) { 738 in_delayed_cksum(m); 739 sw_csum &= ~CSUM_DELAY_DATA; 740 } 741 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 742 743 /* 744 * If small enough for interface, or the interface will take 745 * care of the fragmentation for us, can just send directly. 746 */ 747 if (ip->ip_len <= ifp->if_mtu || (ifp->if_hwassist & CSUM_FRAGMENT && 748 ((ip->ip_off & IP_DF) == 0))) { 749 ip->ip_len = htons(ip->ip_len); 750 ip->ip_off = htons(ip->ip_off); 751 ip->ip_sum = 0; 752 if (sw_csum & CSUM_DELAY_IP) 753 ip->ip_sum = in_cksum(m, hlen); 754 755 /* Record statistics for this interface address. */ 756 if (!(flags & IP_FORWARDING) && ia) { 757 ia->ia_ifa.if_opackets++; 758 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 759 } 760 761#ifdef IPSEC 762 /* clean ipsec history once it goes out of the node */ 763 ipsec_delaux(m); 764#endif 765 766#ifdef MBUF_STRESS_TEST 767 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 768 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 769#endif 770 error = (*ifp->if_output)(ifp, m, 771 (struct sockaddr *)dst, ro->ro_rt); 772 goto done; 773 } 774 775 if (ip->ip_off & IP_DF) { 776 error = EMSGSIZE; 777 /* 778 * This case can happen if the user changed the MTU 779 * of an interface after enabling IP on it. Because 780 * most netifs don't keep track of routes pointing to 781 * them, there is no way for one to update all its 782 * routes when the MTU is changed. 783 */ 784 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 785 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 786 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 787 } 788 ipstat.ips_cantfrag++; 789 goto bad; 790 } 791 792 /* 793 * Too large for interface; fragment if possible. If successful, 794 * on return, m will point to a list of packets to be sent. 795 */ 796 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 797 if (error) 798 goto bad; 799 for (; m; m = m0) { 800 m0 = m->m_nextpkt; 801 m->m_nextpkt = 0; 802#ifdef IPSEC 803 /* clean ipsec history once it goes out of the node */ 804 ipsec_delaux(m); 805#endif 806 if (error == 0) { 807 /* Record statistics for this interface address. */ 808 if (ia != NULL) { 809 ia->ia_ifa.if_opackets++; 810 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 811 } 812 813 error = (*ifp->if_output)(ifp, m, 814 (struct sockaddr *)dst, ro->ro_rt); 815 } else 816 m_freem(m); 817 } 818 819 if (error == 0) 820 ipstat.ips_fragmented++; 821 822done: 823 if (ro == &iproute && ro->ro_rt) { 824 RTFREE(ro->ro_rt); 825 } 826#ifdef IPSEC 827 if (sp != NULL) { 828 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 829 printf("DP ip_output call free SP:%p\n", sp)); 830 key_freesp(sp); 831 } 832#endif 833#ifdef FAST_IPSEC 834 if (sp != NULL) 835 KEY_FREESP(&sp); 836#endif 837 return (error); 838bad: 839 m_freem(m); 840 goto done; 841} 842 843/* 844 * Create a chain of fragments which fit the given mtu. m_frag points to the 845 * mbuf to be fragmented; on return it points to the chain with the fragments. 846 * Return 0 if no error. If error, m_frag may contain a partially built 847 * chain of fragments that should be freed by the caller. 848 * 849 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 850 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 851 */ 852int 853ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 854 u_long if_hwassist_flags, int sw_csum) 855{ 856 int error = 0; 857 int hlen = ip->ip_hl << 2; 858 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 859 int off; 860 struct mbuf *m0 = *m_frag; /* the original packet */ 861 int firstlen; 862 struct mbuf **mnext; 863 int nfrags; 864 865 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 866 ipstat.ips_cantfrag++; 867 return EMSGSIZE; 868 } 869 870 /* 871 * Must be able to put at least 8 bytes per fragment. 872 */ 873 if (len < 8) 874 return EMSGSIZE; 875 876 /* 877 * If the interface will not calculate checksums on 878 * fragmented packets, then do it here. 879 */ 880 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 881 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 882 in_delayed_cksum(m0); 883 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 884 } 885 886 if (len > PAGE_SIZE) { 887 /* 888 * Fragment large datagrams such that each segment 889 * contains a multiple of PAGE_SIZE amount of data, 890 * plus headers. This enables a receiver to perform 891 * page-flipping zero-copy optimizations. 892 * 893 * XXX When does this help given that sender and receiver 894 * could have different page sizes, and also mtu could 895 * be less than the receiver's page size ? 896 */ 897 int newlen; 898 struct mbuf *m; 899 900 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 901 off += m->m_len; 902 903 /* 904 * firstlen (off - hlen) must be aligned on an 905 * 8-byte boundary 906 */ 907 if (off < hlen) 908 goto smart_frag_failure; 909 off = ((off - hlen) & ~7) + hlen; 910 newlen = (~PAGE_MASK) & mtu; 911 if ((newlen + sizeof (struct ip)) > mtu) { 912 /* we failed, go back the default */ 913smart_frag_failure: 914 newlen = len; 915 off = hlen + len; 916 } 917 len = newlen; 918 919 } else { 920 off = hlen + len; 921 } 922 923 firstlen = off - hlen; 924 mnext = &m0->m_nextpkt; /* pointer to next packet */ 925 926 /* 927 * Loop through length of segment after first fragment, 928 * make new header and copy data of each part and link onto chain. 929 * Here, m0 is the original packet, m is the fragment being created. 930 * The fragments are linked off the m_nextpkt of the original 931 * packet, which after processing serves as the first fragment. 932 */ 933 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 934 struct ip *mhip; /* ip header on the fragment */ 935 struct mbuf *m; 936 int mhlen = sizeof (struct ip); 937 938 MGETHDR(m, M_DONTWAIT, MT_HEADER); 939 if (m == NULL) { 940 error = ENOBUFS; 941 ipstat.ips_odropped++; 942 goto done; 943 } 944 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 945 /* 946 * In the first mbuf, leave room for the link header, then 947 * copy the original IP header including options. The payload 948 * goes into an additional mbuf chain returned by m_copy(). 949 */ 950 m->m_data += max_linkhdr; 951 mhip = mtod(m, struct ip *); 952 *mhip = *ip; 953 if (hlen > sizeof (struct ip)) { 954 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 955 mhip->ip_v = IPVERSION; 956 mhip->ip_hl = mhlen >> 2; 957 } 958 m->m_len = mhlen; 959 /* XXX do we need to add ip->ip_off below ? */ 960 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 961 if (off + len >= ip->ip_len) { /* last fragment */ 962 len = ip->ip_len - off; 963 m->m_flags |= M_LASTFRAG; 964 } else 965 mhip->ip_off |= IP_MF; 966 mhip->ip_len = htons((u_short)(len + mhlen)); 967 m->m_next = m_copy(m0, off, len); 968 if (m->m_next == NULL) { /* copy failed */ 969 m_free(m); 970 error = ENOBUFS; /* ??? */ 971 ipstat.ips_odropped++; 972 goto done; 973 } 974 m->m_pkthdr.len = mhlen + len; 975 m->m_pkthdr.rcvif = (struct ifnet *)0; 976#ifdef MAC 977 mac_create_fragment(m0, m); 978#endif 979 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 980 mhip->ip_off = htons(mhip->ip_off); 981 mhip->ip_sum = 0; 982 if (sw_csum & CSUM_DELAY_IP) 983 mhip->ip_sum = in_cksum(m, mhlen); 984 *mnext = m; 985 mnext = &m->m_nextpkt; 986 } 987 ipstat.ips_ofragments += nfrags; 988 989 /* set first marker for fragment chain */ 990 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 991 m0->m_pkthdr.csum_data = nfrags; 992 993 /* 994 * Update first fragment by trimming what's been copied out 995 * and updating header. 996 */ 997 m_adj(m0, hlen + firstlen - ip->ip_len); 998 m0->m_pkthdr.len = hlen + firstlen; 999 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 1000 ip->ip_off |= IP_MF; 1001 ip->ip_off = htons(ip->ip_off); 1002 ip->ip_sum = 0; 1003 if (sw_csum & CSUM_DELAY_IP) 1004 ip->ip_sum = in_cksum(m0, hlen); 1005 1006done: 1007 *m_frag = m0; 1008 return error; 1009} 1010 1011void 1012in_delayed_cksum(struct mbuf *m) 1013{ 1014 struct ip *ip; 1015 u_short csum, offset; 1016 1017 ip = mtod(m, struct ip *); 1018 offset = ip->ip_hl << 2 ; 1019 csum = in_cksum_skip(m, ip->ip_len, offset); 1020 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1021 csum = 0xffff; 1022 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1023 1024 if (offset + sizeof(u_short) > m->m_len) { 1025 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1026 m->m_len, offset, ip->ip_p); 1027 /* 1028 * XXX 1029 * this shouldn't happen, but if it does, the 1030 * correct behavior may be to insert the checksum 1031 * in the existing chain instead of rearranging it. 1032 */ 1033 m = m_pullup(m, offset + sizeof(u_short)); 1034 } 1035 *(u_short *)(m->m_data + offset) = csum; 1036} 1037 1038/* 1039 * Insert IP options into preformed packet. 1040 * Adjust IP destination as required for IP source routing, 1041 * as indicated by a non-zero in_addr at the start of the options. 1042 * 1043 * XXX This routine assumes that the packet has no options in place. 1044 */ 1045static struct mbuf * 1046ip_insertoptions(m, opt, phlen) 1047 register struct mbuf *m; 1048 struct mbuf *opt; 1049 int *phlen; 1050{ 1051 register struct ipoption *p = mtod(opt, struct ipoption *); 1052 struct mbuf *n; 1053 register struct ip *ip = mtod(m, struct ip *); 1054 unsigned optlen; 1055 1056 optlen = opt->m_len - sizeof(p->ipopt_dst); 1057 if (optlen + ip->ip_len > IP_MAXPACKET) { 1058 *phlen = 0; 1059 return (m); /* XXX should fail */ 1060 } 1061 if (p->ipopt_dst.s_addr) 1062 ip->ip_dst = p->ipopt_dst; 1063 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1064 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1065 if (n == NULL) { 1066 *phlen = 0; 1067 return (m); 1068 } 1069 n->m_pkthdr.rcvif = (struct ifnet *)0; 1070#ifdef MAC 1071 mac_create_mbuf_from_mbuf(m, n); 1072#endif 1073 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1074 m->m_len -= sizeof(struct ip); 1075 m->m_data += sizeof(struct ip); 1076 n->m_next = m; 1077 m = n; 1078 m->m_len = optlen + sizeof(struct ip); 1079 m->m_data += max_linkhdr; 1080 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1081 } else { 1082 m->m_data -= optlen; 1083 m->m_len += optlen; 1084 m->m_pkthdr.len += optlen; 1085 bcopy(ip, mtod(m, void *), sizeof(struct ip)); 1086 } 1087 ip = mtod(m, struct ip *); 1088 bcopy(p->ipopt_list, ip + 1, optlen); 1089 *phlen = sizeof(struct ip) + optlen; 1090 ip->ip_v = IPVERSION; 1091 ip->ip_hl = *phlen >> 2; 1092 ip->ip_len += optlen; 1093 return (m); 1094} 1095 1096/* 1097 * Copy options from ip to jp, 1098 * omitting those not copied during fragmentation. 1099 */ 1100int 1101ip_optcopy(ip, jp) 1102 struct ip *ip, *jp; 1103{ 1104 register u_char *cp, *dp; 1105 int opt, optlen, cnt; 1106 1107 cp = (u_char *)(ip + 1); 1108 dp = (u_char *)(jp + 1); 1109 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1110 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1111 opt = cp[0]; 1112 if (opt == IPOPT_EOL) 1113 break; 1114 if (opt == IPOPT_NOP) { 1115 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1116 *dp++ = IPOPT_NOP; 1117 optlen = 1; 1118 continue; 1119 } 1120 1121 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1122 ("ip_optcopy: malformed ipv4 option")); 1123 optlen = cp[IPOPT_OLEN]; 1124 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1125 ("ip_optcopy: malformed ipv4 option")); 1126 1127 /* bogus lengths should have been caught by ip_dooptions */ 1128 if (optlen > cnt) 1129 optlen = cnt; 1130 if (IPOPT_COPIED(opt)) { 1131 bcopy(cp, dp, optlen); 1132 dp += optlen; 1133 } 1134 } 1135 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1136 *dp++ = IPOPT_EOL; 1137 return (optlen); 1138} 1139 1140/* 1141 * IP socket option processing. 1142 */ 1143int 1144ip_ctloutput(so, sopt) 1145 struct socket *so; 1146 struct sockopt *sopt; 1147{ 1148 struct inpcb *inp = sotoinpcb(so); 1149 int error, optval; 1150 1151 error = optval = 0; 1152 if (sopt->sopt_level != IPPROTO_IP) { 1153 return (EINVAL); 1154 } 1155 1156 switch (sopt->sopt_dir) { 1157 case SOPT_SET: 1158 switch (sopt->sopt_name) { 1159 case IP_OPTIONS: 1160#ifdef notyet 1161 case IP_RETOPTS: 1162#endif 1163 { 1164 struct mbuf *m; 1165 if (sopt->sopt_valsize > MLEN) { 1166 error = EMSGSIZE; 1167 break; 1168 } 1169 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1170 if (m == NULL) { 1171 error = ENOBUFS; 1172 break; 1173 } 1174 m->m_len = sopt->sopt_valsize; 1175 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1176 m->m_len); 1177 INP_LOCK(inp); 1178 error = ip_pcbopts(inp, sopt->sopt_name, m); 1179 INP_UNLOCK(inp); 1180 return (error); 1181 } 1182 1183 case IP_TOS: 1184 case IP_TTL: 1185 case IP_RECVOPTS: 1186 case IP_RECVRETOPTS: 1187 case IP_RECVDSTADDR: 1188 case IP_RECVTTL: 1189 case IP_RECVIF: 1190 case IP_FAITH: 1191 case IP_ONESBCAST: 1192 error = sooptcopyin(sopt, &optval, sizeof optval, 1193 sizeof optval); 1194 if (error) 1195 break; 1196 1197 switch (sopt->sopt_name) { 1198 case IP_TOS: 1199 inp->inp_ip_tos = optval; 1200 break; 1201 1202 case IP_TTL: 1203 inp->inp_ip_ttl = optval; 1204 break; 1205#define OPTSET(bit) do { \ 1206 INP_LOCK(inp); \ 1207 if (optval) \ 1208 inp->inp_flags |= bit; \ 1209 else \ 1210 inp->inp_flags &= ~bit; \ 1211 INP_UNLOCK(inp); \ 1212} while (0) 1213 1214 case IP_RECVOPTS: 1215 OPTSET(INP_RECVOPTS); 1216 break; 1217 1218 case IP_RECVRETOPTS: 1219 OPTSET(INP_RECVRETOPTS); 1220 break; 1221 1222 case IP_RECVDSTADDR: 1223 OPTSET(INP_RECVDSTADDR); 1224 break; 1225 1226 case IP_RECVTTL: 1227 OPTSET(INP_RECVTTL); 1228 break; 1229 1230 case IP_RECVIF: 1231 OPTSET(INP_RECVIF); 1232 break; 1233 1234 case IP_FAITH: 1235 OPTSET(INP_FAITH); 1236 break; 1237 1238 case IP_ONESBCAST: 1239 OPTSET(INP_ONESBCAST); 1240 break; 1241 } 1242 break; 1243#undef OPTSET 1244 1245 case IP_MULTICAST_IF: 1246 case IP_MULTICAST_VIF: 1247 case IP_MULTICAST_TTL: 1248 case IP_MULTICAST_LOOP: 1249 case IP_ADD_MEMBERSHIP: 1250 case IP_DROP_MEMBERSHIP: 1251 error = ip_setmoptions(inp, sopt); 1252 break; 1253 1254 case IP_PORTRANGE: 1255 error = sooptcopyin(sopt, &optval, sizeof optval, 1256 sizeof optval); 1257 if (error) 1258 break; 1259 1260 INP_LOCK(inp); 1261 switch (optval) { 1262 case IP_PORTRANGE_DEFAULT: 1263 inp->inp_flags &= ~(INP_LOWPORT); 1264 inp->inp_flags &= ~(INP_HIGHPORT); 1265 break; 1266 1267 case IP_PORTRANGE_HIGH: 1268 inp->inp_flags &= ~(INP_LOWPORT); 1269 inp->inp_flags |= INP_HIGHPORT; 1270 break; 1271 1272 case IP_PORTRANGE_LOW: 1273 inp->inp_flags &= ~(INP_HIGHPORT); 1274 inp->inp_flags |= INP_LOWPORT; 1275 break; 1276 1277 default: 1278 error = EINVAL; 1279 break; 1280 } 1281 INP_UNLOCK(inp); 1282 break; 1283 1284#if defined(IPSEC) || defined(FAST_IPSEC) 1285 case IP_IPSEC_POLICY: 1286 { 1287 caddr_t req; 1288 size_t len = 0; 1289 int priv; 1290 struct mbuf *m; 1291 int optname; 1292 1293 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1294 break; 1295 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1296 break; 1297 priv = (sopt->sopt_td != NULL && 1298 suser(sopt->sopt_td) != 0) ? 0 : 1; 1299 req = mtod(m, caddr_t); 1300 len = m->m_len; 1301 optname = sopt->sopt_name; 1302 error = ipsec4_set_policy(inp, optname, req, len, priv); 1303 m_freem(m); 1304 break; 1305 } 1306#endif /*IPSEC*/ 1307 1308 default: 1309 error = ENOPROTOOPT; 1310 break; 1311 } 1312 break; 1313 1314 case SOPT_GET: 1315 switch (sopt->sopt_name) { 1316 case IP_OPTIONS: 1317 case IP_RETOPTS: 1318 if (inp->inp_options) 1319 error = sooptcopyout(sopt, 1320 mtod(inp->inp_options, 1321 char *), 1322 inp->inp_options->m_len); 1323 else 1324 sopt->sopt_valsize = 0; 1325 break; 1326 1327 case IP_TOS: 1328 case IP_TTL: 1329 case IP_RECVOPTS: 1330 case IP_RECVRETOPTS: 1331 case IP_RECVDSTADDR: 1332 case IP_RECVTTL: 1333 case IP_RECVIF: 1334 case IP_PORTRANGE: 1335 case IP_FAITH: 1336 case IP_ONESBCAST: 1337 switch (sopt->sopt_name) { 1338 1339 case IP_TOS: 1340 optval = inp->inp_ip_tos; 1341 break; 1342 1343 case IP_TTL: 1344 optval = inp->inp_ip_ttl; 1345 break; 1346 1347#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1348 1349 case IP_RECVOPTS: 1350 optval = OPTBIT(INP_RECVOPTS); 1351 break; 1352 1353 case IP_RECVRETOPTS: 1354 optval = OPTBIT(INP_RECVRETOPTS); 1355 break; 1356 1357 case IP_RECVDSTADDR: 1358 optval = OPTBIT(INP_RECVDSTADDR); 1359 break; 1360 1361 case IP_RECVTTL: 1362 optval = OPTBIT(INP_RECVTTL); 1363 break; 1364 1365 case IP_RECVIF: 1366 optval = OPTBIT(INP_RECVIF); 1367 break; 1368 1369 case IP_PORTRANGE: 1370 if (inp->inp_flags & INP_HIGHPORT) 1371 optval = IP_PORTRANGE_HIGH; 1372 else if (inp->inp_flags & INP_LOWPORT) 1373 optval = IP_PORTRANGE_LOW; 1374 else 1375 optval = 0; 1376 break; 1377 1378 case IP_FAITH: 1379 optval = OPTBIT(INP_FAITH); 1380 break; 1381 1382 case IP_ONESBCAST: 1383 optval = OPTBIT(INP_ONESBCAST); 1384 break; 1385 } 1386 error = sooptcopyout(sopt, &optval, sizeof optval); 1387 break; 1388 1389 case IP_MULTICAST_IF: 1390 case IP_MULTICAST_VIF: 1391 case IP_MULTICAST_TTL: 1392 case IP_MULTICAST_LOOP: 1393 case IP_ADD_MEMBERSHIP: 1394 case IP_DROP_MEMBERSHIP: 1395 error = ip_getmoptions(sopt, inp->inp_moptions); 1396 break; 1397 1398#if defined(IPSEC) || defined(FAST_IPSEC) 1399 case IP_IPSEC_POLICY: 1400 { 1401 struct mbuf *m = NULL; 1402 caddr_t req = NULL; 1403 size_t len = 0; 1404 1405 if (m != 0) { 1406 req = mtod(m, caddr_t); 1407 len = m->m_len; 1408 } 1409 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1410 if (error == 0) 1411 error = soopt_mcopyout(sopt, m); /* XXX */ 1412 if (error == 0) 1413 m_freem(m); 1414 break; 1415 } 1416#endif /*IPSEC*/ 1417 1418 default: 1419 error = ENOPROTOOPT; 1420 break; 1421 } 1422 break; 1423 } 1424 return (error); 1425} 1426 1427/* 1428 * Set up IP options in pcb for insertion in output packets. 1429 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1430 * with destination address if source routed. 1431 */ 1432static int 1433ip_pcbopts(struct inpcb *inp, int optname, struct mbuf *m) 1434{ 1435 register int cnt, optlen; 1436 register u_char *cp; 1437 struct mbuf **pcbopt; 1438 u_char opt; 1439 1440 INP_LOCK_ASSERT(inp); 1441 1442 pcbopt = &inp->inp_options; 1443 1444 /* turn off any old options */ 1445 if (*pcbopt) 1446 (void)m_free(*pcbopt); 1447 *pcbopt = 0; 1448 if (m == NULL || m->m_len == 0) { 1449 /* 1450 * Only turning off any previous options. 1451 */ 1452 if (m != NULL) 1453 (void)m_free(m); 1454 return (0); 1455 } 1456 1457 if (m->m_len % sizeof(int32_t)) 1458 goto bad; 1459 /* 1460 * IP first-hop destination address will be stored before 1461 * actual options; move other options back 1462 * and clear it when none present. 1463 */ 1464 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1465 goto bad; 1466 cnt = m->m_len; 1467 m->m_len += sizeof(struct in_addr); 1468 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1469 bcopy(mtod(m, void *), cp, (unsigned)cnt); 1470 bzero(mtod(m, void *), sizeof(struct in_addr)); 1471 1472 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1473 opt = cp[IPOPT_OPTVAL]; 1474 if (opt == IPOPT_EOL) 1475 break; 1476 if (opt == IPOPT_NOP) 1477 optlen = 1; 1478 else { 1479 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1480 goto bad; 1481 optlen = cp[IPOPT_OLEN]; 1482 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1483 goto bad; 1484 } 1485 switch (opt) { 1486 1487 default: 1488 break; 1489 1490 case IPOPT_LSRR: 1491 case IPOPT_SSRR: 1492 /* 1493 * user process specifies route as: 1494 * ->A->B->C->D 1495 * D must be our final destination (but we can't 1496 * check that since we may not have connected yet). 1497 * A is first hop destination, which doesn't appear in 1498 * actual IP option, but is stored before the options. 1499 */ 1500 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1501 goto bad; 1502 m->m_len -= sizeof(struct in_addr); 1503 cnt -= sizeof(struct in_addr); 1504 optlen -= sizeof(struct in_addr); 1505 cp[IPOPT_OLEN] = optlen; 1506 /* 1507 * Move first hop before start of options. 1508 */ 1509 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1510 sizeof(struct in_addr)); 1511 /* 1512 * Then copy rest of options back 1513 * to close up the deleted entry. 1514 */ 1515 bcopy((&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr)), 1516 &cp[IPOPT_OFFSET+1], 1517 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1518 break; 1519 } 1520 } 1521 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1522 goto bad; 1523 *pcbopt = m; 1524 return (0); 1525 1526bad: 1527 (void)m_free(m); 1528 return (EINVAL); 1529} 1530 1531/* 1532 * XXX 1533 * The whole multicast option thing needs to be re-thought. 1534 * Several of these options are equally applicable to non-multicast 1535 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1536 * standard option (IP_TTL). 1537 */ 1538 1539/* 1540 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1541 */ 1542static struct ifnet * 1543ip_multicast_if(a, ifindexp) 1544 struct in_addr *a; 1545 int *ifindexp; 1546{ 1547 int ifindex; 1548 struct ifnet *ifp; 1549 1550 if (ifindexp) 1551 *ifindexp = 0; 1552 if (ntohl(a->s_addr) >> 24 == 0) { 1553 ifindex = ntohl(a->s_addr) & 0xffffff; 1554 if (ifindex < 0 || if_index < ifindex) 1555 return NULL; 1556 ifp = ifnet_byindex(ifindex); 1557 if (ifindexp) 1558 *ifindexp = ifindex; 1559 } else { 1560 INADDR_TO_IFP(*a, ifp); 1561 } 1562 return ifp; 1563} 1564 1565/* 1566 * Set the IP multicast options in response to user setsockopt(). 1567 */ 1568static int 1569ip_setmoptions(struct inpcb *inp, struct sockopt *sopt) 1570{ 1571 int error = 0; 1572 int i; 1573 struct in_addr addr; 1574 struct ip_mreq mreq; 1575 struct ifnet *ifp; 1576 struct ip_moptions *imo; 1577 struct route ro; 1578 struct sockaddr_in *dst; 1579 int ifindex; 1580 int s; 1581 1582 imo = inp->inp_moptions; 1583 if (imo == NULL) { 1584 /* 1585 * No multicast option buffer attached to the pcb; 1586 * allocate one and initialize to default values. 1587 */ 1588 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1589 M_WAITOK); 1590 1591 if (imo == NULL) 1592 return (ENOBUFS); 1593 inp->inp_moptions = imo; 1594 imo->imo_multicast_ifp = NULL; 1595 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1596 imo->imo_multicast_vif = -1; 1597 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1598 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1599 imo->imo_num_memberships = 0; 1600 } 1601 1602 switch (sopt->sopt_name) { 1603 /* store an index number for the vif you wanna use in the send */ 1604 case IP_MULTICAST_VIF: 1605 if (legal_vif_num == 0) { 1606 error = EOPNOTSUPP; 1607 break; 1608 } 1609 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1610 if (error) 1611 break; 1612 if (!legal_vif_num(i) && (i != -1)) { 1613 error = EINVAL; 1614 break; 1615 } 1616 imo->imo_multicast_vif = i; 1617 break; 1618 1619 case IP_MULTICAST_IF: 1620 /* 1621 * Select the interface for outgoing multicast packets. 1622 */ 1623 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1624 if (error) 1625 break; 1626 /* 1627 * INADDR_ANY is used to remove a previous selection. 1628 * When no interface is selected, a default one is 1629 * chosen every time a multicast packet is sent. 1630 */ 1631 if (addr.s_addr == INADDR_ANY) { 1632 imo->imo_multicast_ifp = NULL; 1633 break; 1634 } 1635 /* 1636 * The selected interface is identified by its local 1637 * IP address. Find the interface and confirm that 1638 * it supports multicasting. 1639 */ 1640 s = splimp(); 1641 ifp = ip_multicast_if(&addr, &ifindex); 1642 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1643 splx(s); 1644 error = EADDRNOTAVAIL; 1645 break; 1646 } 1647 imo->imo_multicast_ifp = ifp; 1648 if (ifindex) 1649 imo->imo_multicast_addr = addr; 1650 else 1651 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1652 splx(s); 1653 break; 1654 1655 case IP_MULTICAST_TTL: 1656 /* 1657 * Set the IP time-to-live for outgoing multicast packets. 1658 * The original multicast API required a char argument, 1659 * which is inconsistent with the rest of the socket API. 1660 * We allow either a char or an int. 1661 */ 1662 if (sopt->sopt_valsize == 1) { 1663 u_char ttl; 1664 error = sooptcopyin(sopt, &ttl, 1, 1); 1665 if (error) 1666 break; 1667 imo->imo_multicast_ttl = ttl; 1668 } else { 1669 u_int ttl; 1670 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1671 sizeof ttl); 1672 if (error) 1673 break; 1674 if (ttl > 255) 1675 error = EINVAL; 1676 else 1677 imo->imo_multicast_ttl = ttl; 1678 } 1679 break; 1680 1681 case IP_MULTICAST_LOOP: 1682 /* 1683 * Set the loopback flag for outgoing multicast packets. 1684 * Must be zero or one. The original multicast API required a 1685 * char argument, which is inconsistent with the rest 1686 * of the socket API. We allow either a char or an int. 1687 */ 1688 if (sopt->sopt_valsize == 1) { 1689 u_char loop; 1690 error = sooptcopyin(sopt, &loop, 1, 1); 1691 if (error) 1692 break; 1693 imo->imo_multicast_loop = !!loop; 1694 } else { 1695 u_int loop; 1696 error = sooptcopyin(sopt, &loop, sizeof loop, 1697 sizeof loop); 1698 if (error) 1699 break; 1700 imo->imo_multicast_loop = !!loop; 1701 } 1702 break; 1703 1704 case IP_ADD_MEMBERSHIP: 1705 /* 1706 * Add a multicast group membership. 1707 * Group must be a valid IP multicast address. 1708 */ 1709 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1710 if (error) 1711 break; 1712 1713 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1714 error = EINVAL; 1715 break; 1716 } 1717 s = splimp(); 1718 /* 1719 * If no interface address was provided, use the interface of 1720 * the route to the given multicast address. 1721 */ 1722 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1723 bzero((caddr_t)&ro, sizeof(ro)); 1724 dst = (struct sockaddr_in *)&ro.ro_dst; 1725 dst->sin_len = sizeof(*dst); 1726 dst->sin_family = AF_INET; 1727 dst->sin_addr = mreq.imr_multiaddr; 1728 rtalloc_ign(&ro, RTF_CLONING); 1729 if (ro.ro_rt == NULL) { 1730 error = EADDRNOTAVAIL; 1731 splx(s); 1732 break; 1733 } 1734 ifp = ro.ro_rt->rt_ifp; 1735 RTFREE(ro.ro_rt); 1736 } 1737 else { 1738 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1739 } 1740 1741 /* 1742 * See if we found an interface, and confirm that it 1743 * supports multicast. 1744 */ 1745 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1746 error = EADDRNOTAVAIL; 1747 splx(s); 1748 break; 1749 } 1750 /* 1751 * See if the membership already exists or if all the 1752 * membership slots are full. 1753 */ 1754 for (i = 0; i < imo->imo_num_memberships; ++i) { 1755 if (imo->imo_membership[i]->inm_ifp == ifp && 1756 imo->imo_membership[i]->inm_addr.s_addr 1757 == mreq.imr_multiaddr.s_addr) 1758 break; 1759 } 1760 if (i < imo->imo_num_memberships) { 1761 error = EADDRINUSE; 1762 splx(s); 1763 break; 1764 } 1765 if (i == IP_MAX_MEMBERSHIPS) { 1766 error = ETOOMANYREFS; 1767 splx(s); 1768 break; 1769 } 1770 /* 1771 * Everything looks good; add a new record to the multicast 1772 * address list for the given interface. 1773 */ 1774 if ((imo->imo_membership[i] = 1775 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1776 error = ENOBUFS; 1777 splx(s); 1778 break; 1779 } 1780 ++imo->imo_num_memberships; 1781 splx(s); 1782 break; 1783 1784 case IP_DROP_MEMBERSHIP: 1785 /* 1786 * Drop a multicast group membership. 1787 * Group must be a valid IP multicast address. 1788 */ 1789 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1790 if (error) 1791 break; 1792 1793 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1794 error = EINVAL; 1795 break; 1796 } 1797 1798 s = splimp(); 1799 /* 1800 * If an interface address was specified, get a pointer 1801 * to its ifnet structure. 1802 */ 1803 if (mreq.imr_interface.s_addr == INADDR_ANY) 1804 ifp = NULL; 1805 else { 1806 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1807 if (ifp == NULL) { 1808 error = EADDRNOTAVAIL; 1809 splx(s); 1810 break; 1811 } 1812 } 1813 /* 1814 * Find the membership in the membership array. 1815 */ 1816 for (i = 0; i < imo->imo_num_memberships; ++i) { 1817 if ((ifp == NULL || 1818 imo->imo_membership[i]->inm_ifp == ifp) && 1819 imo->imo_membership[i]->inm_addr.s_addr == 1820 mreq.imr_multiaddr.s_addr) 1821 break; 1822 } 1823 if (i == imo->imo_num_memberships) { 1824 error = EADDRNOTAVAIL; 1825 splx(s); 1826 break; 1827 } 1828 /* 1829 * Give up the multicast address record to which the 1830 * membership points. 1831 */ 1832 in_delmulti(imo->imo_membership[i]); 1833 /* 1834 * Remove the gap in the membership array. 1835 */ 1836 for (++i; i < imo->imo_num_memberships; ++i) 1837 imo->imo_membership[i-1] = imo->imo_membership[i]; 1838 --imo->imo_num_memberships; 1839 splx(s); 1840 break; 1841 1842 default: 1843 error = EOPNOTSUPP; 1844 break; 1845 } 1846 1847 /* 1848 * If all options have default values, no need to keep the mbuf. 1849 */ 1850 if (imo->imo_multicast_ifp == NULL && 1851 imo->imo_multicast_vif == -1 && 1852 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1853 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1854 imo->imo_num_memberships == 0) { 1855 free(inp->inp_moptions, M_IPMOPTS); 1856 inp->inp_moptions = NULL; 1857 } 1858 1859 return (error); 1860} 1861 1862/* 1863 * Return the IP multicast options in response to user getsockopt(). 1864 */ 1865static int 1866ip_getmoptions(sopt, imo) 1867 struct sockopt *sopt; 1868 register struct ip_moptions *imo; 1869{ 1870 struct in_addr addr; 1871 struct in_ifaddr *ia; 1872 int error, optval; 1873 u_char coptval; 1874 1875 error = 0; 1876 switch (sopt->sopt_name) { 1877 case IP_MULTICAST_VIF: 1878 if (imo != NULL) 1879 optval = imo->imo_multicast_vif; 1880 else 1881 optval = -1; 1882 error = sooptcopyout(sopt, &optval, sizeof optval); 1883 break; 1884 1885 case IP_MULTICAST_IF: 1886 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1887 addr.s_addr = INADDR_ANY; 1888 else if (imo->imo_multicast_addr.s_addr) { 1889 /* return the value user has set */ 1890 addr = imo->imo_multicast_addr; 1891 } else { 1892 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1893 addr.s_addr = (ia == NULL) ? INADDR_ANY 1894 : IA_SIN(ia)->sin_addr.s_addr; 1895 } 1896 error = sooptcopyout(sopt, &addr, sizeof addr); 1897 break; 1898 1899 case IP_MULTICAST_TTL: 1900 if (imo == 0) 1901 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1902 else 1903 optval = coptval = imo->imo_multicast_ttl; 1904 if (sopt->sopt_valsize == 1) 1905 error = sooptcopyout(sopt, &coptval, 1); 1906 else 1907 error = sooptcopyout(sopt, &optval, sizeof optval); 1908 break; 1909 1910 case IP_MULTICAST_LOOP: 1911 if (imo == 0) 1912 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1913 else 1914 optval = coptval = imo->imo_multicast_loop; 1915 if (sopt->sopt_valsize == 1) 1916 error = sooptcopyout(sopt, &coptval, 1); 1917 else 1918 error = sooptcopyout(sopt, &optval, sizeof optval); 1919 break; 1920 1921 default: 1922 error = ENOPROTOOPT; 1923 break; 1924 } 1925 return (error); 1926} 1927 1928/* 1929 * Discard the IP multicast options. 1930 */ 1931void 1932ip_freemoptions(imo) 1933 register struct ip_moptions *imo; 1934{ 1935 register int i; 1936 1937 if (imo != NULL) { 1938 for (i = 0; i < imo->imo_num_memberships; ++i) 1939 in_delmulti(imo->imo_membership[i]); 1940 free(imo, M_IPMOPTS); 1941 } 1942} 1943 1944/* 1945 * Routine called from ip_output() to loop back a copy of an IP multicast 1946 * packet to the input queue of a specified interface. Note that this 1947 * calls the output routine of the loopback "driver", but with an interface 1948 * pointer that might NOT be a loopback interface -- evil, but easier than 1949 * replicating that code here. 1950 */ 1951static void 1952ip_mloopback(ifp, m, dst, hlen) 1953 struct ifnet *ifp; 1954 register struct mbuf *m; 1955 register struct sockaddr_in *dst; 1956 int hlen; 1957{ 1958 register struct ip *ip; 1959 struct mbuf *copym; 1960 1961 copym = m_copy(m, 0, M_COPYALL); 1962 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1963 copym = m_pullup(copym, hlen); 1964 if (copym != NULL) { 1965 /* If needed, compute the checksum and mark it as valid. */ 1966 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1967 in_delayed_cksum(copym); 1968 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1969 copym->m_pkthdr.csum_flags |= 1970 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1971 copym->m_pkthdr.csum_data = 0xffff; 1972 } 1973 /* 1974 * We don't bother to fragment if the IP length is greater 1975 * than the interface's MTU. Can this possibly matter? 1976 */ 1977 ip = mtod(copym, struct ip *); 1978 ip->ip_len = htons(ip->ip_len); 1979 ip->ip_off = htons(ip->ip_off); 1980 ip->ip_sum = 0; 1981 ip->ip_sum = in_cksum(copym, hlen); 1982 /* 1983 * NB: 1984 * It's not clear whether there are any lingering 1985 * reentrancy problems in other areas which might 1986 * be exposed by using ip_input directly (in 1987 * particular, everything which modifies the packet 1988 * in-place). Yet another option is using the 1989 * protosw directly to deliver the looped back 1990 * packet. For the moment, we'll err on the side 1991 * of safety by using if_simloop(). 1992 */ 1993#if 1 /* XXX */ 1994 if (dst->sin_family != AF_INET) { 1995 printf("ip_mloopback: bad address family %d\n", 1996 dst->sin_family); 1997 dst->sin_family = AF_INET; 1998 } 1999#endif 2000 2001#ifdef notdef 2002 copym->m_pkthdr.rcvif = ifp; 2003 ip_input(copym); 2004#else 2005 if_simloop(ifp, copym, dst->sin_family, 0); 2006#endif 2007 } 2008} 2009