1 /* $NetBSD: ip_output.c,v 1.208 2011/04/14 15:53:36 yamt 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 1.208 2011/04/14 15:53:36 yamt 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 KERNEL_LOCK(1, NULL); 830 error = 831 (*ifp->if_output)(ifp, m, 832 (m->m_flags & M_MCAST) ? 833 sintocsa(rdst) : sintocsa(dst), 834 rt); 835 KERNEL_UNLOCK_ONE(NULL); 836 } else { 837 error = 838 ip_tso_output(ifp, m, 839 (m->m_flags & M_MCAST) ? 840 sintocsa(rdst) : sintocsa(dst), 841 rt); 842 } 843 goto done; 844 } 845 846 /* 847 * We can't use HW checksumming if we're about to 848 * to fragment the packet. 849 * 850 * XXX Some hardware can do this. 851 */ 852 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 853 if (IN_NEED_CHECKSUM(ifp, 854 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 855 in_delayed_cksum(m); 856 } 857 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 858 } 859 860 /* 861 * Too large for interface; fragment if possible. 862 * Must be able to put at least 8 bytes per fragment. 863 */ 864 if (ntohs(ip->ip_off) & IP_DF) { 865 if (flags & IP_RETURNMTU) 866 *mtu_p = mtu; 867 error = EMSGSIZE; 868 IP_STATINC(IP_STAT_CANTFRAG); 869 goto bad; 870 } 871 872 error = ip_fragment(m, ifp, mtu); 873 if (error) { 874 m = NULL; 875 goto bad; 876 } 877 878 for (; m; m = m0) { 879 m0 = m->m_nextpkt; 880 m->m_nextpkt = 0; 881 if (error == 0) { 882 #if IFA_STATS 883 if (ia) 884 ia->ia_ifa.ifa_data.ifad_outbytes += 885 ntohs(ip->ip_len); 886 #endif 887 #ifdef IPSEC 888 /* clean ipsec history once it goes out of the node */ 889 ipsec_delaux(m); 890 #endif /* IPSEC */ 891 892 #ifdef IPSEC_NAT_T 893 /* 894 * If we get there, the packet has not been handeld by 895 * IPSec whereas it should have. Now that it has been 896 * fragmented, re-inject it in ip_output so that IPsec 897 * processing can occur. 898 */ 899 if (natt_frag) { 900 error = ip_output(m, opt, 901 ro, flags, imo, so, mtu_p); 902 } else 903 #endif /* IPSEC_NAT_T */ 904 { 905 KASSERT((m->m_pkthdr.csum_flags & 906 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 907 KERNEL_LOCK(1, NULL); 908 error = (*ifp->if_output)(ifp, m, 909 (m->m_flags & M_MCAST) ? 910 sintocsa(rdst) : sintocsa(dst), 911 rt); 912 KERNEL_UNLOCK_ONE(NULL); 913 } 914 } else 915 m_freem(m); 916 } 917 918 if (error == 0) 919 IP_STATINC(IP_STAT_FRAGMENTED); 920 done: 921 rtcache_free(&iproute); 922 923 #ifdef IPSEC 924 if (sp != NULL) { 925 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 926 printf("DP ip_output call free SP:%p\n", sp)); 927 key_freesp(sp); 928 } 929 #endif /* IPSEC */ 930 #ifdef FAST_IPSEC 931 if (sp != NULL) 932 KEY_FREESP(&sp); 933 #endif /* FAST_IPSEC */ 934 935 return (error); 936 bad: 937 m_freem(m); 938 goto done; 939 } 940 941 int 942 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 943 { 944 struct ip *ip, *mhip; 945 struct mbuf *m0; 946 int len, hlen, off; 947 int mhlen, firstlen; 948 struct mbuf **mnext; 949 int sw_csum = m->m_pkthdr.csum_flags; 950 int fragments = 0; 951 int s; 952 int error = 0; 953 954 ip = mtod(m, struct ip *); 955 hlen = ip->ip_hl << 2; 956 if (ifp != NULL) 957 sw_csum &= ~ifp->if_csum_flags_tx; 958 959 len = (mtu - hlen) &~ 7; 960 if (len < 8) { 961 m_freem(m); 962 return (EMSGSIZE); 963 } 964 965 firstlen = len; 966 mnext = &m->m_nextpkt; 967 968 /* 969 * Loop through length of segment after first fragment, 970 * make new header and copy data of each part and link onto chain. 971 */ 972 m0 = m; 973 mhlen = sizeof (struct ip); 974 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 975 MGETHDR(m, M_DONTWAIT, MT_HEADER); 976 if (m == 0) { 977 error = ENOBUFS; 978 IP_STATINC(IP_STAT_ODROPPED); 979 goto sendorfree; 980 } 981 MCLAIM(m, m0->m_owner); 982 *mnext = m; 983 mnext = &m->m_nextpkt; 984 m->m_data += max_linkhdr; 985 mhip = mtod(m, struct ip *); 986 *mhip = *ip; 987 /* we must inherit MCAST and BCAST flags */ 988 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 989 if (hlen > sizeof (struct ip)) { 990 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 991 mhip->ip_hl = mhlen >> 2; 992 } 993 m->m_len = mhlen; 994 mhip->ip_off = ((off - hlen) >> 3) + 995 (ntohs(ip->ip_off) & ~IP_MF); 996 if (ip->ip_off & htons(IP_MF)) 997 mhip->ip_off |= IP_MF; 998 if (off + len >= ntohs(ip->ip_len)) 999 len = ntohs(ip->ip_len) - off; 1000 else 1001 mhip->ip_off |= IP_MF; 1002 HTONS(mhip->ip_off); 1003 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 1004 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 1005 if (m->m_next == 0) { 1006 error = ENOBUFS; /* ??? */ 1007 IP_STATINC(IP_STAT_ODROPPED); 1008 goto sendorfree; 1009 } 1010 m->m_pkthdr.len = mhlen + len; 1011 m->m_pkthdr.rcvif = (struct ifnet *)0; 1012 mhip->ip_sum = 0; 1013 if (sw_csum & M_CSUM_IPv4) { 1014 mhip->ip_sum = in_cksum(m, mhlen); 1015 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 1016 } else { 1017 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 1018 m->m_pkthdr.csum_data |= mhlen << 16; 1019 } 1020 IP_STATINC(IP_STAT_OFRAGMENTS); 1021 fragments++; 1022 } 1023 /* 1024 * Update first fragment by trimming what's been copied out 1025 * and updating header, then send each fragment (in order). 1026 */ 1027 m = m0; 1028 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1029 m->m_pkthdr.len = hlen + firstlen; 1030 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1031 ip->ip_off |= htons(IP_MF); 1032 ip->ip_sum = 0; 1033 /* 1034 * We may not use checksums on loopback interfaces 1035 */ 1036 if (__predict_false(ifp == NULL) || 1037 IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 1038 if (sw_csum & M_CSUM_IPv4) { 1039 ip->ip_sum = in_cksum(m, hlen); 1040 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1041 } else { 1042 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1043 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 1044 sizeof(struct ip)); 1045 } 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 IP_STATINC(IP_STAT_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), (void *) &csum); 1095 } else 1096 *(u_int16_t *)(mtod(m, char *) + 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_MOVE_PKTHDR(n, m); 1140 m->m_len -= sizeof(struct ip); 1141 m->m_data += sizeof(struct ip); 1142 n->m_next = m; 1143 m = n; 1144 m->m_len = optlen + sizeof(struct ip); 1145 m->m_data += max_linkhdr; 1146 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 1147 } else { 1148 m->m_data -= optlen; 1149 m->m_len += optlen; 1150 memmove(mtod(m, void *), ip, sizeof(struct ip)); 1151 } 1152 m->m_pkthdr.len += optlen; 1153 ip = mtod(m, struct ip *); 1154 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 1155 *phlen = sizeof(struct ip) + optlen; 1156 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1157 return (m); 1158 } 1159 1160 /* 1161 * Copy options from ip to jp, 1162 * omitting those not copied during fragmentation. 1163 */ 1164 int 1165 ip_optcopy(struct ip *ip, struct ip *jp) 1166 { 1167 u_char *cp, *dp; 1168 int opt, optlen, cnt; 1169 1170 cp = (u_char *)(ip + 1); 1171 dp = (u_char *)(jp + 1); 1172 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1173 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1174 opt = cp[0]; 1175 if (opt == IPOPT_EOL) 1176 break; 1177 if (opt == IPOPT_NOP) { 1178 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1179 *dp++ = IPOPT_NOP; 1180 optlen = 1; 1181 continue; 1182 } 1183 #ifdef DIAGNOSTIC 1184 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1185 panic("malformed IPv4 option passed to ip_optcopy"); 1186 #endif 1187 optlen = cp[IPOPT_OLEN]; 1188 #ifdef DIAGNOSTIC 1189 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1190 panic("malformed IPv4 option passed to ip_optcopy"); 1191 #endif 1192 /* bogus lengths should have been caught by ip_dooptions */ 1193 if (optlen > cnt) 1194 optlen = cnt; 1195 if (IPOPT_COPIED(opt)) { 1196 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1197 dp += optlen; 1198 } 1199 } 1200 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1201 *dp++ = IPOPT_EOL; 1202 return (optlen); 1203 } 1204 1205 /* 1206 * IP socket option processing. 1207 */ 1208 int 1209 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1210 { 1211 struct inpcb *inp = sotoinpcb(so); 1212 int optval = 0; 1213 int error = 0; 1214 #if defined(IPSEC) || defined(FAST_IPSEC) 1215 struct lwp *l = curlwp; /*XXX*/ 1216 #endif 1217 1218 if (sopt->sopt_level != IPPROTO_IP) { 1219 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 1220 return 0; 1221 return ENOPROTOOPT; 1222 } 1223 1224 switch (op) { 1225 case PRCO_SETOPT: 1226 switch (sopt->sopt_name) { 1227 case IP_OPTIONS: 1228 #ifdef notyet 1229 case IP_RETOPTS: 1230 #endif 1231 error = ip_pcbopts(&inp->inp_options, sopt); 1232 break; 1233 1234 case IP_TOS: 1235 case IP_TTL: 1236 case IP_MINTTL: 1237 case IP_RECVOPTS: 1238 case IP_RECVRETOPTS: 1239 case IP_RECVDSTADDR: 1240 case IP_RECVIF: 1241 case IP_RECVTTL: 1242 error = sockopt_getint(sopt, &optval); 1243 if (error) 1244 break; 1245 1246 switch (sopt->sopt_name) { 1247 case IP_TOS: 1248 inp->inp_ip.ip_tos = optval; 1249 break; 1250 1251 case IP_TTL: 1252 inp->inp_ip.ip_ttl = optval; 1253 break; 1254 1255 case IP_MINTTL: 1256 if (optval > 0 && optval <= MAXTTL) 1257 inp->inp_ip_minttl = optval; 1258 else 1259 error = EINVAL; 1260 break; 1261 #define OPTSET(bit) \ 1262 if (optval) \ 1263 inp->inp_flags |= bit; \ 1264 else \ 1265 inp->inp_flags &= ~bit; 1266 1267 case IP_RECVOPTS: 1268 OPTSET(INP_RECVOPTS); 1269 break; 1270 1271 case IP_RECVRETOPTS: 1272 OPTSET(INP_RECVRETOPTS); 1273 break; 1274 1275 case IP_RECVDSTADDR: 1276 OPTSET(INP_RECVDSTADDR); 1277 break; 1278 1279 case IP_RECVIF: 1280 OPTSET(INP_RECVIF); 1281 break; 1282 1283 case IP_RECVTTL: 1284 OPTSET(INP_RECVTTL); 1285 break; 1286 } 1287 break; 1288 #undef OPTSET 1289 1290 case IP_MULTICAST_IF: 1291 case IP_MULTICAST_TTL: 1292 case IP_MULTICAST_LOOP: 1293 case IP_ADD_MEMBERSHIP: 1294 case IP_DROP_MEMBERSHIP: 1295 error = ip_setmoptions(&inp->inp_moptions, sopt); 1296 break; 1297 1298 case IP_PORTRANGE: 1299 error = sockopt_getint(sopt, &optval); 1300 if (error) 1301 break; 1302 1303 /* INP_LOCK(inp); */ 1304 switch (optval) { 1305 case IP_PORTRANGE_DEFAULT: 1306 case IP_PORTRANGE_HIGH: 1307 inp->inp_flags &= ~(INP_LOWPORT); 1308 break; 1309 1310 case IP_PORTRANGE_LOW: 1311 inp->inp_flags |= INP_LOWPORT; 1312 break; 1313 1314 default: 1315 error = EINVAL; 1316 break; 1317 } 1318 /* INP_UNLOCK(inp); */ 1319 break; 1320 1321 #if defined(IPSEC) || defined(FAST_IPSEC) 1322 case IP_IPSEC_POLICY: 1323 { 1324 error = ipsec4_set_policy(inp, sopt->sopt_name, 1325 sopt->sopt_data, sopt->sopt_size, l->l_cred); 1326 break; 1327 } 1328 #endif /*IPSEC*/ 1329 1330 default: 1331 error = ENOPROTOOPT; 1332 break; 1333 } 1334 break; 1335 1336 case PRCO_GETOPT: 1337 switch (sopt->sopt_name) { 1338 case IP_OPTIONS: 1339 case IP_RETOPTS: 1340 if (inp->inp_options) { 1341 struct mbuf *m; 1342 1343 m = m_copym(inp->inp_options, 0, M_COPYALL, 1344 M_DONTWAIT); 1345 if (m == NULL) { 1346 error = ENOBUFS; 1347 break; 1348 } 1349 1350 error = sockopt_setmbuf(sopt, m); 1351 } 1352 break; 1353 1354 case IP_TOS: 1355 case IP_TTL: 1356 case IP_MINTTL: 1357 case IP_RECVOPTS: 1358 case IP_RECVRETOPTS: 1359 case IP_RECVDSTADDR: 1360 case IP_RECVIF: 1361 case IP_RECVTTL: 1362 case IP_ERRORMTU: 1363 switch (sopt->sopt_name) { 1364 case IP_TOS: 1365 optval = inp->inp_ip.ip_tos; 1366 break; 1367 1368 case IP_TTL: 1369 optval = inp->inp_ip.ip_ttl; 1370 break; 1371 1372 case IP_MINTTL: 1373 optval = inp->inp_ip_minttl; 1374 break; 1375 1376 case IP_ERRORMTU: 1377 optval = inp->inp_errormtu; 1378 break; 1379 1380 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1381 1382 case IP_RECVOPTS: 1383 optval = OPTBIT(INP_RECVOPTS); 1384 break; 1385 1386 case IP_RECVRETOPTS: 1387 optval = OPTBIT(INP_RECVRETOPTS); 1388 break; 1389 1390 case IP_RECVDSTADDR: 1391 optval = OPTBIT(INP_RECVDSTADDR); 1392 break; 1393 1394 case IP_RECVIF: 1395 optval = OPTBIT(INP_RECVIF); 1396 break; 1397 1398 case IP_RECVTTL: 1399 optval = OPTBIT(INP_RECVTTL); 1400 break; 1401 } 1402 error = sockopt_setint(sopt, optval); 1403 break; 1404 1405 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1406 case IP_IPSEC_POLICY: 1407 { 1408 struct mbuf *m = NULL; 1409 1410 /* XXX this will return EINVAL as sopt is empty */ 1411 error = ipsec4_get_policy(inp, sopt->sopt_data, 1412 sopt->sopt_size, &m); 1413 if (error == 0) 1414 error = sockopt_setmbuf(sopt, m); 1415 break; 1416 } 1417 #endif /*IPSEC*/ 1418 1419 case IP_MULTICAST_IF: 1420 case IP_MULTICAST_TTL: 1421 case IP_MULTICAST_LOOP: 1422 case IP_ADD_MEMBERSHIP: 1423 case IP_DROP_MEMBERSHIP: 1424 error = ip_getmoptions(inp->inp_moptions, sopt); 1425 break; 1426 1427 case IP_PORTRANGE: 1428 if (inp->inp_flags & INP_LOWPORT) 1429 optval = IP_PORTRANGE_LOW; 1430 else 1431 optval = IP_PORTRANGE_DEFAULT; 1432 1433 error = sockopt_setint(sopt, optval); 1434 1435 break; 1436 1437 default: 1438 error = ENOPROTOOPT; 1439 break; 1440 } 1441 break; 1442 } 1443 return (error); 1444 } 1445 1446 /* 1447 * Set up IP options in pcb for insertion in output packets. 1448 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1449 * with destination address if source routed. 1450 */ 1451 int 1452 ip_pcbopts(struct mbuf **pcbopt, const struct sockopt *sopt) 1453 { 1454 struct mbuf *m; 1455 const u_char *cp; 1456 u_char *dp; 1457 int cnt; 1458 uint8_t optval, olen, offset; 1459 1460 /* turn off any old options */ 1461 if (*pcbopt) 1462 (void)m_free(*pcbopt); 1463 *pcbopt = NULL; 1464 1465 cp = sopt->sopt_data; 1466 cnt = sopt->sopt_size; 1467 1468 if (cnt == 0) 1469 return (0); /* Only turning off any previous options */ 1470 1471 #ifndef __vax__ 1472 if (cnt % sizeof(int32_t)) 1473 return (EINVAL); 1474 #endif 1475 1476 m = m_get(M_DONTWAIT, MT_SOOPTS); 1477 if (m == NULL) 1478 return (ENOBUFS); 1479 1480 dp = mtod(m, u_char *); 1481 memset(dp, 0, sizeof(struct in_addr)); 1482 dp += sizeof(struct in_addr); 1483 m->m_len = sizeof(struct in_addr); 1484 1485 /* 1486 * IP option list according to RFC791. Each option is of the form 1487 * 1488 * [optval] [olen] [(olen - 2) data bytes] 1489 * 1490 * we validate the list and copy options to an mbuf for prepending 1491 * to data packets. The IP first-hop destination address will be 1492 * stored before actual options and is zero if unset. 1493 */ 1494 while (cnt > 0) { 1495 optval = cp[IPOPT_OPTVAL]; 1496 1497 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1498 olen = 1; 1499 } else { 1500 if (cnt < IPOPT_OLEN + 1) 1501 goto bad; 1502 1503 olen = cp[IPOPT_OLEN]; 1504 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1505 goto bad; 1506 } 1507 1508 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1509 /* 1510 * user process specifies route as: 1511 * ->A->B->C->D 1512 * D must be our final destination (but we can't 1513 * check that since we may not have connected yet). 1514 * A is first hop destination, which doesn't appear in 1515 * actual IP option, but is stored before the options. 1516 */ 1517 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1518 goto bad; 1519 1520 offset = cp[IPOPT_OFFSET]; 1521 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1522 sizeof(struct in_addr)); 1523 1524 cp += sizeof(struct in_addr); 1525 cnt -= sizeof(struct in_addr); 1526 olen -= sizeof(struct in_addr); 1527 1528 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1529 goto bad; 1530 1531 memcpy(dp, cp, olen); 1532 dp[IPOPT_OPTVAL] = optval; 1533 dp[IPOPT_OLEN] = olen; 1534 dp[IPOPT_OFFSET] = offset; 1535 break; 1536 } else { 1537 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1538 goto bad; 1539 1540 memcpy(dp, cp, olen); 1541 break; 1542 } 1543 1544 dp += olen; 1545 m->m_len += olen; 1546 1547 if (optval == IPOPT_EOL) 1548 break; 1549 1550 cp += olen; 1551 cnt -= olen; 1552 } 1553 1554 *pcbopt = m; 1555 return (0); 1556 1557 bad: 1558 (void)m_free(m); 1559 return (EINVAL); 1560 } 1561 1562 /* 1563 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1564 */ 1565 static struct ifnet * 1566 ip_multicast_if(struct in_addr *a, int *ifindexp) 1567 { 1568 int ifindex; 1569 struct ifnet *ifp = NULL; 1570 struct in_ifaddr *ia; 1571 1572 if (ifindexp) 1573 *ifindexp = 0; 1574 if (ntohl(a->s_addr) >> 24 == 0) { 1575 ifindex = ntohl(a->s_addr) & 0xffffff; 1576 if (ifindex < 0 || if_indexlim <= ifindex) 1577 return NULL; 1578 ifp = ifindex2ifnet[ifindex]; 1579 if (!ifp) 1580 return NULL; 1581 if (ifindexp) 1582 *ifindexp = ifindex; 1583 } else { 1584 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1585 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1586 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1587 ifp = ia->ia_ifp; 1588 break; 1589 } 1590 } 1591 } 1592 return ifp; 1593 } 1594 1595 static int 1596 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1597 { 1598 u_int tval; 1599 u_char cval; 1600 int error; 1601 1602 if (sopt == NULL) 1603 return EINVAL; 1604 1605 switch (sopt->sopt_size) { 1606 case sizeof(u_char): 1607 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1608 tval = cval; 1609 break; 1610 1611 case sizeof(u_int): 1612 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1613 break; 1614 1615 default: 1616 error = EINVAL; 1617 } 1618 1619 if (error) 1620 return error; 1621 1622 if (tval > maxval) 1623 return EINVAL; 1624 1625 *val = tval; 1626 return 0; 1627 } 1628 1629 /* 1630 * Set the IP multicast options in response to user setsockopt(). 1631 */ 1632 int 1633 ip_setmoptions(struct ip_moptions **imop, const struct sockopt *sopt) 1634 { 1635 int error = 0; 1636 int i; 1637 struct in_addr addr; 1638 struct ip_mreq lmreq, *mreq; 1639 struct ifnet *ifp; 1640 struct ip_moptions *imo = *imop; 1641 int ifindex; 1642 1643 if (imo == NULL) { 1644 /* 1645 * No multicast option buffer attached to the pcb; 1646 * allocate one and initialize to default values. 1647 */ 1648 imo = malloc(sizeof(*imo), M_IPMOPTS, M_NOWAIT); 1649 if (imo == NULL) 1650 return (ENOBUFS); 1651 1652 *imop = imo; 1653 imo->imo_multicast_ifp = NULL; 1654 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1655 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1656 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1657 imo->imo_num_memberships = 0; 1658 } 1659 1660 switch (sopt->sopt_name) { 1661 case IP_MULTICAST_IF: 1662 /* 1663 * Select the interface for outgoing multicast packets. 1664 */ 1665 error = sockopt_get(sopt, &addr, sizeof(addr)); 1666 if (error) 1667 break; 1668 1669 /* 1670 * INADDR_ANY is used to remove a previous selection. 1671 * When no interface is selected, a default one is 1672 * chosen every time a multicast packet is sent. 1673 */ 1674 if (in_nullhost(addr)) { 1675 imo->imo_multicast_ifp = NULL; 1676 break; 1677 } 1678 /* 1679 * The selected interface is identified by its local 1680 * IP address. Find the interface and confirm that 1681 * it supports multicasting. 1682 */ 1683 ifp = ip_multicast_if(&addr, &ifindex); 1684 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1685 error = EADDRNOTAVAIL; 1686 break; 1687 } 1688 imo->imo_multicast_ifp = ifp; 1689 if (ifindex) 1690 imo->imo_multicast_addr = addr; 1691 else 1692 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1693 break; 1694 1695 case IP_MULTICAST_TTL: 1696 /* 1697 * Set the IP time-to-live for outgoing multicast packets. 1698 */ 1699 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 1700 break; 1701 1702 case IP_MULTICAST_LOOP: 1703 /* 1704 * Set the loopback flag for outgoing multicast packets. 1705 * Must be zero or one. 1706 */ 1707 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 1708 break; 1709 1710 case IP_ADD_MEMBERSHIP: 1711 /* 1712 * Add a multicast group membership. 1713 * Group must be a valid IP multicast address. 1714 */ 1715 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1716 if (error) 1717 break; 1718 1719 mreq = &lmreq; 1720 1721 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1722 error = EINVAL; 1723 break; 1724 } 1725 /* 1726 * If no interface address was provided, use the interface of 1727 * the route to the given multicast address. 1728 */ 1729 if (in_nullhost(mreq->imr_interface)) { 1730 struct rtentry *rt; 1731 union { 1732 struct sockaddr dst; 1733 struct sockaddr_in dst4; 1734 } u; 1735 struct route ro; 1736 1737 memset(&ro, 0, sizeof(ro)); 1738 1739 sockaddr_in_init(&u.dst4, &mreq->imr_multiaddr, 0); 1740 rtcache_setdst(&ro, &u.dst); 1741 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 1742 : NULL; 1743 rtcache_free(&ro); 1744 } else { 1745 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1746 } 1747 /* 1748 * See if we found an interface, and confirm that it 1749 * supports multicast. 1750 */ 1751 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1752 error = EADDRNOTAVAIL; 1753 break; 1754 } 1755 /* 1756 * See if the membership already exists or if all the 1757 * membership slots are full. 1758 */ 1759 for (i = 0; i < imo->imo_num_memberships; ++i) { 1760 if (imo->imo_membership[i]->inm_ifp == ifp && 1761 in_hosteq(imo->imo_membership[i]->inm_addr, 1762 mreq->imr_multiaddr)) 1763 break; 1764 } 1765 if (i < imo->imo_num_memberships) { 1766 error = EADDRINUSE; 1767 break; 1768 } 1769 if (i == IP_MAX_MEMBERSHIPS) { 1770 error = ETOOMANYREFS; 1771 break; 1772 } 1773 /* 1774 * Everything looks good; add a new record to the multicast 1775 * address list for the given interface. 1776 */ 1777 if ((imo->imo_membership[i] = 1778 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1779 error = ENOBUFS; 1780 break; 1781 } 1782 ++imo->imo_num_memberships; 1783 break; 1784 1785 case IP_DROP_MEMBERSHIP: 1786 /* 1787 * Drop a multicast group membership. 1788 * Group must be a valid IP multicast address. 1789 */ 1790 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1791 if (error) 1792 break; 1793 1794 mreq = &lmreq; 1795 1796 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1797 error = EINVAL; 1798 break; 1799 } 1800 /* 1801 * If an interface address was specified, get a pointer 1802 * to its ifnet structure. 1803 */ 1804 if (in_nullhost(mreq->imr_interface)) 1805 ifp = NULL; 1806 else { 1807 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1808 if (ifp == NULL) { 1809 error = EADDRNOTAVAIL; 1810 break; 1811 } 1812 } 1813 /* 1814 * Find the membership in the membership array. 1815 */ 1816 for (i = 0; i < imo->imo_num_memberships; ++i) { 1817 if ((ifp == NULL || 1818 imo->imo_membership[i]->inm_ifp == ifp) && 1819 in_hosteq(imo->imo_membership[i]->inm_addr, 1820 mreq->imr_multiaddr)) 1821 break; 1822 } 1823 if (i == imo->imo_num_memberships) { 1824 error = EADDRNOTAVAIL; 1825 break; 1826 } 1827 /* 1828 * Give up the multicast address record to which the 1829 * membership points. 1830 */ 1831 in_delmulti(imo->imo_membership[i]); 1832 /* 1833 * Remove the gap in the membership array. 1834 */ 1835 for (++i; i < imo->imo_num_memberships; ++i) 1836 imo->imo_membership[i-1] = imo->imo_membership[i]; 1837 --imo->imo_num_memberships; 1838 break; 1839 1840 default: 1841 error = EOPNOTSUPP; 1842 break; 1843 } 1844 1845 /* 1846 * If all options have default values, no need to keep the mbuf. 1847 */ 1848 if (imo->imo_multicast_ifp == NULL && 1849 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1850 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1851 imo->imo_num_memberships == 0) { 1852 free(*imop, M_IPMOPTS); 1853 *imop = NULL; 1854 } 1855 1856 return (error); 1857 } 1858 1859 /* 1860 * Return the IP multicast options in response to user getsockopt(). 1861 */ 1862 int 1863 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt) 1864 { 1865 struct in_addr addr; 1866 struct in_ifaddr *ia; 1867 int error; 1868 uint8_t optval; 1869 1870 error = 0; 1871 1872 switch (sopt->sopt_name) { 1873 case IP_MULTICAST_IF: 1874 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1875 addr = zeroin_addr; 1876 else if (imo->imo_multicast_addr.s_addr) { 1877 /* return the value user has set */ 1878 addr = imo->imo_multicast_addr; 1879 } else { 1880 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1881 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1882 } 1883 error = sockopt_set(sopt, &addr, sizeof(addr)); 1884 break; 1885 1886 case IP_MULTICAST_TTL: 1887 optval = imo ? imo->imo_multicast_ttl 1888 : IP_DEFAULT_MULTICAST_TTL; 1889 1890 error = sockopt_set(sopt, &optval, sizeof(optval)); 1891 break; 1892 1893 case IP_MULTICAST_LOOP: 1894 optval = imo ? imo->imo_multicast_loop 1895 : IP_DEFAULT_MULTICAST_LOOP; 1896 1897 error = sockopt_set(sopt, &optval, sizeof(optval)); 1898 break; 1899 1900 default: 1901 error = EOPNOTSUPP; 1902 } 1903 1904 return (error); 1905 } 1906 1907 /* 1908 * Discard the IP multicast options. 1909 */ 1910 void 1911 ip_freemoptions(struct ip_moptions *imo) 1912 { 1913 int i; 1914 1915 if (imo != NULL) { 1916 for (i = 0; i < imo->imo_num_memberships; ++i) 1917 in_delmulti(imo->imo_membership[i]); 1918 free(imo, M_IPMOPTS); 1919 } 1920 } 1921 1922 /* 1923 * Routine called from ip_output() to loop back a copy of an IP multicast 1924 * packet to the input queue of a specified interface. Note that this 1925 * calls the output routine of the loopback "driver", but with an interface 1926 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1927 */ 1928 static void 1929 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1930 { 1931 struct ip *ip; 1932 struct mbuf *copym; 1933 1934 copym = m_copypacket(m, M_DONTWAIT); 1935 if (copym != NULL 1936 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1937 copym = m_pullup(copym, sizeof(struct ip)); 1938 if (copym == NULL) 1939 return; 1940 /* 1941 * We don't bother to fragment if the IP length is greater 1942 * than the interface's MTU. Can this possibly matter? 1943 */ 1944 ip = mtod(copym, struct ip *); 1945 1946 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1947 in_delayed_cksum(copym); 1948 copym->m_pkthdr.csum_flags &= 1949 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1950 } 1951 1952 ip->ip_sum = 0; 1953 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1954 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1955 } 1956