1 /* $NetBSD: ip_output.c,v 1.200 2008/10/12 11:15:54 plunky 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 1.200 2008/10/12 11:15:54 plunky 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 bzero(&state, 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 goto spd_done; 676 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 677 &error, inp); 678 /* 679 * There are four return cases: 680 * sp != NULL apply IPsec policy 681 * sp == NULL, error == 0 no IPsec handling needed 682 * sp == NULL, error == -EINVAL discard packet w/o error 683 * sp == NULL, error != 0 discard packet, report error 684 */ 685 if (sp != NULL) { 686 #ifdef IPSEC_NAT_T 687 /* 688 * NAT-T ESP fragmentation: don't do IPSec processing now, 689 * we'll do it on each fragmented packet. 690 */ 691 if (sp->req->sav && 692 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 693 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 694 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 695 natt_frag = 1; 696 mtu = sp->req->sav->esp_frag; 697 splx(s); 698 goto spd_done; 699 } 700 } 701 #endif /* IPSEC_NAT_T */ 702 703 /* 704 * Do delayed checksums now because we send before 705 * this is done in the normal processing path. 706 */ 707 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 708 in_delayed_cksum(m); 709 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 710 } 711 712 #ifdef __FreeBSD__ 713 ip->ip_len = htons(ip->ip_len); 714 ip->ip_off = htons(ip->ip_off); 715 #endif 716 717 /* NB: callee frees mbuf */ 718 error = ipsec4_process_packet(m, sp->req, flags, 0); 719 /* 720 * Preserve KAME behaviour: ENOENT can be returned 721 * when an SA acquire is in progress. Don't propagate 722 * this to user-level; it confuses applications. 723 * 724 * XXX this will go away when the SADB is redone. 725 */ 726 if (error == ENOENT) 727 error = 0; 728 splx(s); 729 goto done; 730 } else { 731 splx(s); 732 733 if (error != 0) { 734 /* 735 * Hack: -EINVAL is used to signal that a packet 736 * should be silently discarded. This is typically 737 * because we asked key management for an SA and 738 * it was delayed (e.g. kicked up to IKE). 739 */ 740 if (error == -EINVAL) 741 error = 0; 742 goto bad; 743 } else { 744 /* No IPsec processing for this packet. */ 745 } 746 } 747 } 748 spd_done: 749 #endif /* FAST_IPSEC */ 750 751 #ifdef PFIL_HOOKS 752 /* 753 * Run through list of hooks for output packets. 754 */ 755 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 756 goto done; 757 if (m == NULL) 758 goto done; 759 760 ip = mtod(m, struct ip *); 761 hlen = ip->ip_hl << 2; 762 ip_len = ntohs(ip->ip_len); 763 #endif /* PFIL_HOOKS */ 764 765 m->m_pkthdr.csum_data |= hlen << 16; 766 767 #if IFA_STATS 768 /* 769 * search for the source address structure to 770 * maintain output statistics. 771 */ 772 INADDR_TO_IA(ip->ip_src, ia); 773 #endif 774 775 /* Maybe skip checksums on loopback interfaces. */ 776 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 777 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 778 } 779 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 780 /* 781 * If small enough for mtu of path, or if using TCP segmentation 782 * offload, can just send directly. 783 */ 784 if (ip_len <= mtu || 785 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 786 #if IFA_STATS 787 if (ia) 788 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 789 #endif 790 /* 791 * Always initialize the sum to 0! Some HW assisted 792 * checksumming requires this. 793 */ 794 ip->ip_sum = 0; 795 796 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 797 /* 798 * Perform any checksums that the hardware can't do 799 * for us. 800 * 801 * XXX Does any hardware require the {th,uh}_sum 802 * XXX fields to be 0? 803 */ 804 if (sw_csum & M_CSUM_IPv4) { 805 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 806 ip->ip_sum = in_cksum(m, hlen); 807 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 808 } 809 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 810 if (IN_NEED_CHECKSUM(ifp, 811 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 812 in_delayed_cksum(m); 813 } 814 m->m_pkthdr.csum_flags &= 815 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 816 } 817 } 818 819 #ifdef IPSEC 820 /* clean ipsec history once it goes out of the node */ 821 ipsec_delaux(m); 822 #endif 823 824 if (__predict_true( 825 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 826 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 827 error = 828 (*ifp->if_output)(ifp, m, 829 (m->m_flags & M_MCAST) ? 830 sintocsa(rdst) : sintocsa(dst), 831 rt); 832 } else { 833 error = 834 ip_tso_output(ifp, m, 835 (m->m_flags & M_MCAST) ? 836 sintocsa(rdst) : sintocsa(dst), 837 rt); 838 } 839 goto done; 840 } 841 842 /* 843 * We can't use HW checksumming if we're about to 844 * to fragment the packet. 845 * 846 * XXX Some hardware can do this. 847 */ 848 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 849 if (IN_NEED_CHECKSUM(ifp, 850 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 851 in_delayed_cksum(m); 852 } 853 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 854 } 855 856 /* 857 * Too large for interface; fragment if possible. 858 * Must be able to put at least 8 bytes per fragment. 859 */ 860 if (ntohs(ip->ip_off) & IP_DF) { 861 if (flags & IP_RETURNMTU) 862 *mtu_p = mtu; 863 error = EMSGSIZE; 864 IP_STATINC(IP_STAT_CANTFRAG); 865 goto bad; 866 } 867 868 error = ip_fragment(m, ifp, mtu); 869 if (error) { 870 m = NULL; 871 goto bad; 872 } 873 874 for (; m; m = m0) { 875 m0 = m->m_nextpkt; 876 m->m_nextpkt = 0; 877 if (error == 0) { 878 #if IFA_STATS 879 if (ia) 880 ia->ia_ifa.ifa_data.ifad_outbytes += 881 ntohs(ip->ip_len); 882 #endif 883 #ifdef IPSEC 884 /* clean ipsec history once it goes out of the node */ 885 ipsec_delaux(m); 886 #endif /* IPSEC */ 887 888 #ifdef IPSEC_NAT_T 889 /* 890 * If we get there, the packet has not been handeld by 891 * IPSec whereas it should have. Now that it has been 892 * fragmented, re-inject it in ip_output so that IPsec 893 * processing can occur. 894 */ 895 if (natt_frag) { 896 error = ip_output(m, opt, 897 ro, flags, imo, so, mtu_p); 898 } else 899 #endif /* IPSEC_NAT_T */ 900 { 901 KASSERT((m->m_pkthdr.csum_flags & 902 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 903 error = (*ifp->if_output)(ifp, m, 904 (m->m_flags & M_MCAST) ? 905 sintocsa(rdst) : sintocsa(dst), 906 rt); 907 } 908 } else 909 m_freem(m); 910 } 911 912 if (error == 0) 913 IP_STATINC(IP_STAT_FRAGMENTED); 914 done: 915 rtcache_free(&iproute); 916 917 #ifdef IPSEC 918 if (sp != NULL) { 919 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 920 printf("DP ip_output call free SP:%p\n", sp)); 921 key_freesp(sp); 922 } 923 #endif /* IPSEC */ 924 #ifdef FAST_IPSEC 925 if (sp != NULL) 926 KEY_FREESP(&sp); 927 #endif /* FAST_IPSEC */ 928 929 return (error); 930 bad: 931 m_freem(m); 932 goto done; 933 } 934 935 int 936 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 937 { 938 struct ip *ip, *mhip; 939 struct mbuf *m0; 940 int len, hlen, off; 941 int mhlen, firstlen; 942 struct mbuf **mnext; 943 int sw_csum = m->m_pkthdr.csum_flags; 944 int fragments = 0; 945 int s; 946 int error = 0; 947 948 ip = mtod(m, struct ip *); 949 hlen = ip->ip_hl << 2; 950 if (ifp != NULL) 951 sw_csum &= ~ifp->if_csum_flags_tx; 952 953 len = (mtu - hlen) &~ 7; 954 if (len < 8) { 955 m_freem(m); 956 return (EMSGSIZE); 957 } 958 959 firstlen = len; 960 mnext = &m->m_nextpkt; 961 962 /* 963 * Loop through length of segment after first fragment, 964 * make new header and copy data of each part and link onto chain. 965 */ 966 m0 = m; 967 mhlen = sizeof (struct ip); 968 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 969 MGETHDR(m, M_DONTWAIT, MT_HEADER); 970 if (m == 0) { 971 error = ENOBUFS; 972 IP_STATINC(IP_STAT_ODROPPED); 973 goto sendorfree; 974 } 975 MCLAIM(m, m0->m_owner); 976 *mnext = m; 977 mnext = &m->m_nextpkt; 978 m->m_data += max_linkhdr; 979 mhip = mtod(m, struct ip *); 980 *mhip = *ip; 981 /* we must inherit MCAST and BCAST flags */ 982 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 983 if (hlen > sizeof (struct ip)) { 984 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 985 mhip->ip_hl = mhlen >> 2; 986 } 987 m->m_len = mhlen; 988 mhip->ip_off = ((off - hlen) >> 3) + 989 (ntohs(ip->ip_off) & ~IP_MF); 990 if (ip->ip_off & htons(IP_MF)) 991 mhip->ip_off |= IP_MF; 992 if (off + len >= ntohs(ip->ip_len)) 993 len = ntohs(ip->ip_len) - off; 994 else 995 mhip->ip_off |= IP_MF; 996 HTONS(mhip->ip_off); 997 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 998 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 999 if (m->m_next == 0) { 1000 error = ENOBUFS; /* ??? */ 1001 IP_STATINC(IP_STAT_ODROPPED); 1002 goto sendorfree; 1003 } 1004 m->m_pkthdr.len = mhlen + len; 1005 m->m_pkthdr.rcvif = (struct ifnet *)0; 1006 mhip->ip_sum = 0; 1007 if (sw_csum & M_CSUM_IPv4) { 1008 mhip->ip_sum = in_cksum(m, mhlen); 1009 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 1010 } else { 1011 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 1012 m->m_pkthdr.csum_data |= mhlen << 16; 1013 } 1014 IP_STATINC(IP_STAT_OFRAGMENTS); 1015 fragments++; 1016 } 1017 /* 1018 * Update first fragment by trimming what's been copied out 1019 * and updating header, then send each fragment (in order). 1020 */ 1021 m = m0; 1022 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1023 m->m_pkthdr.len = hlen + firstlen; 1024 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1025 ip->ip_off |= htons(IP_MF); 1026 ip->ip_sum = 0; 1027 if (sw_csum & M_CSUM_IPv4) { 1028 ip->ip_sum = in_cksum(m, hlen); 1029 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1030 } else { 1031 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1032 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 1033 sizeof(struct ip)); 1034 } 1035 sendorfree: 1036 /* 1037 * If there is no room for all the fragments, don't queue 1038 * any of them. 1039 */ 1040 if (ifp != NULL) { 1041 s = splnet(); 1042 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1043 error == 0) { 1044 error = ENOBUFS; 1045 IP_STATINC(IP_STAT_ODROPPED); 1046 IFQ_INC_DROPS(&ifp->if_snd); 1047 } 1048 splx(s); 1049 } 1050 if (error) { 1051 for (m = m0; m; m = m0) { 1052 m0 = m->m_nextpkt; 1053 m->m_nextpkt = NULL; 1054 m_freem(m); 1055 } 1056 } 1057 return (error); 1058 } 1059 1060 /* 1061 * Process a delayed payload checksum calculation. 1062 */ 1063 void 1064 in_delayed_cksum(struct mbuf *m) 1065 { 1066 struct ip *ip; 1067 u_int16_t csum, offset; 1068 1069 ip = mtod(m, struct ip *); 1070 offset = ip->ip_hl << 2; 1071 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1072 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1073 csum = 0xffff; 1074 1075 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 1076 1077 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1078 /* This happen when ip options were inserted 1079 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1080 m->m_len, offset, ip->ip_p); 1081 */ 1082 m_copyback(m, offset, sizeof(csum), (void *) &csum); 1083 } else 1084 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 1085 } 1086 1087 /* 1088 * Determine the maximum length of the options to be inserted; 1089 * we would far rather allocate too much space rather than too little. 1090 */ 1091 1092 u_int 1093 ip_optlen(struct inpcb *inp) 1094 { 1095 struct mbuf *m = inp->inp_options; 1096 1097 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1098 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1099 else 1100 return 0; 1101 } 1102 1103 1104 /* 1105 * Insert IP options into preformed packet. 1106 * Adjust IP destination as required for IP source routing, 1107 * as indicated by a non-zero in_addr at the start of the options. 1108 */ 1109 static struct mbuf * 1110 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1111 { 1112 struct ipoption *p = mtod(opt, struct ipoption *); 1113 struct mbuf *n; 1114 struct ip *ip = mtod(m, struct ip *); 1115 unsigned optlen; 1116 1117 optlen = opt->m_len - sizeof(p->ipopt_dst); 1118 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1119 return (m); /* XXX should fail */ 1120 if (!in_nullhost(p->ipopt_dst)) 1121 ip->ip_dst = p->ipopt_dst; 1122 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1123 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1124 if (n == 0) 1125 return (m); 1126 MCLAIM(n, m->m_owner); 1127 M_MOVE_PKTHDR(n, m); 1128 m->m_len -= sizeof(struct ip); 1129 m->m_data += sizeof(struct ip); 1130 n->m_next = m; 1131 m = n; 1132 m->m_len = optlen + sizeof(struct ip); 1133 m->m_data += max_linkhdr; 1134 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 1135 } else { 1136 m->m_data -= optlen; 1137 m->m_len += optlen; 1138 memmove(mtod(m, void *), ip, sizeof(struct ip)); 1139 } 1140 m->m_pkthdr.len += optlen; 1141 ip = mtod(m, struct ip *); 1142 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 1143 *phlen = sizeof(struct ip) + optlen; 1144 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1145 return (m); 1146 } 1147 1148 /* 1149 * Copy options from ip to jp, 1150 * omitting those not copied during fragmentation. 1151 */ 1152 int 1153 ip_optcopy(struct ip *ip, struct ip *jp) 1154 { 1155 u_char *cp, *dp; 1156 int opt, optlen, cnt; 1157 1158 cp = (u_char *)(ip + 1); 1159 dp = (u_char *)(jp + 1); 1160 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1161 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1162 opt = cp[0]; 1163 if (opt == IPOPT_EOL) 1164 break; 1165 if (opt == IPOPT_NOP) { 1166 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1167 *dp++ = IPOPT_NOP; 1168 optlen = 1; 1169 continue; 1170 } 1171 #ifdef DIAGNOSTIC 1172 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1173 panic("malformed IPv4 option passed to ip_optcopy"); 1174 #endif 1175 optlen = cp[IPOPT_OLEN]; 1176 #ifdef DIAGNOSTIC 1177 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1178 panic("malformed IPv4 option passed to ip_optcopy"); 1179 #endif 1180 /* bogus lengths should have been caught by ip_dooptions */ 1181 if (optlen > cnt) 1182 optlen = cnt; 1183 if (IPOPT_COPIED(opt)) { 1184 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1185 dp += optlen; 1186 } 1187 } 1188 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1189 *dp++ = IPOPT_EOL; 1190 return (optlen); 1191 } 1192 1193 /* 1194 * IP socket option processing. 1195 */ 1196 int 1197 ip_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1198 { 1199 struct inpcb *inp = sotoinpcb(so); 1200 int optval = 0; 1201 int error = 0; 1202 #if defined(IPSEC) || defined(FAST_IPSEC) 1203 struct lwp *l = curlwp; /*XXX*/ 1204 #endif 1205 1206 if (sopt->sopt_level != IPPROTO_IP) { 1207 if (sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_NOHEADER) 1208 return 0; 1209 return ENOPROTOOPT; 1210 } 1211 1212 switch (op) { 1213 case PRCO_SETOPT: 1214 switch (sopt->sopt_name) { 1215 case IP_OPTIONS: 1216 #ifdef notyet 1217 case IP_RETOPTS: 1218 #endif 1219 error = ip_pcbopts(&inp->inp_options, sopt); 1220 break; 1221 1222 case IP_TOS: 1223 case IP_TTL: 1224 case IP_RECVOPTS: 1225 case IP_RECVRETOPTS: 1226 case IP_RECVDSTADDR: 1227 case IP_RECVIF: 1228 error = sockopt_getint(sopt, &optval); 1229 if (error) 1230 break; 1231 1232 switch (sopt->sopt_name) { 1233 case IP_TOS: 1234 inp->inp_ip.ip_tos = optval; 1235 break; 1236 1237 case IP_TTL: 1238 inp->inp_ip.ip_ttl = optval; 1239 break; 1240 #define OPTSET(bit) \ 1241 if (optval) \ 1242 inp->inp_flags |= bit; \ 1243 else \ 1244 inp->inp_flags &= ~bit; 1245 1246 case IP_RECVOPTS: 1247 OPTSET(INP_RECVOPTS); 1248 break; 1249 1250 case IP_RECVRETOPTS: 1251 OPTSET(INP_RECVRETOPTS); 1252 break; 1253 1254 case IP_RECVDSTADDR: 1255 OPTSET(INP_RECVDSTADDR); 1256 break; 1257 1258 case IP_RECVIF: 1259 OPTSET(INP_RECVIF); 1260 break; 1261 } 1262 break; 1263 #undef OPTSET 1264 1265 case IP_MULTICAST_IF: 1266 case IP_MULTICAST_TTL: 1267 case IP_MULTICAST_LOOP: 1268 case IP_ADD_MEMBERSHIP: 1269 case IP_DROP_MEMBERSHIP: 1270 error = ip_setmoptions(&inp->inp_moptions, sopt); 1271 break; 1272 1273 case IP_PORTRANGE: 1274 error = sockopt_getint(sopt, &optval); 1275 if (error) 1276 break; 1277 1278 /* INP_LOCK(inp); */ 1279 switch (optval) { 1280 case IP_PORTRANGE_DEFAULT: 1281 case IP_PORTRANGE_HIGH: 1282 inp->inp_flags &= ~(INP_LOWPORT); 1283 break; 1284 1285 case IP_PORTRANGE_LOW: 1286 inp->inp_flags |= INP_LOWPORT; 1287 break; 1288 1289 default: 1290 error = EINVAL; 1291 break; 1292 } 1293 /* INP_UNLOCK(inp); */ 1294 break; 1295 1296 #if defined(IPSEC) || defined(FAST_IPSEC) 1297 case IP_IPSEC_POLICY: 1298 { 1299 int priv = 0; 1300 1301 #ifdef __NetBSD__ 1302 if (l == 0 || kauth_authorize_generic(l->l_cred, 1303 KAUTH_GENERIC_ISSUSER, NULL)) 1304 priv = 0; 1305 else 1306 priv = 1; 1307 #else 1308 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1309 #endif 1310 1311 error = ipsec4_set_policy(inp, sopt->sopt_name, 1312 sopt->sopt_data, sopt->sopt_size, priv); 1313 break; 1314 } 1315 #endif /*IPSEC*/ 1316 1317 default: 1318 error = ENOPROTOOPT; 1319 break; 1320 } 1321 break; 1322 1323 case PRCO_GETOPT: 1324 switch (sopt->sopt_name) { 1325 case IP_OPTIONS: 1326 case IP_RETOPTS: 1327 if (inp->inp_options) { 1328 struct mbuf *m; 1329 1330 m = m_copym(inp->inp_options, 0, M_COPYALL, 1331 M_DONTWAIT); 1332 if (m == NULL) { 1333 error = ENOBUFS; 1334 break; 1335 } 1336 1337 error = sockopt_setmbuf(sopt, m); 1338 } 1339 break; 1340 1341 case IP_TOS: 1342 case IP_TTL: 1343 case IP_RECVOPTS: 1344 case IP_RECVRETOPTS: 1345 case IP_RECVDSTADDR: 1346 case IP_RECVIF: 1347 case IP_ERRORMTU: 1348 switch (sopt->sopt_name) { 1349 case IP_TOS: 1350 optval = inp->inp_ip.ip_tos; 1351 break; 1352 1353 case IP_TTL: 1354 optval = inp->inp_ip.ip_ttl; 1355 break; 1356 1357 case IP_ERRORMTU: 1358 optval = inp->inp_errormtu; 1359 break; 1360 1361 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1362 1363 case IP_RECVOPTS: 1364 optval = OPTBIT(INP_RECVOPTS); 1365 break; 1366 1367 case IP_RECVRETOPTS: 1368 optval = OPTBIT(INP_RECVRETOPTS); 1369 break; 1370 1371 case IP_RECVDSTADDR: 1372 optval = OPTBIT(INP_RECVDSTADDR); 1373 break; 1374 1375 case IP_RECVIF: 1376 optval = OPTBIT(INP_RECVIF); 1377 break; 1378 } 1379 error = sockopt_setint(sopt, optval); 1380 break; 1381 1382 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1383 case IP_IPSEC_POLICY: 1384 { 1385 struct mbuf *m = NULL; 1386 1387 /* XXX this will return EINVAL as sopt is empty */ 1388 error = ipsec4_get_policy(inp, sopt->sopt_data, 1389 sopt->sopt_size, &m); 1390 if (error == 0) 1391 error = sockopt_setmbuf(sopt, m); 1392 break; 1393 } 1394 #endif /*IPSEC*/ 1395 1396 case IP_MULTICAST_IF: 1397 case IP_MULTICAST_TTL: 1398 case IP_MULTICAST_LOOP: 1399 case IP_ADD_MEMBERSHIP: 1400 case IP_DROP_MEMBERSHIP: 1401 error = ip_getmoptions(inp->inp_moptions, sopt); 1402 break; 1403 1404 case IP_PORTRANGE: 1405 if (inp->inp_flags & INP_LOWPORT) 1406 optval = IP_PORTRANGE_LOW; 1407 else 1408 optval = IP_PORTRANGE_DEFAULT; 1409 1410 error = sockopt_setint(sopt, optval); 1411 1412 break; 1413 1414 default: 1415 error = ENOPROTOOPT; 1416 break; 1417 } 1418 break; 1419 } 1420 return (error); 1421 } 1422 1423 /* 1424 * Set up IP options in pcb for insertion in output packets. 1425 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1426 * with destination address if source routed. 1427 */ 1428 int 1429 ip_pcbopts(struct mbuf **pcbopt, const struct sockopt *sopt) 1430 { 1431 struct mbuf *m; 1432 const u_char *cp; 1433 u_char *dp; 1434 int cnt; 1435 uint8_t optval, olen, offset; 1436 1437 /* turn off any old options */ 1438 if (*pcbopt) 1439 (void)m_free(*pcbopt); 1440 *pcbopt = NULL; 1441 1442 cp = sopt->sopt_data; 1443 cnt = sopt->sopt_size; 1444 1445 if (cnt == 0) 1446 return (0); /* Only turning off any previous options */ 1447 1448 #ifndef __vax__ 1449 if (cnt % sizeof(int32_t)) 1450 return (EINVAL); 1451 #endif 1452 1453 m = m_get(M_DONTWAIT, MT_SOOPTS); 1454 if (m == NULL) 1455 return (ENOBUFS); 1456 1457 dp = mtod(m, u_char *); 1458 memset(dp, 0, sizeof(struct in_addr)); 1459 dp += sizeof(struct in_addr); 1460 m->m_len = sizeof(struct in_addr); 1461 1462 /* 1463 * IP option list according to RFC791. Each option is of the form 1464 * 1465 * [optval] [olen] [(olen - 2) data bytes] 1466 * 1467 * we validate the list and copy options to an mbuf for prepending 1468 * to data packets. The IP first-hop destination address will be 1469 * stored before actual options and is zero if unset. 1470 */ 1471 while (cnt > 0) { 1472 optval = cp[IPOPT_OPTVAL]; 1473 1474 if (optval == IPOPT_EOL || optval == IPOPT_NOP) { 1475 olen = 1; 1476 } else { 1477 if (cnt < IPOPT_OLEN + 1) 1478 goto bad; 1479 1480 olen = cp[IPOPT_OLEN]; 1481 if (olen < IPOPT_OLEN + 1 || olen > cnt) 1482 goto bad; 1483 } 1484 1485 if (optval == IPOPT_LSRR || optval == IPOPT_SSRR) { 1486 /* 1487 * user process specifies route as: 1488 * ->A->B->C->D 1489 * D must be our final destination (but we can't 1490 * check that since we may not have connected yet). 1491 * A is first hop destination, which doesn't appear in 1492 * actual IP option, but is stored before the options. 1493 */ 1494 if (olen < IPOPT_OFFSET + 1 + sizeof(struct in_addr)) 1495 goto bad; 1496 1497 offset = cp[IPOPT_OFFSET]; 1498 memcpy(mtod(m, u_char *), cp + IPOPT_OFFSET + 1, 1499 sizeof(struct in_addr)); 1500 1501 cp += sizeof(struct in_addr); 1502 cnt -= sizeof(struct in_addr); 1503 olen -= sizeof(struct in_addr); 1504 1505 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1506 goto bad; 1507 1508 memcpy(dp, cp, olen); 1509 dp[IPOPT_OPTVAL] = optval; 1510 dp[IPOPT_OLEN] = olen; 1511 dp[IPOPT_OFFSET] = offset; 1512 break; 1513 } else { 1514 if (m->m_len + olen > MAX_IPOPTLEN + sizeof(struct in_addr)) 1515 goto bad; 1516 1517 memcpy(dp, cp, olen); 1518 break; 1519 } 1520 1521 dp += olen; 1522 m->m_len += olen; 1523 1524 if (optval == IPOPT_EOL) 1525 break; 1526 1527 cp += olen; 1528 cnt -= olen; 1529 } 1530 1531 *pcbopt = m; 1532 return (0); 1533 1534 bad: 1535 (void)m_free(m); 1536 return (EINVAL); 1537 } 1538 1539 /* 1540 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1541 */ 1542 static struct ifnet * 1543 ip_multicast_if(struct in_addr *a, int *ifindexp) 1544 { 1545 int ifindex; 1546 struct ifnet *ifp = NULL; 1547 struct in_ifaddr *ia; 1548 1549 if (ifindexp) 1550 *ifindexp = 0; 1551 if (ntohl(a->s_addr) >> 24 == 0) { 1552 ifindex = ntohl(a->s_addr) & 0xffffff; 1553 if (ifindex < 0 || if_indexlim <= ifindex) 1554 return NULL; 1555 ifp = ifindex2ifnet[ifindex]; 1556 if (!ifp) 1557 return NULL; 1558 if (ifindexp) 1559 *ifindexp = ifindex; 1560 } else { 1561 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1562 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1563 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1564 ifp = ia->ia_ifp; 1565 break; 1566 } 1567 } 1568 } 1569 return ifp; 1570 } 1571 1572 static int 1573 ip_getoptval(const struct sockopt *sopt, u_int8_t *val, u_int maxval) 1574 { 1575 u_int tval; 1576 u_char cval; 1577 int error; 1578 1579 if (sopt == NULL) 1580 return EINVAL; 1581 1582 switch (sopt->sopt_size) { 1583 case sizeof(u_char): 1584 error = sockopt_get(sopt, &cval, sizeof(u_char)); 1585 tval = cval; 1586 break; 1587 1588 case sizeof(u_int): 1589 error = sockopt_get(sopt, &tval, sizeof(u_int)); 1590 break; 1591 1592 default: 1593 error = EINVAL; 1594 } 1595 1596 if (error) 1597 return error; 1598 1599 if (tval > maxval) 1600 return EINVAL; 1601 1602 *val = tval; 1603 return 0; 1604 } 1605 1606 /* 1607 * Set the IP multicast options in response to user setsockopt(). 1608 */ 1609 int 1610 ip_setmoptions(struct ip_moptions **imop, const struct sockopt *sopt) 1611 { 1612 int error = 0; 1613 int i; 1614 struct in_addr addr; 1615 struct ip_mreq lmreq, *mreq; 1616 struct ifnet *ifp; 1617 struct ip_moptions *imo = *imop; 1618 int ifindex; 1619 1620 if (imo == NULL) { 1621 /* 1622 * No multicast option buffer attached to the pcb; 1623 * allocate one and initialize to default values. 1624 */ 1625 imo = malloc(sizeof(*imo), M_IPMOPTS, M_NOWAIT); 1626 if (imo == NULL) 1627 return (ENOBUFS); 1628 1629 *imop = imo; 1630 imo->imo_multicast_ifp = NULL; 1631 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1632 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1633 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1634 imo->imo_num_memberships = 0; 1635 } 1636 1637 switch (sopt->sopt_name) { 1638 case IP_MULTICAST_IF: 1639 /* 1640 * Select the interface for outgoing multicast packets. 1641 */ 1642 error = sockopt_get(sopt, &addr, sizeof(addr)); 1643 if (error) 1644 break; 1645 1646 /* 1647 * INADDR_ANY is used to remove a previous selection. 1648 * When no interface is selected, a default one is 1649 * chosen every time a multicast packet is sent. 1650 */ 1651 if (in_nullhost(addr)) { 1652 imo->imo_multicast_ifp = NULL; 1653 break; 1654 } 1655 /* 1656 * The selected interface is identified by its local 1657 * IP address. Find the interface and confirm that 1658 * it supports multicasting. 1659 */ 1660 ifp = ip_multicast_if(&addr, &ifindex); 1661 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1662 error = EADDRNOTAVAIL; 1663 break; 1664 } 1665 imo->imo_multicast_ifp = ifp; 1666 if (ifindex) 1667 imo->imo_multicast_addr = addr; 1668 else 1669 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1670 break; 1671 1672 case IP_MULTICAST_TTL: 1673 /* 1674 * Set the IP time-to-live for outgoing multicast packets. 1675 */ 1676 error = ip_getoptval(sopt, &imo->imo_multicast_ttl, MAXTTL); 1677 break; 1678 1679 case IP_MULTICAST_LOOP: 1680 /* 1681 * Set the loopback flag for outgoing multicast packets. 1682 * Must be zero or one. 1683 */ 1684 error = ip_getoptval(sopt, &imo->imo_multicast_loop, 1); 1685 break; 1686 1687 case IP_ADD_MEMBERSHIP: 1688 /* 1689 * Add a multicast group membership. 1690 * Group must be a valid IP multicast address. 1691 */ 1692 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1693 if (error) 1694 break; 1695 1696 mreq = &lmreq; 1697 1698 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1699 error = EINVAL; 1700 break; 1701 } 1702 /* 1703 * If no interface address was provided, use the interface of 1704 * the route to the given multicast address. 1705 */ 1706 if (in_nullhost(mreq->imr_interface)) { 1707 struct rtentry *rt; 1708 union { 1709 struct sockaddr dst; 1710 struct sockaddr_in dst4; 1711 } u; 1712 struct route ro; 1713 1714 memset(&ro, 0, sizeof(ro)); 1715 1716 sockaddr_in_init(&u.dst4, &mreq->imr_multiaddr, 0); 1717 rtcache_setdst(&ro, &u.dst); 1718 ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp 1719 : NULL; 1720 rtcache_free(&ro); 1721 } else { 1722 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1723 } 1724 /* 1725 * See if we found an interface, and confirm that it 1726 * supports multicast. 1727 */ 1728 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1729 error = EADDRNOTAVAIL; 1730 break; 1731 } 1732 /* 1733 * See if the membership already exists or if all the 1734 * membership slots are full. 1735 */ 1736 for (i = 0; i < imo->imo_num_memberships; ++i) { 1737 if (imo->imo_membership[i]->inm_ifp == ifp && 1738 in_hosteq(imo->imo_membership[i]->inm_addr, 1739 mreq->imr_multiaddr)) 1740 break; 1741 } 1742 if (i < imo->imo_num_memberships) { 1743 error = EADDRINUSE; 1744 break; 1745 } 1746 if (i == IP_MAX_MEMBERSHIPS) { 1747 error = ETOOMANYREFS; 1748 break; 1749 } 1750 /* 1751 * Everything looks good; add a new record to the multicast 1752 * address list for the given interface. 1753 */ 1754 if ((imo->imo_membership[i] = 1755 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1756 error = ENOBUFS; 1757 break; 1758 } 1759 ++imo->imo_num_memberships; 1760 break; 1761 1762 case IP_DROP_MEMBERSHIP: 1763 /* 1764 * Drop a multicast group membership. 1765 * Group must be a valid IP multicast address. 1766 */ 1767 error = sockopt_get(sopt, &lmreq, sizeof(lmreq)); 1768 if (error) 1769 break; 1770 1771 mreq = &lmreq; 1772 1773 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1774 error = EINVAL; 1775 break; 1776 } 1777 /* 1778 * If an interface address was specified, get a pointer 1779 * to its ifnet structure. 1780 */ 1781 if (in_nullhost(mreq->imr_interface)) 1782 ifp = NULL; 1783 else { 1784 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1785 if (ifp == NULL) { 1786 error = EADDRNOTAVAIL; 1787 break; 1788 } 1789 } 1790 /* 1791 * Find the membership in the membership array. 1792 */ 1793 for (i = 0; i < imo->imo_num_memberships; ++i) { 1794 if ((ifp == NULL || 1795 imo->imo_membership[i]->inm_ifp == ifp) && 1796 in_hosteq(imo->imo_membership[i]->inm_addr, 1797 mreq->imr_multiaddr)) 1798 break; 1799 } 1800 if (i == imo->imo_num_memberships) { 1801 error = EADDRNOTAVAIL; 1802 break; 1803 } 1804 /* 1805 * Give up the multicast address record to which the 1806 * membership points. 1807 */ 1808 in_delmulti(imo->imo_membership[i]); 1809 /* 1810 * Remove the gap in the membership array. 1811 */ 1812 for (++i; i < imo->imo_num_memberships; ++i) 1813 imo->imo_membership[i-1] = imo->imo_membership[i]; 1814 --imo->imo_num_memberships; 1815 break; 1816 1817 default: 1818 error = EOPNOTSUPP; 1819 break; 1820 } 1821 1822 /* 1823 * If all options have default values, no need to keep the mbuf. 1824 */ 1825 if (imo->imo_multicast_ifp == NULL && 1826 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1827 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1828 imo->imo_num_memberships == 0) { 1829 free(*imop, M_IPMOPTS); 1830 *imop = NULL; 1831 } 1832 1833 return (error); 1834 } 1835 1836 /* 1837 * Return the IP multicast options in response to user getsockopt(). 1838 */ 1839 int 1840 ip_getmoptions(struct ip_moptions *imo, struct sockopt *sopt) 1841 { 1842 struct in_addr addr; 1843 struct in_ifaddr *ia; 1844 int error; 1845 uint8_t optval; 1846 1847 error = 0; 1848 1849 switch (sopt->sopt_name) { 1850 case IP_MULTICAST_IF: 1851 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1852 addr = zeroin_addr; 1853 else if (imo->imo_multicast_addr.s_addr) { 1854 /* return the value user has set */ 1855 addr = imo->imo_multicast_addr; 1856 } else { 1857 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1858 addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1859 } 1860 error = sockopt_set(sopt, &addr, sizeof(addr)); 1861 break; 1862 1863 case IP_MULTICAST_TTL: 1864 optval = imo ? imo->imo_multicast_ttl 1865 : IP_DEFAULT_MULTICAST_TTL; 1866 1867 error = sockopt_set(sopt, &optval, sizeof(optval)); 1868 break; 1869 1870 case IP_MULTICAST_LOOP: 1871 optval = imo ? imo->imo_multicast_loop 1872 : IP_DEFAULT_MULTICAST_LOOP; 1873 1874 error = sockopt_set(sopt, &optval, sizeof(optval)); 1875 break; 1876 1877 default: 1878 error = EOPNOTSUPP; 1879 } 1880 1881 return (error); 1882 } 1883 1884 /* 1885 * Discard the IP multicast options. 1886 */ 1887 void 1888 ip_freemoptions(struct ip_moptions *imo) 1889 { 1890 int i; 1891 1892 if (imo != NULL) { 1893 for (i = 0; i < imo->imo_num_memberships; ++i) 1894 in_delmulti(imo->imo_membership[i]); 1895 free(imo, M_IPMOPTS); 1896 } 1897 } 1898 1899 /* 1900 * Routine called from ip_output() to loop back a copy of an IP multicast 1901 * packet to the input queue of a specified interface. Note that this 1902 * calls the output routine of the loopback "driver", but with an interface 1903 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1904 */ 1905 static void 1906 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1907 { 1908 struct ip *ip; 1909 struct mbuf *copym; 1910 1911 copym = m_copypacket(m, M_DONTWAIT); 1912 if (copym != NULL 1913 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1914 copym = m_pullup(copym, sizeof(struct ip)); 1915 if (copym == NULL) 1916 return; 1917 /* 1918 * We don't bother to fragment if the IP length is greater 1919 * than the interface's MTU. Can this possibly matter? 1920 */ 1921 ip = mtod(copym, struct ip *); 1922 1923 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1924 in_delayed_cksum(copym); 1925 copym->m_pkthdr.csum_flags &= 1926 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1927 } 1928 1929 ip->ip_sum = 0; 1930 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1931 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1932 } 1933