ip_output.c revision 60910
1/* 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 34 * $FreeBSD: head/sys/netinet/ip_output.c 60910 2000-05-25 02:27:14Z jlemon $ 35 */ 36 37#define _IP_VHL 38 39#include "opt_ipfw.h" 40#include "opt_ipdn.h" 41#include "opt_ipdivert.h" 42#include "opt_ipfilter.h" 43#include "opt_ipsec.h" 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/kernel.h> 48#include <sys/malloc.h> 49#include <sys/mbuf.h> 50#include <sys/protosw.h> 51#include <sys/socket.h> 52#include <sys/socketvar.h> 53#include <sys/proc.h> 54 55#include <net/if.h> 56#include <net/route.h> 57 58#include <netinet/in.h> 59#include <netinet/in_systm.h> 60#include <netinet/ip.h> 61#include <netinet/in_pcb.h> 62#include <netinet/in_var.h> 63#include <netinet/ip_var.h> 64 65#include "faith.h" 66 67#ifdef vax 68#include <machine/mtpr.h> 69#endif 70#include <machine/in_cksum.h> 71 72static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 73 74#ifdef IPSEC 75#include <netinet6/ipsec.h> 76#include <netkey/key.h> 77#ifdef IPSEC_DEBUG 78#include <netkey/key_debug.h> 79#else 80#define KEYDEBUG(lev,arg) 81#endif 82#endif /*IPSEC*/ 83 84#include <netinet/ip_fw.h> 85 86#ifdef DUMMYNET 87#include <netinet/ip_dummynet.h> 88#endif 89 90#ifdef IPFIREWALL_FORWARD_DEBUG 91#define print_ip(a) printf("%ld.%ld.%ld.%ld",(ntohl(a.s_addr)>>24)&0xFF,\ 92 (ntohl(a.s_addr)>>16)&0xFF,\ 93 (ntohl(a.s_addr)>>8)&0xFF,\ 94 (ntohl(a.s_addr))&0xFF); 95#endif 96 97u_short ip_id; 98 99static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *)); 100static void ip_mloopback 101 __P((struct ifnet *, struct mbuf *, struct sockaddr_in *, int)); 102static int ip_getmoptions 103 __P((struct sockopt *, struct ip_moptions *)); 104static int ip_pcbopts __P((int, struct mbuf **, struct mbuf *)); 105static int ip_setmoptions 106 __P((struct sockopt *, struct ip_moptions **)); 107 108int ip_optcopy __P((struct ip *, struct ip *)); 109extern int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)); 110 111 112extern struct protosw inetsw[]; 113 114/* 115 * IP output. The packet in mbuf chain m contains a skeletal IP 116 * header (with len, off, ttl, proto, tos, src, dst). 117 * The mbuf chain containing the packet will be freed. 118 * The mbuf opt, if present, will not be freed. 119 */ 120int 121ip_output(m0, opt, ro, flags, imo) 122 struct mbuf *m0; 123 struct mbuf *opt; 124 struct route *ro; 125 int flags; 126 struct ip_moptions *imo; 127{ 128 struct ip *ip, *mhip; 129 struct ifnet *ifp; 130 struct mbuf *m = m0; 131 int hlen = sizeof (struct ip); 132 int len, off, error = 0; 133 struct sockaddr_in *dst; 134 struct in_ifaddr *ia; 135 int isbroadcast, sw_csum; 136#ifdef IPSEC 137 struct route iproute; 138 struct socket *so = NULL; 139 struct secpolicy *sp = NULL; 140#endif 141 u_int16_t divert_cookie; /* firewall cookie */ 142#ifdef IPFIREWALL_FORWARD 143 int fwd_rewrite_src = 0; 144#endif 145 struct ip_fw_chain *rule = NULL; 146 147#ifdef IPDIVERT 148 /* Get and reset firewall cookie */ 149 divert_cookie = ip_divert_cookie; 150 ip_divert_cookie = 0; 151#else 152 divert_cookie = 0; 153#endif 154 155 /* 156 * NOTE: If IP_SOCKINMRCVIF flag is set, 'socket *' is kept in 157 * m->m_pkthdr.rcvif for later IPSEC check. In this case, 158 * m->m_pkthdr will be NULL cleared after the contents is saved in 159 * 'so'. 160 * NULL clearance of rcvif should be natural because the packet should 161 * have been sent from my own socket and has no rcvif in this case. 162 * It is also necessary because someone might consider it as 163 * 'ifnet *', and cause SEGV. 164 */ 165#if defined(IPFIREWALL) && defined(DUMMYNET) 166 /* 167 * dummynet packet are prepended a vestigial mbuf with 168 * m_type = MT_DUMMYNET and m_data pointing to the matching 169 * rule. 170 */ 171 if (m->m_type == MT_DUMMYNET) { 172 /* 173 * the packet was already tagged, so part of the 174 * processing was already done, and we need to go down. 175 * Get parameters from the header. 176 */ 177 rule = (struct ip_fw_chain *)(m->m_data) ; 178 opt = NULL ; 179 ro = & ( ((struct dn_pkt *)m)->ro ) ; 180 imo = NULL ; 181 dst = ((struct dn_pkt *)m)->dn_dst ; 182 ifp = ((struct dn_pkt *)m)->ifp ; 183 flags = ((struct dn_pkt *)m)->flags ; 184 185 m0 = m = m->m_next ; 186#ifdef IPSEC 187 if ((flags & IP_SOCKINMRCVIF) != 0) { 188 so = (struct socket *)m->m_pkthdr.rcvif; 189 m->m_pkthdr.rcvif = NULL; 190 } 191#endif 192 ip = mtod(m, struct ip *); 193 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 194 goto sendit; 195 } else 196 rule = NULL ; 197#endif 198#ifdef IPSEC 199 if ((flags & IP_SOCKINMRCVIF) != 0) { 200 so = (struct socket *)m->m_pkthdr.rcvif; 201 m->m_pkthdr.rcvif = NULL; 202 } 203#endif 204 205#ifdef DIAGNOSTIC 206 if ((m->m_flags & M_PKTHDR) == 0) 207 panic("ip_output no HDR"); 208 if (!ro) 209 panic("ip_output no route, proto = %d", 210 mtod(m, struct ip *)->ip_p); 211#endif 212 if (opt) { 213 m = ip_insertoptions(m, opt, &len); 214 hlen = len; 215 } 216 ip = mtod(m, struct ip *); 217 /* 218 * Fill in IP header. 219 */ 220 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 221 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 222 ip->ip_off &= IP_DF; 223 ip->ip_id = htons(ip_id++); 224 ipstat.ips_localout++; 225 } else { 226 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 227 } 228 229 dst = (struct sockaddr_in *)&ro->ro_dst; 230 /* 231 * If there is a cached route, 232 * check that it is to the same destination 233 * and is still up. If not, free it and try again. 234 */ 235 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 236 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 237 RTFREE(ro->ro_rt); 238 ro->ro_rt = (struct rtentry *)0; 239 } 240 if (ro->ro_rt == 0) { 241 dst->sin_family = AF_INET; 242 dst->sin_len = sizeof(*dst); 243 dst->sin_addr = ip->ip_dst; 244 } 245 /* 246 * If routing to interface only, 247 * short circuit routing lookup. 248 */ 249#define ifatoia(ifa) ((struct in_ifaddr *)(ifa)) 250#define sintosa(sin) ((struct sockaddr *)(sin)) 251 if (flags & IP_ROUTETOIF) { 252 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 253 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 254 ipstat.ips_noroute++; 255 error = ENETUNREACH; 256 goto bad; 257 } 258 ifp = ia->ia_ifp; 259 ip->ip_ttl = 1; 260 isbroadcast = in_broadcast(dst->sin_addr, ifp); 261 } else { 262 /* 263 * If this is the case, we probably don't want to allocate 264 * a protocol-cloned route since we didn't get one from the 265 * ULP. This lets TCP do its thing, while not burdening 266 * forwarding or ICMP with the overhead of cloning a route. 267 * Of course, we still want to do any cloning requested by 268 * the link layer, as this is probably required in all cases 269 * for correct operation (as it is for ARP). 270 */ 271 if (ro->ro_rt == 0) 272 rtalloc_ign(ro, RTF_PRCLONING); 273 if (ro->ro_rt == 0) { 274 ipstat.ips_noroute++; 275 error = EHOSTUNREACH; 276 goto bad; 277 } 278 ia = ifatoia(ro->ro_rt->rt_ifa); 279 ifp = ro->ro_rt->rt_ifp; 280 ro->ro_rt->rt_use++; 281 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 282 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 283 if (ro->ro_rt->rt_flags & RTF_HOST) 284 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 285 else 286 isbroadcast = in_broadcast(dst->sin_addr, ifp); 287 } 288 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 289 struct in_multi *inm; 290 291 m->m_flags |= M_MCAST; 292 /* 293 * IP destination address is multicast. Make sure "dst" 294 * still points to the address in "ro". (It may have been 295 * changed to point to a gateway address, above.) 296 */ 297 dst = (struct sockaddr_in *)&ro->ro_dst; 298 /* 299 * See if the caller provided any multicast options 300 */ 301 if (imo != NULL) { 302 ip->ip_ttl = imo->imo_multicast_ttl; 303 if (imo->imo_multicast_ifp != NULL) 304 ifp = imo->imo_multicast_ifp; 305 if (imo->imo_multicast_vif != -1) 306 ip->ip_src.s_addr = 307 ip_mcast_src(imo->imo_multicast_vif); 308 } else 309 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 310 /* 311 * Confirm that the outgoing interface supports multicast. 312 */ 313 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 314 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 315 ipstat.ips_noroute++; 316 error = ENETUNREACH; 317 goto bad; 318 } 319 } 320 /* 321 * If source address not specified yet, use address 322 * of outgoing interface. 323 */ 324 if (ip->ip_src.s_addr == INADDR_ANY) { 325 register struct in_ifaddr *ia1; 326 327 for (ia1 = in_ifaddrhead.tqh_first; ia1; 328 ia1 = ia1->ia_link.tqe_next) 329 if (ia1->ia_ifp == ifp) { 330 ip->ip_src = IA_SIN(ia1)->sin_addr; 331 break; 332 } 333 } 334 335 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 336 if (inm != NULL && 337 (imo == NULL || imo->imo_multicast_loop)) { 338 /* 339 * If we belong to the destination multicast group 340 * on the outgoing interface, and the caller did not 341 * forbid loopback, loop back a copy. 342 */ 343 ip_mloopback(ifp, m, dst, hlen); 344 } 345 else { 346 /* 347 * If we are acting as a multicast router, perform 348 * multicast forwarding as if the packet had just 349 * arrived on the interface to which we are about 350 * to send. The multicast forwarding function 351 * recursively calls this function, using the 352 * IP_FORWARDING flag to prevent infinite recursion. 353 * 354 * Multicasts that are looped back by ip_mloopback(), 355 * above, will be forwarded by the ip_input() routine, 356 * if necessary. 357 */ 358 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 359 /* 360 * Check if rsvp daemon is running. If not, don't 361 * set ip_moptions. This ensures that the packet 362 * is multicast and not just sent down one link 363 * as prescribed by rsvpd. 364 */ 365 if (!rsvp_on) 366 imo = NULL; 367 if (ip_mforward(ip, ifp, m, imo) != 0) { 368 m_freem(m); 369 goto done; 370 } 371 } 372 } 373 374 /* 375 * Multicasts with a time-to-live of zero may be looped- 376 * back, above, but must not be transmitted on a network. 377 * Also, multicasts addressed to the loopback interface 378 * are not sent -- the above call to ip_mloopback() will 379 * loop back a copy if this host actually belongs to the 380 * destination group on the loopback interface. 381 */ 382 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 383 m_freem(m); 384 goto done; 385 } 386 387 goto sendit; 388 } 389#ifndef notdef 390 /* 391 * If source address not specified yet, use address 392 * of outgoing interface. 393 */ 394 if (ip->ip_src.s_addr == INADDR_ANY) { 395 ip->ip_src = IA_SIN(ia)->sin_addr; 396#ifdef IPFIREWALL_FORWARD 397 /* Keep note that we did this - if the firewall changes 398 * the next-hop, our interface may change, changing the 399 * default source IP. It's a shame so much effort happens 400 * twice. Oh well. 401 */ 402 fwd_rewrite_src++; 403#endif /* IPFIREWALL_FORWARD */ 404 } 405#endif /* notdef */ 406 /* 407 * Verify that we have any chance at all of being able to queue 408 * the packet or packet fragments 409 */ 410 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 411 ifp->if_snd.ifq_maxlen) { 412 error = ENOBUFS; 413 goto bad; 414 } 415 416 /* 417 * Look for broadcast address and 418 * and verify user is allowed to send 419 * such a packet. 420 */ 421 if (isbroadcast) { 422 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 423 error = EADDRNOTAVAIL; 424 goto bad; 425 } 426 if ((flags & IP_ALLOWBROADCAST) == 0) { 427 error = EACCES; 428 goto bad; 429 } 430 /* don't allow broadcast messages to be fragmented */ 431 if ((u_short)ip->ip_len > ifp->if_mtu) { 432 error = EMSGSIZE; 433 goto bad; 434 } 435 m->m_flags |= M_BCAST; 436 } else { 437 m->m_flags &= ~M_BCAST; 438 } 439 440sendit: 441 /* 442 * IpHack's section. 443 * - Xlate: translate packet's addr/port (NAT). 444 * - Firewall: deny/allow/etc. 445 * - Wrap: fake packet's addr/port <unimpl.> 446 * - Encapsulate: put it in another IP and send out. <unimp.> 447 */ 448 if (fr_checkp) { 449 struct mbuf *m1 = m; 450 451 if ((error = (*fr_checkp)(ip, hlen, ifp, 1, &m1)) || !m1) 452 goto done; 453 ip = mtod(m = m1, struct ip *); 454 } 455 456 /* 457 * Check with the firewall... 458 */ 459 if (fw_enable && ip_fw_chk_ptr) { 460 struct sockaddr_in *old = dst; 461 462 off = (*ip_fw_chk_ptr)(&ip, 463 hlen, ifp, &divert_cookie, &m, &rule, &dst); 464 /* 465 * On return we must do the following: 466 * m == NULL -> drop the pkt 467 * 1<=off<= 0xffff -> DIVERT 468 * (off & 0x10000) -> send to a DUMMYNET pipe 469 * (off & 0x20000) -> TEE the packet 470 * dst != old -> IPFIREWALL_FORWARD 471 * off==0, dst==old -> accept 472 * If some of the above modules is not compiled in, then 473 * we should't have to check the corresponding condition 474 * (because the ipfw control socket should not accept 475 * unsupported rules), but better play safe and drop 476 * packets in case of doubt. 477 */ 478 if (!m) { /* firewall said to reject */ 479 error = EACCES; 480 goto done; 481 } 482 if (off == 0 && dst == old) /* common case */ 483 goto pass ; 484#ifdef DUMMYNET 485 if ((off & IP_FW_PORT_DYNT_FLAG) != 0) { 486 /* 487 * pass the pkt to dummynet. Need to include 488 * pipe number, m, ifp, ro, dst because these are 489 * not recomputed in the next pass. 490 * All other parameters have been already used and 491 * so they are not needed anymore. 492 * XXX note: if the ifp or ro entry are deleted 493 * while a pkt is in dummynet, we are in trouble! 494 */ 495 dummynet_io(off & 0xffff, DN_TO_IP_OUT, m,ifp,ro,dst,rule, 496 flags); 497 goto done; 498 } 499#endif 500#ifdef IPDIVERT 501 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 502 struct mbuf *clone = NULL; 503 504 /* Clone packet if we're doing a 'tee' */ 505 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 506 clone = m_dup(m, M_DONTWAIT); 507 508 /* 509 * XXX 510 * delayed checksums are not currently compatible 511 * with divert sockets. 512 */ 513 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 514 in_delayed_cksum(m); 515 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 516 } 517 518 /* Restore packet header fields to original values */ 519 HTONS(ip->ip_len); 520 HTONS(ip->ip_off); 521 522 /* Deliver packet to divert input routine */ 523 ip_divert_cookie = divert_cookie; 524 divert_packet(m, 0, off & 0xffff); 525 526 /* If 'tee', continue with original packet */ 527 if (clone != NULL) { 528 m = clone; 529 ip = mtod(m, struct ip *); 530 goto pass; 531 } 532 goto done; 533 } 534#endif 535 536#ifdef IPFIREWALL_FORWARD 537 /* Here we check dst to make sure it's directly reachable on the 538 * interface we previously thought it was. 539 * If it isn't (which may be likely in some situations) we have 540 * to re-route it (ie, find a route for the next-hop and the 541 * associated interface) and set them here. This is nested 542 * forwarding which in most cases is undesirable, except where 543 * such control is nigh impossible. So we do it here. 544 * And I'm babbling. 545 */ 546 if (off == 0 && old != dst) { 547 struct in_ifaddr *ia; 548 549 /* It's changed... */ 550 /* There must be a better way to do this next line... */ 551 static struct route sro_fwd, *ro_fwd = &sro_fwd; 552#ifdef IPFIREWALL_FORWARD_DEBUG 553 printf("IPFIREWALL_FORWARD: New dst ip: "); 554 print_ip(dst->sin_addr); 555 printf("\n"); 556#endif 557 /* 558 * We need to figure out if we have been forwarded 559 * to a local socket. If so then we should somehow 560 * "loop back" to ip_input, and get directed to the 561 * PCB as if we had received this packet. This is 562 * because it may be dificult to identify the packets 563 * you want to forward until they are being output 564 * and have selected an interface. (e.g. locally 565 * initiated packets) If we used the loopback inteface, 566 * we would not be able to control what happens 567 * as the packet runs through ip_input() as 568 * it is done through a ISR. 569 */ 570 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 571 ia = TAILQ_NEXT(ia, ia_link)) { 572 /* 573 * If the addr to forward to is one 574 * of ours, we pretend to 575 * be the destination for this packet. 576 */ 577 if (IA_SIN(ia)->sin_addr.s_addr == 578 dst->sin_addr.s_addr) 579 break; 580 } 581 if (ia) { 582 /* tell ip_input "dont filter" */ 583 ip_fw_fwd_addr = dst; 584 if (m->m_pkthdr.rcvif == NULL) 585 m->m_pkthdr.rcvif = ifunit("lo0"); 586 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 587 m->m_pkthdr.csum_flags |= 588 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 589 m0->m_pkthdr.csum_data = 0xffff; 590 } 591 m->m_pkthdr.csum_flags |= 592 CSUM_IP_CHECKED | CSUM_IP_VALID; 593 ip->ip_len = htons((u_short)ip->ip_len); 594 ip->ip_off = htons((u_short)ip->ip_off); 595 ip_input(m); 596 goto done; 597 } 598 /* Some of the logic for this was 599 * nicked from above. 600 * 601 * This rewrites the cached route in a local PCB. 602 * Is this what we want to do? 603 */ 604 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 605 606 ro_fwd->ro_rt = 0; 607 rtalloc_ign(ro_fwd, RTF_PRCLONING); 608 609 if (ro_fwd->ro_rt == 0) { 610 ipstat.ips_noroute++; 611 error = EHOSTUNREACH; 612 goto bad; 613 } 614 615 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 616 ifp = ro_fwd->ro_rt->rt_ifp; 617 ro_fwd->ro_rt->rt_use++; 618 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 619 dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway; 620 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 621 isbroadcast = 622 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 623 else 624 isbroadcast = in_broadcast(dst->sin_addr, ifp); 625 RTFREE(ro->ro_rt); 626 ro->ro_rt = ro_fwd->ro_rt; 627 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 628 629 /* 630 * If we added a default src ip earlier, 631 * which would have been gotten from the-then 632 * interface, do it again, from the new one. 633 */ 634 if (fwd_rewrite_src) 635 ip->ip_src = IA_SIN(ia)->sin_addr; 636 goto pass ; 637 } 638#endif /* IPFIREWALL_FORWARD */ 639 /* 640 * if we get here, none of the above matches, and 641 * we have to drop the pkt 642 */ 643 m_freem(m); 644 error = EACCES; /* not sure this is the right error msg */ 645 goto done; 646 } 647 648pass: 649#ifdef IPSEC 650 /* get SP for this packet */ 651 if (so == NULL) 652 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 653 else 654 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 655 656 if (sp == NULL) { 657 ipsecstat.out_inval++; 658 goto bad; 659 } 660 661 error = 0; 662 663 /* check policy */ 664 switch (sp->policy) { 665 case IPSEC_POLICY_DISCARD: 666 /* 667 * This packet is just discarded. 668 */ 669 ipsecstat.out_polvio++; 670 goto bad; 671 672 case IPSEC_POLICY_BYPASS: 673 case IPSEC_POLICY_NONE: 674 /* no need to do IPsec. */ 675 goto skip_ipsec; 676 677 case IPSEC_POLICY_IPSEC: 678 if (sp->req == NULL) { 679 /* XXX should be panic ? */ 680 printf("ip_output: No IPsec request specified.\n"); 681 error = EINVAL; 682 goto bad; 683 } 684 break; 685 686 case IPSEC_POLICY_ENTRUST: 687 default: 688 printf("ip_output: Invalid policy found. %d\n", sp->policy); 689 } 690 { 691 struct ipsec_output_state state; 692 bzero(&state, sizeof(state)); 693 state.m = m; 694 if (flags & IP_ROUTETOIF) { 695 state.ro = &iproute; 696 bzero(&iproute, sizeof(iproute)); 697 } else 698 state.ro = ro; 699 state.dst = (struct sockaddr *)dst; 700 701 ip->ip_sum = 0; 702 703 /* 704 * XXX 705 * delayed checksums are not currently compatible with IPsec 706 */ 707 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 708 in_delayed_cksum(m); 709 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 710 } 711 712 ip->ip_len = htons((u_short)ip->ip_len); 713 ip->ip_off = htons((u_short)ip->ip_off); 714 715 error = ipsec4_output(&state, sp, flags); 716 717 m = state.m; 718 if (flags & IP_ROUTETOIF) { 719 /* 720 * if we have tunnel mode SA, we may need to ignore 721 * IP_ROUTETOIF. 722 */ 723 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 724 flags &= ~IP_ROUTETOIF; 725 ro = state.ro; 726 } 727 } else 728 ro = state.ro; 729 dst = (struct sockaddr_in *)state.dst; 730 if (error) { 731 /* mbuf is already reclaimed in ipsec4_output. */ 732 m0 = NULL; 733 switch (error) { 734 case EHOSTUNREACH: 735 case ENETUNREACH: 736 case EMSGSIZE: 737 case ENOBUFS: 738 case ENOMEM: 739 break; 740 default: 741 printf("ip4_output (ipsec): error code %d\n", error); 742 /*fall through*/ 743 case ENOENT: 744 /* don't show these error codes to the user */ 745 error = 0; 746 break; 747 } 748 goto bad; 749 } 750 } 751 752 /* be sure to update variables that are affected by ipsec4_output() */ 753 ip = mtod(m, struct ip *); 754#ifdef _IP_VHL 755 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 756#else 757 hlen = ip->ip_hl << 2; 758#endif 759 if (ro->ro_rt == NULL) { 760 if ((flags & IP_ROUTETOIF) == 0) { 761 printf("ip_output: " 762 "can't update route after IPsec processing\n"); 763 error = EHOSTUNREACH; /*XXX*/ 764 goto bad; 765 } 766 } else { 767 /* nobody uses ia beyond here */ 768 ifp = ro->ro_rt->rt_ifp; 769 } 770 771 /* make it flipped, again. */ 772 ip->ip_len = ntohs((u_short)ip->ip_len); 773 ip->ip_off = ntohs((u_short)ip->ip_off); 774skip_ipsec: 775#endif /*IPSEC*/ 776 777 sw_csum = m->m_pkthdr.csum_flags | CSUM_IP; 778 m->m_pkthdr.csum_flags = sw_csum & ifp->if_hwassist; 779 sw_csum &= ~ifp->if_hwassist; 780 if (sw_csum & CSUM_DELAY_DATA) { 781 in_delayed_cksum(m); 782 sw_csum &= ~CSUM_DELAY_DATA; 783 } 784 785 /* 786 * If small enough for interface, or the interface will take 787 * care of the fragmentation for us, can just send directly. 788 */ 789 if ((u_short)ip->ip_len <= ifp->if_mtu || 790 ifp->if_hwassist & CSUM_FRAGMENT) { 791 ip->ip_len = htons((u_short)ip->ip_len); 792 ip->ip_off = htons((u_short)ip->ip_off); 793 ip->ip_sum = 0; 794 if (sw_csum & CSUM_DELAY_IP) { 795 if (ip->ip_vhl == IP_VHL_BORING) { 796 ip->ip_sum = in_cksum_hdr(ip); 797 } else { 798 ip->ip_sum = in_cksum(m, hlen); 799 } 800 } 801 error = (*ifp->if_output)(ifp, m, 802 (struct sockaddr *)dst, ro->ro_rt); 803 goto done; 804 } 805 /* 806 * Too large for interface; fragment if possible. 807 * Must be able to put at least 8 bytes per fragment. 808 */ 809 if (ip->ip_off & IP_DF) { 810 error = EMSGSIZE; 811 /* 812 * This case can happen if the user changed the MTU 813 * of an interface after enabling IP on it. Because 814 * most netifs don't keep track of routes pointing to 815 * them, there is no way for one to update all its 816 * routes when the MTU is changed. 817 */ 818 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 819 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 820 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 821 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 822 } 823 ipstat.ips_cantfrag++; 824 goto bad; 825 } 826 len = (ifp->if_mtu - hlen) &~ 7; 827 if (len < 8) { 828 error = EMSGSIZE; 829 goto bad; 830 } 831 832 /* 833 * if the interface will not calculate checksums on 834 * fragmented packets, then do it here. 835 */ 836 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 837 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 838 in_delayed_cksum(m); 839 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 840 } 841 842 { 843 int mhlen, firstlen = len; 844 struct mbuf **mnext = &m->m_nextpkt; 845 int nfrags = 1; 846 847 /* 848 * Loop through length of segment after first fragment, 849 * make new header and copy data of each part and link onto chain. 850 */ 851 m0 = m; 852 mhlen = sizeof (struct ip); 853 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) { 854 MGETHDR(m, M_DONTWAIT, MT_HEADER); 855 if (m == 0) { 856 error = ENOBUFS; 857 ipstat.ips_odropped++; 858 goto sendorfree; 859 } 860 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 861 m->m_data += max_linkhdr; 862 mhip = mtod(m, struct ip *); 863 *mhip = *ip; 864 if (hlen > sizeof (struct ip)) { 865 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 866 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 867 } 868 m->m_len = mhlen; 869 mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF); 870 if (ip->ip_off & IP_MF) 871 mhip->ip_off |= IP_MF; 872 if (off + len >= (u_short)ip->ip_len) 873 len = (u_short)ip->ip_len - off; 874 else 875 mhip->ip_off |= IP_MF; 876 mhip->ip_len = htons((u_short)(len + mhlen)); 877 m->m_next = m_copy(m0, off, len); 878 if (m->m_next == 0) { 879 (void) m_free(m); 880 error = ENOBUFS; /* ??? */ 881 ipstat.ips_odropped++; 882 goto sendorfree; 883 } 884 m->m_pkthdr.len = mhlen + len; 885 m->m_pkthdr.rcvif = (struct ifnet *)0; 886 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 887 mhip->ip_off = htons((u_short)mhip->ip_off); 888 mhip->ip_sum = 0; 889 if (sw_csum & CSUM_DELAY_IP) { 890 if (mhip->ip_vhl == IP_VHL_BORING) { 891 mhip->ip_sum = in_cksum_hdr(mhip); 892 } else { 893 mhip->ip_sum = in_cksum(m, mhlen); 894 } 895 } 896 *mnext = m; 897 mnext = &m->m_nextpkt; 898 nfrags++; 899 } 900 ipstat.ips_ofragments += nfrags; 901 902 /* set first/last markers for fragment chain */ 903 m->m_flags |= M_LASTFRAG; 904 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 905 m0->m_pkthdr.csum_data = nfrags; 906 907 /* 908 * Update first fragment by trimming what's been copied out 909 * and updating header, then send each fragment (in order). 910 */ 911 m = m0; 912 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 913 m->m_pkthdr.len = hlen + firstlen; 914 ip->ip_len = htons((u_short)m->m_pkthdr.len); 915 ip->ip_off = htons((u_short)(ip->ip_off | IP_MF)); 916 ip->ip_sum = 0; 917 if (sw_csum & CSUM_DELAY_IP) { 918 if (ip->ip_vhl == IP_VHL_BORING) { 919 ip->ip_sum = in_cksum_hdr(ip); 920 } else { 921 ip->ip_sum = in_cksum(m, hlen); 922 } 923 } 924sendorfree: 925 for (m = m0; m; m = m0) { 926 m0 = m->m_nextpkt; 927 m->m_nextpkt = 0; 928 if (error == 0) 929 error = (*ifp->if_output)(ifp, m, 930 (struct sockaddr *)dst, ro->ro_rt); 931 else 932 m_freem(m); 933 } 934 935 if (error == 0) 936 ipstat.ips_fragmented++; 937 } 938done: 939#ifdef IPSEC 940 if (ro == &iproute && ro->ro_rt) { 941 RTFREE(ro->ro_rt); 942 ro->ro_rt = NULL; 943 } 944 if (sp != NULL) { 945 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 946 printf("DP ip_output call free SP:%p\n", sp)); 947 key_freesp(sp); 948 } 949#endif /* IPSEC */ 950 return (error); 951bad: 952 m_freem(m0); 953 goto done; 954} 955 956void 957in_delayed_cksum(struct mbuf *m) 958{ 959 struct ip *ip; 960 u_short csum, offset; 961 962 ip = mtod(m, struct ip *); 963 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 964 csum = in_cksum_skip(m, ip->ip_len, offset); 965 offset += m->m_pkthdr.csum_data; /* checksum offset */ 966 967 if (offset + sizeof(u_short) > m->m_len) { 968 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 969 m->m_len, offset, ip->ip_p); 970 /* 971 * XXX 972 * this shouldn't happen, but if it does, the 973 * correct behavior may be to insert the checksum 974 * in the existing chain instead of rearranging it. 975 */ 976 m = m_pullup(m, offset + sizeof(u_short)); 977 } 978 *(u_short *)(m->m_data + offset) = csum; 979} 980 981/* 982 * Insert IP options into preformed packet. 983 * Adjust IP destination as required for IP source routing, 984 * as indicated by a non-zero in_addr at the start of the options. 985 * 986 * XXX This routine assumes that the packet has no options in place. 987 */ 988static struct mbuf * 989ip_insertoptions(m, opt, phlen) 990 register struct mbuf *m; 991 struct mbuf *opt; 992 int *phlen; 993{ 994 register struct ipoption *p = mtod(opt, struct ipoption *); 995 struct mbuf *n; 996 register struct ip *ip = mtod(m, struct ip *); 997 unsigned optlen; 998 999 optlen = opt->m_len - sizeof(p->ipopt_dst); 1000 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) 1001 return (m); /* XXX should fail */ 1002 if (p->ipopt_dst.s_addr) 1003 ip->ip_dst = p->ipopt_dst; 1004 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1005 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1006 if (n == 0) 1007 return (m); 1008 n->m_pkthdr.rcvif = (struct ifnet *)0; 1009 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1010 m->m_len -= sizeof(struct ip); 1011 m->m_data += sizeof(struct ip); 1012 n->m_next = m; 1013 m = n; 1014 m->m_len = optlen + sizeof(struct ip); 1015 m->m_data += max_linkhdr; 1016 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1017 } else { 1018 m->m_data -= optlen; 1019 m->m_len += optlen; 1020 m->m_pkthdr.len += optlen; 1021 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1022 } 1023 ip = mtod(m, struct ip *); 1024 bcopy(p->ipopt_list, ip + 1, optlen); 1025 *phlen = sizeof(struct ip) + optlen; 1026 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1027 ip->ip_len += optlen; 1028 return (m); 1029} 1030 1031/* 1032 * Copy options from ip to jp, 1033 * omitting those not copied during fragmentation. 1034 */ 1035int 1036ip_optcopy(ip, jp) 1037 struct ip *ip, *jp; 1038{ 1039 register u_char *cp, *dp; 1040 int opt, optlen, cnt; 1041 1042 cp = (u_char *)(ip + 1); 1043 dp = (u_char *)(jp + 1); 1044 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1045 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1046 opt = cp[0]; 1047 if (opt == IPOPT_EOL) 1048 break; 1049 if (opt == IPOPT_NOP) { 1050 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1051 *dp++ = IPOPT_NOP; 1052 optlen = 1; 1053 continue; 1054 } else 1055 optlen = cp[IPOPT_OLEN]; 1056 /* bogus lengths should have been caught by ip_dooptions */ 1057 if (optlen > cnt) 1058 optlen = cnt; 1059 if (IPOPT_COPIED(opt)) { 1060 bcopy(cp, dp, optlen); 1061 dp += optlen; 1062 } 1063 } 1064 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1065 *dp++ = IPOPT_EOL; 1066 return (optlen); 1067} 1068 1069/* 1070 * IP socket option processing. 1071 */ 1072int 1073ip_ctloutput(so, sopt) 1074 struct socket *so; 1075 struct sockopt *sopt; 1076{ 1077 struct inpcb *inp = sotoinpcb(so); 1078 int error, optval; 1079 1080 error = optval = 0; 1081 if (sopt->sopt_level != IPPROTO_IP) { 1082 return (EINVAL); 1083 } 1084 1085 switch (sopt->sopt_dir) { 1086 case SOPT_SET: 1087 switch (sopt->sopt_name) { 1088 case IP_OPTIONS: 1089#ifdef notyet 1090 case IP_RETOPTS: 1091#endif 1092 { 1093 struct mbuf *m; 1094 if (sopt->sopt_valsize > MLEN) { 1095 error = EMSGSIZE; 1096 break; 1097 } 1098 MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER); 1099 if (m == 0) { 1100 error = ENOBUFS; 1101 break; 1102 } 1103 m->m_len = sopt->sopt_valsize; 1104 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1105 m->m_len); 1106 1107 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1108 m)); 1109 } 1110 1111 case IP_TOS: 1112 case IP_TTL: 1113 case IP_RECVOPTS: 1114 case IP_RECVRETOPTS: 1115 case IP_RECVDSTADDR: 1116 case IP_RECVIF: 1117#if defined(NFAITH) && NFAITH > 0 1118 case IP_FAITH: 1119#endif 1120 error = sooptcopyin(sopt, &optval, sizeof optval, 1121 sizeof optval); 1122 if (error) 1123 break; 1124 1125 switch (sopt->sopt_name) { 1126 case IP_TOS: 1127 inp->inp_ip_tos = optval; 1128 break; 1129 1130 case IP_TTL: 1131 inp->inp_ip_ttl = optval; 1132 break; 1133#define OPTSET(bit) \ 1134 if (optval) \ 1135 inp->inp_flags |= bit; \ 1136 else \ 1137 inp->inp_flags &= ~bit; 1138 1139 case IP_RECVOPTS: 1140 OPTSET(INP_RECVOPTS); 1141 break; 1142 1143 case IP_RECVRETOPTS: 1144 OPTSET(INP_RECVRETOPTS); 1145 break; 1146 1147 case IP_RECVDSTADDR: 1148 OPTSET(INP_RECVDSTADDR); 1149 break; 1150 1151 case IP_RECVIF: 1152 OPTSET(INP_RECVIF); 1153 break; 1154 1155#if defined(NFAITH) && NFAITH > 0 1156 case IP_FAITH: 1157 OPTSET(INP_FAITH); 1158 break; 1159#endif 1160 } 1161 break; 1162#undef OPTSET 1163 1164 case IP_MULTICAST_IF: 1165 case IP_MULTICAST_VIF: 1166 case IP_MULTICAST_TTL: 1167 case IP_MULTICAST_LOOP: 1168 case IP_ADD_MEMBERSHIP: 1169 case IP_DROP_MEMBERSHIP: 1170 error = ip_setmoptions(sopt, &inp->inp_moptions); 1171 break; 1172 1173 case IP_PORTRANGE: 1174 error = sooptcopyin(sopt, &optval, sizeof optval, 1175 sizeof optval); 1176 if (error) 1177 break; 1178 1179 switch (optval) { 1180 case IP_PORTRANGE_DEFAULT: 1181 inp->inp_flags &= ~(INP_LOWPORT); 1182 inp->inp_flags &= ~(INP_HIGHPORT); 1183 break; 1184 1185 case IP_PORTRANGE_HIGH: 1186 inp->inp_flags &= ~(INP_LOWPORT); 1187 inp->inp_flags |= INP_HIGHPORT; 1188 break; 1189 1190 case IP_PORTRANGE_LOW: 1191 inp->inp_flags &= ~(INP_HIGHPORT); 1192 inp->inp_flags |= INP_LOWPORT; 1193 break; 1194 1195 default: 1196 error = EINVAL; 1197 break; 1198 } 1199 break; 1200 1201#ifdef IPSEC 1202 case IP_IPSEC_POLICY: 1203 { 1204 caddr_t req; 1205 int priv; 1206 struct mbuf *m; 1207 int optname; 1208 1209 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1210 break; 1211 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1212 break; 1213 priv = (sopt->sopt_p != NULL && 1214 suser(sopt->sopt_p) != 0) ? 0 : 1; 1215 req = mtod(m, caddr_t); 1216 optname = sopt->sopt_name; 1217 error = ipsec4_set_policy(inp, optname, req, priv); 1218 m_freem(m); 1219 break; 1220 } 1221#endif /*IPSEC*/ 1222 1223 default: 1224 error = ENOPROTOOPT; 1225 break; 1226 } 1227 break; 1228 1229 case SOPT_GET: 1230 switch (sopt->sopt_name) { 1231 case IP_OPTIONS: 1232 case IP_RETOPTS: 1233 if (inp->inp_options) 1234 error = sooptcopyout(sopt, 1235 mtod(inp->inp_options, 1236 char *), 1237 inp->inp_options->m_len); 1238 else 1239 sopt->sopt_valsize = 0; 1240 break; 1241 1242 case IP_TOS: 1243 case IP_TTL: 1244 case IP_RECVOPTS: 1245 case IP_RECVRETOPTS: 1246 case IP_RECVDSTADDR: 1247 case IP_RECVIF: 1248 case IP_PORTRANGE: 1249#if defined(NFAITH) && NFAITH > 0 1250 case IP_FAITH: 1251#endif 1252 switch (sopt->sopt_name) { 1253 1254 case IP_TOS: 1255 optval = inp->inp_ip_tos; 1256 break; 1257 1258 case IP_TTL: 1259 optval = inp->inp_ip_ttl; 1260 break; 1261 1262#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1263 1264 case IP_RECVOPTS: 1265 optval = OPTBIT(INP_RECVOPTS); 1266 break; 1267 1268 case IP_RECVRETOPTS: 1269 optval = OPTBIT(INP_RECVRETOPTS); 1270 break; 1271 1272 case IP_RECVDSTADDR: 1273 optval = OPTBIT(INP_RECVDSTADDR); 1274 break; 1275 1276 case IP_RECVIF: 1277 optval = OPTBIT(INP_RECVIF); 1278 break; 1279 1280 case IP_PORTRANGE: 1281 if (inp->inp_flags & INP_HIGHPORT) 1282 optval = IP_PORTRANGE_HIGH; 1283 else if (inp->inp_flags & INP_LOWPORT) 1284 optval = IP_PORTRANGE_LOW; 1285 else 1286 optval = 0; 1287 break; 1288 1289#if defined(NFAITH) && NFAITH > 0 1290 case IP_FAITH: 1291 optval = OPTBIT(INP_FAITH); 1292 break; 1293#endif 1294 } 1295 error = sooptcopyout(sopt, &optval, sizeof optval); 1296 break; 1297 1298 case IP_MULTICAST_IF: 1299 case IP_MULTICAST_VIF: 1300 case IP_MULTICAST_TTL: 1301 case IP_MULTICAST_LOOP: 1302 case IP_ADD_MEMBERSHIP: 1303 case IP_DROP_MEMBERSHIP: 1304 error = ip_getmoptions(sopt, inp->inp_moptions); 1305 break; 1306 1307#ifdef IPSEC 1308 case IP_IPSEC_POLICY: 1309 { 1310 struct mbuf *m = NULL; 1311 caddr_t req = NULL; 1312 1313 if (m != 0) 1314 req = mtod(m, caddr_t); 1315 error = ipsec4_get_policy(sotoinpcb(so), req, &m); 1316 if (error == 0) 1317 error = soopt_mcopyout(sopt, m); /* XXX */ 1318 if (error == 0) 1319 m_freem(m); 1320 break; 1321 } 1322#endif /*IPSEC*/ 1323 1324 default: 1325 error = ENOPROTOOPT; 1326 break; 1327 } 1328 break; 1329 } 1330 return (error); 1331} 1332 1333/* 1334 * Set up IP options in pcb for insertion in output packets. 1335 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1336 * with destination address if source routed. 1337 */ 1338static int 1339ip_pcbopts(optname, pcbopt, m) 1340 int optname; 1341 struct mbuf **pcbopt; 1342 register struct mbuf *m; 1343{ 1344 register int cnt, optlen; 1345 register u_char *cp; 1346 u_char opt; 1347 1348 /* turn off any old options */ 1349 if (*pcbopt) 1350 (void)m_free(*pcbopt); 1351 *pcbopt = 0; 1352 if (m == (struct mbuf *)0 || m->m_len == 0) { 1353 /* 1354 * Only turning off any previous options. 1355 */ 1356 if (m) 1357 (void)m_free(m); 1358 return (0); 1359 } 1360 1361#ifndef vax 1362 if (m->m_len % sizeof(int32_t)) 1363 goto bad; 1364#endif 1365 /* 1366 * IP first-hop destination address will be stored before 1367 * actual options; move other options back 1368 * and clear it when none present. 1369 */ 1370 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1371 goto bad; 1372 cnt = m->m_len; 1373 m->m_len += sizeof(struct in_addr); 1374 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1375 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1376 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1377 1378 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1379 opt = cp[IPOPT_OPTVAL]; 1380 if (opt == IPOPT_EOL) 1381 break; 1382 if (opt == IPOPT_NOP) 1383 optlen = 1; 1384 else { 1385 optlen = cp[IPOPT_OLEN]; 1386 if (optlen <= IPOPT_OLEN || optlen > cnt) 1387 goto bad; 1388 } 1389 switch (opt) { 1390 1391 default: 1392 break; 1393 1394 case IPOPT_LSRR: 1395 case IPOPT_SSRR: 1396 /* 1397 * user process specifies route as: 1398 * ->A->B->C->D 1399 * D must be our final destination (but we can't 1400 * check that since we may not have connected yet). 1401 * A is first hop destination, which doesn't appear in 1402 * actual IP option, but is stored before the options. 1403 */ 1404 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1405 goto bad; 1406 m->m_len -= sizeof(struct in_addr); 1407 cnt -= sizeof(struct in_addr); 1408 optlen -= sizeof(struct in_addr); 1409 cp[IPOPT_OLEN] = optlen; 1410 /* 1411 * Move first hop before start of options. 1412 */ 1413 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1414 sizeof(struct in_addr)); 1415 /* 1416 * Then copy rest of options back 1417 * to close up the deleted entry. 1418 */ 1419 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1420 sizeof(struct in_addr)), 1421 (caddr_t)&cp[IPOPT_OFFSET+1], 1422 (unsigned)cnt + sizeof(struct in_addr)); 1423 break; 1424 } 1425 } 1426 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1427 goto bad; 1428 *pcbopt = m; 1429 return (0); 1430 1431bad: 1432 (void)m_free(m); 1433 return (EINVAL); 1434} 1435 1436/* 1437 * XXX 1438 * The whole multicast option thing needs to be re-thought. 1439 * Several of these options are equally applicable to non-multicast 1440 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1441 * standard option (IP_TTL). 1442 */ 1443/* 1444 * Set the IP multicast options in response to user setsockopt(). 1445 */ 1446static int 1447ip_setmoptions(sopt, imop) 1448 struct sockopt *sopt; 1449 struct ip_moptions **imop; 1450{ 1451 int error = 0; 1452 int i; 1453 struct in_addr addr; 1454 struct ip_mreq mreq; 1455 struct ifnet *ifp; 1456 struct ip_moptions *imo = *imop; 1457 struct route ro; 1458 struct sockaddr_in *dst; 1459 int s; 1460 1461 if (imo == NULL) { 1462 /* 1463 * No multicast option buffer attached to the pcb; 1464 * allocate one and initialize to default values. 1465 */ 1466 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1467 M_WAITOK); 1468 1469 if (imo == NULL) 1470 return (ENOBUFS); 1471 *imop = imo; 1472 imo->imo_multicast_ifp = NULL; 1473 imo->imo_multicast_vif = -1; 1474 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1475 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1476 imo->imo_num_memberships = 0; 1477 } 1478 1479 switch (sopt->sopt_name) { 1480 /* store an index number for the vif you wanna use in the send */ 1481 case IP_MULTICAST_VIF: 1482 if (legal_vif_num == 0) { 1483 error = EOPNOTSUPP; 1484 break; 1485 } 1486 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1487 if (error) 1488 break; 1489 if (!legal_vif_num(i) && (i != -1)) { 1490 error = EINVAL; 1491 break; 1492 } 1493 imo->imo_multicast_vif = i; 1494 break; 1495 1496 case IP_MULTICAST_IF: 1497 /* 1498 * Select the interface for outgoing multicast packets. 1499 */ 1500 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1501 if (error) 1502 break; 1503 /* 1504 * INADDR_ANY is used to remove a previous selection. 1505 * When no interface is selected, a default one is 1506 * chosen every time a multicast packet is sent. 1507 */ 1508 if (addr.s_addr == INADDR_ANY) { 1509 imo->imo_multicast_ifp = NULL; 1510 break; 1511 } 1512 /* 1513 * The selected interface is identified by its local 1514 * IP address. Find the interface and confirm that 1515 * it supports multicasting. 1516 */ 1517 s = splimp(); 1518 INADDR_TO_IFP(addr, ifp); 1519 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1520 splx(s); 1521 error = EADDRNOTAVAIL; 1522 break; 1523 } 1524 imo->imo_multicast_ifp = ifp; 1525 splx(s); 1526 break; 1527 1528 case IP_MULTICAST_TTL: 1529 /* 1530 * Set the IP time-to-live for outgoing multicast packets. 1531 * The original multicast API required a char argument, 1532 * which is inconsistent with the rest of the socket API. 1533 * We allow either a char or an int. 1534 */ 1535 if (sopt->sopt_valsize == 1) { 1536 u_char ttl; 1537 error = sooptcopyin(sopt, &ttl, 1, 1); 1538 if (error) 1539 break; 1540 imo->imo_multicast_ttl = ttl; 1541 } else { 1542 u_int ttl; 1543 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1544 sizeof ttl); 1545 if (error) 1546 break; 1547 if (ttl > 255) 1548 error = EINVAL; 1549 else 1550 imo->imo_multicast_ttl = ttl; 1551 } 1552 break; 1553 1554 case IP_MULTICAST_LOOP: 1555 /* 1556 * Set the loopback flag for outgoing multicast packets. 1557 * Must be zero or one. The original multicast API required a 1558 * char argument, which is inconsistent with the rest 1559 * of the socket API. We allow either a char or an int. 1560 */ 1561 if (sopt->sopt_valsize == 1) { 1562 u_char loop; 1563 error = sooptcopyin(sopt, &loop, 1, 1); 1564 if (error) 1565 break; 1566 imo->imo_multicast_loop = !!loop; 1567 } else { 1568 u_int loop; 1569 error = sooptcopyin(sopt, &loop, sizeof loop, 1570 sizeof loop); 1571 if (error) 1572 break; 1573 imo->imo_multicast_loop = !!loop; 1574 } 1575 break; 1576 1577 case IP_ADD_MEMBERSHIP: 1578 /* 1579 * Add a multicast group membership. 1580 * Group must be a valid IP multicast address. 1581 */ 1582 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1583 if (error) 1584 break; 1585 1586 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1587 error = EINVAL; 1588 break; 1589 } 1590 s = splimp(); 1591 /* 1592 * If no interface address was provided, use the interface of 1593 * the route to the given multicast address. 1594 */ 1595 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1596 bzero((caddr_t)&ro, sizeof(ro)); 1597 dst = (struct sockaddr_in *)&ro.ro_dst; 1598 dst->sin_len = sizeof(*dst); 1599 dst->sin_family = AF_INET; 1600 dst->sin_addr = mreq.imr_multiaddr; 1601 rtalloc(&ro); 1602 if (ro.ro_rt == NULL) { 1603 error = EADDRNOTAVAIL; 1604 splx(s); 1605 break; 1606 } 1607 ifp = ro.ro_rt->rt_ifp; 1608 rtfree(ro.ro_rt); 1609 } 1610 else { 1611 INADDR_TO_IFP(mreq.imr_interface, ifp); 1612 } 1613 1614 /* 1615 * See if we found an interface, and confirm that it 1616 * supports multicast. 1617 */ 1618 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1619 error = EADDRNOTAVAIL; 1620 splx(s); 1621 break; 1622 } 1623 /* 1624 * See if the membership already exists or if all the 1625 * membership slots are full. 1626 */ 1627 for (i = 0; i < imo->imo_num_memberships; ++i) { 1628 if (imo->imo_membership[i]->inm_ifp == ifp && 1629 imo->imo_membership[i]->inm_addr.s_addr 1630 == mreq.imr_multiaddr.s_addr) 1631 break; 1632 } 1633 if (i < imo->imo_num_memberships) { 1634 error = EADDRINUSE; 1635 splx(s); 1636 break; 1637 } 1638 if (i == IP_MAX_MEMBERSHIPS) { 1639 error = ETOOMANYREFS; 1640 splx(s); 1641 break; 1642 } 1643 /* 1644 * Everything looks good; add a new record to the multicast 1645 * address list for the given interface. 1646 */ 1647 if ((imo->imo_membership[i] = 1648 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1649 error = ENOBUFS; 1650 splx(s); 1651 break; 1652 } 1653 ++imo->imo_num_memberships; 1654 splx(s); 1655 break; 1656 1657 case IP_DROP_MEMBERSHIP: 1658 /* 1659 * Drop a multicast group membership. 1660 * Group must be a valid IP multicast address. 1661 */ 1662 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1663 if (error) 1664 break; 1665 1666 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1667 error = EINVAL; 1668 break; 1669 } 1670 1671 s = splimp(); 1672 /* 1673 * If an interface address was specified, get a pointer 1674 * to its ifnet structure. 1675 */ 1676 if (mreq.imr_interface.s_addr == INADDR_ANY) 1677 ifp = NULL; 1678 else { 1679 INADDR_TO_IFP(mreq.imr_interface, ifp); 1680 if (ifp == NULL) { 1681 error = EADDRNOTAVAIL; 1682 splx(s); 1683 break; 1684 } 1685 } 1686 /* 1687 * Find the membership in the membership array. 1688 */ 1689 for (i = 0; i < imo->imo_num_memberships; ++i) { 1690 if ((ifp == NULL || 1691 imo->imo_membership[i]->inm_ifp == ifp) && 1692 imo->imo_membership[i]->inm_addr.s_addr == 1693 mreq.imr_multiaddr.s_addr) 1694 break; 1695 } 1696 if (i == imo->imo_num_memberships) { 1697 error = EADDRNOTAVAIL; 1698 splx(s); 1699 break; 1700 } 1701 /* 1702 * Give up the multicast address record to which the 1703 * membership points. 1704 */ 1705 in_delmulti(imo->imo_membership[i]); 1706 /* 1707 * Remove the gap in the membership array. 1708 */ 1709 for (++i; i < imo->imo_num_memberships; ++i) 1710 imo->imo_membership[i-1] = imo->imo_membership[i]; 1711 --imo->imo_num_memberships; 1712 splx(s); 1713 break; 1714 1715 default: 1716 error = EOPNOTSUPP; 1717 break; 1718 } 1719 1720 /* 1721 * If all options have default values, no need to keep the mbuf. 1722 */ 1723 if (imo->imo_multicast_ifp == NULL && 1724 imo->imo_multicast_vif == -1 && 1725 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1726 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1727 imo->imo_num_memberships == 0) { 1728 free(*imop, M_IPMOPTS); 1729 *imop = NULL; 1730 } 1731 1732 return (error); 1733} 1734 1735/* 1736 * Return the IP multicast options in response to user getsockopt(). 1737 */ 1738static int 1739ip_getmoptions(sopt, imo) 1740 struct sockopt *sopt; 1741 register struct ip_moptions *imo; 1742{ 1743 struct in_addr addr; 1744 struct in_ifaddr *ia; 1745 int error, optval; 1746 u_char coptval; 1747 1748 error = 0; 1749 switch (sopt->sopt_name) { 1750 case IP_MULTICAST_VIF: 1751 if (imo != NULL) 1752 optval = imo->imo_multicast_vif; 1753 else 1754 optval = -1; 1755 error = sooptcopyout(sopt, &optval, sizeof optval); 1756 break; 1757 1758 case IP_MULTICAST_IF: 1759 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1760 addr.s_addr = INADDR_ANY; 1761 else { 1762 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1763 addr.s_addr = (ia == NULL) ? INADDR_ANY 1764 : IA_SIN(ia)->sin_addr.s_addr; 1765 } 1766 error = sooptcopyout(sopt, &addr, sizeof addr); 1767 break; 1768 1769 case IP_MULTICAST_TTL: 1770 if (imo == 0) 1771 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1772 else 1773 optval = coptval = imo->imo_multicast_ttl; 1774 if (sopt->sopt_valsize == 1) 1775 error = sooptcopyout(sopt, &coptval, 1); 1776 else 1777 error = sooptcopyout(sopt, &optval, sizeof optval); 1778 break; 1779 1780 case IP_MULTICAST_LOOP: 1781 if (imo == 0) 1782 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1783 else 1784 optval = coptval = imo->imo_multicast_loop; 1785 if (sopt->sopt_valsize == 1) 1786 error = sooptcopyout(sopt, &coptval, 1); 1787 else 1788 error = sooptcopyout(sopt, &optval, sizeof optval); 1789 break; 1790 1791 default: 1792 error = ENOPROTOOPT; 1793 break; 1794 } 1795 return (error); 1796} 1797 1798/* 1799 * Discard the IP multicast options. 1800 */ 1801void 1802ip_freemoptions(imo) 1803 register struct ip_moptions *imo; 1804{ 1805 register int i; 1806 1807 if (imo != NULL) { 1808 for (i = 0; i < imo->imo_num_memberships; ++i) 1809 in_delmulti(imo->imo_membership[i]); 1810 free(imo, M_IPMOPTS); 1811 } 1812} 1813 1814/* 1815 * Routine called from ip_output() to loop back a copy of an IP multicast 1816 * packet to the input queue of a specified interface. Note that this 1817 * calls the output routine of the loopback "driver", but with an interface 1818 * pointer that might NOT be a loopback interface -- evil, but easier than 1819 * replicating that code here. 1820 */ 1821static void 1822ip_mloopback(ifp, m, dst, hlen) 1823 struct ifnet *ifp; 1824 register struct mbuf *m; 1825 register struct sockaddr_in *dst; 1826 int hlen; 1827{ 1828 register struct ip *ip; 1829 struct mbuf *copym; 1830 1831 copym = m_copy(m, 0, M_COPYALL); 1832 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1833 copym = m_pullup(copym, hlen); 1834 if (copym != NULL) { 1835 /* 1836 * We don't bother to fragment if the IP length is greater 1837 * than the interface's MTU. Can this possibly matter? 1838 */ 1839 ip = mtod(copym, struct ip *); 1840 ip->ip_len = htons((u_short)ip->ip_len); 1841 ip->ip_off = htons((u_short)ip->ip_off); 1842 ip->ip_sum = 0; 1843 if (ip->ip_vhl == IP_VHL_BORING) { 1844 ip->ip_sum = in_cksum_hdr(ip); 1845 } else { 1846 ip->ip_sum = in_cksum(copym, hlen); 1847 } 1848 /* 1849 * NB: 1850 * It's not clear whether there are any lingering 1851 * reentrancy problems in other areas which might 1852 * be exposed by using ip_input directly (in 1853 * particular, everything which modifies the packet 1854 * in-place). Yet another option is using the 1855 * protosw directly to deliver the looped back 1856 * packet. For the moment, we'll err on the side 1857 * of safety by using if_simloop(). 1858 */ 1859#if 1 /* XXX */ 1860 if (dst->sin_family != AF_INET) { 1861 printf("ip_mloopback: bad address family %d\n", 1862 dst->sin_family); 1863 dst->sin_family = AF_INET; 1864 } 1865#endif 1866 1867#ifdef notdef 1868 copym->m_pkthdr.rcvif = ifp; 1869 ip_input(copym); 1870#else 1871 /* if the checksum hasn't been computed, mark it as valid */ 1872 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1873 copym->m_pkthdr.csum_flags |= 1874 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1875 copym->m_pkthdr.csum_data = 0xffff; 1876 } 1877 if_simloop(ifp, copym, dst->sin_family, 0); 1878#endif 1879 } 1880} 1881