1 /* $NetBSD: ip_output.c,v 1.205 2009/07/17 22:02:54 minskim 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 * 49 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 50 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 51 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 52 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 53 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 54 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 55 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 56 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 57 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 58 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 59 * POSSIBILITY OF SUCH DAMAGE. 60 */ 61 62 /* 63 * Copyright (c) 1982, 1986, 1988, 1990, 1993 64 * The Regents of the University of California. All rights reserved. 65 * 66 * Redistribution and use in source and binary forms, with or without 67 * modification, are permitted provided that the following conditions 68 * are met: 69 * 1. Redistributions of source code must retain the above copyright 70 * notice, this list of conditions and the following disclaimer. 71 * 2. Redistributions in binary form must reproduce the above copyright 72 * notice, this list of conditions and the following disclaimer in the 73 * documentation and/or other materials provided with the distribution. 74 * 3. Neither the name of the University nor the names of its contributors 75 * may be used to endorse or promote products derived from this software 76 * without specific prior written permission. 77 * 78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 88 * SUCH DAMAGE. 89 * 90 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 91 */ 92 93 #include <sys/cdefs.h> 94 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.205 2009/07/17 22:02:54 minskim Exp $"); 95 96 #include "opt_pfil_hooks.h" 97 #include "opt_inet.h" 98 #include "opt_ipsec.h" 99 #include "opt_mrouting.h" 100 101 #include <sys/param.h> 102 #include <sys/malloc.h> 103 #include <sys/mbuf.h> 104 #include <sys/errno.h> 105 #include <sys/protosw.h> 106 #include <sys/socket.h> 107 #include <sys/socketvar.h> 108 #include <sys/kauth.h> 109 #ifdef FAST_IPSEC 110 #include <sys/domain.h> 111 #endif 112 #include <sys/systm.h> 113 #include <sys/proc.h> 114 115 #include <net/if.h> 116 #include <net/route.h> 117 #include <net/pfil.h> 118 119 #include <netinet/in.h> 120 #include <netinet/in_systm.h> 121 #include <netinet/ip.h> 122 #include <netinet/in_pcb.h> 123 #include <netinet/in_var.h> 124 #include <netinet/ip_var.h> 125 #include <netinet/ip_private.h> 126 #include <netinet/in_offload.h> 127 128 #ifdef MROUTING 129 #include <netinet/ip_mroute.h> 130 #endif 131 132 #include <machine/stdarg.h> 133 134 #ifdef IPSEC 135 #include <netinet6/ipsec.h> 136 #include <netinet6/ipsec_private.h> 137 #include <netkey/key.h> 138 #include <netkey/key_debug.h> 139 #endif /*IPSEC*/ 140 141 #ifdef FAST_IPSEC 142 #include <netipsec/ipsec.h> 143 #include <netipsec/key.h> 144 #include <netipsec/xform.h> 145 #endif /* FAST_IPSEC*/ 146 147 #ifdef IPSEC_NAT_T 148 #include <netinet/udp.h> 149 #endif 150 151 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 152 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 153 static void ip_mloopback(struct ifnet *, struct mbuf *, 154 const struct sockaddr_in *); 155 156 #ifdef PFIL_HOOKS 157 extern struct pfil_head inet_pfil_hook; /* XXX */ 158 #endif 159 160 int ip_do_loopback_cksum = 0; 161 162 /* 163 * IP output. The packet in mbuf chain m contains a skeletal IP 164 * header (with len, off, ttl, proto, tos, src, dst). 165 * The mbuf chain containing the packet will be freed. 166 * The mbuf opt, if present, will not be freed. 167 */ 168 int 169 ip_output(struct mbuf *m0, ...) 170 { 171 struct rtentry *rt; 172 struct ip *ip; 173 struct ifnet *ifp; 174 struct mbuf *m = m0; 175 int hlen = sizeof (struct ip); 176 int len, error = 0; 177 struct route iproute; 178 const struct sockaddr_in *dst; 179 struct in_ifaddr *ia; 180 struct ifaddr *xifa; 181 struct mbuf *opt; 182 struct route *ro; 183 int flags, sw_csum; 184 int *mtu_p; 185 u_long mtu; 186 struct ip_moptions *imo; 187 struct socket *so; 188 va_list ap; 189 #ifdef IPSEC_NAT_T 190 int natt_frag = 0; 191 #endif 192 #ifdef IPSEC 193 struct secpolicy *sp = NULL; 194 #endif /*IPSEC*/ 195 #ifdef FAST_IPSEC 196 struct inpcb *inp; 197 struct secpolicy *sp = NULL; 198 int s; 199 #endif 200 u_int16_t ip_len; 201 union { 202 struct sockaddr dst; 203 struct sockaddr_in dst4; 204 } u; 205 struct sockaddr *rdst = &u.dst; /* real IP destination, as opposed 206 * to the nexthop 207 */ 208 209 len = 0; 210 va_start(ap, m0); 211 opt = va_arg(ap, struct mbuf *); 212 ro = va_arg(ap, struct route *); 213 flags = va_arg(ap, int); 214 imo = va_arg(ap, struct ip_moptions *); 215 so = va_arg(ap, struct socket *); 216 if (flags & IP_RETURNMTU) 217 mtu_p = va_arg(ap, int *); 218 else 219 mtu_p = NULL; 220 va_end(ap); 221 222 MCLAIM(m, &ip_tx_mowner); 223 #ifdef FAST_IPSEC 224 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 225 inp = (struct inpcb *)so->so_pcb; 226 else 227 inp = NULL; 228 #endif /* FAST_IPSEC */ 229 230 #ifdef DIAGNOSTIC 231 if ((m->m_flags & M_PKTHDR) == 0) 232 panic("ip_output: no HDR"); 233 234 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) != 0) { 235 panic("ip_output: IPv6 checksum offload flags: %d", 236 m->m_pkthdr.csum_flags); 237 } 238 239 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) == 240 (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 241 panic("ip_output: conflicting checksum offload flags: %d", 242 m->m_pkthdr.csum_flags); 243 } 244 #endif 245 if (opt) { 246 m = ip_insertoptions(m, opt, &len); 247 if (len >= sizeof(struct ip)) 248 hlen = len; 249 } 250 ip = mtod(m, struct ip *); 251 /* 252 * Fill in IP header. 253 */ 254 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 255 ip->ip_v = IPVERSION; 256 ip->ip_off = htons(0); 257 /* ip->ip_id filled in after we find out source ia */ 258 ip->ip_hl = hlen >> 2; 259 IP_STATINC(IP_STAT_LOCALOUT); 260 } else { 261 hlen = ip->ip_hl << 2; 262 } 263 /* 264 * Route packet. 265 */ 266 memset(&iproute, 0, sizeof(iproute)); 267 if (ro == NULL) 268 ro = &iproute; 269 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 270 dst = satocsin(rtcache_getdst(ro)); 271 /* 272 * If there is a cached route, 273 * check that it is to the same destination 274 * and is still up. If not, free it and try again. 275 * The address family should also be checked in case of sharing the 276 * cache with IPv6. 277 */ 278 if (dst == NULL) 279 ; 280 else if (dst->sin_family != AF_INET || 281 !in_hosteq(dst->sin_addr, ip->ip_dst)) 282 rtcache_free(ro); 283 284 if ((rt = rtcache_validate(ro)) == NULL && 285 (rt = rtcache_update(ro, 1)) == NULL) { 286 dst = &u.dst4; 287 rtcache_setdst(ro, &u.dst); 288 } 289 /* 290 * If routing to interface only, 291 * short circuit routing lookup. 292 */ 293 if (flags & IP_ROUTETOIF) { 294 if ((ia = ifatoia(ifa_ifwithladdr(sintocsa(dst)))) == NULL) { 295 IP_STATINC(IP_STAT_NOROUTE); 296 error = ENETUNREACH; 297 goto bad; 298 } 299 ifp = ia->ia_ifp; 300 mtu = ifp->if_mtu; 301 ip->ip_ttl = 1; 302 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 303 ip->ip_dst.s_addr == INADDR_BROADCAST) && 304 imo != NULL && imo->imo_multicast_ifp != NULL) { 305 ifp = imo->imo_multicast_ifp; 306 mtu = ifp->if_mtu; 307 IFP_TO_IA(ifp, ia); 308 } else { 309 if (rt == NULL) 310 rt = rtcache_init(ro); 311 if (rt == NULL) { 312 IP_STATINC(IP_STAT_NOROUTE); 313 error = EHOSTUNREACH; 314 goto bad; 315 } 316 ia = ifatoia(rt->rt_ifa); 317 ifp = rt->rt_ifp; 318 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 319 mtu = ifp->if_mtu; 320 rt->rt_use++; 321 if (rt->rt_flags & RTF_GATEWAY) 322 dst = satosin(rt->rt_gateway); 323 } 324 if (IN_MULTICAST(ip->ip_dst.s_addr) || 325 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 326 struct in_multi *inm; 327 328 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 329 M_BCAST : M_MCAST; 330 /* 331 * See if the caller provided any multicast options 332 */ 333 if (imo != NULL) 334 ip->ip_ttl = imo->imo_multicast_ttl; 335 else 336 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 337 338 /* 339 * if we don't know the outgoing ifp yet, we can't generate 340 * output 341 */ 342 if (!ifp) { 343 IP_STATINC(IP_STAT_NOROUTE); 344 error = ENETUNREACH; 345 goto bad; 346 } 347 348 /* 349 * If the packet is multicast or broadcast, confirm that 350 * the outgoing interface can transmit it. 351 */ 352 if (((m->m_flags & M_MCAST) && 353 (ifp->if_flags & IFF_MULTICAST) == 0) || 354 ((m->m_flags & M_BCAST) && 355 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 356 IP_STATINC(IP_STAT_NOROUTE); 357 error = ENETUNREACH; 358 goto bad; 359 } 360 /* 361 * If source address not specified yet, use an address 362 * of outgoing interface. 363 */ 364 if (in_nullhost(ip->ip_src)) { 365 struct in_ifaddr *xia; 366 367 IFP_TO_IA(ifp, xia); 368 if (!xia) { 369 error = EADDRNOTAVAIL; 370 goto bad; 371 } 372 xifa = &xia->ia_ifa; 373 if (xifa->ifa_getifa != NULL) { 374 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 375 } 376 ip->ip_src = xia->ia_addr.sin_addr; 377 } 378 379 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 380 if (inm != NULL && 381 (imo == NULL || imo->imo_multicast_loop)) { 382 /* 383 * If we belong to the destination multicast group 384 * on the outgoing interface, and the caller did not 385 * forbid loopback, loop back a copy. 386 */ 387 ip_mloopback(ifp, m, &u.dst4); 388 } 389 #ifdef MROUTING 390 else { 391 /* 392 * If we are acting as a multicast router, perform 393 * multicast forwarding as if the packet had just 394 * arrived on the interface to which we are about 395 * to send. The multicast forwarding function 396 * recursively calls this function, using the 397 * IP_FORWARDING flag to prevent infinite recursion. 398 * 399 * Multicasts that are looped back by ip_mloopback(), 400 * above, will be forwarded by the ip_input() routine, 401 * if necessary. 402 */ 403 extern struct socket *ip_mrouter; 404 405 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 406 if (ip_mforward(m, ifp) != 0) { 407 m_freem(m); 408 goto done; 409 } 410 } 411 } 412 #endif 413 /* 414 * Multicasts with a time-to-live of zero may be looped- 415 * back, above, but must not be transmitted on a network. 416 * Also, multicasts addressed to the loopback interface 417 * are not sent -- the above call to ip_mloopback() will 418 * loop back a copy if this host actually belongs to the 419 * destination group on the loopback interface. 420 */ 421 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 422 m_freem(m); 423 goto done; 424 } 425 426 goto sendit; 427 } 428 /* 429 * If source address not specified yet, use address 430 * of outgoing interface. 431 */ 432 if (in_nullhost(ip->ip_src)) { 433 xifa = &ia->ia_ifa; 434 if (xifa->ifa_getifa != NULL) 435 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 436 ip->ip_src = ia->ia_addr.sin_addr; 437 } 438 439 /* 440 * packets with Class-D address as source are not valid per 441 * RFC 1112 442 */ 443 if (IN_MULTICAST(ip->ip_src.s_addr)) { 444 IP_STATINC(IP_STAT_ODROPPED); 445 error = EADDRNOTAVAIL; 446 goto bad; 447 } 448 449 /* 450 * Look for broadcast address and 451 * and verify user is allowed to send 452 * such a packet. 453 */ 454 if (in_broadcast(dst->sin_addr, ifp)) { 455 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 456 error = EADDRNOTAVAIL; 457 goto bad; 458 } 459 if ((flags & IP_ALLOWBROADCAST) == 0) { 460 error = EACCES; 461 goto bad; 462 } 463 /* don't allow broadcast messages to be fragmented */ 464 if (ntohs(ip->ip_len) > ifp->if_mtu) { 465 error = EMSGSIZE; 466 goto bad; 467 } 468 m->m_flags |= M_BCAST; 469 } else 470 m->m_flags &= ~M_BCAST; 471 472 sendit: 473 if ((flags & (IP_FORWARDING|IP_NOIPNEWID)) == 0) { 474 if (m->m_pkthdr.len < IP_MINFRAGSIZE) { 475 ip->ip_id = 0; 476 } else if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 477 ip->ip_id = ip_newid(ia); 478 } else { 479 480 /* 481 * TSO capable interfaces (typically?) increment 482 * ip_id for each segment. 483 * "allocate" enough ids here to increase the chance 484 * for them to be unique. 485 * 486 * note that the following calculation is not 487 * needed to be precise. wasting some ip_id is fine. 488 */ 489 490 unsigned int segsz = m->m_pkthdr.segsz; 491 unsigned int datasz = ntohs(ip->ip_len) - hlen; 492 unsigned int num = howmany(datasz, segsz); 493 494 ip->ip_id = ip_newid_range(ia, num); 495 } 496 } 497 /* 498 * If we're doing Path MTU Discovery, we need to set DF unless 499 * the route's MTU is locked. 500 */ 501 if ((flags & IP_MTUDISC) != 0 && rt != NULL && 502 (rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 503 ip->ip_off |= htons(IP_DF); 504 505 /* Remember the current ip_len */ 506 ip_len = ntohs(ip->ip_len); 507 508 #ifdef IPSEC 509 /* get SP for this packet */ 510 if (so == NULL) 511 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 512 flags, &error); 513 else { 514 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 515 IPSEC_DIR_OUTBOUND)) 516 goto skip_ipsec; 517 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 518 } 519 520 if (sp == NULL) { 521 IPSEC_STATINC(IPSEC_STAT_IN_INVAL); 522 goto bad; 523 } 524 525 error = 0; 526 527 /* check policy */ 528 switch (sp->policy) { 529 case IPSEC_POLICY_DISCARD: 530 /* 531 * This packet is just discarded. 532 */ 533 IPSEC_STATINC(IPSEC_STAT_OUT_POLVIO); 534 goto bad; 535 536 case IPSEC_POLICY_BYPASS: 537 case IPSEC_POLICY_NONE: 538 /* no need to do IPsec. */ 539 goto skip_ipsec; 540 541 case IPSEC_POLICY_IPSEC: 542 if (sp->req == NULL) { 543 /* XXX should be panic ? */ 544 printf("ip_output: No IPsec request specified.\n"); 545 error = EINVAL; 546 goto bad; 547 } 548 break; 549 550 case IPSEC_POLICY_ENTRUST: 551 default: 552 printf("ip_output: Invalid policy found. %d\n", sp->policy); 553 } 554 555 #ifdef IPSEC_NAT_T 556 /* 557 * NAT-T ESP fragmentation: don't do IPSec processing now, 558 * we'll do it on each fragmented packet. 559 */ 560 if (sp->req->sav && 561 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 562 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 563 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 564 natt_frag = 1; 565 mtu = sp->req->sav->esp_frag; 566 goto skip_ipsec; 567 } 568 } 569 #endif /* IPSEC_NAT_T */ 570 571 /* 572 * ipsec4_output() expects ip_len and ip_off in network 573 * order. They have been set to network order above. 574 */ 575 576 { 577 struct ipsec_output_state state; 578 memset(&state, 0, sizeof(state)); 579 state.m = m; 580 if (flags & IP_ROUTETOIF) { 581 state.ro = &iproute; 582 memset(&iproute, 0, sizeof(iproute)); 583 } else 584 state.ro = ro; 585 state.dst = sintocsa(dst); 586 587 /* 588 * We can't defer the checksum of payload data if 589 * we're about to encrypt/authenticate it. 590 * 591 * XXX When we support crypto offloading functions of 592 * XXX network interfaces, we need to reconsider this, 593 * XXX since it's likely that they'll support checksumming, 594 * XXX as well. 595 */ 596 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 597 in_delayed_cksum(m); 598 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 599 } 600 601 error = ipsec4_output(&state, sp, flags); 602 603 m = state.m; 604 if (flags & IP_ROUTETOIF) { 605 /* 606 * if we have tunnel mode SA, we may need to ignore 607 * IP_ROUTETOIF. 608 */ 609 if (state.ro != &iproute || 610 rtcache_validate(state.ro) != NULL) { 611 flags &= ~IP_ROUTETOIF; 612 ro = state.ro; 613 } 614 } else 615 ro = state.ro; 616 dst = satocsin(state.dst); 617 if (error) { 618 /* mbuf is already reclaimed in ipsec4_output. */ 619 m0 = NULL; 620 switch (error) { 621 case EHOSTUNREACH: 622 case ENETUNREACH: 623 case EMSGSIZE: 624 case ENOBUFS: 625 case ENOMEM: 626 break; 627 default: 628 printf("ip4_output (ipsec): error code %d\n", error); 629 /*fall through*/ 630 case ENOENT: 631 /* don't show these error codes to the user */ 632 error = 0; 633 break; 634 } 635 goto bad; 636 } 637 638 /* be sure to update variables that are affected by ipsec4_output() */ 639 ip = mtod(m, struct ip *); 640 hlen = ip->ip_hl << 2; 641 ip_len = ntohs(ip->ip_len); 642 643 if ((rt = rtcache_validate(ro)) == NULL) { 644 if ((flags & IP_ROUTETOIF) == 0) { 645 printf("ip_output: " 646 "can't update route after IPsec processing\n"); 647 error = EHOSTUNREACH; /*XXX*/ 648 goto bad; 649 } 650 } else { 651 /* nobody uses ia beyond here */ 652 if (state.encap) { 653 ifp = rt->rt_ifp; 654 if ((mtu = rt->rt_rmx.rmx_mtu) == 0) 655 mtu = ifp->if_mtu; 656 } 657 } 658 } 659 skip_ipsec: 660 #endif /*IPSEC*/ 661 #ifdef FAST_IPSEC 662 /* 663 * Check the security policy (SP) for the packet and, if 664 * required, do IPsec-related processing. There are two 665 * cases here; the first time a packet is sent through 666 * it will be untagged and handled by ipsec4_checkpolicy. 667 * If the packet is resubmitted to ip_output (e.g. after 668 * AH, ESP, etc. processing), there will be a tag to bypass 669 * the lookup and related policy checking. 670 */ 671 if (!ipsec_outdone(m)) { 672 s = splsoftnet(); 673 if (inp != NULL && 674 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) { 675 splx(s); 676 goto spd_done; 677 } 678 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 679 &error, inp); 680 /* 681 * There are four return cases: 682 * sp != NULL apply IPsec policy 683 * sp == NULL, error == 0 no IPsec handling needed 684 * sp == NULL, error == -EINVAL discard packet w/o error 685 * sp == NULL, error != 0 discard packet, report error 686 */ 687 if (sp != NULL) { 688 #ifdef IPSEC_NAT_T 689 /* 690 * NAT-T ESP fragmentation: don't do IPSec processing now, 691 * we'll do it on each fragmented packet. 692 */ 693 if (sp->req->sav && 694 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 695 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 696 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 697 natt_frag = 1; 698 mtu = sp->req->sav->esp_frag; 699 splx(s); 700 goto spd_done; 701 } 702 } 703 #endif /* IPSEC_NAT_T */ 704 705 /* 706 * Do delayed checksums now because we send before 707 * this is done in the normal processing path. 708 */ 709 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 710 in_delayed_cksum(m); 711 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 712 } 713 714 #ifdef __FreeBSD__ 715 ip->ip_len = htons(ip->ip_len); 716 ip->ip_off = htons(ip->ip_off); 717 #endif 718 719 /* NB: callee frees mbuf */ 720 error = ipsec4_process_packet(m, sp->req, flags, 0); 721 /* 722 * Preserve KAME behaviour: ENOENT can be returned 723 * when an SA acquire is in progress. Don't propagate 724 * this to user-level; it confuses applications. 725 * 726 * XXX this will go away when the SADB is redone. 727 */ 728 if (error == ENOENT) 729 error = 0; 730 splx(s); 731 goto done; 732 } else { 733 splx(s); 734 735 if (error != 0) { 736 /* 737 * Hack: -EINVAL is used to signal that a packet 738 * should be silently discarded. This is typically 739 * because we asked key management for an SA and 740 * it was delayed (e.g. kicked up to IKE). 741 */ 742 if (error == -EINVAL) 743 error = 0; 744 goto bad; 745 } else { 746 /* No IPsec processing for this packet. */ 747 } 748 } 749 } 750 spd_done: 751 #endif /* FAST_IPSEC */ 752 753 #ifdef PFIL_HOOKS 754 /* 755 * Run through list of hooks for output packets. 756 */ 757 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 758 goto done; 759 if (m == NULL) 760 goto done; 761 762 ip = mtod(m, struct ip *); 763 hlen = ip->ip_hl << 2; 764 ip_len = ntohs(ip->ip_len); 765 #endif /* PFIL_HOOKS */ 766 767 m->m_pkthdr.csum_data |= hlen << 16; 768 769 #if IFA_STATS 770 /* 771 * search for the source address structure to 772 * maintain output statistics. 773 */ 774 INADDR_TO_IA(ip->ip_src, ia); 775 #endif 776 777 /* Maybe skip checksums on loopback interfaces. */ 778 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 779 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 780 } 781 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 782 /* 783 * If small enough for mtu of path, or if using TCP segmentation 784 * offload, can just send directly. 785 */ 786 if (ip_len <= mtu || 787 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 788 #if IFA_STATS 789 if (ia) 790 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 791 #endif 792 /* 793 * Always initialize the sum to 0! Some HW assisted 794 * checksumming requires this. 795 */ 796 ip->ip_sum = 0; 797 798 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 799 /* 800 * Perform any checksums that the hardware can't do 801 * for us. 802 * 803 * XXX Does any hardware require the {th,uh}_sum 804 * XXX fields to be 0? 805 */ 806 if (sw_csum & M_CSUM_IPv4) { 807 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 808 ip->ip_sum = in_cksum(m, hlen); 809 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 810 } 811 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 812 if (IN_NEED_CHECKSUM(ifp, 813 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 814 in_delayed_cksum(m); 815 } 816 m->m_pkthdr.csum_flags &= 817 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 818 } 819 } 820 821 #ifdef IPSEC 822 /* clean ipsec history once it goes out of the node */ 823 ipsec_delaux(m); 824 #endif 825 826 if (__predict_true( 827 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 828 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 829 error = 830 (*ifp->if_output)(ifp, m, 831 (m->m_flags & M_MCAST) ? 832 sintocsa(rdst) : sintocsa(dst), 833 rt); 834 } else { 835 error = 836 ip_tso_output(ifp, m, 837 (m->m_flags & M_MCAST) ? 838 sintocsa(rdst) : sintocsa(dst), 839 rt); 840 } 841 goto done; 842 } 843 844 /* 845 * We can't use HW checksumming if we're about to 846 * to fragment the packet. 847 * 848 * XXX Some hardware can do this. 849 */ 850 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 851 if (IN_NEED_CHECKSUM(ifp, 852 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 853 in_delayed_cksum(m); 854 } 855 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 856 } 857 858 /* 859 * Too large for interface; fragment if possible. 860 * Must be able to put at least 8 bytes per fragment. 861 */ 862 if (ntohs(ip->ip_off) & IP_DF) { 863 if (flags & IP_RETURNMTU) 864 *mtu_p = mtu; 865 error = EMSGSIZE; 866 IP_STATINC(IP_STAT_CANTFRAG); 867 goto bad; 868 } 869 870 error = ip_fragment(m, ifp, mtu); 871 if (error) { 872 m = NULL; 873 goto bad; 874 } 875 876 for (; m; m = m0) { 877 m0 = m->m_nextpkt; 878 m->m_nextpkt = 0; 879 if (error == 0) { 880 #if IFA_STATS 881 if (ia) 882 ia->ia_ifa.ifa_data.ifad_outbytes += 883 ntohs(ip->ip_len); 884 #endif 885 #ifdef IPSEC 886 /* clean ipsec history once it goes out of the node */ 887 ipsec_delaux(m); 888 #endif /* IPSEC */ 889 890 #ifdef IPSEC_NAT_T 891 /* 892 * If we get there, the packet has not been handeld by 893 * IPSec whereas it should have. Now that it has been 894 * fragmented, re-inject it in ip_output so that IPsec 895 * processing can occur. 896 */ 897 if (natt_frag) { 898 error = ip_output(m, opt, 899 ro, flags, imo, so, mtu_p); 900 } else 901 #endif /* IPSEC_NAT_T */ 902 { 903 KASSERT((m->m_pkthdr.csum_flags & 904 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 905 error = (*ifp->if_output)(ifp, m, 906 (m->m_flags & M_MCAST) ? 907 sintocsa(rdst) : sintocsa(dst), 908 rt); 909 } 910 } else 911 m_freem(m); 912 } 913 914 if (error == 0) 915 IP_STATINC(IP_STAT_FRAGMENTED); 916 done: 917 rtcache_free(&iproute); 918 919 #ifdef IPSEC 920 if (sp != NULL) { 921 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 922 printf("DP ip_output call free SP:%p\n", sp)); 923 key_freesp(sp); 924 } 925 #endif /* IPSEC */ 926 #ifdef FAST_IPSEC 927 if (sp != NULL) 928 KEY_FREESP(&sp); 929 #endif /* FAST_IPSEC */ 930 931 return (error); 932 bad: 933 m_freem(m); 934 goto done; 935 } 936 937 int 938 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 939 { 940 struct ip *ip, *mhip; 941 struct mbuf *m0; 942 int len, hlen, off; 943 int mhlen, firstlen; 944 struct mbuf **mnext; 945 int sw_csum = m->m_pkthdr.csum_flags; 946 int fragments = 0; 947 int s; 948 int error = 0; 949 950 ip = mtod(m, struct ip *); 951 hlen = ip->ip_hl << 2; 952 if (ifp != NULL) 953 sw_csum &= ~ifp->if_csum_flags_tx; 954 955 len = (mtu - hlen) &~ 7; 956 if (len < 8) { 957 m_freem(m); 958 return (EMSGSIZE); 959 } 960 961 firstlen = len; 962 mnext = &m->m_nextpkt; 963 964 /* 965 * Loop through length of segment after first fragment, 966 * make new header and copy data of each part and link onto chain. 967 */ 968 m0 = m; 969 mhlen = sizeof (struct ip); 970 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 971 MGETHDR(m, M_DONTWAIT, MT_HEADER); 972 if (m == 0) { 973 error = ENOBUFS; 974 IP_STATINC(IP_STAT_ODROPPED); 975 goto sendorfree; 976 } 977 MCLAIM(m, m0->m_owner); 978 *mnext = m; 979 mnext = &m->m_nextpkt; 980 m->m_data += max_linkhdr; 981 mhip = mtod(m, struct ip *); 982 *mhip = *ip; 983 /* we must inherit MCAST and BCAST flags */ 984 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 985 if (hlen > sizeof (struct ip)) { 986 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 987 mhip->ip_hl = mhlen >> 2; 988 } 989 m->m_len = mhlen; 990 mhip->ip_off = ((off - hlen) >> 3) + 991 (ntohs(ip->ip_off) & ~IP_MF); 992 if (ip->ip_off & htons(IP_MF)) 993 mhip->ip_off |= IP_MF; 994 if (off + len >= ntohs(ip->ip_len)) 995 len = ntohs(ip->ip_len) - off; 996 else 997 mhip->ip_off |= IP_MF; 998 HTONS(mhip->ip_off); 999 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 1000 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 1001 if (m->m_next == 0) { 1002 error = ENOBUFS; /* ??? */ 1003 IP_STATINC(IP_STAT_ODROPPED); 1004 goto sendorfree; 1005 } 1006 m->m_pkthdr.len = mhlen + len; 1007 m->m_pkthdr.rcvif = (struct ifnet *)0; 1008 mhip->ip_sum = 0; 1009 if (sw_csum & M_CSUM_IPv4) { 1010 mhip->ip_sum = in_cksum(m, mhlen); 1011 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 1012 } else { 1013 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 1014 m->m_pkthdr.csum_data |= mhlen << 16; 1015 } 1016 IP_STATINC(IP_STAT_OFRAGMENTS); 1017 fragments++; 1018 } 1019 /* 1020 * Update first fragment by trimming what's been copied out 1021 * and updating header, then send each fragment (in order). 1022 */ 1023 m = m0; 1024 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1025 m->m_pkthdr.len = hlen + firstlen; 1026 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1027 ip->ip_off |= htons(IP_MF); 1028 ip->ip_sum = 0; 1029 if (sw_csum & M_CSUM_IPv4) { 1030 ip->ip_sum = in_cksum(m, hlen); 1031 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1032 } else { 1033 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1034 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 1035 sizeof(struct ip)); 1036 } 1037 sendorfree: 1038 /* 1039 * If there is no room for all the fragments, don't queue 1040 * any of them. 1041 */ 1042 if (ifp != NULL) { 1043 s = splnet(); 1044 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1045 error == 0) { 1046 error = ENOBUFS; 1047 IP_STATINC(IP_STAT_ODROPPED); 1048 IFQ_INC_DROPS(&ifp->if_snd); 1049 } 1050 splx(s); 1051 } 1052 if (error) { 1053 for (m = m0; m; m = m0) { 1054 m0 = m->m_nextpkt; 1055 m->m_nextpkt = NULL; 1056 m_freem(m); 1057 } 1058 } 1059 return (error); 1060 } 1061 1062 /* 1063 * Process a delayed payload checksum calculation. 1064 */ 1065 void 1066 in_delayed_cksum(struct mbuf *m) 1067 { 1068 struct ip *ip; 1069 u_int16_t csum, offset; 1070 1071 ip = mtod(m, struct ip *); 1072 offset = ip->ip_hl << 2; 1073 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1074 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1075 csum = 0xffff; 1076 1077 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 1078 1079 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1080 /* This happen when ip options were inserted 1081 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1082 m->m_len, offset, ip->ip_p); 1083 */ 1084 m_copyback(m, offset, sizeof(csum), (void *) &csum); 1085 } else 1086 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 1087 } 1088 1089 /* 1090 * Determine the maximum length of the options to be inserted; 1091 * we would far rather allocate too much space rather than too little. 1092 */ 1093 1094 u_int 1095 ip_optlen(struct inpcb *inp) 1096 { 1097 struct mbuf *m = inp->inp_options; 1098 1099 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1100 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1101 else 1102 return 0; 1103 } 1104 1105 1106 /* 1107 * Insert IP options into preformed packet. 1108 * Adjust IP destination as required for IP source routing, 1109 * as indicated by a non-zero in_addr at the start of the options. 1110 */ 1111 static struct mbuf * 1112 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1113 { 1114 struct ipoption *p = mtod(opt, struct ipoption *); 1115 struct mbuf *n; 1116 struct ip *ip = mtod(m, struct ip *); 1117 unsigned optlen; 1118 1119 optlen = opt->m_len - sizeof(p->ipopt_dst); 1120 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1121 return (m); /* XXX should fail */ 1122 if (!in_nullhost(p->ipopt_dst)) 1123 ip->ip_dst = p->ipopt_dst; 1124 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1125 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1126 if (n == 0) 1127 return (m); 1128 MCLAIM(n, m->m_owner); 1129 M_MOVE_PKTHDR(n, m); 1130 m->m_len -= sizeof(struct ip); 1131 m->m_data += sizeof(struct ip); 1132 n->m_next = m; 1133 m = n; 1134 m->m_len = optlen + sizeof(struct ip); 1135 m->m_data += max_linkhdr; 1136 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 1137 } else { 1138 m->m_data -= optlen; 1139 m->m_len += optlen; 1140 memmove(mtod(m, void *), ip, sizeof(struct ip)); 1141 } 1142 m->m_pkthdr.len += optlen; 1143 ip = mtod(m, struct ip *); 1144 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 1145 *phlen = sizeof(struct ip) + optlen; 1146 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1147 return (m); 1148 } 1149 1150 /* 1151 * Copy options from ip to jp, 1152 * omitting those not copied during fragmentation. 1153 */ 1154 int 1155 ip_optcopy(struct ip *ip, struct ip *jp) 1156 { 1157 u_char *cp, *dp; 1158 int opt, optlen, cnt; 1159 1160 cp = (u_char *)(ip + 1); 1161 dp = (u_char *)(jp + 1); 1162 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1163 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1164 opt = cp[0]; 1165 if (opt == IPOPT_EOL) 1166 break; 1167 if (opt == IPOPT_NOP) { 1168 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1169 *dp++ = IPOPT_NOP; 1170 optlen = 1; 1171 continue; 1172 } 1173 #ifdef DIAGNOSTIC 1174 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1175 panic("malformed IPv4 option passed to ip_optcopy"); 1176 #endif 1177 optlen = cp[IPOPT_OLEN]; 1178 #ifdef DIAGNOSTIC 1179 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1180 panic("malformed IPv4 option passed to ip_optcopy"); 1181 #endif 1182 /* bogus lengths should have been caught by ip_dooptions */ 1183 if (optlen > cnt) 1184 optlen = cnt; 1185 if (IPOPT_COPIED(opt)) { 1186 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1187 dp += optlen; 1188 } 1189 } 1190 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1191 *dp++ = IPOPT_EOL; 1192 return (optlen); 1193 } 1194 1195 /* 1196 * IP socket option processing. 1197 */ 1198 int 1199 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1200 { 1201 struct inpcb *inp = sotoinpcb(so); 1202 int optval = 0; 1203 int error = 0; 1204 #if defined(IPSEC) || defined(FAST_IPSEC) 1205 struct lwp *l = curlwp; /*XXX*/ 1206 #endif 1207 1208 if (sopt->sopt_level != IPPROTO_IP) { 1209 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 1210 return 0; 1211 return ENOPROTOOPT; 1212 } 1213 1214 switch (op) { 1215 case PRCO_SETOPT: 1216 switch (sopt->sopt_name) { 1217 case IP_OPTIONS: 1218 #ifdef notyet 1219 case IP_RETOPTS: 1220 #endif 1221 error = ip_pcbopts(&inp->inp_options, sopt); 1222 break; 1223 1224 case IP_TOS: 1225 case IP_TTL: 1226 case IP_MINTTL: 1227 case IP_RECVOPTS: 1228 case IP_RECVRETOPTS: 1229 case IP_RECVDSTADDR: 1230 case IP_RECVIF: 1231 case IP_RECVTTL: 1232 error = sockopt_getint(sopt, &optval); 1233 if (error) 1234 break; 1235 1236 switch (sopt->sopt_name) { 1237 case IP_TOS: 1238 inp->inp_ip.ip_tos = optval; 1239 break; 1240 1241 case IP_TTL: 1242 inp->inp_ip.ip_ttl = optval; 1243 break; 1244 1245 case IP_MINTTL: 1246 if (optval > 0 && optval <= MAXTTL) 1247 inp->inp_ip_minttl = optval; 1248 else 1249 error = EINVAL; 1250 break; 1251 #define OPTSET(bit) \ 1252 if (optval) \ 1253 inp->inp_flags |= bit; \ 1254 else \ 1255 inp->inp_flags &= ~bit; 1256 1257 case IP_RECVOPTS: 1258 OPTSET(INP_RECVOPTS); 1259 break; 1260 1261 case IP_RECVRETOPTS: 1262 OPTSET(INP_RECVRETOPTS); 1263 break; 1264 1265 case IP_RECVDSTADDR: 1266 OPTSET(INP_RECVDSTADDR); 1267 break; 1268 1269 case IP_RECVIF: 1270 OPTSET(INP_RECVIF); 1271 break; 1272 1273 case IP_RECVTTL: 1274 OPTSET(INP_RECVTTL); 1275 break; 1276 } 1277 break; 1278 #undef OPTSET 1279 1280 case IP_MULTICAST_IF: 1281 case IP_MULTICAST_TTL: 1282 case IP_MULTICAST_LOOP: 1283 case IP_ADD_MEMBERSHIP: 1284 case IP_DROP_MEMBERSHIP: 1285 error = ip_setmoptions(&inp->inp_moptions, sopt); 1286 break; 1287 1288 case IP_PORTRANGE: 1289 error = sockopt_getint(sopt, &optval); 1290 if (error) 1291 break; 1292 1293 /* INP_LOCK(inp); */ 1294 switch (optval) { 1295 case IP_PORTRANGE_DEFAULT: 1296 case IP_PORTRANGE_HIGH: 1297 inp->inp_flags &= ~(INP_LOWPORT); 1298 break; 1299 1300 case IP_PORTRANGE_LOW: 1301 inp->inp_flags |= INP_LOWPORT; 1302 break; 1303 1304 default: 1305 error = EINVAL; 1306 break; 1307 } 1308 /* INP_UNLOCK(inp); */ 1309 break; 1310 1311 #if defined(IPSEC) || defined(FAST_IPSEC) 1312 case IP_IPSEC_POLICY: 1313 { 1314 error = ipsec4_set_policy(inp, sopt->sopt_name, 1315 sopt->sopt_data, sopt->sopt_size, l->l_cred); 1316 break; 1317 } 1318 #endif /*IPSEC*/ 1319 1320 default: 1321 error = ENOPROTOOPT; 1322 break; 1323 } 1324 break; 1325 1326 case PRCO_GETOPT: 1327 switch (sopt->sopt_name) { 1328 case IP_OPTIONS: 1329 case IP_RETOPTS: 1330 if (inp->inp_options) { 1331 struct mbuf *m; 1332 1333 m = m_copym(inp->inp_options, 0, M_COPYALL, 1334 M_DONTWAIT); 1335 if (m == NULL) { 1336 error = ENOBUFS; 1337 break; 1338 } 1339 1340 error = sockopt_setmbuf(sopt, m); 1341 } 1342 break; 1343 1344 case IP_TOS: 1345 case IP_TTL: 1346 case IP_MINTTL: 1347 case IP_RECVOPTS: 1348 case IP_RECVRETOPTS: 1349 case IP_RECVDSTADDR: 1350 case IP_RECVIF: 1351 case IP_RECVTTL: 1352 case IP_ERRORMTU: 1353 switch (sopt->sopt_name) { 1354 case IP_TOS: 1355 optval = inp->inp_ip.ip_tos; 1356 break; 1357 1358 case IP_TTL: 1359 optval = inp->inp_ip.ip_ttl; 1360 break; 1361 1362 case IP_MINTTL: 1363 optval = inp->inp_ip_minttl; 1364 break; 1365 1366 case IP_ERRORMTU: 1367 optval = inp->inp_errormtu; 1368 break; 1369 1370 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1371 1372 case IP_RECVOPTS: 1373 optval = OPTBIT(INP_RECVOPTS); 1374 break; 1375 1376 case IP_RECVRETOPTS: 1377 optval = OPTBIT(INP_RECVRETOPTS); 1378 break; 1379 1380 case IP_RECVDSTADDR: 1381 optval = OPTBIT(INP_RECVDSTADDR); 1382 break; 1383 1384 case IP_RECVIF: 1385 optval = OPTBIT(INP_RECVIF); 1386 break; 1387 1388 case IP_RECVTTL: 1389 optval = OPTBIT(INP_RECVTTL); 1390 break; 1391 } 1392 error = sockopt_setint(sopt, optval); 1393 break; 1394 1395 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1396 case IP_IPSEC_POLICY: 1397 { 1398 struct mbuf *m = NULL; 1399 1400 /* XXX this will return EINVAL as sopt is empty */ 1401 error = ipsec4_get_policy(inp, sopt->sopt_data, 1402 sopt->sopt_size, &m); 1403 if (error == 0) 1404 error = sockopt_setmbuf(sopt, m); 1405 break; 1406 } 1407 #endif /*IPSEC*/ 1408 1409 case IP_MULTICAST_IF: 1410 case IP_MULTICAST_TTL: 1411 case IP_MULTICAST_LOOP: 1412 case IP_ADD_MEMBERSHIP: 1413 case IP_DROP_MEMBERSHIP: 1414 error = ip_getmoptions(inp->inp_moptions, sopt); 1415 break; 1416 1417 case IP_PORTRANGE: 1418 if (inp->inp_flags & INP_LOWPORT) 1419 optval = IP_PORTRANGE_LOW; 1420 else 1421 optval = IP_PORTRANGE_DEFAULT; 1422 1423 error = sockopt_setint(sopt, optval); 1424 1425 break; 1426 1427 default: 1428 error = ENOPROTOOPT; 1429 break; 1430 } 1431 break; 1432 } 1433 return (error); 1434 } 1435 1436 /* 1437 * Set up IP options in pcb for insertion in output packets. 1438 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1439 * with destination address if source routed. 1440 */ 1441 int 1442 ip_pcbopts(struct mbuf **pcbopt, const struct sockopt *sopt) 1443 { 1444 struct mbuf *m; 1445 const u_char *cp; 1446 u_char *dp; 1447 int cnt; 1448 uint8_t optval, olen, offset; 1449 1450 /* turn off any old options */ 1451 if (*pcbopt) 1452 (void)m_free(*pcbopt); 1453 *pcbopt = NULL; 1454 1455 cp = sopt->sopt_data; 1456 cnt = sopt->sopt_size; 1457 1458 if (cnt == 0) 1459 return (0); /* Only turning off any previous options */ 1460 1461 #ifndef __vax__ 1462 if (cnt % sizeof(int32_t)) 1463 return (EINVAL); 1464 #endif 1465 1466 m = m_get(M_DONTWAIT, MT_SOOPTS); 1467 if (m == NULL) 1468 return (ENOBUFS); 1469 1470 dp = mtod(m, u_char *); 1471 memset(dp, 0, sizeof(struct in_addr)); 1472 dp += sizeof(struct in_addr); 1473 m->m_len = sizeof(struct in_addr); 1474 1475 /* 1476 * IP option list according to RFC791. Each option is of the form 1477 * 1478 * [optval] [olen] [(olen - 2) data bytes] 1479 * 1480 * we validate the list and copy options to an mbuf for prepending 1481 * to data packets. The IP first-hop destination address will be 1482 * stored before actual options and is zero if unset. 1483 */ 1484 while (cnt > 0) { 1485 optval = cp[IPOPT_OPTVAL]; 1486 1487 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1488 olen = 1; 1489 } else { 1490 if (cnt < IPOPT_OLEN + 1) 1491 goto bad; 1492 1493 olen = cp[IPOPT_OLEN]; 1494 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1495 goto bad; 1496 } 1497 1498 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1499 /* 1500 * user process specifies route as: 1501 * ->A->B->C->D 1502 * D must be our final destination (but we can't 1503 * check that since we may not have connected yet). 1504 * A is first hop destination, which doesn't appear in 1505 * actual IP option, but is stored before the options. 1506 */ 1507 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1508 goto bad; 1509 1510 offset = cp[IPOPT_OFFSET]; 1511 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1512 sizeof(struct in_addr)); 1513 1514 cp += sizeof(struct in_addr); 1515 cnt -= sizeof(struct in_addr); 1516 olen -= sizeof(struct in_addr); 1517 1518 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1519 goto bad; 1520 1521 memcpy(dp, cp, olen); 1522 dp[IPOPT_OPTVAL] = optval; 1523 dp[IPOPT_OLEN] = olen; 1524 dp[IPOPT_OFFSET] = offset; 1525 break; 1526 } else { 1527 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1528 goto bad; 1529 1530 memcpy(dp, cp, olen); 1531 break; 1532 } 1533 1534 dp += olen; 1535 m->m_len += olen; 1536 1537 if (optval == IPOPT_EOL) 1538 break; 1539 1540 cp += olen; 1541 cnt -= olen; 1542 } 1543 1544 *pcbopt = m; 1545 return (0); 1546 1547 bad: 1548 (void)m_free(m); 1549 return (EINVAL); 1550 } 1551 1552 /* 1553 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1554 */ 1555 static struct ifnet * 1556 ip_multicast_if(struct in_addr *a, int *ifindexp) 1557 { 1558 int ifindex; 1559 struct ifnet *ifp = NULL; 1560 struct in_ifaddr *ia; 1561 1562 if (ifindexp) 1563 *ifindexp = 0; 1564 if (ntohl(a->s_addr) >> 24 == 0) { 1565 ifindex = ntohl(a->s_addr) & 0xffffff; 1566 if (ifindex < 0 || if_indexlim <= ifindex) 1567 return NULL; 1568 ifp = ifindex2ifnet[ifindex]; 1569 if (!ifp) 1570 return NULL; 1571 if (ifindexp) 1572 *ifindexp = ifindex; 1573 } else { 1574 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1575 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1576 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1577 ifp = ia->ia_ifp; 1578 break; 1579 } 1580 } 1581 } 1582 return ifp; 1583 } 1584 1585 static int 1586 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1587 { 1588 u_int tval; 1589 u_char cval; 1590 int error; 1591 1592 if (sopt == NULL) 1593 return EINVAL; 1594 1595 switch (sopt->sopt_size) { 1596 case sizeof(u_char): 1597 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1598 tval = cval; 1599 break; 1600 1601 case sizeof(u_int): 1602 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1603 break; 1604 1605 default: 1606 error = EINVAL; 1607 } 1608 1609 if (error) 1610 return error; 1611 1612 if (tval > maxval) 1613 return EINVAL; 1614 1615 *val = tval; 1616 return 0; 1617 } 1618 1619 /* 1620 * Set the IP multicast options in response to user setsockopt(). 1621 */ 1622 int 1623 ip_setmoptions(struct ip_moptions **imop, const struct sockopt *sopt) 1624 { 1625 int error = 0; 1626 int i; 1627 struct in_addr addr; 1628 struct ip_mreq lmreq, *mreq; 1629 struct ifnet *ifp; 1630 struct ip_moptions *imo = *imop; 1631 int ifindex; 1632 1633 if (imo == NULL) { 1634 /* 1635 * No multicast option buffer attached to the pcb; 1636 * allocate one and initialize to default values. 1637 */ 1638 imo = malloc(sizeof(*imo), M_IPMOPTS, M_NOWAIT); 1639 if (imo == NULL) 1640 return (ENOBUFS); 1641 1642 *imop = imo; 1643 imo->imo_multicast_ifp = NULL; 1644 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1645 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1646 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1647 imo->imo_num_memberships = 0; 1648 } 1649 1650 switch (sopt->sopt_name) { 1651 case IP_MULTICAST_IF: 1652 /* 1653 * Select the interface for outgoing multicast packets. 1654 */ 1655 error = sockopt_get(sopt, &addr, sizeof(addr)); 1656 if (error) 1657 break; 1658 1659 /* 1660 * INADDR_ANY is used to remove a previous selection. 1661 * When no interface is selected, a default one is 1662 * chosen every time a multicast packet is sent. 1663 */ 1664 if (in_nullhost(addr)) { 1665 imo->imo_multicast_ifp = NULL; 1666 break; 1667 } 1668 /* 1669 * The selected interface is identified by its local 1670 * IP address. Find the interface and confirm that 1671 * it supports multicasting. 1672 */ 1673 ifp = ip_multicast_if(&addr, &ifindex); 1674 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1675 error = EADDRNOTAVAIL; 1676 break; 1677 } 1678 imo->imo_multicast_ifp = ifp; 1679 if (ifindex) 1680 imo->imo_multicast_addr = addr; 1681 else 1682 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1683 break; 1684 1685 case IP_MULTICAST_TTL: 1686 /* 1687 * Set the IP time-to-live for outgoing multicast packets. 1688 */ 1689 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 1690 break; 1691 1692 case IP_MULTICAST_LOOP: 1693 /* 1694 * Set the loopback flag for outgoing multicast packets. 1695 * Must be zero or one. 1696 */ 1697 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 1698 break; 1699 1700 case IP_ADD_MEMBERSHIP: 1701 /* 1702 * Add a multicast group membership. 1703 * Group must be a valid IP multicast address. 1704 */ 1705 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1706 if (error) 1707 break; 1708 1709 mreq = &lmreq; 1710 1711 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1712 error = EINVAL; 1713 break; 1714 } 1715 /* 1716 * If no interface address was provided, use the interface of 1717 * the route to the given multicast address. 1718 */ 1719 if (in_nullhost(mreq->imr_interface)) { 1720 struct rtentry *rt; 1721 union { 1722 struct sockaddr dst; 1723 struct sockaddr_in dst4; 1724 } u; 1725 struct route ro; 1726 1727 memset(&ro, 0, sizeof(ro)); 1728 1729 sockaddr_in_init(&u.dst4, &mreq->imr_multiaddr, 0); 1730 rtcache_setdst(&ro, &u.dst); 1731 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 1732 : NULL; 1733 rtcache_free(&ro); 1734 } else { 1735 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1736 } 1737 /* 1738 * See if we found an interface, and confirm that it 1739 * supports multicast. 1740 */ 1741 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1742 error = EADDRNOTAVAIL; 1743 break; 1744 } 1745 /* 1746 * See if the membership already exists or if all the 1747 * membership slots are full. 1748 */ 1749 for (i = 0; i < imo->imo_num_memberships; ++i) { 1750 if (imo->imo_membership[i]->inm_ifp == ifp && 1751 in_hosteq(imo->imo_membership[i]->inm_addr, 1752 mreq->imr_multiaddr)) 1753 break; 1754 } 1755 if (i < imo->imo_num_memberships) { 1756 error = EADDRINUSE; 1757 break; 1758 } 1759 if (i == IP_MAX_MEMBERSHIPS) { 1760 error = ETOOMANYREFS; 1761 break; 1762 } 1763 /* 1764 * Everything looks good; add a new record to the multicast 1765 * address list for the given interface. 1766 */ 1767 if ((imo->imo_membership[i] = 1768 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1769 error = ENOBUFS; 1770 break; 1771 } 1772 ++imo->imo_num_memberships; 1773 break; 1774 1775 case IP_DROP_MEMBERSHIP: 1776 /* 1777 * Drop a multicast group membership. 1778 * Group must be a valid IP multicast address. 1779 */ 1780 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1781 if (error) 1782 break; 1783 1784 mreq = &lmreq; 1785 1786 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1787 error = EINVAL; 1788 break; 1789 } 1790 /* 1791 * If an interface address was specified, get a pointer 1792 * to its ifnet structure. 1793 */ 1794 if (in_nullhost(mreq->imr_interface)) 1795 ifp = NULL; 1796 else { 1797 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1798 if (ifp == NULL) { 1799 error = EADDRNOTAVAIL; 1800 break; 1801 } 1802 } 1803 /* 1804 * Find the membership in the membership array. 1805 */ 1806 for (i = 0; i < imo->imo_num_memberships; ++i) { 1807 if ((ifp == NULL || 1808 imo->imo_membership[i]->inm_ifp == ifp) && 1809 in_hosteq(imo->imo_membership[i]->inm_addr, 1810 mreq->imr_multiaddr)) 1811 break; 1812 } 1813 if (i == imo->imo_num_memberships) { 1814 error = EADDRNOTAVAIL; 1815 break; 1816 } 1817 /* 1818 * Give up the multicast address record to which the 1819 * membership points. 1820 */ 1821 in_delmulti(imo->imo_membership[i]); 1822 /* 1823 * Remove the gap in the membership array. 1824 */ 1825 for (++i; i < imo->imo_num_memberships; ++i) 1826 imo->imo_membership[i-1] = imo->imo_membership[i]; 1827 --imo->imo_num_memberships; 1828 break; 1829 1830 default: 1831 error = EOPNOTSUPP; 1832 break; 1833 } 1834 1835 /* 1836 * If all options have default values, no need to keep the mbuf. 1837 */ 1838 if (imo->imo_multicast_ifp == NULL && 1839 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1840 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1841 imo->imo_num_memberships == 0) { 1842 free(*imop, M_IPMOPTS); 1843 *imop = NULL; 1844 } 1845 1846 return (error); 1847 } 1848 1849 /* 1850 * Return the IP multicast options in response to user getsockopt(). 1851 */ 1852 int 1853 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt) 1854 { 1855 struct in_addr addr; 1856 struct in_ifaddr *ia; 1857 int error; 1858 uint8_t optval; 1859 1860 error = 0; 1861 1862 switch (sopt->sopt_name) { 1863 case IP_MULTICAST_IF: 1864 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1865 addr = zeroin_addr; 1866 else if (imo->imo_multicast_addr.s_addr) { 1867 /* return the value user has set */ 1868 addr = imo->imo_multicast_addr; 1869 } else { 1870 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1871 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1872 } 1873 error = sockopt_set(sopt, &addr, sizeof(addr)); 1874 break; 1875 1876 case IP_MULTICAST_TTL: 1877 optval = imo ? imo->imo_multicast_ttl 1878 : IP_DEFAULT_MULTICAST_TTL; 1879 1880 error = sockopt_set(sopt, &optval, sizeof(optval)); 1881 break; 1882 1883 case IP_MULTICAST_LOOP: 1884 optval = imo ? imo->imo_multicast_loop 1885 : IP_DEFAULT_MULTICAST_LOOP; 1886 1887 error = sockopt_set(sopt, &optval, sizeof(optval)); 1888 break; 1889 1890 default: 1891 error = EOPNOTSUPP; 1892 } 1893 1894 return (error); 1895 } 1896 1897 /* 1898 * Discard the IP multicast options. 1899 */ 1900 void 1901 ip_freemoptions(struct ip_moptions *imo) 1902 { 1903 int i; 1904 1905 if (imo != NULL) { 1906 for (i = 0; i < imo->imo_num_memberships; ++i) 1907 in_delmulti(imo->imo_membership[i]); 1908 free(imo, M_IPMOPTS); 1909 } 1910 } 1911 1912 /* 1913 * Routine called from ip_output() to loop back a copy of an IP multicast 1914 * packet to the input queue of a specified interface. Note that this 1915 * calls the output routine of the loopback "driver", but with an interface 1916 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1917 */ 1918 static void 1919 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1920 { 1921 struct ip *ip; 1922 struct mbuf *copym; 1923 1924 copym = m_copypacket(m, M_DONTWAIT); 1925 if (copym != NULL 1926 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1927 copym = m_pullup(copym, sizeof(struct ip)); 1928 if (copym == NULL) 1929 return; 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, sintocsa(dst), NULL); 1945 } 1946