ip_output.c revision 58895
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 58895 2000-04-01 18:51:03Z 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 void in_delayed_cksum(struct mbuf *m); 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 *)); 110extern int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)); 111 112 113extern struct protosw inetsw[]; 114 115/* 116 * IP output. The packet in mbuf chain m contains a skeletal IP 117 * header (with len, off, ttl, proto, tos, src, dst). 118 * The mbuf chain containing the packet will be freed. 119 * The mbuf opt, if present, will not be freed. 120 */ 121int 122ip_output(m0, opt, ro, flags, imo) 123 struct mbuf *m0; 124 struct mbuf *opt; 125 struct route *ro; 126 int flags; 127 struct ip_moptions *imo; 128{ 129 struct ip *ip, *mhip; 130 struct ifnet *ifp; 131 struct mbuf *m = m0; 132 int hlen = sizeof (struct ip); 133 int len, off, error = 0; 134 struct sockaddr_in *dst; 135 struct in_ifaddr *ia; 136 int isbroadcast, sw_csum; 137#ifdef IPSEC 138 struct route iproute; 139 struct socket *so = NULL; 140 struct secpolicy *sp = NULL; 141#endif 142 u_int16_t divert_cookie; /* firewall cookie */ 143#ifdef IPFIREWALL_FORWARD 144 int fwd_rewrite_src = 0; 145#endif 146 struct ip_fw_chain *rule = NULL; 147 148#ifdef IPDIVERT 149 /* Get and reset firewall cookie */ 150 divert_cookie = ip_divert_cookie; 151 ip_divert_cookie = 0; 152#else 153 divert_cookie = 0; 154#endif 155 156 /* 157 * NOTE: If IP_SOCKINMRCVIF flag is set, 'socket *' is kept in 158 * m->m_pkthdr.rcvif for later IPSEC check. In this case, 159 * m->m_pkthdr will be NULL cleared after the contents is saved in 160 * 'so'. 161 * NULL clearance of rcvif should be natural because the packet should 162 * have been sent from my own socket and has no rcvif in this case. 163 * It is also necessary because someone might consider it as 164 * 'ifnet *', and cause SEGV. 165 */ 166#if defined(IPFIREWALL) && defined(DUMMYNET) 167 /* 168 * dummynet packet are prepended a vestigial mbuf with 169 * m_type = MT_DUMMYNET and m_data pointing to the matching 170 * rule. 171 */ 172 if (m->m_type == MT_DUMMYNET) { 173 /* 174 * the packet was already tagged, so part of the 175 * processing was already done, and we need to go down. 176 * Get parameters from the header. 177 */ 178 rule = (struct ip_fw_chain *)(m->m_data) ; 179 opt = NULL ; 180 ro = & ( ((struct dn_pkt *)m)->ro ) ; 181 imo = NULL ; 182 dst = ((struct dn_pkt *)m)->dn_dst ; 183 ifp = ((struct dn_pkt *)m)->ifp ; 184 flags = ((struct dn_pkt *)m)->flags ; 185 186 m0 = m = m->m_next ; 187#ifdef IPSEC 188 if ((flags & IP_SOCKINMRCVIF) != 0) { 189 so = (struct socket *)m->m_pkthdr.rcvif; 190 m->m_pkthdr.rcvif = NULL; 191 } 192#endif 193 ip = mtod(m, struct ip *); 194 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 195 goto sendit; 196 } else 197 rule = NULL ; 198#endif 199#ifdef IPSEC 200 if ((flags & IP_SOCKINMRCVIF) != 0) { 201 so = (struct socket *)m->m_pkthdr.rcvif; 202 m->m_pkthdr.rcvif = NULL; 203 } 204#endif 205 206#ifdef DIAGNOSTIC 207 if ((m->m_flags & M_PKTHDR) == 0) 208 panic("ip_output no HDR"); 209 if (!ro) 210 panic("ip_output no route, proto = %d", 211 mtod(m, struct ip *)->ip_p); 212#endif 213 if (opt) { 214 m = ip_insertoptions(m, opt, &len); 215 hlen = len; 216 } 217 ip = mtod(m, struct ip *); 218 /* 219 * Fill in IP header. 220 */ 221 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 222 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 223 ip->ip_off &= IP_DF; 224 ip->ip_id = htons(ip_id++); 225 ipstat.ips_localout++; 226 } else { 227 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 228 } 229 230 dst = (struct sockaddr_in *)&ro->ro_dst; 231 /* 232 * If there is a cached route, 233 * check that it is to the same destination 234 * and is still up. If not, free it and try again. 235 */ 236 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 237 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 238 RTFREE(ro->ro_rt); 239 ro->ro_rt = (struct rtentry *)0; 240 } 241 if (ro->ro_rt == 0) { 242 dst->sin_family = AF_INET; 243 dst->sin_len = sizeof(*dst); 244 dst->sin_addr = ip->ip_dst; 245 } 246 /* 247 * If routing to interface only, 248 * short circuit routing lookup. 249 */ 250#define ifatoia(ifa) ((struct in_ifaddr *)(ifa)) 251#define sintosa(sin) ((struct sockaddr *)(sin)) 252 if (flags & IP_ROUTETOIF) { 253 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 254 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 255 ipstat.ips_noroute++; 256 error = ENETUNREACH; 257 goto bad; 258 } 259 ifp = ia->ia_ifp; 260 ip->ip_ttl = 1; 261 isbroadcast = in_broadcast(dst->sin_addr, ifp); 262 } else { 263 /* 264 * If this is the case, we probably don't want to allocate 265 * a protocol-cloned route since we didn't get one from the 266 * ULP. This lets TCP do its thing, while not burdening 267 * forwarding or ICMP with the overhead of cloning a route. 268 * Of course, we still want to do any cloning requested by 269 * the link layer, as this is probably required in all cases 270 * for correct operation (as it is for ARP). 271 */ 272 if (ro->ro_rt == 0) 273 rtalloc_ign(ro, RTF_PRCLONING); 274 if (ro->ro_rt == 0) { 275 ipstat.ips_noroute++; 276 error = EHOSTUNREACH; 277 goto bad; 278 } 279 ia = ifatoia(ro->ro_rt->rt_ifa); 280 ifp = ro->ro_rt->rt_ifp; 281 ro->ro_rt->rt_use++; 282 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 283 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 284 if (ro->ro_rt->rt_flags & RTF_HOST) 285 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 286 else 287 isbroadcast = in_broadcast(dst->sin_addr, ifp); 288 } 289 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 290 struct in_multi *inm; 291 292 m->m_flags |= M_MCAST; 293 /* 294 * IP destination address is multicast. Make sure "dst" 295 * still points to the address in "ro". (It may have been 296 * changed to point to a gateway address, above.) 297 */ 298 dst = (struct sockaddr_in *)&ro->ro_dst; 299 /* 300 * See if the caller provided any multicast options 301 */ 302 if (imo != NULL) { 303 ip->ip_ttl = imo->imo_multicast_ttl; 304 if (imo->imo_multicast_ifp != NULL) 305 ifp = imo->imo_multicast_ifp; 306 if (imo->imo_multicast_vif != -1) 307 ip->ip_src.s_addr = 308 ip_mcast_src(imo->imo_multicast_vif); 309 } else 310 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 311 /* 312 * Confirm that the outgoing interface supports multicast. 313 */ 314 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 315 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 316 ipstat.ips_noroute++; 317 error = ENETUNREACH; 318 goto bad; 319 } 320 } 321 /* 322 * If source address not specified yet, use address 323 * of outgoing interface. 324 */ 325 if (ip->ip_src.s_addr == INADDR_ANY) { 326 register struct in_ifaddr *ia1; 327 328 for (ia1 = in_ifaddrhead.tqh_first; ia1; 329 ia1 = ia1->ia_link.tqe_next) 330 if (ia1->ia_ifp == ifp) { 331 ip->ip_src = IA_SIN(ia1)->sin_addr; 332 break; 333 } 334 } 335 336 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 337 if (inm != NULL && 338 (imo == NULL || imo->imo_multicast_loop)) { 339 /* 340 * If we belong to the destination multicast group 341 * on the outgoing interface, and the caller did not 342 * forbid loopback, loop back a copy. 343 */ 344 ip_mloopback(ifp, m, dst, hlen); 345 } 346 else { 347 /* 348 * If we are acting as a multicast router, perform 349 * multicast forwarding as if the packet had just 350 * arrived on the interface to which we are about 351 * to send. The multicast forwarding function 352 * recursively calls this function, using the 353 * IP_FORWARDING flag to prevent infinite recursion. 354 * 355 * Multicasts that are looped back by ip_mloopback(), 356 * above, will be forwarded by the ip_input() routine, 357 * if necessary. 358 */ 359 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 360 /* 361 * Check if rsvp daemon is running. If not, don't 362 * set ip_moptions. This ensures that the packet 363 * is multicast and not just sent down one link 364 * as prescribed by rsvpd. 365 */ 366 if (!rsvp_on) 367 imo = NULL; 368 if (ip_mforward(ip, ifp, m, imo) != 0) { 369 m_freem(m); 370 goto done; 371 } 372 } 373 } 374 375 /* 376 * Multicasts with a time-to-live of zero may be looped- 377 * back, above, but must not be transmitted on a network. 378 * Also, multicasts addressed to the loopback interface 379 * are not sent -- the above call to ip_mloopback() will 380 * loop back a copy if this host actually belongs to the 381 * destination group on the loopback interface. 382 */ 383 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 384 m_freem(m); 385 goto done; 386 } 387 388 goto sendit; 389 } 390#ifndef notdef 391 /* 392 * If source address not specified yet, use address 393 * of outgoing interface. 394 */ 395 if (ip->ip_src.s_addr == INADDR_ANY) { 396 ip->ip_src = IA_SIN(ia)->sin_addr; 397#ifdef IPFIREWALL_FORWARD 398 /* Keep note that we did this - if the firewall changes 399 * the next-hop, our interface may change, changing the 400 * default source IP. It's a shame so much effort happens 401 * twice. Oh well. 402 */ 403 fwd_rewrite_src++; 404#endif /* IPFIREWALL_FORWARD */ 405 } 406#endif /* notdef */ 407 /* 408 * Verify that we have any chance at all of being able to queue 409 * the packet or packet fragments 410 */ 411 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 412 ifp->if_snd.ifq_maxlen) { 413 error = ENOBUFS; 414 goto bad; 415 } 416 417 /* 418 * Look for broadcast address and 419 * and verify user is allowed to send 420 * such a packet. 421 */ 422 if (isbroadcast) { 423 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 424 error = EADDRNOTAVAIL; 425 goto bad; 426 } 427 if ((flags & IP_ALLOWBROADCAST) == 0) { 428 error = EACCES; 429 goto bad; 430 } 431 /* don't allow broadcast messages to be fragmented */ 432 if ((u_short)ip->ip_len > ifp->if_mtu) { 433 error = EMSGSIZE; 434 goto bad; 435 } 436 m->m_flags |= M_BCAST; 437 } else { 438 m->m_flags &= ~M_BCAST; 439 } 440 441sendit: 442 /* 443 * IpHack's section. 444 * - Xlate: translate packet's addr/port (NAT). 445 * - Firewall: deny/allow/etc. 446 * - Wrap: fake packet's addr/port <unimpl.> 447 * - Encapsulate: put it in another IP and send out. <unimp.> 448 */ 449 if (fr_checkp) { 450 struct mbuf *m1 = m; 451 452 if ((error = (*fr_checkp)(ip, hlen, ifp, 1, &m1)) || !m1) 453 goto done; 454 ip = mtod(m = m1, struct ip *); 455 } 456 457 /* 458 * Check with the firewall... 459 */ 460 if (fw_enable && ip_fw_chk_ptr) { 461 struct sockaddr_in *old = dst; 462 463 off = (*ip_fw_chk_ptr)(&ip, 464 hlen, ifp, &divert_cookie, &m, &rule, &dst); 465 /* 466 * On return we must do the following: 467 * m == NULL -> drop the pkt 468 * 1<=off<= 0xffff -> DIVERT 469 * (off & 0x10000) -> send to a DUMMYNET pipe 470 * (off & 0x20000) -> TEE the packet 471 * dst != old -> IPFIREWALL_FORWARD 472 * off==0, dst==old -> accept 473 * If some of the above modules is not compiled in, then 474 * we should't have to check the corresponding condition 475 * (because the ipfw control socket should not accept 476 * unsupported rules), but better play safe and drop 477 * packets in case of doubt. 478 */ 479 if (!m) { /* firewall said to reject */ 480 error = EACCES; 481 goto done; 482 } 483 if (off == 0 && dst == old) /* common case */ 484 goto pass ; 485#ifdef DUMMYNET 486 if ((off & IP_FW_PORT_DYNT_FLAG) != 0) { 487 /* 488 * pass the pkt to dummynet. Need to include 489 * pipe number, m, ifp, ro, dst because these are 490 * not recomputed in the next pass. 491 * All other parameters have been already used and 492 * so they are not needed anymore. 493 * XXX note: if the ifp or ro entry are deleted 494 * while a pkt is in dummynet, we are in trouble! 495 */ 496 dummynet_io(off & 0xffff, DN_TO_IP_OUT, m,ifp,ro,dst,rule, 497 flags); 498 goto done; 499 } 500#endif 501#ifdef IPDIVERT 502 if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 503 struct mbuf *clone = NULL; 504 505 /* Clone packet if we're doing a 'tee' */ 506 if ((off & IP_FW_PORT_TEE_FLAG) != 0) 507 clone = m_dup(m, M_DONTWAIT); 508 509 /* 510 * XXX 511 * delayed checksums are not currently compatible 512 * with divert sockets. 513 */ 514 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 515 in_delayed_cksum(m); 516 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 517 } 518 519 /* Restore packet header fields to original values */ 520 HTONS(ip->ip_len); 521 HTONS(ip->ip_off); 522 523 /* Deliver packet to divert input routine */ 524 ip_divert_cookie = divert_cookie; 525 divert_packet(m, 0, off & 0xffff); 526 527 /* If 'tee', continue with original packet */ 528 if (clone != NULL) { 529 m = clone; 530 ip = mtod(m, struct ip *); 531 goto pass; 532 } 533 goto done; 534 } 535#endif 536 537#ifdef IPFIREWALL_FORWARD 538 /* Here we check dst to make sure it's directly reachable on the 539 * interface we previously thought it was. 540 * If it isn't (which may be likely in some situations) we have 541 * to re-route it (ie, find a route for the next-hop and the 542 * associated interface) and set them here. This is nested 543 * forwarding which in most cases is undesirable, except where 544 * such control is nigh impossible. So we do it here. 545 * And I'm babbling. 546 */ 547 if (off == 0 && old != dst) { 548 struct in_ifaddr *ia; 549 550 /* It's changed... */ 551 /* There must be a better way to do this next line... */ 552 static struct route sro_fwd, *ro_fwd = &sro_fwd; 553#ifdef IPFIREWALL_FORWARD_DEBUG 554 printf("IPFIREWALL_FORWARD: New dst ip: "); 555 print_ip(dst->sin_addr); 556 printf("\n"); 557#endif 558 /* 559 * We need to figure out if we have been forwarded 560 * to a local socket. If so then we should somehow 561 * "loop back" to ip_input, and get directed to the 562 * PCB as if we had received this packet. This is 563 * because it may be dificult to identify the packets 564 * you want to forward until they are being output 565 * and have selected an interface. (e.g. locally 566 * initiated packets) If we used the loopback inteface, 567 * we would not be able to control what happens 568 * as the packet runs through ip_input() as 569 * it is done through a ISR. 570 */ 571 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 572 ia = TAILQ_NEXT(ia, ia_link)) { 573 /* 574 * If the addr to forward to is one 575 * of ours, we pretend to 576 * be the destination for this packet. 577 */ 578 if (IA_SIN(ia)->sin_addr.s_addr == 579 dst->sin_addr.s_addr) 580 break; 581 } 582 if (ia) { 583 /* tell ip_input "dont filter" */ 584 ip_fw_fwd_addr = dst; 585 if (m->m_pkthdr.rcvif == NULL) 586 m->m_pkthdr.rcvif = ifunit("lo0"); 587 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 588 m->m_pkthdr.csum_flags |= 589 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 590 m0->m_pkthdr.csum_data = 0xffff; 591 } 592 m->m_pkthdr.csum_flags |= 593 CSUM_IP_CHECKED | CSUM_IP_VALID; 594 ip->ip_len = htons((u_short)ip->ip_len); 595 ip->ip_off = htons((u_short)ip->ip_off); 596 ip_input(m); 597 goto done; 598 } 599 /* Some of the logic for this was 600 * nicked from above. 601 * 602 * This rewrites the cached route in a local PCB. 603 * Is this what we want to do? 604 */ 605 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 606 607 ro_fwd->ro_rt = 0; 608 rtalloc_ign(ro_fwd, RTF_PRCLONING); 609 610 if (ro_fwd->ro_rt == 0) { 611 ipstat.ips_noroute++; 612 error = EHOSTUNREACH; 613 goto bad; 614 } 615 616 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 617 ifp = ro_fwd->ro_rt->rt_ifp; 618 ro_fwd->ro_rt->rt_use++; 619 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 620 dst = (struct sockaddr_in *)ro_fwd->ro_rt->rt_gateway; 621 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 622 isbroadcast = 623 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 624 else 625 isbroadcast = in_broadcast(dst->sin_addr, ifp); 626 RTFREE(ro->ro_rt); 627 ro->ro_rt = ro_fwd->ro_rt; 628 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 629 630 /* 631 * If we added a default src ip earlier, 632 * which would have been gotten from the-then 633 * interface, do it again, from the new one. 634 */ 635 if (fwd_rewrite_src) 636 ip->ip_src = IA_SIN(ia)->sin_addr; 637 goto pass ; 638 } 639#endif /* IPFIREWALL_FORWARD */ 640 /* 641 * if we get here, none of the above matches, and 642 * we have to drop the pkt 643 */ 644 m_freem(m); 645 error = EACCES; /* not sure this is the right error msg */ 646 goto done; 647 } 648 649pass: 650#ifdef IPSEC 651 /* get SP for this packet */ 652 if (so == NULL) 653 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 654 else 655 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 656 657 if (sp == NULL) { 658 ipsecstat.out_inval++; 659 goto bad; 660 } 661 662 error = 0; 663 664 /* check policy */ 665 switch (sp->policy) { 666 case IPSEC_POLICY_DISCARD: 667 /* 668 * This packet is just discarded. 669 */ 670 ipsecstat.out_polvio++; 671 goto bad; 672 673 case IPSEC_POLICY_BYPASS: 674 case IPSEC_POLICY_NONE: 675 /* no need to do IPsec. */ 676 goto skip_ipsec; 677 678 case IPSEC_POLICY_IPSEC: 679 if (sp->req == NULL) { 680 /* XXX should be panic ? */ 681 printf("ip_output: No IPsec request specified.\n"); 682 error = EINVAL; 683 goto bad; 684 } 685 break; 686 687 case IPSEC_POLICY_ENTRUST: 688 default: 689 printf("ip_output: Invalid policy found. %d\n", sp->policy); 690 } 691 692 ip->ip_len = htons((u_short)ip->ip_len); 693 ip->ip_off = htons((u_short)ip->ip_off); 694 ip->ip_sum = 0; 695 696 { 697 struct ipsec_output_state state; 698 bzero(&state, sizeof(state)); 699 state.m = m; 700 if (flags & IP_ROUTETOIF) { 701 state.ro = &iproute; 702 bzero(&iproute, sizeof(iproute)); 703 } else 704 state.ro = ro; 705 state.dst = (struct sockaddr *)dst; 706 707 /* 708 * XXX 709 * delayed checksums are not currently compatible with IPsec 710 */ 711 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 712 in_delayed_cksum(m); 713 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 714 } 715 716 error = ipsec4_output(&state, sp, flags); 717 718 m = state.m; 719 if (flags & IP_ROUTETOIF) { 720 /* 721 * if we have tunnel mode SA, we may need to ignore 722 * IP_ROUTETOIF. 723 */ 724 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 725 flags &= ~IP_ROUTETOIF; 726 ro = state.ro; 727 } 728 } else 729 ro = state.ro; 730 dst = (struct sockaddr_in *)state.dst; 731 if (error) { 732 /* mbuf is already reclaimed in ipsec4_output. */ 733 m0 = NULL; 734 switch (error) { 735 case EHOSTUNREACH: 736 case ENETUNREACH: 737 case EMSGSIZE: 738 case ENOBUFS: 739 case ENOMEM: 740 break; 741 default: 742 printf("ip4_output (ipsec): error code %d\n", error); 743 /*fall through*/ 744 case ENOENT: 745 /* don't show these error codes to the user */ 746 error = 0; 747 break; 748 } 749 goto bad; 750 } 751 } 752 753 /* be sure to update variables that are affected by ipsec4_output() */ 754 ip = mtod(m, struct ip *); 755#ifdef _IP_VHL 756 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 757#else 758 hlen = ip->ip_hl << 2; 759#endif 760 if (ro->ro_rt == NULL) { 761 if ((flags & IP_ROUTETOIF) == 0) { 762 printf("ip_output: " 763 "can't update route after IPsec processing\n"); 764 error = EHOSTUNREACH; /*XXX*/ 765 goto bad; 766 } 767 } else { 768 /* nobody uses ia beyond here */ 769 ifp = ro->ro_rt->rt_ifp; 770 } 771 772 /* make it flipped, again. */ 773 ip->ip_len = ntohs((u_short)ip->ip_len); 774 ip->ip_off = ntohs((u_short)ip->ip_off); 775skip_ipsec: 776#endif /*IPSEC*/ 777 778 sw_csum = m->m_pkthdr.csum_flags | CSUM_IP; 779 m->m_pkthdr.csum_flags = sw_csum & ifp->if_hwassist; 780 sw_csum &= ~ifp->if_hwassist; 781 if (sw_csum & CSUM_DELAY_DATA) { 782 in_delayed_cksum(m); 783 sw_csum &= ~CSUM_DELAY_DATA; 784 } 785 786 /* 787 * If small enough for interface, or the interface will take 788 * care of the fragmentation for us, can just send directly. 789 */ 790 if ((u_short)ip->ip_len <= ifp->if_mtu || 791 ifp->if_hwassist & CSUM_FRAGMENT) { 792 ip->ip_len = htons((u_short)ip->ip_len); 793 ip->ip_off = htons((u_short)ip->ip_off); 794 ip->ip_sum = 0; 795 if (sw_csum & CSUM_DELAY_IP) { 796 if (ip->ip_vhl == IP_VHL_BORING) { 797 ip->ip_sum = in_cksum_hdr(ip); 798 } else { 799 ip->ip_sum = in_cksum(m, hlen); 800 } 801 } 802 error = (*ifp->if_output)(ifp, m, 803 (struct sockaddr *)dst, ro->ro_rt); 804 goto done; 805 } 806 /* 807 * Too large for interface; fragment if possible. 808 * Must be able to put at least 8 bytes per fragment. 809 */ 810 if (ip->ip_off & IP_DF) { 811 error = EMSGSIZE; 812 /* 813 * This case can happen if the user changed the MTU 814 * of an interface after enabling IP on it. Because 815 * most netifs don't keep track of routes pointing to 816 * them, there is no way for one to update all its 817 * routes when the MTU is changed. 818 */ 819 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 820 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 821 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 822 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 823 } 824 ipstat.ips_cantfrag++; 825 goto bad; 826 } 827 len = (ifp->if_mtu - hlen) &~ 7; 828 if (len < 8) { 829 error = EMSGSIZE; 830 goto bad; 831 } 832 833 /* 834 * if the interface will not calculate checksums on 835 * fragmented packets, then do it here. 836 */ 837 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 838 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 839 in_delayed_cksum(m); 840 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 841 } 842 843 { 844 int mhlen, firstlen = len; 845 struct mbuf **mnext = &m->m_nextpkt; 846 int nfrags = 1; 847 848 /* 849 * Loop through length of segment after first fragment, 850 * make new header and copy data of each part and link onto chain. 851 */ 852 m0 = m; 853 mhlen = sizeof (struct ip); 854 for (off = hlen + len; off < (u_short)ip->ip_len; off += len) { 855 MGETHDR(m, M_DONTWAIT, MT_HEADER); 856 if (m == 0) { 857 error = ENOBUFS; 858 ipstat.ips_odropped++; 859 goto sendorfree; 860 } 861 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 862 m->m_data += max_linkhdr; 863 mhip = mtod(m, struct ip *); 864 *mhip = *ip; 865 if (hlen > sizeof (struct ip)) { 866 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 867 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 868 } 869 m->m_len = mhlen; 870 mhip->ip_off = ((off - hlen) >> 3) + (ip->ip_off & ~IP_MF); 871 if (ip->ip_off & IP_MF) 872 mhip->ip_off |= IP_MF; 873 if (off + len >= (u_short)ip->ip_len) 874 len = (u_short)ip->ip_len - off; 875 else 876 mhip->ip_off |= IP_MF; 877 mhip->ip_len = htons((u_short)(len + mhlen)); 878 m->m_next = m_copy(m0, off, len); 879 if (m->m_next == 0) { 880 (void) m_free(m); 881 error = ENOBUFS; /* ??? */ 882 ipstat.ips_odropped++; 883 goto sendorfree; 884 } 885 m->m_pkthdr.len = mhlen + len; 886 m->m_pkthdr.rcvif = (struct ifnet *)0; 887 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 888 mhip->ip_off = htons((u_short)mhip->ip_off); 889 mhip->ip_sum = 0; 890 if (sw_csum & CSUM_DELAY_IP) { 891 if (mhip->ip_vhl == IP_VHL_BORING) { 892 mhip->ip_sum = in_cksum_hdr(mhip); 893 } else { 894 mhip->ip_sum = in_cksum(m, mhlen); 895 } 896 } 897 *mnext = m; 898 mnext = &m->m_nextpkt; 899 nfrags++; 900 } 901 ipstat.ips_ofragments += nfrags; 902 903 /* set first/last markers for fragment chain */ 904 m->m_flags |= M_LASTFRAG; 905 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 906 m0->m_pkthdr.csum_data = nfrags; 907 908 /* 909 * Update first fragment by trimming what's been copied out 910 * and updating header, then send each fragment (in order). 911 */ 912 m = m0; 913 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 914 m->m_pkthdr.len = hlen + firstlen; 915 ip->ip_len = htons((u_short)m->m_pkthdr.len); 916 ip->ip_off = htons((u_short)(ip->ip_off | IP_MF)); 917 ip->ip_sum = 0; 918 if (sw_csum & CSUM_DELAY_IP) { 919 if (ip->ip_vhl == IP_VHL_BORING) { 920 ip->ip_sum = in_cksum_hdr(ip); 921 } else { 922 ip->ip_sum = in_cksum(m, hlen); 923 } 924 } 925sendorfree: 926 for (m = m0; m; m = m0) { 927 m0 = m->m_nextpkt; 928 m->m_nextpkt = 0; 929 if (error == 0) 930 error = (*ifp->if_output)(ifp, m, 931 (struct sockaddr *)dst, ro->ro_rt); 932 else 933 m_freem(m); 934 } 935 936 if (error == 0) 937 ipstat.ips_fragmented++; 938 } 939done: 940#ifdef IPSEC 941 if (ro == &iproute && ro->ro_rt) { 942 RTFREE(ro->ro_rt); 943 ro->ro_rt = NULL; 944 } 945 if (sp != NULL) { 946 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 947 printf("DP ip_output call free SP:%p\n", sp)); 948 key_freesp(sp); 949 } 950#endif /* IPSEC */ 951 return (error); 952bad: 953 m_freem(m0); 954 goto done; 955} 956 957static void 958in_delayed_cksum(struct mbuf *m) 959{ 960 struct ip *ip; 961 u_short csum, offset; 962 963 ip = mtod(m, struct ip *); 964 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 965 csum = in_cksum_skip(m, ip->ip_len, offset); 966 offset += m->m_pkthdr.csum_data; /* checksum offset */ 967 968 if (offset + sizeof(u_short) > m->m_len) { 969 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 970 m->m_len, offset, ip->ip_p); 971 /* 972 * XXX 973 * this shouldn't happen, but if it does, the 974 * correct behavior may be to insert the checksum 975 * in the existing chain instead of rearranging it. 976 */ 977 m = m_pullup(m, offset + sizeof(u_short)); 978 } 979 *(u_short *)(m->m_data + offset) = csum; 980} 981 982/* 983 * Insert IP options into preformed packet. 984 * Adjust IP destination as required for IP source routing, 985 * as indicated by a non-zero in_addr at the start of the options. 986 * 987 * XXX This routine assumes that the packet has no options in place. 988 */ 989static struct mbuf * 990ip_insertoptions(m, opt, phlen) 991 register struct mbuf *m; 992 struct mbuf *opt; 993 int *phlen; 994{ 995 register struct ipoption *p = mtod(opt, struct ipoption *); 996 struct mbuf *n; 997 register struct ip *ip = mtod(m, struct ip *); 998 unsigned optlen; 999 1000 optlen = opt->m_len - sizeof(p->ipopt_dst); 1001 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) 1002 return (m); /* XXX should fail */ 1003 if (p->ipopt_dst.s_addr) 1004 ip->ip_dst = p->ipopt_dst; 1005 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1006 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1007 if (n == 0) 1008 return (m); 1009 n->m_pkthdr.rcvif = (struct ifnet *)0; 1010 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1011 m->m_len -= sizeof(struct ip); 1012 m->m_data += sizeof(struct ip); 1013 n->m_next = m; 1014 m = n; 1015 m->m_len = optlen + sizeof(struct ip); 1016 m->m_data += max_linkhdr; 1017 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1018 } else { 1019 m->m_data -= optlen; 1020 m->m_len += optlen; 1021 m->m_pkthdr.len += optlen; 1022 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1023 } 1024 ip = mtod(m, struct ip *); 1025 bcopy(p->ipopt_list, ip + 1, optlen); 1026 *phlen = sizeof(struct ip) + optlen; 1027 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1028 ip->ip_len += optlen; 1029 return (m); 1030} 1031 1032/* 1033 * Copy options from ip to jp, 1034 * omitting those not copied during fragmentation. 1035 */ 1036int 1037ip_optcopy(ip, jp) 1038 struct ip *ip, *jp; 1039{ 1040 register u_char *cp, *dp; 1041 int opt, optlen, cnt; 1042 1043 cp = (u_char *)(ip + 1); 1044 dp = (u_char *)(jp + 1); 1045 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1046 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1047 opt = cp[0]; 1048 if (opt == IPOPT_EOL) 1049 break; 1050 if (opt == IPOPT_NOP) { 1051 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1052 *dp++ = IPOPT_NOP; 1053 optlen = 1; 1054 continue; 1055 } else 1056 optlen = cp[IPOPT_OLEN]; 1057 /* bogus lengths should have been caught by ip_dooptions */ 1058 if (optlen > cnt) 1059 optlen = cnt; 1060 if (IPOPT_COPIED(opt)) { 1061 bcopy(cp, dp, optlen); 1062 dp += optlen; 1063 } 1064 } 1065 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1066 *dp++ = IPOPT_EOL; 1067 return (optlen); 1068} 1069 1070/* 1071 * IP socket option processing. 1072 */ 1073int 1074ip_ctloutput(so, sopt) 1075 struct socket *so; 1076 struct sockopt *sopt; 1077{ 1078 struct inpcb *inp = sotoinpcb(so); 1079 int error, optval; 1080 1081 error = optval = 0; 1082 if (sopt->sopt_level != IPPROTO_IP) { 1083 return (EINVAL); 1084 } 1085 1086 switch (sopt->sopt_dir) { 1087 case SOPT_SET: 1088 switch (sopt->sopt_name) { 1089 case IP_OPTIONS: 1090#ifdef notyet 1091 case IP_RETOPTS: 1092#endif 1093 { 1094 struct mbuf *m; 1095 if (sopt->sopt_valsize > MLEN) { 1096 error = EMSGSIZE; 1097 break; 1098 } 1099 MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER); 1100 if (m == 0) { 1101 error = ENOBUFS; 1102 break; 1103 } 1104 m->m_len = sopt->sopt_valsize; 1105 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1106 m->m_len); 1107 1108 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1109 m)); 1110 } 1111 1112 case IP_TOS: 1113 case IP_TTL: 1114 case IP_RECVOPTS: 1115 case IP_RECVRETOPTS: 1116 case IP_RECVDSTADDR: 1117 case IP_RECVIF: 1118#if defined(NFAITH) && NFAITH > 0 1119 case IP_FAITH: 1120#endif 1121 error = sooptcopyin(sopt, &optval, sizeof optval, 1122 sizeof optval); 1123 if (error) 1124 break; 1125 1126 switch (sopt->sopt_name) { 1127 case IP_TOS: 1128 inp->inp_ip_tos = optval; 1129 break; 1130 1131 case IP_TTL: 1132 inp->inp_ip_ttl = optval; 1133 break; 1134#define OPTSET(bit) \ 1135 if (optval) \ 1136 inp->inp_flags |= bit; \ 1137 else \ 1138 inp->inp_flags &= ~bit; 1139 1140 case IP_RECVOPTS: 1141 OPTSET(INP_RECVOPTS); 1142 break; 1143 1144 case IP_RECVRETOPTS: 1145 OPTSET(INP_RECVRETOPTS); 1146 break; 1147 1148 case IP_RECVDSTADDR: 1149 OPTSET(INP_RECVDSTADDR); 1150 break; 1151 1152 case IP_RECVIF: 1153 OPTSET(INP_RECVIF); 1154 break; 1155 1156#if defined(NFAITH) && NFAITH > 0 1157 case IP_FAITH: 1158 OPTSET(INP_FAITH); 1159 break; 1160#endif 1161 } 1162 break; 1163#undef OPTSET 1164 1165 case IP_MULTICAST_IF: 1166 case IP_MULTICAST_VIF: 1167 case IP_MULTICAST_TTL: 1168 case IP_MULTICAST_LOOP: 1169 case IP_ADD_MEMBERSHIP: 1170 case IP_DROP_MEMBERSHIP: 1171 error = ip_setmoptions(sopt, &inp->inp_moptions); 1172 break; 1173 1174 case IP_PORTRANGE: 1175 error = sooptcopyin(sopt, &optval, sizeof optval, 1176 sizeof optval); 1177 if (error) 1178 break; 1179 1180 switch (optval) { 1181 case IP_PORTRANGE_DEFAULT: 1182 inp->inp_flags &= ~(INP_LOWPORT); 1183 inp->inp_flags &= ~(INP_HIGHPORT); 1184 break; 1185 1186 case IP_PORTRANGE_HIGH: 1187 inp->inp_flags &= ~(INP_LOWPORT); 1188 inp->inp_flags |= INP_HIGHPORT; 1189 break; 1190 1191 case IP_PORTRANGE_LOW: 1192 inp->inp_flags &= ~(INP_HIGHPORT); 1193 inp->inp_flags |= INP_LOWPORT; 1194 break; 1195 1196 default: 1197 error = EINVAL; 1198 break; 1199 } 1200 break; 1201 1202#ifdef IPSEC 1203 case IP_IPSEC_POLICY: 1204 { 1205 caddr_t req; 1206 int priv; 1207 struct mbuf *m; 1208 int optname; 1209 1210 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1211 break; 1212 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1213 break; 1214 priv = (sopt->sopt_p != NULL && 1215 suser(sopt->sopt_p) != 0) ? 0 : 1; 1216 req = mtod(m, caddr_t); 1217 optname = sopt->sopt_name; 1218 error = ipsec4_set_policy(inp, optname, req, priv); 1219 m_freem(m); 1220 break; 1221 } 1222#endif /*IPSEC*/ 1223 1224 default: 1225 error = ENOPROTOOPT; 1226 break; 1227 } 1228 break; 1229 1230 case SOPT_GET: 1231 switch (sopt->sopt_name) { 1232 case IP_OPTIONS: 1233 case IP_RETOPTS: 1234 if (inp->inp_options) 1235 error = sooptcopyout(sopt, 1236 mtod(inp->inp_options, 1237 char *), 1238 inp->inp_options->m_len); 1239 else 1240 sopt->sopt_valsize = 0; 1241 break; 1242 1243 case IP_TOS: 1244 case IP_TTL: 1245 case IP_RECVOPTS: 1246 case IP_RECVRETOPTS: 1247 case IP_RECVDSTADDR: 1248 case IP_RECVIF: 1249 case IP_PORTRANGE: 1250#if defined(NFAITH) && NFAITH > 0 1251 case IP_FAITH: 1252#endif 1253 switch (sopt->sopt_name) { 1254 1255 case IP_TOS: 1256 optval = inp->inp_ip_tos; 1257 break; 1258 1259 case IP_TTL: 1260 optval = inp->inp_ip_ttl; 1261 break; 1262 1263#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1264 1265 case IP_RECVOPTS: 1266 optval = OPTBIT(INP_RECVOPTS); 1267 break; 1268 1269 case IP_RECVRETOPTS: 1270 optval = OPTBIT(INP_RECVRETOPTS); 1271 break; 1272 1273 case IP_RECVDSTADDR: 1274 optval = OPTBIT(INP_RECVDSTADDR); 1275 break; 1276 1277 case IP_RECVIF: 1278 optval = OPTBIT(INP_RECVIF); 1279 break; 1280 1281 case IP_PORTRANGE: 1282 if (inp->inp_flags & INP_HIGHPORT) 1283 optval = IP_PORTRANGE_HIGH; 1284 else if (inp->inp_flags & INP_LOWPORT) 1285 optval = IP_PORTRANGE_LOW; 1286 else 1287 optval = 0; 1288 break; 1289 1290#if defined(NFAITH) && NFAITH > 0 1291 case IP_FAITH: 1292 optval = OPTBIT(INP_FAITH); 1293 break; 1294#endif 1295 } 1296 error = sooptcopyout(sopt, &optval, sizeof optval); 1297 break; 1298 1299 case IP_MULTICAST_IF: 1300 case IP_MULTICAST_VIF: 1301 case IP_MULTICAST_TTL: 1302 case IP_MULTICAST_LOOP: 1303 case IP_ADD_MEMBERSHIP: 1304 case IP_DROP_MEMBERSHIP: 1305 error = ip_getmoptions(sopt, inp->inp_moptions); 1306 break; 1307 1308#ifdef IPSEC 1309 case IP_IPSEC_POLICY: 1310 { 1311 struct mbuf *m = NULL; 1312 caddr_t req = NULL; 1313 1314 if (m != 0) 1315 req = mtod(m, caddr_t); 1316 error = ipsec4_get_policy(sotoinpcb(so), req, &m); 1317 if (error == 0) 1318 error = soopt_mcopyout(sopt, m); /* XXX */ 1319 if (error == 0) 1320 m_freem(m); 1321 break; 1322 } 1323#endif /*IPSEC*/ 1324 1325 default: 1326 error = ENOPROTOOPT; 1327 break; 1328 } 1329 break; 1330 } 1331 return (error); 1332} 1333 1334/* 1335 * Set up IP options in pcb for insertion in output packets. 1336 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1337 * with destination address if source routed. 1338 */ 1339static int 1340ip_pcbopts(optname, pcbopt, m) 1341 int optname; 1342 struct mbuf **pcbopt; 1343 register struct mbuf *m; 1344{ 1345 register int cnt, optlen; 1346 register u_char *cp; 1347 u_char opt; 1348 1349 /* turn off any old options */ 1350 if (*pcbopt) 1351 (void)m_free(*pcbopt); 1352 *pcbopt = 0; 1353 if (m == (struct mbuf *)0 || m->m_len == 0) { 1354 /* 1355 * Only turning off any previous options. 1356 */ 1357 if (m) 1358 (void)m_free(m); 1359 return (0); 1360 } 1361 1362#ifndef vax 1363 if (m->m_len % sizeof(int32_t)) 1364 goto bad; 1365#endif 1366 /* 1367 * IP first-hop destination address will be stored before 1368 * actual options; move other options back 1369 * and clear it when none present. 1370 */ 1371 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1372 goto bad; 1373 cnt = m->m_len; 1374 m->m_len += sizeof(struct in_addr); 1375 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1376 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1377 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1378 1379 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1380 opt = cp[IPOPT_OPTVAL]; 1381 if (opt == IPOPT_EOL) 1382 break; 1383 if (opt == IPOPT_NOP) 1384 optlen = 1; 1385 else { 1386 optlen = cp[IPOPT_OLEN]; 1387 if (optlen <= IPOPT_OLEN || optlen > cnt) 1388 goto bad; 1389 } 1390 switch (opt) { 1391 1392 default: 1393 break; 1394 1395 case IPOPT_LSRR: 1396 case IPOPT_SSRR: 1397 /* 1398 * user process specifies route as: 1399 * ->A->B->C->D 1400 * D must be our final destination (but we can't 1401 * check that since we may not have connected yet). 1402 * A is first hop destination, which doesn't appear in 1403 * actual IP option, but is stored before the options. 1404 */ 1405 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1406 goto bad; 1407 m->m_len -= sizeof(struct in_addr); 1408 cnt -= sizeof(struct in_addr); 1409 optlen -= sizeof(struct in_addr); 1410 cp[IPOPT_OLEN] = optlen; 1411 /* 1412 * Move first hop before start of options. 1413 */ 1414 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1415 sizeof(struct in_addr)); 1416 /* 1417 * Then copy rest of options back 1418 * to close up the deleted entry. 1419 */ 1420 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1421 sizeof(struct in_addr)), 1422 (caddr_t)&cp[IPOPT_OFFSET+1], 1423 (unsigned)cnt + sizeof(struct in_addr)); 1424 break; 1425 } 1426 } 1427 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1428 goto bad; 1429 *pcbopt = m; 1430 return (0); 1431 1432bad: 1433 (void)m_free(m); 1434 return (EINVAL); 1435} 1436 1437/* 1438 * XXX 1439 * The whole multicast option thing needs to be re-thought. 1440 * Several of these options are equally applicable to non-multicast 1441 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1442 * standard option (IP_TTL). 1443 */ 1444/* 1445 * Set the IP multicast options in response to user setsockopt(). 1446 */ 1447static int 1448ip_setmoptions(sopt, imop) 1449 struct sockopt *sopt; 1450 struct ip_moptions **imop; 1451{ 1452 int error = 0; 1453 int i; 1454 struct in_addr addr; 1455 struct ip_mreq mreq; 1456 struct ifnet *ifp; 1457 struct ip_moptions *imo = *imop; 1458 struct route ro; 1459 struct sockaddr_in *dst; 1460 int s; 1461 1462 if (imo == NULL) { 1463 /* 1464 * No multicast option buffer attached to the pcb; 1465 * allocate one and initialize to default values. 1466 */ 1467 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1468 M_WAITOK); 1469 1470 if (imo == NULL) 1471 return (ENOBUFS); 1472 *imop = imo; 1473 imo->imo_multicast_ifp = NULL; 1474 imo->imo_multicast_vif = -1; 1475 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1476 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1477 imo->imo_num_memberships = 0; 1478 } 1479 1480 switch (sopt->sopt_name) { 1481 /* store an index number for the vif you wanna use in the send */ 1482 case IP_MULTICAST_VIF: 1483 if (legal_vif_num == 0) { 1484 error = EOPNOTSUPP; 1485 break; 1486 } 1487 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1488 if (error) 1489 break; 1490 if (!legal_vif_num(i) && (i != -1)) { 1491 error = EINVAL; 1492 break; 1493 } 1494 imo->imo_multicast_vif = i; 1495 break; 1496 1497 case IP_MULTICAST_IF: 1498 /* 1499 * Select the interface for outgoing multicast packets. 1500 */ 1501 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1502 if (error) 1503 break; 1504 /* 1505 * INADDR_ANY is used to remove a previous selection. 1506 * When no interface is selected, a default one is 1507 * chosen every time a multicast packet is sent. 1508 */ 1509 if (addr.s_addr == INADDR_ANY) { 1510 imo->imo_multicast_ifp = NULL; 1511 break; 1512 } 1513 /* 1514 * The selected interface is identified by its local 1515 * IP address. Find the interface and confirm that 1516 * it supports multicasting. 1517 */ 1518 s = splimp(); 1519 INADDR_TO_IFP(addr, ifp); 1520 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1521 splx(s); 1522 error = EADDRNOTAVAIL; 1523 break; 1524 } 1525 imo->imo_multicast_ifp = ifp; 1526 splx(s); 1527 break; 1528 1529 case IP_MULTICAST_TTL: 1530 /* 1531 * Set the IP time-to-live for outgoing multicast packets. 1532 * The original multicast API required a char argument, 1533 * which is inconsistent with the rest of the socket API. 1534 * We allow either a char or an int. 1535 */ 1536 if (sopt->sopt_valsize == 1) { 1537 u_char ttl; 1538 error = sooptcopyin(sopt, &ttl, 1, 1); 1539 if (error) 1540 break; 1541 imo->imo_multicast_ttl = ttl; 1542 } else { 1543 u_int ttl; 1544 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1545 sizeof ttl); 1546 if (error) 1547 break; 1548 if (ttl > 255) 1549 error = EINVAL; 1550 else 1551 imo->imo_multicast_ttl = ttl; 1552 } 1553 break; 1554 1555 case IP_MULTICAST_LOOP: 1556 /* 1557 * Set the loopback flag for outgoing multicast packets. 1558 * Must be zero or one. The original multicast API required a 1559 * char argument, which is inconsistent with the rest 1560 * of the socket API. We allow either a char or an int. 1561 */ 1562 if (sopt->sopt_valsize == 1) { 1563 u_char loop; 1564 error = sooptcopyin(sopt, &loop, 1, 1); 1565 if (error) 1566 break; 1567 imo->imo_multicast_loop = !!loop; 1568 } else { 1569 u_int loop; 1570 error = sooptcopyin(sopt, &loop, sizeof loop, 1571 sizeof loop); 1572 if (error) 1573 break; 1574 imo->imo_multicast_loop = !!loop; 1575 } 1576 break; 1577 1578 case IP_ADD_MEMBERSHIP: 1579 /* 1580 * Add a multicast group membership. 1581 * Group must be a valid IP multicast address. 1582 */ 1583 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1584 if (error) 1585 break; 1586 1587 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1588 error = EINVAL; 1589 break; 1590 } 1591 s = splimp(); 1592 /* 1593 * If no interface address was provided, use the interface of 1594 * the route to the given multicast address. 1595 */ 1596 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1597 bzero((caddr_t)&ro, sizeof(ro)); 1598 dst = (struct sockaddr_in *)&ro.ro_dst; 1599 dst->sin_len = sizeof(*dst); 1600 dst->sin_family = AF_INET; 1601 dst->sin_addr = mreq.imr_multiaddr; 1602 rtalloc(&ro); 1603 if (ro.ro_rt == NULL) { 1604 error = EADDRNOTAVAIL; 1605 splx(s); 1606 break; 1607 } 1608 ifp = ro.ro_rt->rt_ifp; 1609 rtfree(ro.ro_rt); 1610 } 1611 else { 1612 INADDR_TO_IFP(mreq.imr_interface, ifp); 1613 } 1614 1615 /* 1616 * See if we found an interface, and confirm that it 1617 * supports multicast. 1618 */ 1619 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1620 error = EADDRNOTAVAIL; 1621 splx(s); 1622 break; 1623 } 1624 /* 1625 * See if the membership already exists or if all the 1626 * membership slots are full. 1627 */ 1628 for (i = 0; i < imo->imo_num_memberships; ++i) { 1629 if (imo->imo_membership[i]->inm_ifp == ifp && 1630 imo->imo_membership[i]->inm_addr.s_addr 1631 == mreq.imr_multiaddr.s_addr) 1632 break; 1633 } 1634 if (i < imo->imo_num_memberships) { 1635 error = EADDRINUSE; 1636 splx(s); 1637 break; 1638 } 1639 if (i == IP_MAX_MEMBERSHIPS) { 1640 error = ETOOMANYREFS; 1641 splx(s); 1642 break; 1643 } 1644 /* 1645 * Everything looks good; add a new record to the multicast 1646 * address list for the given interface. 1647 */ 1648 if ((imo->imo_membership[i] = 1649 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1650 error = ENOBUFS; 1651 splx(s); 1652 break; 1653 } 1654 ++imo->imo_num_memberships; 1655 splx(s); 1656 break; 1657 1658 case IP_DROP_MEMBERSHIP: 1659 /* 1660 * Drop a multicast group membership. 1661 * Group must be a valid IP multicast address. 1662 */ 1663 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1664 if (error) 1665 break; 1666 1667 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1668 error = EINVAL; 1669 break; 1670 } 1671 1672 s = splimp(); 1673 /* 1674 * If an interface address was specified, get a pointer 1675 * to its ifnet structure. 1676 */ 1677 if (mreq.imr_interface.s_addr == INADDR_ANY) 1678 ifp = NULL; 1679 else { 1680 INADDR_TO_IFP(mreq.imr_interface, ifp); 1681 if (ifp == NULL) { 1682 error = EADDRNOTAVAIL; 1683 splx(s); 1684 break; 1685 } 1686 } 1687 /* 1688 * Find the membership in the membership array. 1689 */ 1690 for (i = 0; i < imo->imo_num_memberships; ++i) { 1691 if ((ifp == NULL || 1692 imo->imo_membership[i]->inm_ifp == ifp) && 1693 imo->imo_membership[i]->inm_addr.s_addr == 1694 mreq.imr_multiaddr.s_addr) 1695 break; 1696 } 1697 if (i == imo->imo_num_memberships) { 1698 error = EADDRNOTAVAIL; 1699 splx(s); 1700 break; 1701 } 1702 /* 1703 * Give up the multicast address record to which the 1704 * membership points. 1705 */ 1706 in_delmulti(imo->imo_membership[i]); 1707 /* 1708 * Remove the gap in the membership array. 1709 */ 1710 for (++i; i < imo->imo_num_memberships; ++i) 1711 imo->imo_membership[i-1] = imo->imo_membership[i]; 1712 --imo->imo_num_memberships; 1713 splx(s); 1714 break; 1715 1716 default: 1717 error = EOPNOTSUPP; 1718 break; 1719 } 1720 1721 /* 1722 * If all options have default values, no need to keep the mbuf. 1723 */ 1724 if (imo->imo_multicast_ifp == NULL && 1725 imo->imo_multicast_vif == -1 && 1726 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1727 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1728 imo->imo_num_memberships == 0) { 1729 free(*imop, M_IPMOPTS); 1730 *imop = NULL; 1731 } 1732 1733 return (error); 1734} 1735 1736/* 1737 * Return the IP multicast options in response to user getsockopt(). 1738 */ 1739static int 1740ip_getmoptions(sopt, imo) 1741 struct sockopt *sopt; 1742 register struct ip_moptions *imo; 1743{ 1744 struct in_addr addr; 1745 struct in_ifaddr *ia; 1746 int error, optval; 1747 u_char coptval; 1748 1749 error = 0; 1750 switch (sopt->sopt_name) { 1751 case IP_MULTICAST_VIF: 1752 if (imo != NULL) 1753 optval = imo->imo_multicast_vif; 1754 else 1755 optval = -1; 1756 error = sooptcopyout(sopt, &optval, sizeof optval); 1757 break; 1758 1759 case IP_MULTICAST_IF: 1760 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1761 addr.s_addr = INADDR_ANY; 1762 else { 1763 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1764 addr.s_addr = (ia == NULL) ? INADDR_ANY 1765 : IA_SIN(ia)->sin_addr.s_addr; 1766 } 1767 error = sooptcopyout(sopt, &addr, sizeof addr); 1768 break; 1769 1770 case IP_MULTICAST_TTL: 1771 if (imo == 0) 1772 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1773 else 1774 optval = coptval = imo->imo_multicast_ttl; 1775 if (sopt->sopt_valsize == 1) 1776 error = sooptcopyout(sopt, &coptval, 1); 1777 else 1778 error = sooptcopyout(sopt, &optval, sizeof optval); 1779 break; 1780 1781 case IP_MULTICAST_LOOP: 1782 if (imo == 0) 1783 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1784 else 1785 optval = coptval = imo->imo_multicast_loop; 1786 if (sopt->sopt_valsize == 1) 1787 error = sooptcopyout(sopt, &coptval, 1); 1788 else 1789 error = sooptcopyout(sopt, &optval, sizeof optval); 1790 break; 1791 1792 default: 1793 error = ENOPROTOOPT; 1794 break; 1795 } 1796 return (error); 1797} 1798 1799/* 1800 * Discard the IP multicast options. 1801 */ 1802void 1803ip_freemoptions(imo) 1804 register struct ip_moptions *imo; 1805{ 1806 register int i; 1807 1808 if (imo != NULL) { 1809 for (i = 0; i < imo->imo_num_memberships; ++i) 1810 in_delmulti(imo->imo_membership[i]); 1811 free(imo, M_IPMOPTS); 1812 } 1813} 1814 1815/* 1816 * Routine called from ip_output() to loop back a copy of an IP multicast 1817 * packet to the input queue of a specified interface. Note that this 1818 * calls the output routine of the loopback "driver", but with an interface 1819 * pointer that might NOT be a loopback interface -- evil, but easier than 1820 * replicating that code here. 1821 */ 1822static void 1823ip_mloopback(ifp, m, dst, hlen) 1824 struct ifnet *ifp; 1825 register struct mbuf *m; 1826 register struct sockaddr_in *dst; 1827 int hlen; 1828{ 1829 register struct ip *ip; 1830 struct mbuf *copym; 1831 1832 copym = m_copy(m, 0, M_COPYALL); 1833 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1834 copym = m_pullup(copym, hlen); 1835 if (copym != NULL) { 1836 /* 1837 * We don't bother to fragment if the IP length is greater 1838 * than the interface's MTU. Can this possibly matter? 1839 */ 1840 ip = mtod(copym, struct ip *); 1841 ip->ip_len = htons((u_short)ip->ip_len); 1842 ip->ip_off = htons((u_short)ip->ip_off); 1843 ip->ip_sum = 0; 1844 if (ip->ip_vhl == IP_VHL_BORING) { 1845 ip->ip_sum = in_cksum_hdr(ip); 1846 } else { 1847 ip->ip_sum = in_cksum(copym, hlen); 1848 } 1849 /* 1850 * NB: 1851 * It's not clear whether there are any lingering 1852 * reentrancy problems in other areas which might 1853 * be exposed by using ip_input directly (in 1854 * particular, everything which modifies the packet 1855 * in-place). Yet another option is using the 1856 * protosw directly to deliver the looped back 1857 * packet. For the moment, we'll err on the side 1858 * of safety by using if_simloop(). 1859 */ 1860#if 1 /* XXX */ 1861 if (dst->sin_family != AF_INET) { 1862 printf("ip_mloopback: bad address family %d\n", 1863 dst->sin_family); 1864 dst->sin_family = AF_INET; 1865 } 1866#endif 1867 1868#ifdef notdef 1869 copym->m_pkthdr.rcvif = ifp; 1870 ip_input(copym); 1871#else 1872 if_simloop(ifp, copym, (struct sockaddr *)dst, 0); 1873#endif 1874 } 1875} 1876