ip_output.c revision 67833
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 67833 2000-10-29 01:05:09Z joe $ 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#include "opt_pfil_hooks.h" 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/kernel.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/protosw.h> 52#include <sys/socket.h> 53#include <sys/socketvar.h> 54#include <sys/proc.h> 55 56#include <net/if.h> 57#include <net/route.h> 58 59#include <netinet/in.h> 60#include <netinet/in_systm.h> 61#include <netinet/ip.h> 62#include <netinet/in_pcb.h> 63#include <netinet/in_var.h> 64#include <netinet/ip_var.h> 65 66#include "faith.h" 67 68#ifdef vax 69#include <machine/mtpr.h> 70#endif 71#include <machine/in_cksum.h> 72 73static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 74 75#ifdef IPSEC 76#include <netinet6/ipsec.h> 77#include <netkey/key.h> 78#ifdef IPSEC_DEBUG 79#include <netkey/key_debug.h> 80#else 81#define KEYDEBUG(lev,arg) 82#endif 83#endif /*IPSEC*/ 84 85#include <netinet/ip_fw.h> 86 87#ifdef DUMMYNET 88#include <netinet/ip_dummynet.h> 89#endif 90 91#ifdef IPFIREWALL_FORWARD_DEBUG 92#define print_ip(a) printf("%ld.%ld.%ld.%ld",(ntohl(a.s_addr)>>24)&0xFF,\ 93 (ntohl(a.s_addr)>>16)&0xFF,\ 94 (ntohl(a.s_addr)>>8)&0xFF,\ 95 (ntohl(a.s_addr))&0xFF); 96#endif 97 98u_short ip_id; 99 100static struct mbuf *ip_insertoptions __P((struct mbuf *, struct mbuf *, int *)); 101static void ip_mloopback 102 __P((struct ifnet *, struct mbuf *, struct sockaddr_in *, int)); 103static int ip_getmoptions 104 __P((struct sockopt *, struct ip_moptions *)); 105static int ip_pcbopts __P((int, struct mbuf **, struct mbuf *)); 106static int ip_setmoptions 107 __P((struct sockopt *, struct ip_moptions **)); 108 109int ip_optcopy __P((struct ip *, struct ip *)); 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 PFIL_HOOKS 143 struct packet_filter_hook *pfh; 144 struct mbuf *m1; 145 int rv; 146#endif /* PFIL_HOOKS */ 147#ifdef IPFIREWALL_FORWARD 148 int fwd_rewrite_src = 0; 149#endif 150 struct ip_fw_chain *rule = NULL; 151 152#ifdef IPDIVERT 153 /* Get and reset firewall cookie */ 154 divert_cookie = ip_divert_cookie; 155 ip_divert_cookie = 0; 156#else 157 divert_cookie = 0; 158#endif 159 160#if defined(IPFIREWALL) && defined(DUMMYNET) 161 /* 162 * dummynet packet are prepended a vestigial mbuf with 163 * m_type = MT_DUMMYNET and m_data pointing to the matching 164 * rule. 165 */ 166 if (m->m_type == MT_DUMMYNET) { 167 /* 168 * the packet was already tagged, so part of the 169 * processing was already done, and we need to go down. 170 * Get parameters from the header. 171 */ 172 rule = (struct ip_fw_chain *)(m->m_data) ; 173 opt = NULL ; 174 ro = & ( ((struct dn_pkt *)m)->ro ) ; 175 imo = NULL ; 176 dst = ((struct dn_pkt *)m)->dn_dst ; 177 ifp = ((struct dn_pkt *)m)->ifp ; 178 flags = ((struct dn_pkt *)m)->flags ; 179 180 m0 = m = m->m_next ; 181#ifdef IPSEC 182 so = ipsec_getsocket(m); 183 ipsec_setsocket(m, NULL); 184#endif 185 ip = mtod(m, struct ip *); 186 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 187 goto sendit; 188 } else 189 rule = NULL ; 190#endif 191#ifdef IPSEC 192 so = ipsec_getsocket(m); 193 ipsec_setsocket(m, NULL); 194#endif 195 196#ifdef DIAGNOSTIC 197 if ((m->m_flags & M_PKTHDR) == 0) 198 panic("ip_output no HDR"); 199 if (!ro) 200 panic("ip_output no route, proto = %d", 201 mtod(m, struct ip *)->ip_p); 202#endif 203 if (opt) { 204 m = ip_insertoptions(m, opt, &len); 205 hlen = len; 206 } 207 ip = mtod(m, struct ip *); 208 /* 209 * Fill in IP header. 210 */ 211 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 212 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 213 ip->ip_off &= IP_DF; 214 ip->ip_id = htons(ip_id++); 215 ipstat.ips_localout++; 216 } else { 217 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 218 } 219 220 dst = (struct sockaddr_in *)&ro->ro_dst; 221 /* 222 * If there is a cached route, 223 * check that it is to the same destination 224 * and is still up. If not, free it and try again. 225 */ 226 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 227 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 228 RTFREE(ro->ro_rt); 229 ro->ro_rt = (struct rtentry *)0; 230 } 231 if (ro->ro_rt == 0) { 232 dst->sin_family = AF_INET; 233 dst->sin_len = sizeof(*dst); 234 dst->sin_addr = ip->ip_dst; 235 } 236 /* 237 * If routing to interface only, 238 * short circuit routing lookup. 239 */ 240#define ifatoia(ifa) ((struct in_ifaddr *)(ifa)) 241#define sintosa(sin) ((struct sockaddr *)(sin)) 242 if (flags & IP_ROUTETOIF) { 243 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 244 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 245 ipstat.ips_noroute++; 246 error = ENETUNREACH; 247 goto bad; 248 } 249 ifp = ia->ia_ifp; 250 ip->ip_ttl = 1; 251 isbroadcast = in_broadcast(dst->sin_addr, ifp); 252 } else { 253 /* 254 * If this is the case, we probably don't want to allocate 255 * a protocol-cloned route since we didn't get one from the 256 * ULP. This lets TCP do its thing, while not burdening 257 * forwarding or ICMP with the overhead of cloning a route. 258 * Of course, we still want to do any cloning requested by 259 * the link layer, as this is probably required in all cases 260 * for correct operation (as it is for ARP). 261 */ 262 if (ro->ro_rt == 0) 263 rtalloc_ign(ro, RTF_PRCLONING); 264 if (ro->ro_rt == 0) { 265 ipstat.ips_noroute++; 266 error = EHOSTUNREACH; 267 goto bad; 268 } 269 ia = ifatoia(ro->ro_rt->rt_ifa); 270 ifp = ro->ro_rt->rt_ifp; 271 ro->ro_rt->rt_use++; 272 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 273 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 274 if (ro->ro_rt->rt_flags & RTF_HOST) 275 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 276 else 277 isbroadcast = in_broadcast(dst->sin_addr, ifp); 278 } 279 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 280 struct in_multi *inm; 281 282 m->m_flags |= M_MCAST; 283 /* 284 * IP destination address is multicast. Make sure "dst" 285 * still points to the address in "ro". (It may have been 286 * changed to point to a gateway address, above.) 287 */ 288 dst = (struct sockaddr_in *)&ro->ro_dst; 289 /* 290 * See if the caller provided any multicast options 291 */ 292 if (imo != NULL) { 293 ip->ip_ttl = imo->imo_multicast_ttl; 294 if (imo->imo_multicast_ifp != NULL) 295 ifp = imo->imo_multicast_ifp; 296 if (imo->imo_multicast_vif != -1) 297 ip->ip_src.s_addr = 298 ip_mcast_src(imo->imo_multicast_vif); 299 } else 300 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 301 /* 302 * Confirm that the outgoing interface supports multicast. 303 */ 304 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 305 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 306 ipstat.ips_noroute++; 307 error = ENETUNREACH; 308 goto bad; 309 } 310 } 311 /* 312 * If source address not specified yet, use address 313 * of outgoing interface. 314 */ 315 if (ip->ip_src.s_addr == INADDR_ANY) { 316 register struct in_ifaddr *ia1; 317 318 for (ia1 = in_ifaddrhead.tqh_first; ia1; 319 ia1 = ia1->ia_link.tqe_next) 320 if (ia1->ia_ifp == ifp) { 321 ip->ip_src = IA_SIN(ia1)->sin_addr; 322 break; 323 } 324 } 325 326 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 327 if (inm != NULL && 328 (imo == NULL || imo->imo_multicast_loop)) { 329 /* 330 * If we belong to the destination multicast group 331 * on the outgoing interface, and the caller did not 332 * forbid loopback, loop back a copy. 333 */ 334 ip_mloopback(ifp, m, dst, hlen); 335 } 336 else { 337 /* 338 * If we are acting as a multicast router, perform 339 * multicast forwarding as if the packet had just 340 * arrived on the interface to which we are about 341 * to send. The multicast forwarding function 342 * recursively calls this function, using the 343 * IP_FORWARDING flag to prevent infinite recursion. 344 * 345 * Multicasts that are looped back by ip_mloopback(), 346 * above, will be forwarded by the ip_input() routine, 347 * if necessary. 348 */ 349 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 350 /* 351 * Check if rsvp daemon is running. If not, don't 352 * set ip_moptions. This ensures that the packet 353 * is multicast and not just sent down one link 354 * as prescribed by rsvpd. 355 */ 356 if (!rsvp_on) 357 imo = NULL; 358 if (ip_mforward(ip, ifp, m, imo) != 0) { 359 m_freem(m); 360 goto done; 361 } 362 } 363 } 364 365 /* 366 * Multicasts with a time-to-live of zero may be looped- 367 * back, above, but must not be transmitted on a network. 368 * Also, multicasts addressed to the loopback interface 369 * are not sent -- the above call to ip_mloopback() will 370 * loop back a copy if this host actually belongs to the 371 * destination group on the loopback interface. 372 */ 373 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 374 m_freem(m); 375 goto done; 376 } 377 378 goto sendit; 379 } 380#ifndef notdef 381 /* 382 * If source address not specified yet, use address 383 * of outgoing interface. 384 */ 385 if (ip->ip_src.s_addr == INADDR_ANY) { 386 ip->ip_src = IA_SIN(ia)->sin_addr; 387#ifdef IPFIREWALL_FORWARD 388 /* Keep note that we did this - if the firewall changes 389 * the next-hop, our interface may change, changing the 390 * default source IP. It's a shame so much effort happens 391 * twice. Oh well. 392 */ 393 fwd_rewrite_src++; 394#endif /* IPFIREWALL_FORWARD */ 395 } 396#endif /* notdef */ 397 /* 398 * Verify that we have any chance at all of being able to queue 399 * the packet or packet fragments 400 */ 401 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 402 ifp->if_snd.ifq_maxlen) { 403 error = ENOBUFS; 404 goto bad; 405 } 406 407 /* 408 * Look for broadcast address and 409 * and verify user is allowed to send 410 * such a packet. 411 */ 412 if (isbroadcast) { 413 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 414 error = EADDRNOTAVAIL; 415 goto bad; 416 } 417 if ((flags & IP_ALLOWBROADCAST) == 0) { 418 error = EACCES; 419 goto bad; 420 } 421 /* don't allow broadcast messages to be fragmented */ 422 if ((u_short)ip->ip_len > ifp->if_mtu) { 423 error = EMSGSIZE; 424 goto bad; 425 } 426 m->m_flags |= M_BCAST; 427 } else { 428 m->m_flags &= ~M_BCAST; 429 } 430 431sendit: 432 /* 433 * IpHack's section. 434 * - Xlate: translate packet's addr/port (NAT). 435 * - Firewall: deny/allow/etc. 436 * - Wrap: fake packet's addr/port <unimpl.> 437 * - Encapsulate: put it in another IP and send out. <unimp.> 438 */ 439#ifdef PFIL_HOOKS 440 /* 441 * Run through list of hooks for output packets. 442 */ 443 m1 = m; 444 pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 445 for (; pfh; pfh = pfh->pfil_link.tqe_next) 446 if (pfh->pfil_func) { 447 rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1); 448 if (rv) { 449 error = EHOSTUNREACH; 450 goto done; 451 } 452 m = m1; 453 if (m == NULL) 454 goto done; 455 ip = mtod(m, struct ip *); 456 } 457#endif /* PFIL_HOOKS */ 458 459 /* 460 * Check with the firewall... 461 */ 462 if (fw_enable && ip_fw_chk_ptr) { 463 struct sockaddr_in *old = dst; 464 465 off = (*ip_fw_chk_ptr)(&ip, 466 hlen, ifp, &divert_cookie, &m, &rule, &dst); 467 /* 468 * On return we must do the following: 469 * m == NULL -> drop the pkt 470 * 1<=off<= 0xffff -> DIVERT 471 * (off & 0x10000) -> send to a DUMMYNET pipe 472 * (off & 0x20000) -> TEE the packet 473 * dst != old -> IPFIREWALL_FORWARD 474 * off==0, dst==old -> accept 475 * If some of the above modules is not compiled in, then 476 * we should't have to check the corresponding condition 477 * (because the ipfw control socket should not accept 478 * unsupported rules), but better play safe and drop 479 * packets in case of doubt. 480 */ 481 if (!m) { /* firewall said to reject */ 482 error = EACCES; 483 goto done; 484 } 485 if (off == 0 && dst == old) /* common case */ 486 goto pass ; 487#ifdef DUMMYNET 488 if ((off & IP_FW_PORT_DYNT_FLAG) != 0) { 489 /* 490 * pass the pkt to dummynet. Need to include 491 * pipe number, m, ifp, ro, dst because these are 492 * not recomputed in the next pass. 493 * All other parameters have been already used and 494 * so they are not needed anymore. 495 * XXX note: if the ifp or ro entry are deleted 496 * while a pkt is in dummynet, we are in trouble! 497 */ 498 dummynet_io(off & 0xffff, DN_TO_IP_OUT, m,ifp,ro,dst,rule, 499 flags); 500 goto done; 501 } 502#endif 503#ifdef IPDIVERT 504 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 505 struct mbuf *clone = NULL; 506 507 /* Clone packet if we're doing a 'tee' */ 508 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 509 clone = m_dup(m, M_DONTWAIT); 510 511 /* 512 * XXX 513 * delayed checksums are not currently compatible 514 * with divert sockets. 515 */ 516 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 517 in_delayed_cksum(m); 518 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 519 } 520 521 /* Restore packet header fields to original values */ 522 HTONS(ip->ip_len); 523 HTONS(ip->ip_off); 524 525 /* Deliver packet to divert input routine */ 526 ip_divert_cookie = divert_cookie; 527 divert_packet(m, 0, off & 0xffff); 528 529 /* If 'tee', continue with original packet */ 530 if (clone != NULL) { 531 m = clone; 532 ip = mtod(m, struct ip *); 533 goto pass; 534 } 535 goto done; 536 } 537#endif 538 539#ifdef IPFIREWALL_FORWARD 540 /* Here we check dst to make sure it's directly reachable on the 541 * interface we previously thought it was. 542 * If it isn't (which may be likely in some situations) we have 543 * to re-route it (ie, find a route for the next-hop and the 544 * associated interface) and set them here. This is nested 545 * forwarding which in most cases is undesirable, except where 546 * such control is nigh impossible. So we do it here. 547 * And I'm babbling. 548 */ 549 if (off == 0 && old != dst) { 550 struct in_ifaddr *ia; 551 552 /* It's changed... */ 553 /* There must be a better way to do this next line... */ 554 static struct route sro_fwd, *ro_fwd = &sro_fwd; 555#ifdef IPFIREWALL_FORWARD_DEBUG 556 printf("IPFIREWALL_FORWARD: New dst ip: "); 557 print_ip(dst->sin_addr); 558 printf("\n"); 559#endif 560 /* 561 * We need to figure out if we have been forwarded 562 * to a local socket. If so then we should somehow 563 * "loop back" to ip_input, and get directed to the 564 * PCB as if we had received this packet. This is 565 * because it may be dificult to identify the packets 566 * you want to forward until they are being output 567 * and have selected an interface. (e.g. locally 568 * initiated packets) If we used the loopback inteface, 569 * we would not be able to control what happens 570 * as the packet runs through ip_input() as 571 * it is done through a ISR. 572 */ 573 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 574 ia = TAILQ_NEXT(ia, ia_link)) { 575 /* 576 * If the addr to forward to is one 577 * of ours, we pretend to 578 * be the destination for this packet. 579 */ 580 if (IA_SIN(ia)->sin_addr.s_addr == 581 dst->sin_addr.s_addr) 582 break; 583 } 584 if (ia) { 585 /* tell ip_input "dont filter" */ 586 ip_fw_fwd_addr = dst; 587 if (m->m_pkthdr.rcvif == NULL) 588 m->m_pkthdr.rcvif = ifunit("lo0"); 589 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 590 m->m_pkthdr.csum_flags |= 591 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 592 m0->m_pkthdr.csum_data = 0xffff; 593 } 594 m->m_pkthdr.csum_flags |= 595 CSUM_IP_CHECKED | CSUM_IP_VALID; 596 HTONS(ip->ip_len); 597 HTONS(ip->ip_off); 598 ip_input(m); 599 goto done; 600 } 601 /* Some of the logic for this was 602 * nicked from above. 603 * 604 * This rewrites the cached route in a local PCB. 605 * Is this what we want to do? 606 */ 607 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 608 609 ro_fwd->ro_rt = 0; 610 rtalloc_ign(ro_fwd, RTF_PRCLONING); 611 612 if (ro_fwd->ro_rt == 0) { 613 ipstat.ips_noroute++; 614 error = EHOSTUNREACH; 615 goto bad; 616 } 617 618 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 619 ifp = ro_fwd->ro_rt->rt_ifp; 620 ro_fwd->ro_rt->rt_use++; 621 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 622 dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway; 623 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 624 isbroadcast = 625 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 626 else 627 isbroadcast = in_broadcast(dst->sin_addr, ifp); 628 RTFREE(ro->ro_rt); 629 ro->ro_rt = ro_fwd->ro_rt; 630 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 631 632 /* 633 * If we added a default src ip earlier, 634 * which would have been gotten from the-then 635 * interface, do it again, from the new one. 636 */ 637 if (fwd_rewrite_src) 638 ip->ip_src = IA_SIN(ia)->sin_addr; 639 goto pass ; 640 } 641#endif /* IPFIREWALL_FORWARD */ 642 /* 643 * if we get here, none of the above matches, and 644 * we have to drop the pkt 645 */ 646 m_freem(m); 647 error = EACCES; /* not sure this is the right error msg */ 648 goto done; 649 } 650 651pass: 652#ifdef IPSEC 653 /* get SP for this packet */ 654 if (so == NULL) 655 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 656 else 657 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 658 659 if (sp == NULL) { 660 ipsecstat.out_inval++; 661 goto bad; 662 } 663 664 error = 0; 665 666 /* check policy */ 667 switch (sp->policy) { 668 case IPSEC_POLICY_DISCARD: 669 /* 670 * This packet is just discarded. 671 */ 672 ipsecstat.out_polvio++; 673 goto bad; 674 675 case IPSEC_POLICY_BYPASS: 676 case IPSEC_POLICY_NONE: 677 /* no need to do IPsec. */ 678 goto skip_ipsec; 679 680 case IPSEC_POLICY_IPSEC: 681 if (sp->req == NULL) { 682 /* XXX should be panic ? */ 683 printf("ip_output: No IPsec request specified.\n"); 684 error = EINVAL; 685 goto bad; 686 } 687 break; 688 689 case IPSEC_POLICY_ENTRUST: 690 default: 691 printf("ip_output: Invalid policy found. %d\n", sp->policy); 692 } 693 { 694 struct ipsec_output_state state; 695 bzero(&state, sizeof(state)); 696 state.m = m; 697 if (flags & IP_ROUTETOIF) { 698 state.ro = &iproute; 699 bzero(&iproute, sizeof(iproute)); 700 } else 701 state.ro = ro; 702 state.dst = (struct sockaddr *)dst; 703 704 ip->ip_sum = 0; 705 706 /* 707 * XXX 708 * delayed checksums are not currently compatible with IPsec 709 */ 710 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 711 in_delayed_cksum(m); 712 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 713 } 714 715 HTONS(ip->ip_len); 716 HTONS(ip->ip_off); 717 718 error = ipsec4_output(&state, sp, flags); 719 720 m = state.m; 721 if (flags & IP_ROUTETOIF) { 722 /* 723 * if we have tunnel mode SA, we may need to ignore 724 * IP_ROUTETOIF. 725 */ 726 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 727 flags &= ~IP_ROUTETOIF; 728 ro = state.ro; 729 } 730 } else 731 ro = state.ro; 732 dst = (struct sockaddr_in *)state.dst; 733 if (error) { 734 /* mbuf is already reclaimed in ipsec4_output. */ 735 m0 = NULL; 736 switch (error) { 737 case EHOSTUNREACH: 738 case ENETUNREACH: 739 case EMSGSIZE: 740 case ENOBUFS: 741 case ENOMEM: 742 break; 743 default: 744 printf("ip4_output (ipsec): error code %d\n", error); 745 /*fall through*/ 746 case ENOENT: 747 /* don't show these error codes to the user */ 748 error = 0; 749 break; 750 } 751 goto bad; 752 } 753 } 754 755 /* be sure to update variables that are affected by ipsec4_output() */ 756 ip = mtod(m, struct ip *); 757#ifdef _IP_VHL 758 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 759#else 760 hlen = ip->ip_hl << 2; 761#endif 762 if (ro->ro_rt == NULL) { 763 if ((flags & IP_ROUTETOIF) == 0) { 764 printf("ip_output: " 765 "can't update route after IPsec processing\n"); 766 error = EHOSTUNREACH; /*XXX*/ 767 goto bad; 768 } 769 } else { 770 /* nobody uses ia beyond here */ 771 ifp = ro->ro_rt->rt_ifp; 772 } 773 774 /* make it flipped, again. */ 775 NTOHS(ip->ip_len); 776 NTOHS(ip->ip_off); 777skip_ipsec: 778#endif /*IPSEC*/ 779 780 sw_csum = m->m_pkthdr.csum_flags | CSUM_IP; 781 m->m_pkthdr.csum_flags = sw_csum & ifp->if_hwassist; 782 sw_csum &= ~ifp->if_hwassist; 783 if (sw_csum & CSUM_DELAY_DATA) { 784 in_delayed_cksum(m); 785 sw_csum &= ~CSUM_DELAY_DATA; 786 } 787 788 /* 789 * If small enough for interface, or the interface will take 790 * care of the fragmentation for us, can just send directly. 791 */ 792 if ((u_short)ip->ip_len <= ifp->if_mtu || 793 ifp->if_hwassist & CSUM_FRAGMENT) { 794 HTONS(ip->ip_len); 795 HTONS(ip->ip_off); 796 ip->ip_sum = 0; 797 if (sw_csum & CSUM_DELAY_IP) { 798 if (ip->ip_vhl == IP_VHL_BORING) { 799 ip->ip_sum = in_cksum_hdr(ip); 800 } else { 801 ip->ip_sum = in_cksum(m, hlen); 802 } 803 } 804 805 /* Record statistics for this interface address. */ 806 if (!(flags & IP_FORWARDING)) { 807 ia->ia_ifa.if_opackets++; 808 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 809 } 810 811 error = (*ifp->if_output)(ifp, m, 812 (struct sockaddr *)dst, ro->ro_rt); 813 goto done; 814 } 815 /* 816 * Too large for interface; fragment if possible. 817 * Must be able to put at least 8 bytes per fragment. 818 */ 819 if (ip->ip_off & IP_DF) { 820 error = EMSGSIZE; 821 /* 822 * This case can happen if the user changed the MTU 823 * of an interface after enabling IP on it. Because 824 * most netifs don't keep track of routes pointing to 825 * them, there is no way for one to update all its 826 * routes when the MTU is changed. 827 */ 828 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 829 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 830 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 831 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 832 } 833 ipstat.ips_cantfrag++; 834 goto bad; 835 } 836 len = (ifp->if_mtu - hlen) &~ 7; 837 if (len < 8) { 838 error = EMSGSIZE; 839 goto bad; 840 } 841 842 /* 843 * if the interface will not calculate checksums on 844 * fragmented packets, then do it here. 845 */ 846 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 847 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 848 in_delayed_cksum(m); 849 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 850 } 851 852 { 853 int mhlen, firstlen = len; 854 struct mbuf **mnext = &m->m_nextpkt; 855 int nfrags = 1; 856 857 /* 858 * Loop through length of segment after first fragment, 859 * make new header and copy data of each part and link onto chain. 860 */ 861 m0 = m; 862 mhlen = sizeof (struct ip); 863 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) { 864 MGETHDR(m, M_DONTWAIT, MT_HEADER); 865 if (m == 0) { 866 error = ENOBUFS; 867 ipstat.ips_odropped++; 868 goto sendorfree; 869 } 870 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 871 m->m_data += max_linkhdr; 872 mhip = mtod(m, struct ip *); 873 *mhip = *ip; 874 if (hlen > sizeof (struct ip)) { 875 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 876 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 877 } 878 m->m_len = mhlen; 879 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 880 if (off + len >= (u_short)ip->ip_len) 881 len = (u_short)ip->ip_len - off; 882 else 883 mhip->ip_off |= IP_MF; 884 mhip->ip_len = htons((u_short)(len + mhlen)); 885 m->m_next = m_copy(m0, off, len); 886 if (m->m_next == 0) { 887 (void) m_free(m); 888 error = ENOBUFS; /* ??? */ 889 ipstat.ips_odropped++; 890 goto sendorfree; 891 } 892 m->m_pkthdr.len = mhlen + len; 893 m->m_pkthdr.rcvif = (struct ifnet *)0; 894 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 895 HTONS(mhip->ip_off); 896 mhip->ip_sum = 0; 897 if (sw_csum & CSUM_DELAY_IP) { 898 if (mhip->ip_vhl == IP_VHL_BORING) { 899 mhip->ip_sum = in_cksum_hdr(mhip); 900 } else { 901 mhip->ip_sum = in_cksum(m, mhlen); 902 } 903 } 904 *mnext = m; 905 mnext = &m->m_nextpkt; 906 nfrags++; 907 } 908 ipstat.ips_ofragments += nfrags; 909 910 /* set first/last markers for fragment chain */ 911 m->m_flags |= M_LASTFRAG; 912 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 913 m0->m_pkthdr.csum_data = nfrags; 914 915 /* 916 * Update first fragment by trimming what's been copied out 917 * and updating header, then send each fragment (in order). 918 */ 919 m = m0; 920 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 921 m->m_pkthdr.len = hlen + firstlen; 922 ip->ip_len = htons((u_short)m->m_pkthdr.len); 923 ip->ip_off |= IP_MF; 924 HTONS(ip->ip_off); 925 ip->ip_sum = 0; 926 if (sw_csum & CSUM_DELAY_IP) { 927 if (ip->ip_vhl == IP_VHL_BORING) { 928 ip->ip_sum = in_cksum_hdr(ip); 929 } else { 930 ip->ip_sum = in_cksum(m, hlen); 931 } 932 } 933sendorfree: 934 for (m = m0; m; m = m0) { 935 m0 = m->m_nextpkt; 936 m->m_nextpkt = 0; 937 if (error == 0) { 938 /* Record statistics for this interface address. */ 939 ia->ia_ifa.if_opackets++; 940 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 941 942 error = (*ifp->if_output)(ifp, m, 943 (struct sockaddr *)dst, ro->ro_rt); 944 } else 945 m_freem(m); 946 } 947 948 if (error == 0) 949 ipstat.ips_fragmented++; 950 } 951done: 952#ifdef IPSEC 953 if (ro == &iproute && ro->ro_rt) { 954 RTFREE(ro->ro_rt); 955 ro->ro_rt = NULL; 956 } 957 if (sp != NULL) { 958 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 959 printf("DP ip_output call free SP:%p\n", sp)); 960 key_freesp(sp); 961 } 962#endif /* IPSEC */ 963 return (error); 964bad: 965 m_freem(m0); 966 goto done; 967} 968 969void 970in_delayed_cksum(struct mbuf *m) 971{ 972 struct ip *ip; 973 u_short csum, offset; 974 975 ip = mtod(m, struct ip *); 976 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 977 csum = in_cksum_skip(m, ip->ip_len, offset); 978 offset += m->m_pkthdr.csum_data; /* checksum offset */ 979 980 if (offset + sizeof(u_short) > m->m_len) { 981 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 982 m->m_len, offset, ip->ip_p); 983 /* 984 * XXX 985 * this shouldn't happen, but if it does, the 986 * correct behavior may be to insert the checksum 987 * in the existing chain instead of rearranging it. 988 */ 989 m = m_pullup(m, offset + sizeof(u_short)); 990 } 991 *(u_short *)(m->m_data + offset) = csum; 992} 993 994/* 995 * Insert IP options into preformed packet. 996 * Adjust IP destination as required for IP source routing, 997 * as indicated by a non-zero in_addr at the start of the options. 998 * 999 * XXX This routine assumes that the packet has no options in place. 1000 */ 1001static struct mbuf * 1002ip_insertoptions(m, opt, phlen) 1003 register struct mbuf *m; 1004 struct mbuf *opt; 1005 int *phlen; 1006{ 1007 register struct ipoption *p = mtod(opt, struct ipoption *); 1008 struct mbuf *n; 1009 register struct ip *ip = mtod(m, struct ip *); 1010 unsigned optlen; 1011 1012 optlen = opt->m_len - sizeof(p->ipopt_dst); 1013 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) 1014 return (m); /* XXX should fail */ 1015 if (p->ipopt_dst.s_addr) 1016 ip->ip_dst = p->ipopt_dst; 1017 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1018 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1019 if (n == 0) 1020 return (m); 1021 n->m_pkthdr.rcvif = (struct ifnet *)0; 1022 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1023 m->m_len -= sizeof(struct ip); 1024 m->m_data += sizeof(struct ip); 1025 n->m_next = m; 1026 m = n; 1027 m->m_len = optlen + sizeof(struct ip); 1028 m->m_data += max_linkhdr; 1029 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1030 } else { 1031 m->m_data -= optlen; 1032 m->m_len += optlen; 1033 m->m_pkthdr.len += optlen; 1034 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1035 } 1036 ip = mtod(m, struct ip *); 1037 bcopy(p->ipopt_list, ip + 1, optlen); 1038 *phlen = sizeof(struct ip) + optlen; 1039 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1040 ip->ip_len += optlen; 1041 return (m); 1042} 1043 1044/* 1045 * Copy options from ip to jp, 1046 * omitting those not copied during fragmentation. 1047 */ 1048int 1049ip_optcopy(ip, jp) 1050 struct ip *ip, *jp; 1051{ 1052 register u_char *cp, *dp; 1053 int opt, optlen, cnt; 1054 1055 cp = (u_char *)(ip + 1); 1056 dp = (u_char *)(jp + 1); 1057 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1058 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1059 opt = cp[0]; 1060 if (opt == IPOPT_EOL) 1061 break; 1062 if (opt == IPOPT_NOP) { 1063 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1064 *dp++ = IPOPT_NOP; 1065 optlen = 1; 1066 continue; 1067 } 1068#ifdef DIAGNOSTIC 1069 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1070 panic("malformed IPv4 option passed to ip_optcopy"); 1071#endif 1072 optlen = cp[IPOPT_OLEN]; 1073#ifdef DIAGNOSTIC 1074 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1075 panic("malformed IPv4 option passed to ip_optcopy"); 1076#endif 1077 /* bogus lengths should have been caught by ip_dooptions */ 1078 if (optlen > cnt) 1079 optlen = cnt; 1080 if (IPOPT_COPIED(opt)) { 1081 bcopy(cp, dp, optlen); 1082 dp += optlen; 1083 } 1084 } 1085 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1086 *dp++ = IPOPT_EOL; 1087 return (optlen); 1088} 1089 1090/* 1091 * IP socket option processing. 1092 */ 1093int 1094ip_ctloutput(so, sopt) 1095 struct socket *so; 1096 struct sockopt *sopt; 1097{ 1098 struct inpcb *inp = sotoinpcb(so); 1099 int error, optval; 1100 1101 error = optval = 0; 1102 if (sopt->sopt_level != IPPROTO_IP) { 1103 return (EINVAL); 1104 } 1105 1106 switch (sopt->sopt_dir) { 1107 case SOPT_SET: 1108 switch (sopt->sopt_name) { 1109 case IP_OPTIONS: 1110#ifdef notyet 1111 case IP_RETOPTS: 1112#endif 1113 { 1114 struct mbuf *m; 1115 if (sopt->sopt_valsize > MLEN) { 1116 error = EMSGSIZE; 1117 break; 1118 } 1119 MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER); 1120 if (m == 0) { 1121 error = ENOBUFS; 1122 break; 1123 } 1124 m->m_len = sopt->sopt_valsize; 1125 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1126 m->m_len); 1127 1128 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1129 m)); 1130 } 1131 1132 case IP_TOS: 1133 case IP_TTL: 1134 case IP_RECVOPTS: 1135 case IP_RECVRETOPTS: 1136 case IP_RECVDSTADDR: 1137 case IP_RECVIF: 1138#if defined(NFAITH) && NFAITH > 0 1139 case IP_FAITH: 1140#endif 1141 error = sooptcopyin(sopt, &optval, sizeof optval, 1142 sizeof optval); 1143 if (error) 1144 break; 1145 1146 switch (sopt->sopt_name) { 1147 case IP_TOS: 1148 inp->inp_ip_tos = optval; 1149 break; 1150 1151 case IP_TTL: 1152 inp->inp_ip_ttl = optval; 1153 break; 1154#define OPTSET(bit) \ 1155 if (optval) \ 1156 inp->inp_flags |= bit; \ 1157 else \ 1158 inp->inp_flags &= ~bit; 1159 1160 case IP_RECVOPTS: 1161 OPTSET(INP_RECVOPTS); 1162 break; 1163 1164 case IP_RECVRETOPTS: 1165 OPTSET(INP_RECVRETOPTS); 1166 break; 1167 1168 case IP_RECVDSTADDR: 1169 OPTSET(INP_RECVDSTADDR); 1170 break; 1171 1172 case IP_RECVIF: 1173 OPTSET(INP_RECVIF); 1174 break; 1175 1176#if defined(NFAITH) && NFAITH > 0 1177 case IP_FAITH: 1178 OPTSET(INP_FAITH); 1179 break; 1180#endif 1181 } 1182 break; 1183#undef OPTSET 1184 1185 case IP_MULTICAST_IF: 1186 case IP_MULTICAST_VIF: 1187 case IP_MULTICAST_TTL: 1188 case IP_MULTICAST_LOOP: 1189 case IP_ADD_MEMBERSHIP: 1190 case IP_DROP_MEMBERSHIP: 1191 error = ip_setmoptions(sopt, &inp->inp_moptions); 1192 break; 1193 1194 case IP_PORTRANGE: 1195 error = sooptcopyin(sopt, &optval, sizeof optval, 1196 sizeof optval); 1197 if (error) 1198 break; 1199 1200 switch (optval) { 1201 case IP_PORTRANGE_DEFAULT: 1202 inp->inp_flags &= ~(INP_LOWPORT); 1203 inp->inp_flags &= ~(INP_HIGHPORT); 1204 break; 1205 1206 case IP_PORTRANGE_HIGH: 1207 inp->inp_flags &= ~(INP_LOWPORT); 1208 inp->inp_flags |= INP_HIGHPORT; 1209 break; 1210 1211 case IP_PORTRANGE_LOW: 1212 inp->inp_flags &= ~(INP_HIGHPORT); 1213 inp->inp_flags |= INP_LOWPORT; 1214 break; 1215 1216 default: 1217 error = EINVAL; 1218 break; 1219 } 1220 break; 1221 1222#ifdef IPSEC 1223 case IP_IPSEC_POLICY: 1224 { 1225 caddr_t req; 1226 size_t len = 0; 1227 int priv; 1228 struct mbuf *m; 1229 int optname; 1230 1231 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1232 break; 1233 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1234 break; 1235 priv = (sopt->sopt_p != NULL && 1236 suser(sopt->sopt_p) != 0) ? 0 : 1; 1237 req = mtod(m, caddr_t); 1238 len = m->m_len; 1239 optname = sopt->sopt_name; 1240 error = ipsec4_set_policy(inp, optname, req, len, priv); 1241 m_freem(m); 1242 break; 1243 } 1244#endif /*IPSEC*/ 1245 1246 default: 1247 error = ENOPROTOOPT; 1248 break; 1249 } 1250 break; 1251 1252 case SOPT_GET: 1253 switch (sopt->sopt_name) { 1254 case IP_OPTIONS: 1255 case IP_RETOPTS: 1256 if (inp->inp_options) 1257 error = sooptcopyout(sopt, 1258 mtod(inp->inp_options, 1259 char *), 1260 inp->inp_options->m_len); 1261 else 1262 sopt->sopt_valsize = 0; 1263 break; 1264 1265 case IP_TOS: 1266 case IP_TTL: 1267 case IP_RECVOPTS: 1268 case IP_RECVRETOPTS: 1269 case IP_RECVDSTADDR: 1270 case IP_RECVIF: 1271 case IP_PORTRANGE: 1272#if defined(NFAITH) && NFAITH > 0 1273 case IP_FAITH: 1274#endif 1275 switch (sopt->sopt_name) { 1276 1277 case IP_TOS: 1278 optval = inp->inp_ip_tos; 1279 break; 1280 1281 case IP_TTL: 1282 optval = inp->inp_ip_ttl; 1283 break; 1284 1285#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1286 1287 case IP_RECVOPTS: 1288 optval = OPTBIT(INP_RECVOPTS); 1289 break; 1290 1291 case IP_RECVRETOPTS: 1292 optval = OPTBIT(INP_RECVRETOPTS); 1293 break; 1294 1295 case IP_RECVDSTADDR: 1296 optval = OPTBIT(INP_RECVDSTADDR); 1297 break; 1298 1299 case IP_RECVIF: 1300 optval = OPTBIT(INP_RECVIF); 1301 break; 1302 1303 case IP_PORTRANGE: 1304 if (inp->inp_flags & INP_HIGHPORT) 1305 optval = IP_PORTRANGE_HIGH; 1306 else if (inp->inp_flags & INP_LOWPORT) 1307 optval = IP_PORTRANGE_LOW; 1308 else 1309 optval = 0; 1310 break; 1311 1312#if defined(NFAITH) && NFAITH > 0 1313 case IP_FAITH: 1314 optval = OPTBIT(INP_FAITH); 1315 break; 1316#endif 1317 } 1318 error = sooptcopyout(sopt, &optval, sizeof optval); 1319 break; 1320 1321 case IP_MULTICAST_IF: 1322 case IP_MULTICAST_VIF: 1323 case IP_MULTICAST_TTL: 1324 case IP_MULTICAST_LOOP: 1325 case IP_ADD_MEMBERSHIP: 1326 case IP_DROP_MEMBERSHIP: 1327 error = ip_getmoptions(sopt, inp->inp_moptions); 1328 break; 1329 1330#ifdef IPSEC 1331 case IP_IPSEC_POLICY: 1332 { 1333 struct mbuf *m = NULL; 1334 caddr_t req = NULL; 1335 size_t len = 0; 1336 1337 if (m != 0) { 1338 req = mtod(m, caddr_t); 1339 len = m->m_len; 1340 } 1341 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1342 if (error == 0) 1343 error = soopt_mcopyout(sopt, m); /* XXX */ 1344 if (error == 0) 1345 m_freem(m); 1346 break; 1347 } 1348#endif /*IPSEC*/ 1349 1350 default: 1351 error = ENOPROTOOPT; 1352 break; 1353 } 1354 break; 1355 } 1356 return (error); 1357} 1358 1359/* 1360 * Set up IP options in pcb for insertion in output packets. 1361 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1362 * with destination address if source routed. 1363 */ 1364static int 1365ip_pcbopts(optname, pcbopt, m) 1366 int optname; 1367 struct mbuf **pcbopt; 1368 register struct mbuf *m; 1369{ 1370 register int cnt, optlen; 1371 register u_char *cp; 1372 u_char opt; 1373 1374 /* turn off any old options */ 1375 if (*pcbopt) 1376 (void)m_free(*pcbopt); 1377 *pcbopt = 0; 1378 if (m == (struct mbuf *)0 || m->m_len == 0) { 1379 /* 1380 * Only turning off any previous options. 1381 */ 1382 if (m) 1383 (void)m_free(m); 1384 return (0); 1385 } 1386 1387#ifndef vax 1388 if (m->m_len % sizeof(int32_t)) 1389 goto bad; 1390#endif 1391 /* 1392 * IP first-hop destination address will be stored before 1393 * actual options; move other options back 1394 * and clear it when none present. 1395 */ 1396 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1397 goto bad; 1398 cnt = m->m_len; 1399 m->m_len += sizeof(struct in_addr); 1400 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1401 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1402 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1403 1404 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1405 opt = cp[IPOPT_OPTVAL]; 1406 if (opt == IPOPT_EOL) 1407 break; 1408 if (opt == IPOPT_NOP) 1409 optlen = 1; 1410 else { 1411 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1412 goto bad; 1413 optlen = cp[IPOPT_OLEN]; 1414 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1415 goto bad; 1416 } 1417 switch (opt) { 1418 1419 default: 1420 break; 1421 1422 case IPOPT_LSRR: 1423 case IPOPT_SSRR: 1424 /* 1425 * user process specifies route as: 1426 * ->A->B->C->D 1427 * D must be our final destination (but we can't 1428 * check that since we may not have connected yet). 1429 * A is first hop destination, which doesn't appear in 1430 * actual IP option, but is stored before the options. 1431 */ 1432 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1433 goto bad; 1434 m->m_len -= sizeof(struct in_addr); 1435 cnt -= sizeof(struct in_addr); 1436 optlen -= sizeof(struct in_addr); 1437 cp[IPOPT_OLEN] = optlen; 1438 /* 1439 * Move first hop before start of options. 1440 */ 1441 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1442 sizeof(struct in_addr)); 1443 /* 1444 * Then copy rest of options back 1445 * to close up the deleted entry. 1446 */ 1447 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1448 sizeof(struct in_addr)), 1449 (caddr_t)&cp[IPOPT_OFFSET+1], 1450 (unsigned)cnt + sizeof(struct in_addr)); 1451 break; 1452 } 1453 } 1454 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1455 goto bad; 1456 *pcbopt = m; 1457 return (0); 1458 1459bad: 1460 (void)m_free(m); 1461 return (EINVAL); 1462} 1463 1464/* 1465 * XXX 1466 * The whole multicast option thing needs to be re-thought. 1467 * Several of these options are equally applicable to non-multicast 1468 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1469 * standard option (IP_TTL). 1470 */ 1471/* 1472 * Set the IP multicast options in response to user setsockopt(). 1473 */ 1474static int 1475ip_setmoptions(sopt, imop) 1476 struct sockopt *sopt; 1477 struct ip_moptions **imop; 1478{ 1479 int error = 0; 1480 int i; 1481 struct in_addr addr; 1482 struct ip_mreq mreq; 1483 struct ifnet *ifp; 1484 struct ip_moptions *imo = *imop; 1485 struct route ro; 1486 struct sockaddr_in *dst; 1487 int s; 1488 1489 if (imo == NULL) { 1490 /* 1491 * No multicast option buffer attached to the pcb; 1492 * allocate one and initialize to default values. 1493 */ 1494 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1495 M_WAITOK); 1496 1497 if (imo == NULL) 1498 return (ENOBUFS); 1499 *imop = imo; 1500 imo->imo_multicast_ifp = NULL; 1501 imo->imo_multicast_vif = -1; 1502 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1503 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1504 imo->imo_num_memberships = 0; 1505 } 1506 1507 switch (sopt->sopt_name) { 1508 /* store an index number for the vif you wanna use in the send */ 1509 case IP_MULTICAST_VIF: 1510 if (legal_vif_num == 0) { 1511 error = EOPNOTSUPP; 1512 break; 1513 } 1514 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1515 if (error) 1516 break; 1517 if (!legal_vif_num(i) && (i != -1)) { 1518 error = EINVAL; 1519 break; 1520 } 1521 imo->imo_multicast_vif = i; 1522 break; 1523 1524 case IP_MULTICAST_IF: 1525 /* 1526 * Select the interface for outgoing multicast packets. 1527 */ 1528 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1529 if (error) 1530 break; 1531 /* 1532 * INADDR_ANY is used to remove a previous selection. 1533 * When no interface is selected, a default one is 1534 * chosen every time a multicast packet is sent. 1535 */ 1536 if (addr.s_addr == INADDR_ANY) { 1537 imo->imo_multicast_ifp = NULL; 1538 break; 1539 } 1540 /* 1541 * The selected interface is identified by its local 1542 * IP address. Find the interface and confirm that 1543 * it supports multicasting. 1544 */ 1545 s = splimp(); 1546 INADDR_TO_IFP(addr, ifp); 1547 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1548 splx(s); 1549 error = EADDRNOTAVAIL; 1550 break; 1551 } 1552 imo->imo_multicast_ifp = ifp; 1553 splx(s); 1554 break; 1555 1556 case IP_MULTICAST_TTL: 1557 /* 1558 * Set the IP time-to-live for outgoing multicast packets. 1559 * The original multicast API required a char argument, 1560 * which is inconsistent with the rest of the socket API. 1561 * We allow either a char or an int. 1562 */ 1563 if (sopt->sopt_valsize == 1) { 1564 u_char ttl; 1565 error = sooptcopyin(sopt, &ttl, 1, 1); 1566 if (error) 1567 break; 1568 imo->imo_multicast_ttl = ttl; 1569 } else { 1570 u_int ttl; 1571 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1572 sizeof ttl); 1573 if (error) 1574 break; 1575 if (ttl > 255) 1576 error = EINVAL; 1577 else 1578 imo->imo_multicast_ttl = ttl; 1579 } 1580 break; 1581 1582 case IP_MULTICAST_LOOP: 1583 /* 1584 * Set the loopback flag for outgoing multicast packets. 1585 * Must be zero or one. The original multicast API required a 1586 * char argument, which is inconsistent with the rest 1587 * of the socket API. We allow either a char or an int. 1588 */ 1589 if (sopt->sopt_valsize == 1) { 1590 u_char loop; 1591 error = sooptcopyin(sopt, &loop, 1, 1); 1592 if (error) 1593 break; 1594 imo->imo_multicast_loop = !!loop; 1595 } else { 1596 u_int loop; 1597 error = sooptcopyin(sopt, &loop, sizeof loop, 1598 sizeof loop); 1599 if (error) 1600 break; 1601 imo->imo_multicast_loop = !!loop; 1602 } 1603 break; 1604 1605 case IP_ADD_MEMBERSHIP: 1606 /* 1607 * Add a multicast group membership. 1608 * Group must be a valid IP multicast address. 1609 */ 1610 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1611 if (error) 1612 break; 1613 1614 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1615 error = EINVAL; 1616 break; 1617 } 1618 s = splimp(); 1619 /* 1620 * If no interface address was provided, use the interface of 1621 * the route to the given multicast address. 1622 */ 1623 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1624 bzero((caddr_t)&ro, sizeof(ro)); 1625 dst = (struct sockaddr_in *)&ro.ro_dst; 1626 dst->sin_len = sizeof(*dst); 1627 dst->sin_family = AF_INET; 1628 dst->sin_addr = mreq.imr_multiaddr; 1629 rtalloc(&ro); 1630 if (ro.ro_rt == NULL) { 1631 error = EADDRNOTAVAIL; 1632 splx(s); 1633 break; 1634 } 1635 ifp = ro.ro_rt->rt_ifp; 1636 rtfree(ro.ro_rt); 1637 } 1638 else { 1639 INADDR_TO_IFP(mreq.imr_interface, ifp); 1640 } 1641 1642 /* 1643 * See if we found an interface, and confirm that it 1644 * supports multicast. 1645 */ 1646 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1647 error = EADDRNOTAVAIL; 1648 splx(s); 1649 break; 1650 } 1651 /* 1652 * See if the membership already exists or if all the 1653 * membership slots are full. 1654 */ 1655 for (i = 0; i < imo->imo_num_memberships; ++i) { 1656 if (imo->imo_membership[i]->inm_ifp == ifp && 1657 imo->imo_membership[i]->inm_addr.s_addr 1658 == mreq.imr_multiaddr.s_addr) 1659 break; 1660 } 1661 if (i < imo->imo_num_memberships) { 1662 error = EADDRINUSE; 1663 splx(s); 1664 break; 1665 } 1666 if (i == IP_MAX_MEMBERSHIPS) { 1667 error = ETOOMANYREFS; 1668 splx(s); 1669 break; 1670 } 1671 /* 1672 * Everything looks good; add a new record to the multicast 1673 * address list for the given interface. 1674 */ 1675 if ((imo->imo_membership[i] = 1676 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1677 error = ENOBUFS; 1678 splx(s); 1679 break; 1680 } 1681 ++imo->imo_num_memberships; 1682 splx(s); 1683 break; 1684 1685 case IP_DROP_MEMBERSHIP: 1686 /* 1687 * Drop a multicast group membership. 1688 * Group must be a valid IP multicast address. 1689 */ 1690 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1691 if (error) 1692 break; 1693 1694 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1695 error = EINVAL; 1696 break; 1697 } 1698 1699 s = splimp(); 1700 /* 1701 * If an interface address was specified, get a pointer 1702 * to its ifnet structure. 1703 */ 1704 if (mreq.imr_interface.s_addr == INADDR_ANY) 1705 ifp = NULL; 1706 else { 1707 INADDR_TO_IFP(mreq.imr_interface, ifp); 1708 if (ifp == NULL) { 1709 error = EADDRNOTAVAIL; 1710 splx(s); 1711 break; 1712 } 1713 } 1714 /* 1715 * Find the membership in the membership array. 1716 */ 1717 for (i = 0; i < imo->imo_num_memberships; ++i) { 1718 if ((ifp == NULL || 1719 imo->imo_membership[i]->inm_ifp == ifp) && 1720 imo->imo_membership[i]->inm_addr.s_addr == 1721 mreq.imr_multiaddr.s_addr) 1722 break; 1723 } 1724 if (i == imo->imo_num_memberships) { 1725 error = EADDRNOTAVAIL; 1726 splx(s); 1727 break; 1728 } 1729 /* 1730 * Give up the multicast address record to which the 1731 * membership points. 1732 */ 1733 in_delmulti(imo->imo_membership[i]); 1734 /* 1735 * Remove the gap in the membership array. 1736 */ 1737 for (++i; i < imo->imo_num_memberships; ++i) 1738 imo->imo_membership[i-1] = imo->imo_membership[i]; 1739 --imo->imo_num_memberships; 1740 splx(s); 1741 break; 1742 1743 default: 1744 error = EOPNOTSUPP; 1745 break; 1746 } 1747 1748 /* 1749 * If all options have default values, no need to keep the mbuf. 1750 */ 1751 if (imo->imo_multicast_ifp == NULL && 1752 imo->imo_multicast_vif == -1 && 1753 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1754 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1755 imo->imo_num_memberships == 0) { 1756 free(*imop, M_IPMOPTS); 1757 *imop = NULL; 1758 } 1759 1760 return (error); 1761} 1762 1763/* 1764 * Return the IP multicast options in response to user getsockopt(). 1765 */ 1766static int 1767ip_getmoptions(sopt, imo) 1768 struct sockopt *sopt; 1769 register struct ip_moptions *imo; 1770{ 1771 struct in_addr addr; 1772 struct in_ifaddr *ia; 1773 int error, optval; 1774 u_char coptval; 1775 1776 error = 0; 1777 switch (sopt->sopt_name) { 1778 case IP_MULTICAST_VIF: 1779 if (imo != NULL) 1780 optval = imo->imo_multicast_vif; 1781 else 1782 optval = -1; 1783 error = sooptcopyout(sopt, &optval, sizeof optval); 1784 break; 1785 1786 case IP_MULTICAST_IF: 1787 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1788 addr.s_addr = INADDR_ANY; 1789 else { 1790 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1791 addr.s_addr = (ia == NULL) ? INADDR_ANY 1792 : IA_SIN(ia)->sin_addr.s_addr; 1793 } 1794 error = sooptcopyout(sopt, &addr, sizeof addr); 1795 break; 1796 1797 case IP_MULTICAST_TTL: 1798 if (imo == 0) 1799 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1800 else 1801 optval = coptval = imo->imo_multicast_ttl; 1802 if (sopt->sopt_valsize == 1) 1803 error = sooptcopyout(sopt, &coptval, 1); 1804 else 1805 error = sooptcopyout(sopt, &optval, sizeof optval); 1806 break; 1807 1808 case IP_MULTICAST_LOOP: 1809 if (imo == 0) 1810 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1811 else 1812 optval = coptval = imo->imo_multicast_loop; 1813 if (sopt->sopt_valsize == 1) 1814 error = sooptcopyout(sopt, &coptval, 1); 1815 else 1816 error = sooptcopyout(sopt, &optval, sizeof optval); 1817 break; 1818 1819 default: 1820 error = ENOPROTOOPT; 1821 break; 1822 } 1823 return (error); 1824} 1825 1826/* 1827 * Discard the IP multicast options. 1828 */ 1829void 1830ip_freemoptions(imo) 1831 register struct ip_moptions *imo; 1832{ 1833 register int i; 1834 1835 if (imo != NULL) { 1836 for (i = 0; i < imo->imo_num_memberships; ++i) 1837 in_delmulti(imo->imo_membership[i]); 1838 free(imo, M_IPMOPTS); 1839 } 1840} 1841 1842/* 1843 * Routine called from ip_output() to loop back a copy of an IP multicast 1844 * packet to the input queue of a specified interface. Note that this 1845 * calls the output routine of the loopback "driver", but with an interface 1846 * pointer that might NOT be a loopback interface -- evil, but easier than 1847 * replicating that code here. 1848 */ 1849static void 1850ip_mloopback(ifp, m, dst, hlen) 1851 struct ifnet *ifp; 1852 register struct mbuf *m; 1853 register struct sockaddr_in *dst; 1854 int hlen; 1855{ 1856 register struct ip *ip; 1857 struct mbuf *copym; 1858 1859 copym = m_copy(m, 0, M_COPYALL); 1860 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1861 copym = m_pullup(copym, hlen); 1862 if (copym != NULL) { 1863 /* 1864 * We don't bother to fragment if the IP length is greater 1865 * than the interface's MTU. Can this possibly matter? 1866 */ 1867 ip = mtod(copym, struct ip *); 1868 HTONS(ip->ip_len); 1869 HTONS(ip->ip_off); 1870 ip->ip_sum = 0; 1871 if (ip->ip_vhl == IP_VHL_BORING) { 1872 ip->ip_sum = in_cksum_hdr(ip); 1873 } else { 1874 ip->ip_sum = in_cksum(copym, hlen); 1875 } 1876 /* 1877 * NB: 1878 * It's not clear whether there are any lingering 1879 * reentrancy problems in other areas which might 1880 * be exposed by using ip_input directly (in 1881 * particular, everything which modifies the packet 1882 * in-place). Yet another option is using the 1883 * protosw directly to deliver the looped back 1884 * packet. For the moment, we'll err on the side 1885 * of safety by using if_simloop(). 1886 */ 1887#if 1 /* XXX */ 1888 if (dst->sin_family != AF_INET) { 1889 printf("ip_mloopback: bad address family %d\n", 1890 dst->sin_family); 1891 dst->sin_family = AF_INET; 1892 } 1893#endif 1894 1895#ifdef notdef 1896 copym->m_pkthdr.rcvif = ifp; 1897 ip_input(copym); 1898#else 1899 /* if the checksum hasn't been computed, mark it as valid */ 1900 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1901 copym->m_pkthdr.csum_flags |= 1902 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1903 copym->m_pkthdr.csum_data = 0xffff; 1904 } 1905 if_simloop(ifp, copym, dst->sin_family, 0); 1906#endif 1907 } 1908} 1909