ip_output.c revision 1.149.2.3
1 /* $NetBSD: ip_output.c,v 1.149.2.3 2005/10/21 18:55:52 riz Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1982, 1986, 1988, 1990, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. Neither the name of the University nor the names of its contributors 82 * may be used to endorse or promote products derived from this software 83 * without specific prior written permission. 84 * 85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 95 * SUCH DAMAGE. 96 * 97 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 98 */ 99 100 #include <sys/cdefs.h> 101 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.149.2.3 2005/10/21 18:55:52 riz Exp $"); 102 103 #include "opt_pfil_hooks.h" 104 #include "opt_inet.h" 105 #include "opt_ipsec.h" 106 #include "opt_mrouting.h" 107 108 #include <sys/param.h> 109 #include <sys/malloc.h> 110 #include <sys/mbuf.h> 111 #include <sys/errno.h> 112 #include <sys/protosw.h> 113 #include <sys/socket.h> 114 #include <sys/socketvar.h> 115 #ifdef FAST_IPSEC 116 #include <sys/domain.h> 117 #endif 118 #include <sys/systm.h> 119 #include <sys/proc.h> 120 121 #include <net/if.h> 122 #include <net/route.h> 123 #include <net/pfil.h> 124 125 #include <netinet/in.h> 126 #include <netinet/in_systm.h> 127 #include <netinet/ip.h> 128 #include <netinet/in_pcb.h> 129 #include <netinet/in_var.h> 130 #include <netinet/ip_var.h> 131 132 #ifdef MROUTING 133 #include <netinet/ip_mroute.h> 134 #endif 135 136 #include <machine/stdarg.h> 137 138 #ifdef IPSEC 139 #include <netinet6/ipsec.h> 140 #include <netkey/key.h> 141 #include <netkey/key_debug.h> 142 #ifdef IPSEC_NAT_T 143 #include <netinet/udp.h> 144 #endif 145 #endif /*IPSEC*/ 146 147 #ifdef FAST_IPSEC 148 #include <netipsec/ipsec.h> 149 #include <netipsec/key.h> 150 #include <netipsec/xform.h> 151 #endif /* FAST_IPSEC*/ 152 153 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 154 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 155 static void ip_mloopback(struct ifnet *, struct mbuf *, struct sockaddr_in *); 156 static int ip_getoptval(struct mbuf *, u_int8_t *, u_int); 157 158 #ifdef PFIL_HOOKS 159 extern struct pfil_head inet_pfil_hook; /* XXX */ 160 #endif 161 162 int udp_do_loopback_cksum = 0; 163 int tcp_do_loopback_cksum = 0; 164 int ip_do_loopback_cksum = 0; 165 166 #define IN_NEED_CHECKSUM(ifp, csum_flags) \ 167 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 168 (((csum_flags) & M_CSUM_UDPv4) != 0 && udp_do_loopback_cksum) || \ 169 (((csum_flags) & M_CSUM_TCPv4) != 0 && tcp_do_loopback_cksum) || \ 170 (((csum_flags) & M_CSUM_IPv4) != 0 && ip_do_loopback_cksum))) 171 172 /* 173 * IP output. The packet in mbuf chain m contains a skeletal IP 174 * header (with len, off, ttl, proto, tos, src, dst). 175 * The mbuf chain containing the packet will be freed. 176 * The mbuf opt, if present, will not be freed. 177 */ 178 int 179 ip_output(struct mbuf *m0, ...) 180 { 181 struct ip *ip; 182 struct ifnet *ifp; 183 struct mbuf *m = m0; 184 int hlen = sizeof (struct ip); 185 int len, error = 0; 186 struct route iproute; 187 struct sockaddr_in *dst; 188 struct in_ifaddr *ia; 189 struct mbuf *opt; 190 struct route *ro; 191 int flags, sw_csum; 192 int *mtu_p; 193 u_long mtu; 194 struct ip_moptions *imo; 195 struct socket *so; 196 va_list ap; 197 #ifdef IPSEC 198 struct secpolicy *sp = NULL; 199 #ifdef IPSEC_NAT_T 200 int natt_frag = 0; 201 #endif 202 #endif /*IPSEC*/ 203 #ifdef FAST_IPSEC 204 struct inpcb *inp; 205 struct m_tag *mtag; 206 struct secpolicy *sp = NULL; 207 struct tdb_ident *tdbi; 208 int s; 209 #endif 210 u_int16_t ip_len; 211 212 len = 0; 213 va_start(ap, m0); 214 opt = va_arg(ap, struct mbuf *); 215 ro = va_arg(ap, struct route *); 216 flags = va_arg(ap, int); 217 imo = va_arg(ap, struct ip_moptions *); 218 so = va_arg(ap, struct socket *); 219 if (flags & IP_RETURNMTU) 220 mtu_p = va_arg(ap, int *); 221 else 222 mtu_p = NULL; 223 va_end(ap); 224 225 MCLAIM(m, &ip_tx_mowner); 226 #ifdef FAST_IPSEC 227 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 228 inp = (struct inpcb *)so->so_pcb; 229 else 230 inp = NULL; 231 #endif /* FAST_IPSEC */ 232 233 #ifdef DIAGNOSTIC 234 if ((m->m_flags & M_PKTHDR) == 0) 235 panic("ip_output no HDR"); 236 #endif 237 if (opt) { 238 m = ip_insertoptions(m, opt, &len); 239 if (len >= sizeof(struct ip)) 240 hlen = len; 241 } 242 ip = mtod(m, struct ip *); 243 /* 244 * Fill in IP header. 245 */ 246 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 247 ip->ip_v = IPVERSION; 248 ip->ip_off = htons(0); 249 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 250 ip->ip_id = ip_newid(); 251 } else { 252 253 /* 254 * TSO capable interfaces (typically?) increment 255 * ip_id for each segment. 256 * "allocate" enough ids here to increase the chance 257 * for them to be unique. 258 * 259 * note that the following calculation is not 260 * needed to be precise. wasting some ip_id is fine. 261 */ 262 263 unsigned int segsz = m->m_pkthdr.segsz; 264 unsigned int datasz = ntohs(ip->ip_len) - hlen; 265 unsigned int num = howmany(datasz, segsz); 266 267 ip->ip_id = ip_newid_range(num); 268 } 269 ip->ip_hl = hlen >> 2; 270 ipstat.ips_localout++; 271 } else { 272 hlen = ip->ip_hl << 2; 273 } 274 /* 275 * Route packet. 276 */ 277 if (ro == 0) { 278 ro = &iproute; 279 bzero((caddr_t)ro, sizeof (*ro)); 280 } 281 dst = satosin(&ro->ro_dst); 282 /* 283 * If there is a cached route, 284 * check that it is to the same destination 285 * and is still up. If not, free it and try again. 286 * The address family should also be checked in case of sharing the 287 * cache with IPv6. 288 */ 289 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 290 dst->sin_family != AF_INET || 291 !in_hosteq(dst->sin_addr, ip->ip_dst))) { 292 RTFREE(ro->ro_rt); 293 ro->ro_rt = (struct rtentry *)0; 294 } 295 if (ro->ro_rt == 0) { 296 bzero(dst, sizeof(*dst)); 297 dst->sin_family = AF_INET; 298 dst->sin_len = sizeof(*dst); 299 dst->sin_addr = ip->ip_dst; 300 } 301 /* 302 * If routing to interface only, 303 * short circuit routing lookup. 304 */ 305 if (flags & IP_ROUTETOIF) { 306 if ((ia = ifatoia(ifa_ifwithladdr(sintosa(dst)))) == 0) { 307 ipstat.ips_noroute++; 308 error = ENETUNREACH; 309 goto bad; 310 } 311 ifp = ia->ia_ifp; 312 mtu = ifp->if_mtu; 313 ip->ip_ttl = 1; 314 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 315 ip->ip_dst.s_addr == INADDR_BROADCAST) && 316 imo != NULL && imo->imo_multicast_ifp != NULL) { 317 ifp = imo->imo_multicast_ifp; 318 mtu = ifp->if_mtu; 319 IFP_TO_IA(ifp, ia); 320 } else { 321 if (ro->ro_rt == 0) 322 rtalloc(ro); 323 if (ro->ro_rt == 0) { 324 ipstat.ips_noroute++; 325 error = EHOSTUNREACH; 326 goto bad; 327 } 328 ia = ifatoia(ro->ro_rt->rt_ifa); 329 ifp = ro->ro_rt->rt_ifp; 330 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 331 mtu = ifp->if_mtu; 332 ro->ro_rt->rt_use++; 333 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 334 dst = satosin(ro->ro_rt->rt_gateway); 335 } 336 if (IN_MULTICAST(ip->ip_dst.s_addr) || 337 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 338 struct in_multi *inm; 339 340 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 341 M_BCAST : M_MCAST; 342 /* 343 * IP destination address is multicast. Make sure "dst" 344 * still points to the address in "ro". (It may have been 345 * changed to point to a gateway address, above.) 346 */ 347 dst = satosin(&ro->ro_dst); 348 /* 349 * See if the caller provided any multicast options 350 */ 351 if (imo != NULL) 352 ip->ip_ttl = imo->imo_multicast_ttl; 353 else 354 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 355 356 /* 357 * if we don't know the outgoing ifp yet, we can't generate 358 * output 359 */ 360 if (!ifp) { 361 ipstat.ips_noroute++; 362 error = ENETUNREACH; 363 goto bad; 364 } 365 366 /* 367 * If the packet is multicast or broadcast, confirm that 368 * the outgoing interface can transmit it. 369 */ 370 if (((m->m_flags & M_MCAST) && 371 (ifp->if_flags & IFF_MULTICAST) == 0) || 372 ((m->m_flags & M_BCAST) && 373 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 374 ipstat.ips_noroute++; 375 error = ENETUNREACH; 376 goto bad; 377 } 378 /* 379 * If source address not specified yet, use an address 380 * of outgoing interface. 381 */ 382 if (in_nullhost(ip->ip_src)) { 383 struct in_ifaddr *ia; 384 385 IFP_TO_IA(ifp, ia); 386 if (!ia) { 387 error = EADDRNOTAVAIL; 388 goto bad; 389 } 390 ip->ip_src = ia->ia_addr.sin_addr; 391 } 392 393 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 394 if (inm != NULL && 395 (imo == NULL || imo->imo_multicast_loop)) { 396 /* 397 * If we belong to the destination multicast group 398 * on the outgoing interface, and the caller did not 399 * forbid loopback, loop back a copy. 400 */ 401 ip_mloopback(ifp, m, dst); 402 } 403 #ifdef MROUTING 404 else { 405 /* 406 * If we are acting as a multicast router, perform 407 * multicast forwarding as if the packet had just 408 * arrived on the interface to which we are about 409 * to send. The multicast forwarding function 410 * recursively calls this function, using the 411 * IP_FORWARDING flag to prevent infinite recursion. 412 * 413 * Multicasts that are looped back by ip_mloopback(), 414 * above, will be forwarded by the ip_input() routine, 415 * if necessary. 416 */ 417 extern struct socket *ip_mrouter; 418 419 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 420 if (ip_mforward(m, ifp) != 0) { 421 m_freem(m); 422 goto done; 423 } 424 } 425 } 426 #endif 427 /* 428 * Multicasts with a time-to-live of zero may be looped- 429 * back, above, but must not be transmitted on a network. 430 * Also, multicasts addressed to the loopback interface 431 * are not sent -- the above call to ip_mloopback() will 432 * loop back a copy if this host actually belongs to the 433 * destination group on the loopback interface. 434 */ 435 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 436 m_freem(m); 437 goto done; 438 } 439 440 goto sendit; 441 } 442 #ifndef notdef 443 /* 444 * If source address not specified yet, use address 445 * of outgoing interface. 446 */ 447 if (in_nullhost(ip->ip_src)) 448 ip->ip_src = ia->ia_addr.sin_addr; 449 #endif 450 451 /* 452 * packets with Class-D address as source are not valid per 453 * RFC 1112 454 */ 455 if (IN_MULTICAST(ip->ip_src.s_addr)) { 456 ipstat.ips_odropped++; 457 error = EADDRNOTAVAIL; 458 goto bad; 459 } 460 461 /* 462 * Look for broadcast address and 463 * and verify user is allowed to send 464 * such a packet. 465 */ 466 if (in_broadcast(dst->sin_addr, ifp)) { 467 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 468 error = EADDRNOTAVAIL; 469 goto bad; 470 } 471 if ((flags & IP_ALLOWBROADCAST) == 0) { 472 error = EACCES; 473 goto bad; 474 } 475 /* don't allow broadcast messages to be fragmented */ 476 if (ntohs(ip->ip_len) > ifp->if_mtu) { 477 error = EMSGSIZE; 478 goto bad; 479 } 480 m->m_flags |= M_BCAST; 481 } else 482 m->m_flags &= ~M_BCAST; 483 484 sendit: 485 /* 486 * If we're doing Path MTU Discovery, we need to set DF unless 487 * the route's MTU is locked. 488 */ 489 if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL && 490 (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 491 ip->ip_off |= htons(IP_DF); 492 493 /* Remember the current ip_len */ 494 ip_len = ntohs(ip->ip_len); 495 496 #ifdef IPSEC 497 /* get SP for this packet */ 498 if (so == NULL) 499 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 500 flags, &error); 501 else { 502 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 503 IPSEC_DIR_OUTBOUND)) 504 goto skip_ipsec; 505 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 506 } 507 508 if (sp == NULL) { 509 ipsecstat.out_inval++; 510 goto bad; 511 } 512 513 error = 0; 514 515 /* check policy */ 516 switch (sp->policy) { 517 case IPSEC_POLICY_DISCARD: 518 /* 519 * This packet is just discarded. 520 */ 521 ipsecstat.out_polvio++; 522 goto bad; 523 524 case IPSEC_POLICY_BYPASS: 525 case IPSEC_POLICY_NONE: 526 /* no need to do IPsec. */ 527 goto skip_ipsec; 528 529 case IPSEC_POLICY_IPSEC: 530 if (sp->req == NULL) { 531 /* XXX should be panic ? */ 532 printf("ip_output: No IPsec request specified.\n"); 533 error = EINVAL; 534 goto bad; 535 } 536 break; 537 538 case IPSEC_POLICY_ENTRUST: 539 default: 540 printf("ip_output: Invalid policy found. %d\n", sp->policy); 541 } 542 543 #ifdef IPSEC_NAT_T 544 /* 545 * NAT-T ESP fragmentation: don't do IPSec processing now, 546 * we'll do it on each fragmented packet. 547 */ 548 if (sp->req->sav && 549 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 550 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 551 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 552 natt_frag = 1; 553 mtu = sp->req->sav->esp_frag; 554 goto skip_ipsec; 555 } 556 } 557 #endif /* IPSEC_NAT_T */ 558 559 /* 560 * ipsec4_output() expects ip_len and ip_off in network 561 * order. They have been set to network order above. 562 */ 563 564 { 565 struct ipsec_output_state state; 566 bzero(&state, sizeof(state)); 567 state.m = m; 568 if (flags & IP_ROUTETOIF) { 569 state.ro = &iproute; 570 bzero(&iproute, sizeof(iproute)); 571 } else 572 state.ro = ro; 573 state.dst = (struct sockaddr *)dst; 574 575 /* 576 * We can't defer the checksum of payload data if 577 * we're about to encrypt/authenticate it. 578 * 579 * XXX When we support crypto offloading functions of 580 * XXX network interfaces, we need to reconsider this, 581 * XXX since it's likely that they'll support checksumming, 582 * XXX as well. 583 */ 584 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 585 in_delayed_cksum(m); 586 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 587 } 588 589 error = ipsec4_output(&state, sp, flags); 590 591 m = state.m; 592 if (flags & IP_ROUTETOIF) { 593 /* 594 * if we have tunnel mode SA, we may need to ignore 595 * IP_ROUTETOIF. 596 */ 597 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 598 flags &= ~IP_ROUTETOIF; 599 ro = state.ro; 600 } 601 } else 602 ro = state.ro; 603 dst = (struct sockaddr_in *)state.dst; 604 if (error) { 605 /* mbuf is already reclaimed in ipsec4_output. */ 606 m0 = NULL; 607 switch (error) { 608 case EHOSTUNREACH: 609 case ENETUNREACH: 610 case EMSGSIZE: 611 case ENOBUFS: 612 case ENOMEM: 613 break; 614 default: 615 printf("ip4_output (ipsec): error code %d\n", error); 616 /*fall through*/ 617 case ENOENT: 618 /* don't show these error codes to the user */ 619 error = 0; 620 break; 621 } 622 goto bad; 623 } 624 625 /* be sure to update variables that are affected by ipsec4_output() */ 626 ip = mtod(m, struct ip *); 627 hlen = ip->ip_hl << 2; 628 ip_len = ntohs(ip->ip_len); 629 630 if (ro->ro_rt == NULL) { 631 if ((flags & IP_ROUTETOIF) == 0) { 632 printf("ip_output: " 633 "can't update route after IPsec processing\n"); 634 error = EHOSTUNREACH; /*XXX*/ 635 goto bad; 636 } 637 } else { 638 /* nobody uses ia beyond here */ 639 if (state.encap) { 640 ifp = ro->ro_rt->rt_ifp; 641 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 642 mtu = ifp->if_mtu; 643 } 644 } 645 } 646 skip_ipsec: 647 #endif /*IPSEC*/ 648 #ifdef FAST_IPSEC 649 /* 650 * Check the security policy (SP) for the packet and, if 651 * required, do IPsec-related processing. There are two 652 * cases here; the first time a packet is sent through 653 * it will be untagged and handled by ipsec4_checkpolicy. 654 * If the packet is resubmitted to ip_output (e.g. after 655 * AH, ESP, etc. processing), there will be a tag to bypass 656 * the lookup and related policy checking. 657 */ 658 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 659 s = splsoftnet(); 660 if (mtag != NULL) { 661 tdbi = (struct tdb_ident *)(mtag + 1); 662 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 663 if (sp == NULL) 664 error = -EINVAL; /* force silent drop */ 665 m_tag_delete(m, mtag); 666 } else { 667 if (inp != NULL && 668 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 669 goto spd_done; 670 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 671 &error, inp); 672 } 673 /* 674 * There are four return cases: 675 * sp != NULL apply IPsec policy 676 * sp == NULL, error == 0 no IPsec handling needed 677 * sp == NULL, error == -EINVAL discard packet w/o error 678 * sp == NULL, error != 0 discard packet, report error 679 */ 680 if (sp != NULL) { 681 /* Loop detection, check if ipsec processing already done */ 682 IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 683 for (mtag = m_tag_first(m); mtag != NULL; 684 mtag = m_tag_next(m, mtag)) { 685 #ifdef MTAG_ABI_COMPAT 686 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 687 continue; 688 #endif 689 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 690 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 691 continue; 692 /* 693 * Check if policy has an SA associated with it. 694 * This can happen when an SP has yet to acquire 695 * an SA; e.g. on first reference. If it occurs, 696 * then we let ipsec4_process_packet do its thing. 697 */ 698 if (sp->req->sav == NULL) 699 break; 700 tdbi = (struct tdb_ident *)(mtag + 1); 701 if (tdbi->spi == sp->req->sav->spi && 702 tdbi->proto == sp->req->sav->sah->saidx.proto && 703 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 704 sizeof (union sockaddr_union)) == 0) { 705 /* 706 * No IPsec processing is needed, free 707 * reference to SP. 708 * 709 * NB: null pointer to avoid free at 710 * done: below. 711 */ 712 KEY_FREESP(&sp), sp = NULL; 713 splx(s); 714 goto spd_done; 715 } 716 } 717 718 /* 719 * Do delayed checksums now because we send before 720 * this is done in the normal processing path. 721 */ 722 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 723 in_delayed_cksum(m); 724 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 725 } 726 727 #ifdef __FreeBSD__ 728 ip->ip_len = htons(ip->ip_len); 729 ip->ip_off = htons(ip->ip_off); 730 #endif 731 732 /* NB: callee frees mbuf */ 733 error = ipsec4_process_packet(m, sp->req, flags, 0); 734 /* 735 * Preserve KAME behaviour: ENOENT can be returned 736 * when an SA acquire is in progress. Don't propagate 737 * this to user-level; it confuses applications. 738 * 739 * XXX this will go away when the SADB is redone. 740 */ 741 if (error == ENOENT) 742 error = 0; 743 splx(s); 744 goto done; 745 } else { 746 splx(s); 747 748 if (error != 0) { 749 /* 750 * Hack: -EINVAL is used to signal that a packet 751 * should be silently discarded. This is typically 752 * because we asked key management for an SA and 753 * it was delayed (e.g. kicked up to IKE). 754 */ 755 if (error == -EINVAL) 756 error = 0; 757 goto bad; 758 } else { 759 /* No IPsec processing for this packet. */ 760 } 761 #ifdef notyet 762 /* 763 * If deferred crypto processing is needed, check that 764 * the interface supports it. 765 */ 766 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 767 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 768 /* notify IPsec to do its own crypto */ 769 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 770 error = EHOSTUNREACH; 771 goto bad; 772 } 773 #endif 774 } 775 spd_done: 776 #endif /* FAST_IPSEC */ 777 778 #ifdef PFIL_HOOKS 779 /* 780 * Run through list of hooks for output packets. 781 */ 782 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 783 goto done; 784 if (m == NULL) 785 goto done; 786 787 ip = mtod(m, struct ip *); 788 hlen = ip->ip_hl << 2; 789 #endif /* PFIL_HOOKS */ 790 791 m->m_pkthdr.csum_data |= hlen << 16; 792 793 #if IFA_STATS 794 /* 795 * search for the source address structure to 796 * maintain output statistics. 797 */ 798 INADDR_TO_IA(ip->ip_src, ia); 799 #endif 800 801 /* Maybe skip checksums on loopback interfaces. */ 802 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 803 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 804 } 805 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 806 /* 807 * If small enough for mtu of path, or if using TCP segmentation 808 * offload, can just send directly. 809 */ 810 if (ip_len <= mtu || 811 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 812 #if IFA_STATS 813 if (ia) 814 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 815 #endif 816 /* 817 * Always initialize the sum to 0! Some HW assisted 818 * checksumming requires this. 819 */ 820 ip->ip_sum = 0; 821 822 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 823 /* 824 * Perform any checksums that the hardware can't do 825 * for us. 826 * 827 * XXX Does any hardware require the {th,uh}_sum 828 * XXX fields to be 0? 829 */ 830 if (sw_csum & M_CSUM_IPv4) { 831 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 832 ip->ip_sum = in_cksum(m, hlen); 833 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 834 } 835 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 836 if (IN_NEED_CHECKSUM(ifp, 837 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 838 in_delayed_cksum(m); 839 } 840 m->m_pkthdr.csum_flags &= 841 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 842 } 843 } 844 845 #ifdef IPSEC 846 /* clean ipsec history once it goes out of the node */ 847 ipsec_delaux(m); 848 #endif 849 error = (*ifp->if_output)(ifp, m, sintosa(dst), ro->ro_rt); 850 goto done; 851 } 852 853 /* 854 * We can't use HW checksumming if we're about to 855 * to fragment the packet. 856 * 857 * XXX Some hardware can do this. 858 */ 859 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 860 if (IN_NEED_CHECKSUM(ifp, 861 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 862 in_delayed_cksum(m); 863 } 864 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 865 } 866 867 /* 868 * Too large for interface; fragment if possible. 869 * Must be able to put at least 8 bytes per fragment. 870 */ 871 if (ntohs(ip->ip_off) & IP_DF) { 872 if (flags & IP_RETURNMTU) 873 *mtu_p = mtu; 874 error = EMSGSIZE; 875 ipstat.ips_cantfrag++; 876 goto bad; 877 } 878 879 error = ip_fragment(m, ifp, mtu); 880 if (error) { 881 m = NULL; 882 goto bad; 883 } 884 885 for (; m; m = m0) { 886 m0 = m->m_nextpkt; 887 m->m_nextpkt = 0; 888 if (error == 0) { 889 #if IFA_STATS 890 if (ia) 891 ia->ia_ifa.ifa_data.ifad_outbytes += 892 ntohs(ip->ip_len); 893 #endif 894 #ifdef IPSEC 895 /* clean ipsec history once it goes out of the node */ 896 ipsec_delaux(m); 897 898 #ifdef IPSEC_NAT_T 899 /* 900 * If we get there, the packet has not been handeld by 901 * IPSec whereas it should have. Now that it has been 902 * fragmented, re-inject it in ip_output so that IPsec 903 * processing can occur. 904 */ 905 if (natt_frag) { 906 error = ip_output(m, opt, 907 ro, flags, imo, so, mtu_p); 908 } else 909 #endif /* IPSEC_NAT_T */ 910 #endif /* IPSEC */ 911 { 912 KASSERT((m->m_pkthdr.csum_flags & 913 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 914 error = (*ifp->if_output)(ifp, m, sintosa(dst), 915 ro->ro_rt); 916 } 917 } else 918 m_freem(m); 919 } 920 921 if (error == 0) 922 ipstat.ips_fragmented++; 923 done: 924 if (ro == &iproute && (flags & IP_ROUTETOIF) == 0 && ro->ro_rt) { 925 RTFREE(ro->ro_rt); 926 ro->ro_rt = 0; 927 } 928 929 #ifdef IPSEC 930 if (sp != NULL) { 931 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 932 printf("DP ip_output call free SP:%p\n", sp)); 933 key_freesp(sp); 934 } 935 #endif /* IPSEC */ 936 #ifdef FAST_IPSEC 937 if (sp != NULL) 938 KEY_FREESP(&sp); 939 #endif /* FAST_IPSEC */ 940 941 return (error); 942 bad: 943 m_freem(m); 944 goto done; 945 } 946 947 int 948 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 949 { 950 struct ip *ip, *mhip; 951 struct mbuf *m0; 952 int len, hlen, off; 953 int mhlen, firstlen; 954 struct mbuf **mnext; 955 int sw_csum = m->m_pkthdr.csum_flags; 956 int fragments = 0; 957 int s; 958 int error = 0; 959 960 ip = mtod(m, struct ip *); 961 hlen = ip->ip_hl << 2; 962 if (ifp != NULL) 963 sw_csum &= ~ifp->if_csum_flags_tx; 964 965 len = (mtu - hlen) &~ 7; 966 if (len < 8) { 967 m_freem(m); 968 return (EMSGSIZE); 969 } 970 971 firstlen = len; 972 mnext = &m->m_nextpkt; 973 974 /* 975 * Loop through length of segment after first fragment, 976 * make new header and copy data of each part and link onto chain. 977 */ 978 m0 = m; 979 mhlen = sizeof (struct ip); 980 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 981 MGETHDR(m, M_DONTWAIT, MT_HEADER); 982 if (m == 0) { 983 error = ENOBUFS; 984 ipstat.ips_odropped++; 985 goto sendorfree; 986 } 987 MCLAIM(m, m0->m_owner); 988 *mnext = m; 989 mnext = &m->m_nextpkt; 990 m->m_data += max_linkhdr; 991 mhip = mtod(m, struct ip *); 992 *mhip = *ip; 993 /* we must inherit MCAST and BCAST flags */ 994 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 995 if (hlen > sizeof (struct ip)) { 996 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 997 mhip->ip_hl = mhlen >> 2; 998 } 999 m->m_len = mhlen; 1000 mhip->ip_off = ((off - hlen) >> 3) + 1001 (ntohs(ip->ip_off) & ~IP_MF); 1002 if (ip->ip_off & htons(IP_MF)) 1003 mhip->ip_off |= IP_MF; 1004 if (off + len >= ntohs(ip->ip_len)) 1005 len = ntohs(ip->ip_len) - off; 1006 else 1007 mhip->ip_off |= IP_MF; 1008 HTONS(mhip->ip_off); 1009 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 1010 m->m_next = m_copy(m0, off, len); 1011 if (m->m_next == 0) { 1012 error = ENOBUFS; /* ??? */ 1013 ipstat.ips_odropped++; 1014 goto sendorfree; 1015 } 1016 m->m_pkthdr.len = mhlen + len; 1017 m->m_pkthdr.rcvif = (struct ifnet *)0; 1018 mhip->ip_sum = 0; 1019 if (sw_csum & M_CSUM_IPv4) { 1020 mhip->ip_sum = in_cksum(m, mhlen); 1021 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 1022 } else { 1023 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 1024 m->m_pkthdr.csum_data |= mhlen << 16; 1025 } 1026 ipstat.ips_ofragments++; 1027 fragments++; 1028 } 1029 /* 1030 * Update first fragment by trimming what's been copied out 1031 * and updating header, then send each fragment (in order). 1032 */ 1033 m = m0; 1034 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1035 m->m_pkthdr.len = hlen + firstlen; 1036 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1037 ip->ip_off |= htons(IP_MF); 1038 ip->ip_sum = 0; 1039 if (sw_csum & M_CSUM_IPv4) { 1040 ip->ip_sum = in_cksum(m, hlen); 1041 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1042 } else { 1043 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1044 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 1045 sizeof(struct ip)); 1046 } 1047 sendorfree: 1048 /* 1049 * If there is no room for all the fragments, don't queue 1050 * any of them. 1051 */ 1052 if (ifp != NULL) { 1053 s = splnet(); 1054 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1055 error == 0) { 1056 error = ENOBUFS; 1057 ipstat.ips_odropped++; 1058 IFQ_INC_DROPS(&ifp->if_snd); 1059 } 1060 splx(s); 1061 } 1062 if (error) { 1063 for (m = m0; m; m = m0) { 1064 m0 = m->m_nextpkt; 1065 m->m_nextpkt = NULL; 1066 m_freem(m); 1067 } 1068 } 1069 return (error); 1070 } 1071 1072 /* 1073 * Process a delayed payload checksum calculation. 1074 */ 1075 void 1076 in_delayed_cksum(struct mbuf *m) 1077 { 1078 struct ip *ip; 1079 u_int16_t csum, offset; 1080 1081 ip = mtod(m, struct ip *); 1082 offset = ip->ip_hl << 2; 1083 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1084 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1085 csum = 0xffff; 1086 1087 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 1088 1089 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1090 /* This happen when ip options were inserted 1091 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1092 m->m_len, offset, ip->ip_p); 1093 */ 1094 m_copyback(m, offset, sizeof(csum), (caddr_t) &csum); 1095 } else 1096 *(u_int16_t *)(mtod(m, caddr_t) + offset) = csum; 1097 } 1098 1099 /* 1100 * Determine the maximum length of the options to be inserted; 1101 * we would far rather allocate too much space rather than too little. 1102 */ 1103 1104 u_int 1105 ip_optlen(struct inpcb *inp) 1106 { 1107 struct mbuf *m = inp->inp_options; 1108 1109 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1110 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1111 else 1112 return 0; 1113 } 1114 1115 1116 /* 1117 * Insert IP options into preformed packet. 1118 * Adjust IP destination as required for IP source routing, 1119 * as indicated by a non-zero in_addr at the start of the options. 1120 */ 1121 static struct mbuf * 1122 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1123 { 1124 struct ipoption *p = mtod(opt, struct ipoption *); 1125 struct mbuf *n; 1126 struct ip *ip = mtod(m, struct ip *); 1127 unsigned optlen; 1128 1129 optlen = opt->m_len - sizeof(p->ipopt_dst); 1130 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1131 return (m); /* XXX should fail */ 1132 if (!in_nullhost(p->ipopt_dst)) 1133 ip->ip_dst = p->ipopt_dst; 1134 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1135 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1136 if (n == 0) 1137 return (m); 1138 MCLAIM(n, m->m_owner); 1139 M_COPY_PKTHDR(n, m); 1140 m_tag_delete_chain(m, NULL); 1141 m->m_flags &= ~M_PKTHDR; 1142 m->m_len -= sizeof(struct ip); 1143 m->m_data += sizeof(struct ip); 1144 n->m_next = m; 1145 m = n; 1146 m->m_len = optlen + sizeof(struct ip); 1147 m->m_data += max_linkhdr; 1148 bcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1149 } else { 1150 m->m_data -= optlen; 1151 m->m_len += optlen; 1152 memmove(mtod(m, caddr_t), ip, sizeof(struct ip)); 1153 } 1154 m->m_pkthdr.len += optlen; 1155 ip = mtod(m, struct ip *); 1156 bcopy((caddr_t)p->ipopt_list, (caddr_t)(ip + 1), (unsigned)optlen); 1157 *phlen = sizeof(struct ip) + optlen; 1158 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1159 return (m); 1160 } 1161 1162 /* 1163 * Copy options from ip to jp, 1164 * omitting those not copied during fragmentation. 1165 */ 1166 int 1167 ip_optcopy(struct ip *ip, struct ip *jp) 1168 { 1169 u_char *cp, *dp; 1170 int opt, optlen, cnt; 1171 1172 cp = (u_char *)(ip + 1); 1173 dp = (u_char *)(jp + 1); 1174 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1175 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1176 opt = cp[0]; 1177 if (opt == IPOPT_EOL) 1178 break; 1179 if (opt == IPOPT_NOP) { 1180 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1181 *dp++ = IPOPT_NOP; 1182 optlen = 1; 1183 continue; 1184 } 1185 #ifdef DIAGNOSTIC 1186 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1187 panic("malformed IPv4 option passed to ip_optcopy"); 1188 #endif 1189 optlen = cp[IPOPT_OLEN]; 1190 #ifdef DIAGNOSTIC 1191 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1192 panic("malformed IPv4 option passed to ip_optcopy"); 1193 #endif 1194 /* bogus lengths should have been caught by ip_dooptions */ 1195 if (optlen > cnt) 1196 optlen = cnt; 1197 if (IPOPT_COPIED(opt)) { 1198 bcopy((caddr_t)cp, (caddr_t)dp, (unsigned)optlen); 1199 dp += optlen; 1200 } 1201 } 1202 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1203 *dp++ = IPOPT_EOL; 1204 return (optlen); 1205 } 1206 1207 /* 1208 * IP socket option processing. 1209 */ 1210 int 1211 ip_ctloutput(int op, struct socket *so, int level, int optname, 1212 struct mbuf **mp) 1213 { 1214 struct inpcb *inp = sotoinpcb(so); 1215 struct mbuf *m = *mp; 1216 int optval = 0; 1217 int error = 0; 1218 #if defined(IPSEC) || defined(FAST_IPSEC) 1219 struct proc *p = curproc; /*XXX*/ 1220 #endif 1221 1222 if (level != IPPROTO_IP) { 1223 error = EINVAL; 1224 if (op == PRCO_SETOPT && *mp) 1225 (void) m_free(*mp); 1226 } else switch (op) { 1227 1228 case PRCO_SETOPT: 1229 switch (optname) { 1230 case IP_OPTIONS: 1231 #ifdef notyet 1232 case IP_RETOPTS: 1233 return (ip_pcbopts(optname, &inp->inp_options, m)); 1234 #else 1235 return (ip_pcbopts(&inp->inp_options, m)); 1236 #endif 1237 1238 case IP_TOS: 1239 case IP_TTL: 1240 case IP_RECVOPTS: 1241 case IP_RECVRETOPTS: 1242 case IP_RECVDSTADDR: 1243 case IP_RECVIF: 1244 if (m == NULL || m->m_len != sizeof(int)) 1245 error = EINVAL; 1246 else { 1247 optval = *mtod(m, int *); 1248 switch (optname) { 1249 1250 case IP_TOS: 1251 inp->inp_ip.ip_tos = optval; 1252 break; 1253 1254 case IP_TTL: 1255 inp->inp_ip.ip_ttl = optval; 1256 break; 1257 #define OPTSET(bit) \ 1258 if (optval) \ 1259 inp->inp_flags |= bit; \ 1260 else \ 1261 inp->inp_flags &= ~bit; 1262 1263 case IP_RECVOPTS: 1264 OPTSET(INP_RECVOPTS); 1265 break; 1266 1267 case IP_RECVRETOPTS: 1268 OPTSET(INP_RECVRETOPTS); 1269 break; 1270 1271 case IP_RECVDSTADDR: 1272 OPTSET(INP_RECVDSTADDR); 1273 break; 1274 1275 case IP_RECVIF: 1276 OPTSET(INP_RECVIF); 1277 break; 1278 } 1279 } 1280 break; 1281 #undef OPTSET 1282 1283 case IP_MULTICAST_IF: 1284 case IP_MULTICAST_TTL: 1285 case IP_MULTICAST_LOOP: 1286 case IP_ADD_MEMBERSHIP: 1287 case IP_DROP_MEMBERSHIP: 1288 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1289 break; 1290 1291 case IP_PORTRANGE: 1292 if (m == 0 || m->m_len != sizeof(int)) 1293 error = EINVAL; 1294 else { 1295 optval = *mtod(m, int *); 1296 1297 switch (optval) { 1298 1299 case IP_PORTRANGE_DEFAULT: 1300 case IP_PORTRANGE_HIGH: 1301 inp->inp_flags &= ~(INP_LOWPORT); 1302 break; 1303 1304 case IP_PORTRANGE_LOW: 1305 inp->inp_flags |= INP_LOWPORT; 1306 break; 1307 1308 default: 1309 error = EINVAL; 1310 break; 1311 } 1312 } 1313 break; 1314 1315 #if defined(IPSEC) || defined(FAST_IPSEC) 1316 case IP_IPSEC_POLICY: 1317 { 1318 caddr_t req = NULL; 1319 size_t len = 0; 1320 int priv = 0; 1321 1322 #ifdef __NetBSD__ 1323 if (p == 0 || suser(p->p_ucred, &p->p_acflag)) 1324 priv = 0; 1325 else 1326 priv = 1; 1327 #else 1328 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1329 #endif 1330 if (m) { 1331 req = mtod(m, caddr_t); 1332 len = m->m_len; 1333 } 1334 error = ipsec4_set_policy(inp, optname, req, len, priv); 1335 break; 1336 } 1337 #endif /*IPSEC*/ 1338 1339 default: 1340 error = ENOPROTOOPT; 1341 break; 1342 } 1343 if (m) 1344 (void)m_free(m); 1345 break; 1346 1347 case PRCO_GETOPT: 1348 switch (optname) { 1349 case IP_OPTIONS: 1350 case IP_RETOPTS: 1351 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1352 MCLAIM(m, so->so_mowner); 1353 if (inp->inp_options) { 1354 m->m_len = inp->inp_options->m_len; 1355 bcopy(mtod(inp->inp_options, caddr_t), 1356 mtod(m, caddr_t), (unsigned)m->m_len); 1357 } else 1358 m->m_len = 0; 1359 break; 1360 1361 case IP_TOS: 1362 case IP_TTL: 1363 case IP_RECVOPTS: 1364 case IP_RECVRETOPTS: 1365 case IP_RECVDSTADDR: 1366 case IP_RECVIF: 1367 case IP_ERRORMTU: 1368 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1369 MCLAIM(m, so->so_mowner); 1370 m->m_len = sizeof(int); 1371 switch (optname) { 1372 1373 case IP_TOS: 1374 optval = inp->inp_ip.ip_tos; 1375 break; 1376 1377 case IP_TTL: 1378 optval = inp->inp_ip.ip_ttl; 1379 break; 1380 1381 case IP_ERRORMTU: 1382 optval = inp->inp_errormtu; 1383 break; 1384 1385 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1386 1387 case IP_RECVOPTS: 1388 optval = OPTBIT(INP_RECVOPTS); 1389 break; 1390 1391 case IP_RECVRETOPTS: 1392 optval = OPTBIT(INP_RECVRETOPTS); 1393 break; 1394 1395 case IP_RECVDSTADDR: 1396 optval = OPTBIT(INP_RECVDSTADDR); 1397 break; 1398 1399 case IP_RECVIF: 1400 optval = OPTBIT(INP_RECVIF); 1401 break; 1402 } 1403 *mtod(m, int *) = optval; 1404 break; 1405 1406 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1407 /* XXX: code broken */ 1408 case IP_IPSEC_POLICY: 1409 { 1410 caddr_t req = NULL; 1411 size_t len = 0; 1412 1413 if (m) { 1414 req = mtod(m, caddr_t); 1415 len = m->m_len; 1416 } 1417 error = ipsec4_get_policy(inp, req, len, mp); 1418 break; 1419 } 1420 #endif /*IPSEC*/ 1421 1422 case IP_MULTICAST_IF: 1423 case IP_MULTICAST_TTL: 1424 case IP_MULTICAST_LOOP: 1425 case IP_ADD_MEMBERSHIP: 1426 case IP_DROP_MEMBERSHIP: 1427 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1428 if (*mp) 1429 MCLAIM(*mp, so->so_mowner); 1430 break; 1431 1432 case IP_PORTRANGE: 1433 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1434 MCLAIM(m, so->so_mowner); 1435 m->m_len = sizeof(int); 1436 1437 if (inp->inp_flags & INP_LOWPORT) 1438 optval = IP_PORTRANGE_LOW; 1439 else 1440 optval = IP_PORTRANGE_DEFAULT; 1441 1442 *mtod(m, int *) = optval; 1443 break; 1444 1445 default: 1446 error = ENOPROTOOPT; 1447 break; 1448 } 1449 break; 1450 } 1451 return (error); 1452 } 1453 1454 /* 1455 * Set up IP options in pcb for insertion in output packets. 1456 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1457 * with destination address if source routed. 1458 */ 1459 int 1460 #ifdef notyet 1461 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m) 1462 #else 1463 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m) 1464 #endif 1465 { 1466 int cnt, optlen; 1467 u_char *cp; 1468 u_char opt; 1469 1470 /* turn off any old options */ 1471 if (*pcbopt) 1472 (void)m_free(*pcbopt); 1473 *pcbopt = 0; 1474 if (m == (struct mbuf *)0 || m->m_len == 0) { 1475 /* 1476 * Only turning off any previous options. 1477 */ 1478 if (m) 1479 (void)m_free(m); 1480 return (0); 1481 } 1482 1483 #ifndef __vax__ 1484 if (m->m_len % sizeof(int32_t)) 1485 goto bad; 1486 #endif 1487 /* 1488 * IP first-hop destination address will be stored before 1489 * actual options; move other options back 1490 * and clear it when none present. 1491 */ 1492 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1493 goto bad; 1494 cnt = m->m_len; 1495 m->m_len += sizeof(struct in_addr); 1496 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1497 memmove(cp, mtod(m, caddr_t), (unsigned)cnt); 1498 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1499 1500 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1501 opt = cp[IPOPT_OPTVAL]; 1502 if (opt == IPOPT_EOL) 1503 break; 1504 if (opt == IPOPT_NOP) 1505 optlen = 1; 1506 else { 1507 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1508 goto bad; 1509 optlen = cp[IPOPT_OLEN]; 1510 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1511 goto bad; 1512 } 1513 switch (opt) { 1514 1515 default: 1516 break; 1517 1518 case IPOPT_LSRR: 1519 case IPOPT_SSRR: 1520 /* 1521 * user process specifies route as: 1522 * ->A->B->C->D 1523 * D must be our final destination (but we can't 1524 * check that since we may not have connected yet). 1525 * A is first hop destination, which doesn't appear in 1526 * actual IP option, but is stored before the options. 1527 */ 1528 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1529 goto bad; 1530 m->m_len -= sizeof(struct in_addr); 1531 cnt -= sizeof(struct in_addr); 1532 optlen -= sizeof(struct in_addr); 1533 cp[IPOPT_OLEN] = optlen; 1534 /* 1535 * Move first hop before start of options. 1536 */ 1537 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1538 sizeof(struct in_addr)); 1539 /* 1540 * Then copy rest of options back 1541 * to close up the deleted entry. 1542 */ 1543 (void)memmove(&cp[IPOPT_OFFSET+1], 1544 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1545 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1546 break; 1547 } 1548 } 1549 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1550 goto bad; 1551 *pcbopt = m; 1552 return (0); 1553 1554 bad: 1555 (void)m_free(m); 1556 return (EINVAL); 1557 } 1558 1559 /* 1560 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1561 */ 1562 static struct ifnet * 1563 ip_multicast_if(struct in_addr *a, int *ifindexp) 1564 { 1565 int ifindex; 1566 struct ifnet *ifp = NULL; 1567 struct in_ifaddr *ia; 1568 1569 if (ifindexp) 1570 *ifindexp = 0; 1571 if (ntohl(a->s_addr) >> 24 == 0) { 1572 ifindex = ntohl(a->s_addr) & 0xffffff; 1573 if (ifindex < 0 || if_indexlim <= ifindex) 1574 return NULL; 1575 ifp = ifindex2ifnet[ifindex]; 1576 if (!ifp) 1577 return NULL; 1578 if (ifindexp) 1579 *ifindexp = ifindex; 1580 } else { 1581 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1582 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1583 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1584 ifp = ia->ia_ifp; 1585 break; 1586 } 1587 } 1588 } 1589 return ifp; 1590 } 1591 1592 static int 1593 ip_getoptval(struct mbuf *m, u_int8_t *val, u_int maxval) 1594 { 1595 u_int tval; 1596 1597 if (m == NULL) 1598 return EINVAL; 1599 1600 switch (m->m_len) { 1601 case sizeof(u_char): 1602 tval = *(mtod(m, u_char *)); 1603 break; 1604 case sizeof(u_int): 1605 tval = *(mtod(m, u_int *)); 1606 break; 1607 default: 1608 return EINVAL; 1609 } 1610 1611 if (tval > maxval) 1612 return EINVAL; 1613 1614 *val = tval; 1615 return 0; 1616 } 1617 1618 /* 1619 * Set the IP multicast options in response to user setsockopt(). 1620 */ 1621 int 1622 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m) 1623 { 1624 int error = 0; 1625 int i; 1626 struct in_addr addr; 1627 struct ip_mreq *mreq; 1628 struct ifnet *ifp; 1629 struct ip_moptions *imo = *imop; 1630 struct route ro; 1631 struct sockaddr_in *dst; 1632 int ifindex; 1633 1634 if (imo == NULL) { 1635 /* 1636 * No multicast option buffer attached to the pcb; 1637 * allocate one and initialize to default values. 1638 */ 1639 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1640 M_WAITOK); 1641 1642 if (imo == NULL) 1643 return (ENOBUFS); 1644 *imop = imo; 1645 imo->imo_multicast_ifp = NULL; 1646 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1647 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1648 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1649 imo->imo_num_memberships = 0; 1650 } 1651 1652 switch (optname) { 1653 1654 case IP_MULTICAST_IF: 1655 /* 1656 * Select the interface for outgoing multicast packets. 1657 */ 1658 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1659 error = EINVAL; 1660 break; 1661 } 1662 addr = *(mtod(m, struct in_addr *)); 1663 /* 1664 * INADDR_ANY is used to remove a previous selection. 1665 * When no interface is selected, a default one is 1666 * chosen every time a multicast packet is sent. 1667 */ 1668 if (in_nullhost(addr)) { 1669 imo->imo_multicast_ifp = NULL; 1670 break; 1671 } 1672 /* 1673 * The selected interface is identified by its local 1674 * IP address. Find the interface and confirm that 1675 * it supports multicasting. 1676 */ 1677 ifp = ip_multicast_if(&addr, &ifindex); 1678 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1679 error = EADDRNOTAVAIL; 1680 break; 1681 } 1682 imo->imo_multicast_ifp = ifp; 1683 if (ifindex) 1684 imo->imo_multicast_addr = addr; 1685 else 1686 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1687 break; 1688 1689 case IP_MULTICAST_TTL: 1690 /* 1691 * Set the IP time-to-live for outgoing multicast packets. 1692 */ 1693 error = ip_getoptval(m, &imo->imo_multicast_ttl, MAXTTL); 1694 break; 1695 1696 case IP_MULTICAST_LOOP: 1697 /* 1698 * Set the loopback flag for outgoing multicast packets. 1699 * Must be zero or one. 1700 */ 1701 error = ip_getoptval(m, &imo->imo_multicast_loop, 1); 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 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1710 error = EINVAL; 1711 break; 1712 } 1713 mreq = mtod(m, struct ip_mreq *); 1714 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1715 error = EINVAL; 1716 break; 1717 } 1718 /* 1719 * If no interface address was provided, use the interface of 1720 * the route to the given multicast address. 1721 */ 1722 if (in_nullhost(mreq->imr_interface)) { 1723 bzero((caddr_t)&ro, sizeof(ro)); 1724 ro.ro_rt = NULL; 1725 dst = satosin(&ro.ro_dst); 1726 dst->sin_len = sizeof(*dst); 1727 dst->sin_family = AF_INET; 1728 dst->sin_addr = mreq->imr_multiaddr; 1729 rtalloc(&ro); 1730 if (ro.ro_rt == NULL) { 1731 error = EADDRNOTAVAIL; 1732 break; 1733 } 1734 ifp = ro.ro_rt->rt_ifp; 1735 rtfree(ro.ro_rt); 1736 } else { 1737 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1738 } 1739 /* 1740 * See if we found an interface, and confirm that it 1741 * supports multicast. 1742 */ 1743 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1744 error = EADDRNOTAVAIL; 1745 break; 1746 } 1747 /* 1748 * See if the membership already exists or if all the 1749 * membership slots are full. 1750 */ 1751 for (i = 0; i < imo->imo_num_memberships; ++i) { 1752 if (imo->imo_membership[i]->inm_ifp == ifp && 1753 in_hosteq(imo->imo_membership[i]->inm_addr, 1754 mreq->imr_multiaddr)) 1755 break; 1756 } 1757 if (i < imo->imo_num_memberships) { 1758 error = EADDRINUSE; 1759 break; 1760 } 1761 if (i == IP_MAX_MEMBERSHIPS) { 1762 error = ETOOMANYREFS; 1763 break; 1764 } 1765 /* 1766 * Everything looks good; add a new record to the multicast 1767 * address list for the given interface. 1768 */ 1769 if ((imo->imo_membership[i] = 1770 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1771 error = ENOBUFS; 1772 break; 1773 } 1774 ++imo->imo_num_memberships; 1775 break; 1776 1777 case IP_DROP_MEMBERSHIP: 1778 /* 1779 * Drop a multicast group membership. 1780 * Group must be a valid IP multicast address. 1781 */ 1782 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1783 error = EINVAL; 1784 break; 1785 } 1786 mreq = mtod(m, struct ip_mreq *); 1787 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1788 error = EINVAL; 1789 break; 1790 } 1791 /* 1792 * If an interface address was specified, get a pointer 1793 * to its ifnet structure. 1794 */ 1795 if (in_nullhost(mreq->imr_interface)) 1796 ifp = NULL; 1797 else { 1798 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1799 if (ifp == NULL) { 1800 error = EADDRNOTAVAIL; 1801 break; 1802 } 1803 } 1804 /* 1805 * Find the membership in the membership array. 1806 */ 1807 for (i = 0; i < imo->imo_num_memberships; ++i) { 1808 if ((ifp == NULL || 1809 imo->imo_membership[i]->inm_ifp == ifp) && 1810 in_hosteq(imo->imo_membership[i]->inm_addr, 1811 mreq->imr_multiaddr)) 1812 break; 1813 } 1814 if (i == imo->imo_num_memberships) { 1815 error = EADDRNOTAVAIL; 1816 break; 1817 } 1818 /* 1819 * Give up the multicast address record to which the 1820 * membership points. 1821 */ 1822 in_delmulti(imo->imo_membership[i]); 1823 /* 1824 * Remove the gap in the membership array. 1825 */ 1826 for (++i; i < imo->imo_num_memberships; ++i) 1827 imo->imo_membership[i-1] = imo->imo_membership[i]; 1828 --imo->imo_num_memberships; 1829 break; 1830 1831 default: 1832 error = EOPNOTSUPP; 1833 break; 1834 } 1835 1836 /* 1837 * If all options have default values, no need to keep the mbuf. 1838 */ 1839 if (imo->imo_multicast_ifp == NULL && 1840 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1841 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1842 imo->imo_num_memberships == 0) { 1843 free(*imop, M_IPMOPTS); 1844 *imop = NULL; 1845 } 1846 1847 return (error); 1848 } 1849 1850 /* 1851 * Return the IP multicast options in response to user getsockopt(). 1852 */ 1853 int 1854 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp) 1855 { 1856 u_char *ttl; 1857 u_char *loop; 1858 struct in_addr *addr; 1859 struct in_ifaddr *ia; 1860 1861 *mp = m_get(M_WAIT, MT_SOOPTS); 1862 1863 switch (optname) { 1864 1865 case IP_MULTICAST_IF: 1866 addr = mtod(*mp, struct in_addr *); 1867 (*mp)->m_len = sizeof(struct in_addr); 1868 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1869 *addr = zeroin_addr; 1870 else if (imo->imo_multicast_addr.s_addr) { 1871 /* return the value user has set */ 1872 *addr = imo->imo_multicast_addr; 1873 } else { 1874 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1875 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1876 } 1877 return (0); 1878 1879 case IP_MULTICAST_TTL: 1880 ttl = mtod(*mp, u_char *); 1881 (*mp)->m_len = 1; 1882 *ttl = imo ? imo->imo_multicast_ttl 1883 : IP_DEFAULT_MULTICAST_TTL; 1884 return (0); 1885 1886 case IP_MULTICAST_LOOP: 1887 loop = mtod(*mp, u_char *); 1888 (*mp)->m_len = 1; 1889 *loop = imo ? imo->imo_multicast_loop 1890 : IP_DEFAULT_MULTICAST_LOOP; 1891 return (0); 1892 1893 default: 1894 return (EOPNOTSUPP); 1895 } 1896 } 1897 1898 /* 1899 * Discard the IP multicast options. 1900 */ 1901 void 1902 ip_freemoptions(struct ip_moptions *imo) 1903 { 1904 int i; 1905 1906 if (imo != NULL) { 1907 for (i = 0; i < imo->imo_num_memberships; ++i) 1908 in_delmulti(imo->imo_membership[i]); 1909 free(imo, M_IPMOPTS); 1910 } 1911 } 1912 1913 /* 1914 * Routine called from ip_output() to loop back a copy of an IP multicast 1915 * packet to the input queue of a specified interface. Note that this 1916 * calls the output routine of the loopback "driver", but with an interface 1917 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1918 */ 1919 static void 1920 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst) 1921 { 1922 struct ip *ip; 1923 struct mbuf *copym; 1924 1925 copym = m_copy(m, 0, M_COPYALL); 1926 if (copym != NULL 1927 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1928 copym = m_pullup(copym, sizeof(struct ip)); 1929 if (copym != NULL) { 1930 /* 1931 * We don't bother to fragment if the IP length is greater 1932 * than the interface's MTU. Can this possibly matter? 1933 */ 1934 ip = mtod(copym, struct ip *); 1935 1936 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1937 in_delayed_cksum(copym); 1938 copym->m_pkthdr.csum_flags &= 1939 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1940 } 1941 1942 ip->ip_sum = 0; 1943 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1944 (void) looutput(ifp, copym, sintosa(dst), NULL); 1945 } 1946 } 1947
Indexes created Wed Oct 01 12:09:54 GMT 2025