1 /* $NetBSD: ip_output.c,v 1.184 2007/09/19 04:33:43 dyoung Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 /*- 33 * Copyright (c) 1998 The NetBSD Foundation, Inc. 34 * All rights reserved. 35 * 36 * This code is derived from software contributed to The NetBSD Foundation 37 * by Public Access Networks Corporation ("Panix"). It was developed under 38 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. All advertising materials mentioning features or use of this software 49 * must display the following acknowledgement: 50 * This product includes software developed by the NetBSD 51 * Foundation, Inc. and its contributors. 52 * 4. Neither the name of The NetBSD Foundation nor the names of its 53 * contributors may be used to endorse or promote products derived 54 * from this software without specific prior written permission. 55 * 56 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 57 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 58 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 59 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 60 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 61 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 62 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 63 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 64 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 65 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 66 * POSSIBILITY OF SUCH DAMAGE. 67 */ 68 69 /* 70 * Copyright (c) 1982, 1986, 1988, 1990, 1993 71 * The Regents of the University of California. All rights reserved. 72 * 73 * Redistribution and use in source and binary forms, with or without 74 * modification, are permitted provided that the following conditions 75 * are met: 76 * 1. Redistributions of source code must retain the above copyright 77 * notice, this list of conditions and the following disclaimer. 78 * 2. Redistributions in binary form must reproduce the above copyright 79 * notice, this list of conditions and the following disclaimer in the 80 * documentation and/or other materials provided with the distribution. 81 * 3. Neither the name of the University nor the names of its contributors 82 * may be used to endorse or promote products derived from this software 83 * without specific prior written permission. 84 * 85 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 86 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 87 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 88 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 89 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 90 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 91 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 92 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 93 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 94 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 95 * SUCH DAMAGE. 96 * 97 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 98 */ 99 100 #include <sys/cdefs.h> 101 __KERNEL_RCSID(0, "$NetBSD: ip_output.c,v 1.184 2007/09/19 04:33:43 dyoung Exp $"); 102 103 #include "opt_pfil_hooks.h" 104 #include "opt_inet.h" 105 #include "opt_ipsec.h" 106 #include "opt_mrouting.h" 107 108 #include <sys/param.h> 109 #include <sys/malloc.h> 110 #include <sys/mbuf.h> 111 #include <sys/errno.h> 112 #include <sys/protosw.h> 113 #include <sys/socket.h> 114 #include <sys/socketvar.h> 115 #include <sys/kauth.h> 116 #ifdef FAST_IPSEC 117 #include <sys/domain.h> 118 #endif 119 #include <sys/systm.h> 120 #include <sys/proc.h> 121 122 #include <net/if.h> 123 #include <net/route.h> 124 #include <net/pfil.h> 125 126 #include <netinet/in.h> 127 #include <netinet/in_systm.h> 128 #include <netinet/ip.h> 129 #include <netinet/in_pcb.h> 130 #include <netinet/in_var.h> 131 #include <netinet/ip_var.h> 132 #include <netinet/in_offload.h> 133 134 #ifdef MROUTING 135 #include <netinet/ip_mroute.h> 136 #endif 137 138 #include <machine/stdarg.h> 139 140 #ifdef IPSEC 141 #include <netinet6/ipsec.h> 142 #include <netkey/key.h> 143 #include <netkey/key_debug.h> 144 #endif /*IPSEC*/ 145 146 #ifdef FAST_IPSEC 147 #include <netipsec/ipsec.h> 148 #include <netipsec/key.h> 149 #include <netipsec/xform.h> 150 #endif /* FAST_IPSEC*/ 151 152 #ifdef IPSEC_NAT_T 153 #include <netinet/udp.h> 154 #endif 155 156 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 157 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 158 static void ip_mloopback(struct ifnet *, struct mbuf *, 159 const struct sockaddr_in *); 160 static int ip_getoptval(struct mbuf *, u_int8_t *, u_int); 161 162 #ifdef PFIL_HOOKS 163 extern struct pfil_head inet_pfil_hook; /* XXX */ 164 #endif 165 166 int ip_do_loopback_cksum = 0; 167 168 #define IN_NEED_CHECKSUM(ifp, csum_flags) \ 169 (__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \ 170 (((csum_flags) & M_CSUM_UDPv4) != 0 && udp_do_loopback_cksum) || \ 171 (((csum_flags) & M_CSUM_TCPv4) != 0 && tcp_do_loopback_cksum) || \ 172 (((csum_flags) & M_CSUM_IPv4) != 0 && ip_do_loopback_cksum))) 173 174 /* 175 * IP output. The packet in mbuf chain m contains a skeletal IP 176 * header (with len, off, ttl, proto, tos, src, dst). 177 * The mbuf chain containing the packet will be freed. 178 * The mbuf opt, if present, will not be freed. 179 */ 180 int 181 ip_output(struct mbuf *m0, ...) 182 { 183 struct ip *ip; 184 struct ifnet *ifp; 185 struct mbuf *m = m0; 186 int hlen = sizeof (struct ip); 187 int len, error = 0; 188 struct route iproute; 189 const struct sockaddr_in *dst; 190 struct in_ifaddr *ia; 191 struct ifaddr *xifa; 192 struct mbuf *opt; 193 struct route *ro; 194 int flags, sw_csum; 195 int *mtu_p; 196 u_long mtu; 197 struct ip_moptions *imo; 198 struct socket *so; 199 va_list ap; 200 #ifdef IPSEC_NAT_T 201 int natt_frag = 0; 202 #endif 203 #ifdef IPSEC 204 struct secpolicy *sp = NULL; 205 #endif /*IPSEC*/ 206 #ifdef FAST_IPSEC 207 struct inpcb *inp; 208 struct m_tag *mtag; 209 struct secpolicy *sp = NULL; 210 struct tdb_ident *tdbi; 211 int s; 212 #endif 213 u_int16_t ip_len; 214 union { 215 struct sockaddr dst; 216 struct sockaddr_in dst4; 217 } u; 218 struct sockaddr *rdst = &u.dst; /* real IP destination, as opposed 219 * to the nexthop 220 */ 221 222 len = 0; 223 va_start(ap, m0); 224 opt = va_arg(ap, struct mbuf *); 225 ro = va_arg(ap, struct route *); 226 flags = va_arg(ap, int); 227 imo = va_arg(ap, struct ip_moptions *); 228 so = va_arg(ap, struct socket *); 229 if (flags & IP_RETURNMTU) 230 mtu_p = va_arg(ap, int *); 231 else 232 mtu_p = NULL; 233 va_end(ap); 234 235 MCLAIM(m, &ip_tx_mowner); 236 #ifdef FAST_IPSEC 237 if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET) 238 inp = (struct inpcb *)so->so_pcb; 239 else 240 inp = NULL; 241 #endif /* FAST_IPSEC */ 242 243 #ifdef DIAGNOSTIC 244 if ((m->m_flags & M_PKTHDR) == 0) 245 panic("ip_output: no HDR"); 246 247 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) != 0) { 248 panic("ip_output: IPv6 checksum offload flags: %d", 249 m->m_pkthdr.csum_flags); 250 } 251 252 if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) == 253 (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 254 panic("ip_output: conflicting checksum offload flags: %d", 255 m->m_pkthdr.csum_flags); 256 } 257 #endif 258 if (opt) { 259 m = ip_insertoptions(m, opt, &len); 260 if (len >= sizeof(struct ip)) 261 hlen = len; 262 } 263 ip = mtod(m, struct ip *); 264 /* 265 * Fill in IP header. 266 */ 267 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 268 ip->ip_v = IPVERSION; 269 ip->ip_off = htons(0); 270 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 271 ip->ip_id = ip_newid(); 272 } else { 273 274 /* 275 * TSO capable interfaces (typically?) increment 276 * ip_id for each segment. 277 * "allocate" enough ids here to increase the chance 278 * for them to be unique. 279 * 280 * note that the following calculation is not 281 * needed to be precise. wasting some ip_id is fine. 282 */ 283 284 unsigned int segsz = m->m_pkthdr.segsz; 285 unsigned int datasz = ntohs(ip->ip_len) - hlen; 286 unsigned int num = howmany(datasz, segsz); 287 288 ip->ip_id = ip_newid_range(num); 289 } 290 ip->ip_hl = hlen >> 2; 291 ipstat.ips_localout++; 292 } else { 293 hlen = ip->ip_hl << 2; 294 } 295 /* 296 * Route packet. 297 */ 298 memset(&iproute, 0, sizeof(iproute)); 299 if (ro == NULL) 300 ro = &iproute; 301 sockaddr_in_init(&u.dst4, &ip->ip_dst, 0); 302 dst = satocsin(rtcache_getdst(ro)); 303 /* 304 * If there is a cached route, 305 * check that it is to the same destination 306 * and is still up. If not, free it and try again. 307 * The address family should also be checked in case of sharing the 308 * cache with IPv6. 309 */ 310 if (dst == NULL) 311 ; 312 else if (dst->sin_family != AF_INET || 313 !in_hosteq(dst->sin_addr, ip->ip_dst)) 314 rtcache_free(ro); 315 else 316 rtcache_check(ro); 317 if (ro->ro_rt == NULL) { 318 dst = &u.dst4; 319 rtcache_setdst(ro, &u.dst); 320 } 321 /* 322 * If routing to interface only, 323 * short circuit routing lookup. 324 */ 325 if (flags & IP_ROUTETOIF) { 326 if ((ia = ifatoia(ifa_ifwithladdr(sintocsa(dst)))) == NULL) { 327 ipstat.ips_noroute++; 328 error = ENETUNREACH; 329 goto bad; 330 } 331 ifp = ia->ia_ifp; 332 mtu = ifp->if_mtu; 333 ip->ip_ttl = 1; 334 } else if ((IN_MULTICAST(ip->ip_dst.s_addr) || 335 ip->ip_dst.s_addr == INADDR_BROADCAST) && 336 imo != NULL && imo->imo_multicast_ifp != NULL) { 337 ifp = imo->imo_multicast_ifp; 338 mtu = ifp->if_mtu; 339 IFP_TO_IA(ifp, ia); 340 } else { 341 if (ro->ro_rt == NULL) 342 rtcache_init(ro); 343 if (ro->ro_rt == NULL) { 344 ipstat.ips_noroute++; 345 error = EHOSTUNREACH; 346 goto bad; 347 } 348 ia = ifatoia(ro->ro_rt->rt_ifa); 349 ifp = ro->ro_rt->rt_ifp; 350 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 351 mtu = ifp->if_mtu; 352 ro->ro_rt->rt_use++; 353 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 354 dst = satosin(ro->ro_rt->rt_gateway); 355 } 356 if (IN_MULTICAST(ip->ip_dst.s_addr) || 357 (ip->ip_dst.s_addr == INADDR_BROADCAST)) { 358 struct in_multi *inm; 359 360 m->m_flags |= (ip->ip_dst.s_addr == INADDR_BROADCAST) ? 361 M_BCAST : M_MCAST; 362 /* 363 * See if the caller provided any multicast options 364 */ 365 if (imo != NULL) 366 ip->ip_ttl = imo->imo_multicast_ttl; 367 else 368 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 369 370 /* 371 * if we don't know the outgoing ifp yet, we can't generate 372 * output 373 */ 374 if (!ifp) { 375 ipstat.ips_noroute++; 376 error = ENETUNREACH; 377 goto bad; 378 } 379 380 /* 381 * If the packet is multicast or broadcast, confirm that 382 * the outgoing interface can transmit it. 383 */ 384 if (((m->m_flags & M_MCAST) && 385 (ifp->if_flags & IFF_MULTICAST) == 0) || 386 ((m->m_flags & M_BCAST) && 387 (ifp->if_flags & (IFF_BROADCAST|IFF_POINTOPOINT)) == 0)) { 388 ipstat.ips_noroute++; 389 error = ENETUNREACH; 390 goto bad; 391 } 392 /* 393 * If source address not specified yet, use an address 394 * of outgoing interface. 395 */ 396 if (in_nullhost(ip->ip_src)) { 397 struct in_ifaddr *xia; 398 399 IFP_TO_IA(ifp, xia); 400 if (!xia) { 401 error = EADDRNOTAVAIL; 402 goto bad; 403 } 404 xifa = &xia->ia_ifa; 405 if (xifa->ifa_getifa != NULL) { 406 xia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 407 } 408 ip->ip_src = xia->ia_addr.sin_addr; 409 } 410 411 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 412 if (inm != NULL && 413 (imo == NULL || imo->imo_multicast_loop)) { 414 /* 415 * If we belong to the destination multicast group 416 * on the outgoing interface, and the caller did not 417 * forbid loopback, loop back a copy. 418 */ 419 ip_mloopback(ifp, m, &u.dst4); 420 } 421 #ifdef MROUTING 422 else { 423 /* 424 * If we are acting as a multicast router, perform 425 * multicast forwarding as if the packet had just 426 * arrived on the interface to which we are about 427 * to send. The multicast forwarding function 428 * recursively calls this function, using the 429 * IP_FORWARDING flag to prevent infinite recursion. 430 * 431 * Multicasts that are looped back by ip_mloopback(), 432 * above, will be forwarded by the ip_input() routine, 433 * if necessary. 434 */ 435 extern struct socket *ip_mrouter; 436 437 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 438 if (ip_mforward(m, ifp) != 0) { 439 m_freem(m); 440 goto done; 441 } 442 } 443 } 444 #endif 445 /* 446 * Multicasts with a time-to-live of zero may be looped- 447 * back, above, but must not be transmitted on a network. 448 * Also, multicasts addressed to the loopback interface 449 * are not sent -- the above call to ip_mloopback() will 450 * loop back a copy if this host actually belongs to the 451 * destination group on the loopback interface. 452 */ 453 if (ip->ip_ttl == 0 || (ifp->if_flags & IFF_LOOPBACK) != 0) { 454 m_freem(m); 455 goto done; 456 } 457 458 goto sendit; 459 } 460 /* 461 * If source address not specified yet, use address 462 * of outgoing interface. 463 */ 464 if (in_nullhost(ip->ip_src)) { 465 xifa = &ia->ia_ifa; 466 if (xifa->ifa_getifa != NULL) 467 ia = ifatoia((*xifa->ifa_getifa)(xifa, rdst)); 468 ip->ip_src = ia->ia_addr.sin_addr; 469 } 470 471 /* 472 * packets with Class-D address as source are not valid per 473 * RFC 1112 474 */ 475 if (IN_MULTICAST(ip->ip_src.s_addr)) { 476 ipstat.ips_odropped++; 477 error = EADDRNOTAVAIL; 478 goto bad; 479 } 480 481 /* 482 * Look for broadcast address and 483 * and verify user is allowed to send 484 * such a packet. 485 */ 486 if (in_broadcast(dst->sin_addr, ifp)) { 487 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 488 error = EADDRNOTAVAIL; 489 goto bad; 490 } 491 if ((flags & IP_ALLOWBROADCAST) == 0) { 492 error = EACCES; 493 goto bad; 494 } 495 /* don't allow broadcast messages to be fragmented */ 496 if (ntohs(ip->ip_len) > ifp->if_mtu) { 497 error = EMSGSIZE; 498 goto bad; 499 } 500 m->m_flags |= M_BCAST; 501 } else 502 m->m_flags &= ~M_BCAST; 503 504 sendit: 505 /* 506 * If we're doing Path MTU Discovery, we need to set DF unless 507 * the route's MTU is locked. 508 */ 509 if ((flags & IP_MTUDISC) != 0 && ro->ro_rt != NULL && 510 (ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0) 511 ip->ip_off |= htons(IP_DF); 512 513 /* Remember the current ip_len */ 514 ip_len = ntohs(ip->ip_len); 515 516 #ifdef IPSEC 517 /* get SP for this packet */ 518 if (so == NULL) 519 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 520 flags, &error); 521 else { 522 if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp, 523 IPSEC_DIR_OUTBOUND)) 524 goto skip_ipsec; 525 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 526 } 527 528 if (sp == NULL) { 529 ipsecstat.out_inval++; 530 goto bad; 531 } 532 533 error = 0; 534 535 /* check policy */ 536 switch (sp->policy) { 537 case IPSEC_POLICY_DISCARD: 538 /* 539 * This packet is just discarded. 540 */ 541 ipsecstat.out_polvio++; 542 goto bad; 543 544 case IPSEC_POLICY_BYPASS: 545 case IPSEC_POLICY_NONE: 546 /* no need to do IPsec. */ 547 goto skip_ipsec; 548 549 case IPSEC_POLICY_IPSEC: 550 if (sp->req == NULL) { 551 /* XXX should be panic ? */ 552 printf("ip_output: No IPsec request specified.\n"); 553 error = EINVAL; 554 goto bad; 555 } 556 break; 557 558 case IPSEC_POLICY_ENTRUST: 559 default: 560 printf("ip_output: Invalid policy found. %d\n", sp->policy); 561 } 562 563 #ifdef IPSEC_NAT_T 564 /* 565 * NAT-T ESP fragmentation: don't do IPSec processing now, 566 * we'll do it on each fragmented packet. 567 */ 568 if (sp->req->sav && 569 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 570 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 571 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 572 natt_frag = 1; 573 mtu = sp->req->sav->esp_frag; 574 goto skip_ipsec; 575 } 576 } 577 #endif /* IPSEC_NAT_T */ 578 579 /* 580 * ipsec4_output() expects ip_len and ip_off in network 581 * order. They have been set to network order above. 582 */ 583 584 { 585 struct ipsec_output_state state; 586 bzero(&state, sizeof(state)); 587 state.m = m; 588 if (flags & IP_ROUTETOIF) { 589 state.ro = &iproute; 590 memset(&iproute, 0, sizeof(iproute)); 591 } else 592 state.ro = ro; 593 state.dst = sintocsa(dst); 594 595 /* 596 * We can't defer the checksum of payload data if 597 * we're about to encrypt/authenticate it. 598 * 599 * XXX When we support crypto offloading functions of 600 * XXX network interfaces, we need to reconsider this, 601 * XXX since it's likely that they'll support checksumming, 602 * XXX as well. 603 */ 604 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 605 in_delayed_cksum(m); 606 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 607 } 608 609 error = ipsec4_output(&state, sp, flags); 610 611 m = state.m; 612 if (flags & IP_ROUTETOIF) { 613 /* 614 * if we have tunnel mode SA, we may need to ignore 615 * IP_ROUTETOIF. 616 */ 617 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 618 flags &= ~IP_ROUTETOIF; 619 ro = state.ro; 620 } 621 } else 622 ro = state.ro; 623 dst = satocsin(state.dst); 624 if (error) { 625 /* mbuf is already reclaimed in ipsec4_output. */ 626 m0 = NULL; 627 switch (error) { 628 case EHOSTUNREACH: 629 case ENETUNREACH: 630 case EMSGSIZE: 631 case ENOBUFS: 632 case ENOMEM: 633 break; 634 default: 635 printf("ip4_output (ipsec): error code %d\n", error); 636 /*fall through*/ 637 case ENOENT: 638 /* don't show these error codes to the user */ 639 error = 0; 640 break; 641 } 642 goto bad; 643 } 644 645 /* be sure to update variables that are affected by ipsec4_output() */ 646 ip = mtod(m, struct ip *); 647 hlen = ip->ip_hl << 2; 648 ip_len = ntohs(ip->ip_len); 649 650 if (ro->ro_rt == NULL) { 651 if ((flags & IP_ROUTETOIF) == 0) { 652 printf("ip_output: " 653 "can't update route after IPsec processing\n"); 654 error = EHOSTUNREACH; /*XXX*/ 655 goto bad; 656 } 657 } else { 658 /* nobody uses ia beyond here */ 659 if (state.encap) { 660 ifp = ro->ro_rt->rt_ifp; 661 if ((mtu = ro->ro_rt->rt_rmx.rmx_mtu) == 0) 662 mtu = ifp->if_mtu; 663 } 664 } 665 } 666 skip_ipsec: 667 #endif /*IPSEC*/ 668 #ifdef FAST_IPSEC 669 /* 670 * Check the security policy (SP) for the packet and, if 671 * required, do IPsec-related processing. There are two 672 * cases here; the first time a packet is sent through 673 * it will be untagged and handled by ipsec4_checkpolicy. 674 * If the packet is resubmitted to ip_output (e.g. after 675 * AH, ESP, etc. processing), there will be a tag to bypass 676 * the lookup and related policy checking. 677 */ 678 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 679 s = splsoftnet(); 680 if (mtag != NULL) { 681 tdbi = (struct tdb_ident *)(mtag + 1); 682 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 683 if (sp == NULL) 684 error = -EINVAL; /* force silent drop */ 685 m_tag_delete(m, mtag); 686 } else { 687 if (inp != NULL && 688 IPSEC_PCB_SKIP_IPSEC(inp->inp_sp, IPSEC_DIR_OUTBOUND)) 689 goto spd_done; 690 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 691 &error, inp); 692 } 693 /* 694 * There are four return cases: 695 * sp != NULL apply IPsec policy 696 * sp == NULL, error == 0 no IPsec handling needed 697 * sp == NULL, error == -EINVAL discard packet w/o error 698 * sp == NULL, error != 0 discard packet, report error 699 */ 700 if (sp != NULL) { 701 #ifdef IPSEC_NAT_T 702 /* 703 * NAT-T ESP fragmentation: don't do IPSec processing now, 704 * we'll do it on each fragmented packet. 705 */ 706 if (sp->req->sav && 707 ((sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP) || 708 (sp->req->sav->natt_type & UDP_ENCAP_ESPINUDP_NON_IKE))) { 709 if (ntohs(ip->ip_len) > sp->req->sav->esp_frag) { 710 natt_frag = 1; 711 mtu = sp->req->sav->esp_frag; 712 goto spd_done; 713 } 714 } 715 #endif /* IPSEC_NAT_T */ 716 /* Loop detection, check if ipsec processing already done */ 717 IPSEC_ASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 718 for (mtag = m_tag_first(m); mtag != NULL; 719 mtag = m_tag_next(m, mtag)) { 720 #ifdef MTAG_ABI_COMPAT 721 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 722 continue; 723 #endif 724 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 725 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 726 continue; 727 /* 728 * Check if policy has an SA associated with it. 729 * This can happen when an SP has yet to acquire 730 * an SA; e.g. on first reference. If it occurs, 731 * then we let ipsec4_process_packet do its thing. 732 */ 733 if (sp->req->sav == NULL) 734 break; 735 tdbi = (struct tdb_ident *)(mtag + 1); 736 if (tdbi->spi == sp->req->sav->spi && 737 tdbi->proto == sp->req->sav->sah->saidx.proto && 738 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 739 sizeof (union sockaddr_union)) == 0) { 740 /* 741 * No IPsec processing is needed, free 742 * reference to SP. 743 * 744 * NB: null pointer to avoid free at 745 * done: below. 746 */ 747 KEY_FREESP(&sp), sp = NULL; 748 splx(s); 749 goto spd_done; 750 } 751 } 752 753 /* 754 * Do delayed checksums now because we send before 755 * this is done in the normal processing path. 756 */ 757 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 758 in_delayed_cksum(m); 759 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 760 } 761 762 #ifdef __FreeBSD__ 763 ip->ip_len = htons(ip->ip_len); 764 ip->ip_off = htons(ip->ip_off); 765 #endif 766 767 /* NB: callee frees mbuf */ 768 error = ipsec4_process_packet(m, sp->req, flags, 0); 769 /* 770 * Preserve KAME behaviour: ENOENT can be returned 771 * when an SA acquire is in progress. Don't propagate 772 * this to user-level; it confuses applications. 773 * 774 * XXX this will go away when the SADB is redone. 775 */ 776 if (error == ENOENT) 777 error = 0; 778 splx(s); 779 goto done; 780 } else { 781 splx(s); 782 783 if (error != 0) { 784 /* 785 * Hack: -EINVAL is used to signal that a packet 786 * should be silently discarded. This is typically 787 * because we asked key management for an SA and 788 * it was delayed (e.g. kicked up to IKE). 789 */ 790 if (error == -EINVAL) 791 error = 0; 792 goto bad; 793 } else { 794 /* No IPsec processing for this packet. */ 795 } 796 #ifdef notyet 797 /* 798 * If deferred crypto processing is needed, check that 799 * the interface supports it. 800 */ 801 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 802 if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 803 /* notify IPsec to do its own crypto */ 804 ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 805 error = EHOSTUNREACH; 806 goto bad; 807 } 808 #endif 809 } 810 spd_done: 811 #endif /* FAST_IPSEC */ 812 813 #ifdef PFIL_HOOKS 814 /* 815 * Run through list of hooks for output packets. 816 */ 817 if ((error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT)) != 0) 818 goto done; 819 if (m == NULL) 820 goto done; 821 822 ip = mtod(m, struct ip *); 823 hlen = ip->ip_hl << 2; 824 ip_len = ntohs(ip->ip_len); 825 #endif /* PFIL_HOOKS */ 826 827 m->m_pkthdr.csum_data |= hlen << 16; 828 829 #if IFA_STATS 830 /* 831 * search for the source address structure to 832 * maintain output statistics. 833 */ 834 INADDR_TO_IA(ip->ip_src, ia); 835 #endif 836 837 /* Maybe skip checksums on loopback interfaces. */ 838 if (IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)) { 839 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 840 } 841 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx; 842 /* 843 * If small enough for mtu of path, or if using TCP segmentation 844 * offload, can just send directly. 845 */ 846 if (ip_len <= mtu || 847 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) != 0) { 848 #if IFA_STATS 849 if (ia) 850 ia->ia_ifa.ifa_data.ifad_outbytes += ip_len; 851 #endif 852 /* 853 * Always initialize the sum to 0! Some HW assisted 854 * checksumming requires this. 855 */ 856 ip->ip_sum = 0; 857 858 if ((m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0) { 859 /* 860 * Perform any checksums that the hardware can't do 861 * for us. 862 * 863 * XXX Does any hardware require the {th,uh}_sum 864 * XXX fields to be 0? 865 */ 866 if (sw_csum & M_CSUM_IPv4) { 867 KASSERT(IN_NEED_CHECKSUM(ifp, M_CSUM_IPv4)); 868 ip->ip_sum = in_cksum(m, hlen); 869 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 870 } 871 if (sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 872 if (IN_NEED_CHECKSUM(ifp, 873 sw_csum & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 874 in_delayed_cksum(m); 875 } 876 m->m_pkthdr.csum_flags &= 877 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 878 } 879 } 880 881 #ifdef IPSEC 882 /* clean ipsec history once it goes out of the node */ 883 ipsec_delaux(m); 884 #endif 885 886 if (__predict_true( 887 (m->m_pkthdr.csum_flags & M_CSUM_TSOv4) == 0 || 888 (ifp->if_capenable & IFCAP_TSOv4) != 0)) { 889 error = 890 (*ifp->if_output)(ifp, m, 891 (m->m_flags & M_MCAST) ? 892 sintocsa(rdst) : sintocsa(dst), 893 ro->ro_rt); 894 } else { 895 error = 896 ip_tso_output(ifp, m, 897 (m->m_flags & M_MCAST) ? 898 sintocsa(rdst) : sintocsa(dst), 899 ro->ro_rt); 900 } 901 goto done; 902 } 903 904 /* 905 * We can't use HW checksumming if we're about to 906 * to fragment the packet. 907 * 908 * XXX Some hardware can do this. 909 */ 910 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 911 if (IN_NEED_CHECKSUM(ifp, 912 m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4))) { 913 in_delayed_cksum(m); 914 } 915 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 916 } 917 918 /* 919 * Too large for interface; fragment if possible. 920 * Must be able to put at least 8 bytes per fragment. 921 */ 922 if (ntohs(ip->ip_off) & IP_DF) { 923 if (flags & IP_RETURNMTU) 924 *mtu_p = mtu; 925 error = EMSGSIZE; 926 ipstat.ips_cantfrag++; 927 goto bad; 928 } 929 930 error = ip_fragment(m, ifp, mtu); 931 if (error) { 932 m = NULL; 933 goto bad; 934 } 935 936 for (; m; m = m0) { 937 m0 = m->m_nextpkt; 938 m->m_nextpkt = 0; 939 if (error == 0) { 940 #if IFA_STATS 941 if (ia) 942 ia->ia_ifa.ifa_data.ifad_outbytes += 943 ntohs(ip->ip_len); 944 #endif 945 #ifdef IPSEC 946 /* clean ipsec history once it goes out of the node */ 947 ipsec_delaux(m); 948 #endif /* IPSEC */ 949 950 #ifdef IPSEC_NAT_T 951 /* 952 * If we get there, the packet has not been handeld by 953 * IPSec whereas it should have. Now that it has been 954 * fragmented, re-inject it in ip_output so that IPsec 955 * processing can occur. 956 */ 957 if (natt_frag) { 958 error = ip_output(m, opt, 959 ro, flags, imo, so, mtu_p); 960 } else 961 #endif /* IPSEC_NAT_T */ 962 { 963 KASSERT((m->m_pkthdr.csum_flags & 964 (M_CSUM_UDPv4 | M_CSUM_TCPv4)) == 0); 965 error = (*ifp->if_output)(ifp, m, 966 (m->m_flags & M_MCAST) ? 967 sintocsa(rdst) : sintocsa(dst), 968 ro->ro_rt); 969 } 970 } else 971 m_freem(m); 972 } 973 974 if (error == 0) 975 ipstat.ips_fragmented++; 976 done: 977 rtcache_free(&iproute); 978 979 #ifdef IPSEC 980 if (sp != NULL) { 981 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 982 printf("DP ip_output call free SP:%p\n", sp)); 983 key_freesp(sp); 984 } 985 #endif /* IPSEC */ 986 #ifdef FAST_IPSEC 987 if (sp != NULL) 988 KEY_FREESP(&sp); 989 #endif /* FAST_IPSEC */ 990 991 return (error); 992 bad: 993 m_freem(m); 994 goto done; 995 } 996 997 int 998 ip_fragment(struct mbuf *m, struct ifnet *ifp, u_long mtu) 999 { 1000 struct ip *ip, *mhip; 1001 struct mbuf *m0; 1002 int len, hlen, off; 1003 int mhlen, firstlen; 1004 struct mbuf **mnext; 1005 int sw_csum = m->m_pkthdr.csum_flags; 1006 int fragments = 0; 1007 int s; 1008 int error = 0; 1009 1010 ip = mtod(m, struct ip *); 1011 hlen = ip->ip_hl << 2; 1012 if (ifp != NULL) 1013 sw_csum &= ~ifp->if_csum_flags_tx; 1014 1015 len = (mtu - hlen) &~ 7; 1016 if (len < 8) { 1017 m_freem(m); 1018 return (EMSGSIZE); 1019 } 1020 1021 firstlen = len; 1022 mnext = &m->m_nextpkt; 1023 1024 /* 1025 * Loop through length of segment after first fragment, 1026 * make new header and copy data of each part and link onto chain. 1027 */ 1028 m0 = m; 1029 mhlen = sizeof (struct ip); 1030 for (off = hlen + len; off < ntohs(ip->ip_len); off += len) { 1031 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1032 if (m == 0) { 1033 error = ENOBUFS; 1034 ipstat.ips_odropped++; 1035 goto sendorfree; 1036 } 1037 MCLAIM(m, m0->m_owner); 1038 *mnext = m; 1039 mnext = &m->m_nextpkt; 1040 m->m_data += max_linkhdr; 1041 mhip = mtod(m, struct ip *); 1042 *mhip = *ip; 1043 /* we must inherit MCAST and BCAST flags */ 1044 m->m_flags |= m0->m_flags & (M_MCAST|M_BCAST); 1045 if (hlen > sizeof (struct ip)) { 1046 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1047 mhip->ip_hl = mhlen >> 2; 1048 } 1049 m->m_len = mhlen; 1050 mhip->ip_off = ((off - hlen) >> 3) + 1051 (ntohs(ip->ip_off) & ~IP_MF); 1052 if (ip->ip_off & htons(IP_MF)) 1053 mhip->ip_off |= IP_MF; 1054 if (off + len >= ntohs(ip->ip_len)) 1055 len = ntohs(ip->ip_len) - off; 1056 else 1057 mhip->ip_off |= IP_MF; 1058 HTONS(mhip->ip_off); 1059 mhip->ip_len = htons((u_int16_t)(len + mhlen)); 1060 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 1061 if (m->m_next == 0) { 1062 error = ENOBUFS; /* ??? */ 1063 ipstat.ips_odropped++; 1064 goto sendorfree; 1065 } 1066 m->m_pkthdr.len = mhlen + len; 1067 m->m_pkthdr.rcvif = (struct ifnet *)0; 1068 mhip->ip_sum = 0; 1069 if (sw_csum & M_CSUM_IPv4) { 1070 mhip->ip_sum = in_cksum(m, mhlen); 1071 KASSERT((m->m_pkthdr.csum_flags & M_CSUM_IPv4) == 0); 1072 } else { 1073 m->m_pkthdr.csum_flags |= M_CSUM_IPv4; 1074 m->m_pkthdr.csum_data |= mhlen << 16; 1075 } 1076 ipstat.ips_ofragments++; 1077 fragments++; 1078 } 1079 /* 1080 * Update first fragment by trimming what's been copied out 1081 * and updating header, then send each fragment (in order). 1082 */ 1083 m = m0; 1084 m_adj(m, hlen + firstlen - ntohs(ip->ip_len)); 1085 m->m_pkthdr.len = hlen + firstlen; 1086 ip->ip_len = htons((u_int16_t)m->m_pkthdr.len); 1087 ip->ip_off |= htons(IP_MF); 1088 ip->ip_sum = 0; 1089 if (sw_csum & M_CSUM_IPv4) { 1090 ip->ip_sum = in_cksum(m, hlen); 1091 m->m_pkthdr.csum_flags &= ~M_CSUM_IPv4; 1092 } else { 1093 KASSERT(m->m_pkthdr.csum_flags & M_CSUM_IPv4); 1094 KASSERT(M_CSUM_DATA_IPv4_IPHL(m->m_pkthdr.csum_data) >= 1095 sizeof(struct ip)); 1096 } 1097 sendorfree: 1098 /* 1099 * If there is no room for all the fragments, don't queue 1100 * any of them. 1101 */ 1102 if (ifp != NULL) { 1103 s = splnet(); 1104 if (ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len < fragments && 1105 error == 0) { 1106 error = ENOBUFS; 1107 ipstat.ips_odropped++; 1108 IFQ_INC_DROPS(&ifp->if_snd); 1109 } 1110 splx(s); 1111 } 1112 if (error) { 1113 for (m = m0; m; m = m0) { 1114 m0 = m->m_nextpkt; 1115 m->m_nextpkt = NULL; 1116 m_freem(m); 1117 } 1118 } 1119 return (error); 1120 } 1121 1122 /* 1123 * Process a delayed payload checksum calculation. 1124 */ 1125 void 1126 in_delayed_cksum(struct mbuf *m) 1127 { 1128 struct ip *ip; 1129 u_int16_t csum, offset; 1130 1131 ip = mtod(m, struct ip *); 1132 offset = ip->ip_hl << 2; 1133 csum = in4_cksum(m, 0, offset, ntohs(ip->ip_len) - offset); 1134 if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv4) != 0) 1135 csum = 0xffff; 1136 1137 offset += M_CSUM_DATA_IPv4_OFFSET(m->m_pkthdr.csum_data); 1138 1139 if ((offset + sizeof(u_int16_t)) > m->m_len) { 1140 /* This happen when ip options were inserted 1141 printf("in_delayed_cksum: pullup len %d off %d proto %d\n", 1142 m->m_len, offset, ip->ip_p); 1143 */ 1144 m_copyback(m, offset, sizeof(csum), (void *) &csum); 1145 } else 1146 *(u_int16_t *)(mtod(m, char *) + offset) = csum; 1147 } 1148 1149 /* 1150 * Determine the maximum length of the options to be inserted; 1151 * we would far rather allocate too much space rather than too little. 1152 */ 1153 1154 u_int 1155 ip_optlen(struct inpcb *inp) 1156 { 1157 struct mbuf *m = inp->inp_options; 1158 1159 if (m && m->m_len > offsetof(struct ipoption, ipopt_dst)) 1160 return (m->m_len - offsetof(struct ipoption, ipopt_dst)); 1161 else 1162 return 0; 1163 } 1164 1165 1166 /* 1167 * Insert IP options into preformed packet. 1168 * Adjust IP destination as required for IP source routing, 1169 * as indicated by a non-zero in_addr at the start of the options. 1170 */ 1171 static struct mbuf * 1172 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1173 { 1174 struct ipoption *p = mtod(opt, struct ipoption *); 1175 struct mbuf *n; 1176 struct ip *ip = mtod(m, struct ip *); 1177 unsigned optlen; 1178 1179 optlen = opt->m_len - sizeof(p->ipopt_dst); 1180 if (optlen + ntohs(ip->ip_len) > IP_MAXPACKET) 1181 return (m); /* XXX should fail */ 1182 if (!in_nullhost(p->ipopt_dst)) 1183 ip->ip_dst = p->ipopt_dst; 1184 if (M_READONLY(m) || M_LEADINGSPACE(m) < optlen) { 1185 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1186 if (n == 0) 1187 return (m); 1188 MCLAIM(n, m->m_owner); 1189 M_MOVE_PKTHDR(n, m); 1190 m->m_len -= sizeof(struct ip); 1191 m->m_data += sizeof(struct ip); 1192 n->m_next = m; 1193 m = n; 1194 m->m_len = optlen + sizeof(struct ip); 1195 m->m_data += max_linkhdr; 1196 bcopy((void *)ip, mtod(m, void *), sizeof(struct ip)); 1197 } else { 1198 m->m_data -= optlen; 1199 m->m_len += optlen; 1200 memmove(mtod(m, void *), ip, sizeof(struct ip)); 1201 } 1202 m->m_pkthdr.len += optlen; 1203 ip = mtod(m, struct ip *); 1204 bcopy((void *)p->ipopt_list, (void *)(ip + 1), (unsigned)optlen); 1205 *phlen = sizeof(struct ip) + optlen; 1206 ip->ip_len = htons(ntohs(ip->ip_len) + optlen); 1207 return (m); 1208 } 1209 1210 /* 1211 * Copy options from ip to jp, 1212 * omitting those not copied during fragmentation. 1213 */ 1214 int 1215 ip_optcopy(struct ip *ip, struct ip *jp) 1216 { 1217 u_char *cp, *dp; 1218 int opt, optlen, cnt; 1219 1220 cp = (u_char *)(ip + 1); 1221 dp = (u_char *)(jp + 1); 1222 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1223 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1224 opt = cp[0]; 1225 if (opt == IPOPT_EOL) 1226 break; 1227 if (opt == IPOPT_NOP) { 1228 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1229 *dp++ = IPOPT_NOP; 1230 optlen = 1; 1231 continue; 1232 } 1233 #ifdef DIAGNOSTIC 1234 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1235 panic("malformed IPv4 option passed to ip_optcopy"); 1236 #endif 1237 optlen = cp[IPOPT_OLEN]; 1238 #ifdef DIAGNOSTIC 1239 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1240 panic("malformed IPv4 option passed to ip_optcopy"); 1241 #endif 1242 /* bogus lengths should have been caught by ip_dooptions */ 1243 if (optlen > cnt) 1244 optlen = cnt; 1245 if (IPOPT_COPIED(opt)) { 1246 bcopy((void *)cp, (void *)dp, (unsigned)optlen); 1247 dp += optlen; 1248 } 1249 } 1250 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1251 *dp++ = IPOPT_EOL; 1252 return (optlen); 1253 } 1254 1255 /* 1256 * IP socket option processing. 1257 */ 1258 int 1259 ip_ctloutput(int op, struct socket *so, int level, int optname, 1260 struct mbuf **mp) 1261 { 1262 struct inpcb *inp = sotoinpcb(so); 1263 struct mbuf *m = *mp; 1264 int optval = 0; 1265 int error = 0; 1266 #if defined(IPSEC) || defined(FAST_IPSEC) 1267 struct lwp *l = curlwp; /*XXX*/ 1268 #endif 1269 1270 if (level != IPPROTO_IP) { 1271 if (op == PRCO_SETOPT && *mp) 1272 (void) m_free(*mp); 1273 if (level == SOL_SOCKET && optname == SO_NOHEADER) 1274 return 0; 1275 return ENOPROTOOPT; 1276 } 1277 1278 switch (op) { 1279 1280 case PRCO_SETOPT: 1281 switch (optname) { 1282 case IP_OPTIONS: 1283 #ifdef notyet 1284 case IP_RETOPTS: 1285 return (ip_pcbopts(optname, &inp->inp_options, m)); 1286 #else 1287 return (ip_pcbopts(&inp->inp_options, m)); 1288 #endif 1289 1290 case IP_TOS: 1291 case IP_TTL: 1292 case IP_RECVOPTS: 1293 case IP_RECVRETOPTS: 1294 case IP_RECVDSTADDR: 1295 case IP_RECVIF: 1296 if (m == NULL || m->m_len != sizeof(int)) 1297 error = EINVAL; 1298 else { 1299 optval = *mtod(m, int *); 1300 switch (optname) { 1301 1302 case IP_TOS: 1303 inp->inp_ip.ip_tos = optval; 1304 break; 1305 1306 case IP_TTL: 1307 inp->inp_ip.ip_ttl = optval; 1308 break; 1309 #define OPTSET(bit) \ 1310 if (optval) \ 1311 inp->inp_flags |= bit; \ 1312 else \ 1313 inp->inp_flags &= ~bit; 1314 1315 case IP_RECVOPTS: 1316 OPTSET(INP_RECVOPTS); 1317 break; 1318 1319 case IP_RECVRETOPTS: 1320 OPTSET(INP_RECVRETOPTS); 1321 break; 1322 1323 case IP_RECVDSTADDR: 1324 OPTSET(INP_RECVDSTADDR); 1325 break; 1326 1327 case IP_RECVIF: 1328 OPTSET(INP_RECVIF); 1329 break; 1330 } 1331 } 1332 break; 1333 #undef OPTSET 1334 1335 case IP_MULTICAST_IF: 1336 case IP_MULTICAST_TTL: 1337 case IP_MULTICAST_LOOP: 1338 case IP_ADD_MEMBERSHIP: 1339 case IP_DROP_MEMBERSHIP: 1340 error = ip_setmoptions(optname, &inp->inp_moptions, m); 1341 break; 1342 1343 case IP_PORTRANGE: 1344 if (m == 0 || m->m_len != sizeof(int)) 1345 error = EINVAL; 1346 else { 1347 optval = *mtod(m, int *); 1348 1349 switch (optval) { 1350 1351 case IP_PORTRANGE_DEFAULT: 1352 case IP_PORTRANGE_HIGH: 1353 inp->inp_flags &= ~(INP_LOWPORT); 1354 break; 1355 1356 case IP_PORTRANGE_LOW: 1357 inp->inp_flags |= INP_LOWPORT; 1358 break; 1359 1360 default: 1361 error = EINVAL; 1362 break; 1363 } 1364 } 1365 break; 1366 1367 #if defined(IPSEC) || defined(FAST_IPSEC) 1368 case IP_IPSEC_POLICY: 1369 { 1370 void *req = NULL; 1371 size_t len = 0; 1372 int priv = 0; 1373 1374 #ifdef __NetBSD__ 1375 if (l == 0 || kauth_authorize_generic(l->l_cred, 1376 KAUTH_GENERIC_ISSUSER, NULL)) 1377 priv = 0; 1378 else 1379 priv = 1; 1380 #else 1381 priv = (in6p->in6p_socket->so_state & SS_PRIV); 1382 #endif 1383 if (m) { 1384 req = mtod(m, void *); 1385 len = m->m_len; 1386 } 1387 error = ipsec4_set_policy(inp, optname, req, len, priv); 1388 break; 1389 } 1390 #endif /*IPSEC*/ 1391 1392 default: 1393 error = ENOPROTOOPT; 1394 break; 1395 } 1396 if (m) 1397 (void)m_free(m); 1398 break; 1399 1400 case PRCO_GETOPT: 1401 switch (optname) { 1402 case IP_OPTIONS: 1403 case IP_RETOPTS: 1404 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1405 MCLAIM(m, so->so_mowner); 1406 if (inp->inp_options) { 1407 m->m_len = inp->inp_options->m_len; 1408 bcopy(mtod(inp->inp_options, void *), 1409 mtod(m, void *), (unsigned)m->m_len); 1410 } else 1411 m->m_len = 0; 1412 break; 1413 1414 case IP_TOS: 1415 case IP_TTL: 1416 case IP_RECVOPTS: 1417 case IP_RECVRETOPTS: 1418 case IP_RECVDSTADDR: 1419 case IP_RECVIF: 1420 case IP_ERRORMTU: 1421 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1422 MCLAIM(m, so->so_mowner); 1423 m->m_len = sizeof(int); 1424 switch (optname) { 1425 1426 case IP_TOS: 1427 optval = inp->inp_ip.ip_tos; 1428 break; 1429 1430 case IP_TTL: 1431 optval = inp->inp_ip.ip_ttl; 1432 break; 1433 1434 case IP_ERRORMTU: 1435 optval = inp->inp_errormtu; 1436 break; 1437 1438 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1439 1440 case IP_RECVOPTS: 1441 optval = OPTBIT(INP_RECVOPTS); 1442 break; 1443 1444 case IP_RECVRETOPTS: 1445 optval = OPTBIT(INP_RECVRETOPTS); 1446 break; 1447 1448 case IP_RECVDSTADDR: 1449 optval = OPTBIT(INP_RECVDSTADDR); 1450 break; 1451 1452 case IP_RECVIF: 1453 optval = OPTBIT(INP_RECVIF); 1454 break; 1455 } 1456 *mtod(m, int *) = optval; 1457 break; 1458 1459 #if 0 /* defined(IPSEC) || defined(FAST_IPSEC) */ 1460 /* XXX: code broken */ 1461 case IP_IPSEC_POLICY: 1462 { 1463 void *req = NULL; 1464 size_t len = 0; 1465 1466 if (m) { 1467 req = mtod(m, void *); 1468 len = m->m_len; 1469 } 1470 error = ipsec4_get_policy(inp, req, len, mp); 1471 break; 1472 } 1473 #endif /*IPSEC*/ 1474 1475 case IP_MULTICAST_IF: 1476 case IP_MULTICAST_TTL: 1477 case IP_MULTICAST_LOOP: 1478 case IP_ADD_MEMBERSHIP: 1479 case IP_DROP_MEMBERSHIP: 1480 error = ip_getmoptions(optname, inp->inp_moptions, mp); 1481 if (*mp) 1482 MCLAIM(*mp, so->so_mowner); 1483 break; 1484 1485 case IP_PORTRANGE: 1486 *mp = m = m_get(M_WAIT, MT_SOOPTS); 1487 MCLAIM(m, so->so_mowner); 1488 m->m_len = sizeof(int); 1489 1490 if (inp->inp_flags & INP_LOWPORT) 1491 optval = IP_PORTRANGE_LOW; 1492 else 1493 optval = IP_PORTRANGE_DEFAULT; 1494 1495 *mtod(m, int *) = optval; 1496 break; 1497 1498 default: 1499 error = ENOPROTOOPT; 1500 break; 1501 } 1502 break; 1503 } 1504 return (error); 1505 } 1506 1507 /* 1508 * Set up IP options in pcb for insertion in output packets. 1509 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1510 * with destination address if source routed. 1511 */ 1512 int 1513 #ifdef notyet 1514 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m) 1515 #else 1516 ip_pcbopts(struct mbuf **pcbopt, struct mbuf *m) 1517 #endif 1518 { 1519 int cnt, optlen; 1520 u_char *cp; 1521 u_char opt; 1522 1523 /* turn off any old options */ 1524 if (*pcbopt) 1525 (void)m_free(*pcbopt); 1526 *pcbopt = 0; 1527 if (m == (struct mbuf *)0 || m->m_len == 0) { 1528 /* 1529 * Only turning off any previous options. 1530 */ 1531 if (m) 1532 (void)m_free(m); 1533 return (0); 1534 } 1535 1536 #ifndef __vax__ 1537 if (m->m_len % sizeof(int32_t)) 1538 goto bad; 1539 #endif 1540 /* 1541 * IP first-hop destination address will be stored before 1542 * actual options; move other options back 1543 * and clear it when none present. 1544 */ 1545 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1546 goto bad; 1547 cnt = m->m_len; 1548 m->m_len += sizeof(struct in_addr); 1549 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1550 memmove(cp, mtod(m, void *), (unsigned)cnt); 1551 bzero(mtod(m, void *), sizeof(struct in_addr)); 1552 1553 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1554 opt = cp[IPOPT_OPTVAL]; 1555 if (opt == IPOPT_EOL) 1556 break; 1557 if (opt == IPOPT_NOP) 1558 optlen = 1; 1559 else { 1560 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1561 goto bad; 1562 optlen = cp[IPOPT_OLEN]; 1563 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1564 goto bad; 1565 } 1566 switch (opt) { 1567 1568 default: 1569 break; 1570 1571 case IPOPT_LSRR: 1572 case IPOPT_SSRR: 1573 /* 1574 * user process specifies route as: 1575 * ->A->B->C->D 1576 * D must be our final destination (but we can't 1577 * check that since we may not have connected yet). 1578 * A is first hop destination, which doesn't appear in 1579 * actual IP option, but is stored before the options. 1580 */ 1581 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1582 goto bad; 1583 m->m_len -= sizeof(struct in_addr); 1584 cnt -= sizeof(struct in_addr); 1585 optlen -= sizeof(struct in_addr); 1586 cp[IPOPT_OLEN] = optlen; 1587 /* 1588 * Move first hop before start of options. 1589 */ 1590 bcopy((void *)&cp[IPOPT_OFFSET+1], mtod(m, void *), 1591 sizeof(struct in_addr)); 1592 /* 1593 * Then copy rest of options back 1594 * to close up the deleted entry. 1595 */ 1596 (void)memmove(&cp[IPOPT_OFFSET+1], 1597 &cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1598 (unsigned)cnt - (IPOPT_MINOFF - 1)); 1599 break; 1600 } 1601 } 1602 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1603 goto bad; 1604 *pcbopt = m; 1605 return (0); 1606 1607 bad: 1608 (void)m_free(m); 1609 return (EINVAL); 1610 } 1611 1612 /* 1613 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1614 */ 1615 static struct ifnet * 1616 ip_multicast_if(struct in_addr *a, int *ifindexp) 1617 { 1618 int ifindex; 1619 struct ifnet *ifp = NULL; 1620 struct in_ifaddr *ia; 1621 1622 if (ifindexp) 1623 *ifindexp = 0; 1624 if (ntohl(a->s_addr) >> 24 == 0) { 1625 ifindex = ntohl(a->s_addr) & 0xffffff; 1626 if (ifindex < 0 || if_indexlim <= ifindex) 1627 return NULL; 1628 ifp = ifindex2ifnet[ifindex]; 1629 if (!ifp) 1630 return NULL; 1631 if (ifindexp) 1632 *ifindexp = ifindex; 1633 } else { 1634 LIST_FOREACH(ia, &IN_IFADDR_HASH(a->s_addr), ia_hash) { 1635 if (in_hosteq(ia->ia_addr.sin_addr, *a) && 1636 (ia->ia_ifp->if_flags & IFF_MULTICAST) != 0) { 1637 ifp = ia->ia_ifp; 1638 break; 1639 } 1640 } 1641 } 1642 return ifp; 1643 } 1644 1645 static int 1646 ip_getoptval(struct mbuf *m, u_int8_t *val, u_int maxval) 1647 { 1648 u_int tval; 1649 1650 if (m == NULL) 1651 return EINVAL; 1652 1653 switch (m->m_len) { 1654 case sizeof(u_char): 1655 tval = *(mtod(m, u_char *)); 1656 break; 1657 case sizeof(u_int): 1658 tval = *(mtod(m, u_int *)); 1659 break; 1660 default: 1661 return EINVAL; 1662 } 1663 1664 if (tval > maxval) 1665 return EINVAL; 1666 1667 *val = tval; 1668 return 0; 1669 } 1670 1671 /* 1672 * Set the IP multicast options in response to user setsockopt(). 1673 */ 1674 int 1675 ip_setmoptions(int optname, struct ip_moptions **imop, struct mbuf *m) 1676 { 1677 int error = 0; 1678 int i; 1679 struct in_addr addr; 1680 struct ip_mreq *mreq; 1681 struct ifnet *ifp; 1682 struct ip_moptions *imo = *imop; 1683 int ifindex; 1684 1685 if (imo == NULL) { 1686 /* 1687 * No multicast option buffer attached to the pcb; 1688 * allocate one and initialize to default values. 1689 */ 1690 imo = (struct ip_moptions *)malloc(sizeof(*imo), M_IPMOPTS, 1691 M_WAITOK); 1692 1693 if (imo == NULL) 1694 return (ENOBUFS); 1695 *imop = imo; 1696 imo->imo_multicast_ifp = NULL; 1697 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1698 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1699 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1700 imo->imo_num_memberships = 0; 1701 } 1702 1703 switch (optname) { 1704 1705 case IP_MULTICAST_IF: 1706 /* 1707 * Select the interface for outgoing multicast packets. 1708 */ 1709 if (m == NULL || m->m_len != sizeof(struct in_addr)) { 1710 error = EINVAL; 1711 break; 1712 } 1713 addr = *(mtod(m, struct in_addr *)); 1714 /* 1715 * INADDR_ANY is used to remove a previous selection. 1716 * When no interface is selected, a default one is 1717 * chosen every time a multicast packet is sent. 1718 */ 1719 if (in_nullhost(addr)) { 1720 imo->imo_multicast_ifp = NULL; 1721 break; 1722 } 1723 /* 1724 * The selected interface is identified by its local 1725 * IP address. Find the interface and confirm that 1726 * it supports multicasting. 1727 */ 1728 ifp = ip_multicast_if(&addr, &ifindex); 1729 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1730 error = EADDRNOTAVAIL; 1731 break; 1732 } 1733 imo->imo_multicast_ifp = ifp; 1734 if (ifindex) 1735 imo->imo_multicast_addr = addr; 1736 else 1737 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1738 break; 1739 1740 case IP_MULTICAST_TTL: 1741 /* 1742 * Set the IP time-to-live for outgoing multicast packets. 1743 */ 1744 error = ip_getoptval(m, &imo->imo_multicast_ttl, MAXTTL); 1745 break; 1746 1747 case IP_MULTICAST_LOOP: 1748 /* 1749 * Set the loopback flag for outgoing multicast packets. 1750 * Must be zero or one. 1751 */ 1752 error = ip_getoptval(m, &imo->imo_multicast_loop, 1); 1753 break; 1754 1755 case IP_ADD_MEMBERSHIP: 1756 /* 1757 * Add a multicast group membership. 1758 * Group must be a valid IP multicast address. 1759 */ 1760 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1761 error = EINVAL; 1762 break; 1763 } 1764 mreq = mtod(m, struct ip_mreq *); 1765 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1766 error = EINVAL; 1767 break; 1768 } 1769 /* 1770 * If no interface address was provided, use the interface of 1771 * the route to the given multicast address. 1772 */ 1773 if (in_nullhost(mreq->imr_interface)) { 1774 union { 1775 struct sockaddr dst; 1776 struct sockaddr_in dst4; 1777 } u; 1778 struct route ro; 1779 1780 memset(&ro, 0, sizeof(ro)); 1781 1782 sockaddr_in_init(&u.dst4, &mreq->imr_multiaddr, 0); 1783 rtcache_setdst(&ro, &u.dst); 1784 rtcache_init(&ro); 1785 ifp = (ro.ro_rt != NULL) ? ro.ro_rt->rt_ifp : NULL; 1786 rtcache_free(&ro); 1787 } else { 1788 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1789 } 1790 /* 1791 * See if we found an interface, and confirm that it 1792 * supports multicast. 1793 */ 1794 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1795 error = EADDRNOTAVAIL; 1796 break; 1797 } 1798 /* 1799 * See if the membership already exists or if all the 1800 * membership slots are full. 1801 */ 1802 for (i = 0; i < imo->imo_num_memberships; ++i) { 1803 if (imo->imo_membership[i]->inm_ifp == ifp && 1804 in_hosteq(imo->imo_membership[i]->inm_addr, 1805 mreq->imr_multiaddr)) 1806 break; 1807 } 1808 if (i < imo->imo_num_memberships) { 1809 error = EADDRINUSE; 1810 break; 1811 } 1812 if (i == IP_MAX_MEMBERSHIPS) { 1813 error = ETOOMANYREFS; 1814 break; 1815 } 1816 /* 1817 * Everything looks good; add a new record to the multicast 1818 * address list for the given interface. 1819 */ 1820 if ((imo->imo_membership[i] = 1821 in_addmulti(&mreq->imr_multiaddr, ifp)) == NULL) { 1822 error = ENOBUFS; 1823 break; 1824 } 1825 ++imo->imo_num_memberships; 1826 break; 1827 1828 case IP_DROP_MEMBERSHIP: 1829 /* 1830 * Drop a multicast group membership. 1831 * Group must be a valid IP multicast address. 1832 */ 1833 if (m == NULL || m->m_len != sizeof(struct ip_mreq)) { 1834 error = EINVAL; 1835 break; 1836 } 1837 mreq = mtod(m, struct ip_mreq *); 1838 if (!IN_MULTICAST(mreq->imr_multiaddr.s_addr)) { 1839 error = EINVAL; 1840 break; 1841 } 1842 /* 1843 * If an interface address was specified, get a pointer 1844 * to its ifnet structure. 1845 */ 1846 if (in_nullhost(mreq->imr_interface)) 1847 ifp = NULL; 1848 else { 1849 ifp = ip_multicast_if(&mreq->imr_interface, NULL); 1850 if (ifp == NULL) { 1851 error = EADDRNOTAVAIL; 1852 break; 1853 } 1854 } 1855 /* 1856 * Find the membership in the membership array. 1857 */ 1858 for (i = 0; i < imo->imo_num_memberships; ++i) { 1859 if ((ifp == NULL || 1860 imo->imo_membership[i]->inm_ifp == ifp) && 1861 in_hosteq(imo->imo_membership[i]->inm_addr, 1862 mreq->imr_multiaddr)) 1863 break; 1864 } 1865 if (i == imo->imo_num_memberships) { 1866 error = EADDRNOTAVAIL; 1867 break; 1868 } 1869 /* 1870 * Give up the multicast address record to which the 1871 * membership points. 1872 */ 1873 in_delmulti(imo->imo_membership[i]); 1874 /* 1875 * Remove the gap in the membership array. 1876 */ 1877 for (++i; i < imo->imo_num_memberships; ++i) 1878 imo->imo_membership[i-1] = imo->imo_membership[i]; 1879 --imo->imo_num_memberships; 1880 break; 1881 1882 default: 1883 error = EOPNOTSUPP; 1884 break; 1885 } 1886 1887 /* 1888 * If all options have default values, no need to keep the mbuf. 1889 */ 1890 if (imo->imo_multicast_ifp == NULL && 1891 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1892 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1893 imo->imo_num_memberships == 0) { 1894 free(*imop, M_IPMOPTS); 1895 *imop = NULL; 1896 } 1897 1898 return (error); 1899 } 1900 1901 /* 1902 * Return the IP multicast options in response to user getsockopt(). 1903 */ 1904 int 1905 ip_getmoptions(int optname, struct ip_moptions *imo, struct mbuf **mp) 1906 { 1907 u_char *ttl; 1908 u_char *loop; 1909 struct in_addr *addr; 1910 struct in_ifaddr *ia; 1911 1912 *mp = m_get(M_WAIT, MT_SOOPTS); 1913 1914 switch (optname) { 1915 1916 case IP_MULTICAST_IF: 1917 addr = mtod(*mp, struct in_addr *); 1918 (*mp)->m_len = sizeof(struct in_addr); 1919 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1920 *addr = zeroin_addr; 1921 else if (imo->imo_multicast_addr.s_addr) { 1922 /* return the value user has set */ 1923 *addr = imo->imo_multicast_addr; 1924 } else { 1925 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1926 *addr = ia ? ia->ia_addr.sin_addr : zeroin_addr; 1927 } 1928 return (0); 1929 1930 case IP_MULTICAST_TTL: 1931 ttl = mtod(*mp, u_char *); 1932 (*mp)->m_len = 1; 1933 *ttl = imo ? imo->imo_multicast_ttl 1934 : IP_DEFAULT_MULTICAST_TTL; 1935 return (0); 1936 1937 case IP_MULTICAST_LOOP: 1938 loop = mtod(*mp, u_char *); 1939 (*mp)->m_len = 1; 1940 *loop = imo ? imo->imo_multicast_loop 1941 : IP_DEFAULT_MULTICAST_LOOP; 1942 return (0); 1943 1944 default: 1945 return (EOPNOTSUPP); 1946 } 1947 } 1948 1949 /* 1950 * Discard the IP multicast options. 1951 */ 1952 void 1953 ip_freemoptions(struct ip_moptions *imo) 1954 { 1955 int i; 1956 1957 if (imo != NULL) { 1958 for (i = 0; i < imo->imo_num_memberships; ++i) 1959 in_delmulti(imo->imo_membership[i]); 1960 free(imo, M_IPMOPTS); 1961 } 1962 } 1963 1964 /* 1965 * Routine called from ip_output() to loop back a copy of an IP multicast 1966 * packet to the input queue of a specified interface. Note that this 1967 * calls the output routine of the loopback "driver", but with an interface 1968 * pointer that might NOT be lo0ifp -- easier than replicating that code here. 1969 */ 1970 static void 1971 ip_mloopback(struct ifnet *ifp, struct mbuf *m, const struct sockaddr_in *dst) 1972 { 1973 struct ip *ip; 1974 struct mbuf *copym; 1975 1976 copym = m_copypacket(m, M_DONTWAIT); 1977 if (copym != NULL 1978 && (copym->m_flags & M_EXT || copym->m_len < sizeof(struct ip))) 1979 copym = m_pullup(copym, sizeof(struct ip)); 1980 if (copym == NULL) 1981 return; 1982 /* 1983 * We don't bother to fragment if the IP length is greater 1984 * than the interface's MTU. Can this possibly matter? 1985 */ 1986 ip = mtod(copym, struct ip *); 1987 1988 if (copym->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1989 in_delayed_cksum(copym); 1990 copym->m_pkthdr.csum_flags &= 1991 ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1992 } 1993 1994 ip->ip_sum = 0; 1995 ip->ip_sum = in_cksum(copym, ip->ip_hl << 2); 1996 (void)looutput(ifp, copym, sintocsa(dst), NULL); 1997 } 1998