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. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 30 * $FreeBSD: src/sys/netinet/ip_output.c,v 1.99.2.37 2003/04/15 06:44:45 silby Exp $ 31 */ 32 33 #define _IP_VHL 34 35 #include "opt_ipdn.h" 36 #include "opt_ipdivert.h" 37 #include "opt_ipsec.h" 38 #include "opt_mbuf_stress_test.h" 39 #include "opt_mpls.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/kernel.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/protosw.h> 47 #include <sys/socket.h> 48 #include <sys/socketvar.h> 49 #include <sys/proc.h> 50 #include <sys/priv.h> 51 #include <sys/sysctl.h> 52 #include <sys/in_cksum.h> 53 #include <sys/lock.h> 54 55 #include <sys/thread2.h> 56 #include <sys/mplock2.h> 57 #include <sys/msgport2.h> 58 59 #include <net/if.h> 60 #include <net/netisr.h> 61 #include <net/pfil.h> 62 #include <net/route.h> 63 64 #include <netinet/in.h> 65 #include <netinet/in_systm.h> 66 #include <netinet/ip.h> 67 #include <netinet/in_pcb.h> 68 #include <netinet/in_var.h> 69 #include <netinet/ip_var.h> 70 71 #include <netproto/mpls/mpls_var.h> 72 73 static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 74 75 #ifdef IPSEC 76 #include <netinet6/ipsec.h> 77 #include <netproto/key/key.h> 78 #ifdef IPSEC_DEBUG 79 #include <netproto/key/key_debug.h> 80 #else 81 #define KEYDEBUG(lev,arg) 82 #endif 83 #endif /*IPSEC*/ 84 85 #ifdef FAST_IPSEC 86 #include <netproto/ipsec/ipsec.h> 87 #include <netproto/ipsec/xform.h> 88 #include <netproto/ipsec/key.h> 89 #endif /*FAST_IPSEC*/ 90 91 #include <net/ipfw/ip_fw.h> 92 #include <net/dummynet/ip_dummynet.h> 93 94 #define print_ip(x, a, y) kprintf("%s %d.%d.%d.%d%s",\ 95 x, (ntohl(a.s_addr)>>24)&0xFF,\ 96 (ntohl(a.s_addr)>>16)&0xFF,\ 97 (ntohl(a.s_addr)>>8)&0xFF,\ 98 (ntohl(a.s_addr))&0xFF, y); 99 100 u_short ip_id; 101 102 #ifdef MBUF_STRESS_TEST 103 int mbuf_frag_size = 0; 104 SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 105 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 106 #endif 107 108 static struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 109 static struct ifnet *ip_multicast_if(struct in_addr *, int *); 110 static void ip_mloopback 111 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 112 static int ip_getmoptions 113 (struct sockopt *, struct ip_moptions *); 114 static int ip_pcbopts(int, struct mbuf **, struct mbuf *); 115 static int ip_setmoptions 116 (struct sockopt *, struct ip_moptions **); 117 118 int ip_optcopy(struct ip *, struct ip *); 119 120 extern int route_assert_owner_access; 121 122 extern struct protosw inetsw[]; 123 124 static int 125 ip_localforward(struct mbuf *m, const struct sockaddr_in *dst, int hlen) 126 { 127 struct in_ifaddr_container *iac; 128 129 /* 130 * We need to figure out if we have been forwarded to a local 131 * socket. If so, then we should somehow "loop back" to 132 * ip_input(), and get directed to the PCB as if we had received 133 * this packet. This is because it may be difficult to identify 134 * the packets you want to forward until they are being output 135 * and have selected an interface (e.g. locally initiated 136 * packets). If we used the loopback inteface, we would not be 137 * able to control what happens as the packet runs through 138 * ip_input() as it is done through a ISR. 139 */ 140 LIST_FOREACH(iac, INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 141 /* 142 * If the addr to forward to is one of ours, we pretend 143 * to be the destination for this packet. 144 */ 145 if (IA_SIN(iac->ia)->sin_addr.s_addr == dst->sin_addr.s_addr) 146 break; 147 } 148 if (iac != NULL) { 149 struct ip *ip; 150 151 if (m->m_pkthdr.rcvif == NULL) 152 m->m_pkthdr.rcvif = ifunit("lo0"); 153 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 154 m->m_pkthdr.csum_flags |= CSUM_DATA_VALID | 155 CSUM_PSEUDO_HDR; 156 m->m_pkthdr.csum_data = 0xffff; 157 } 158 m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID; 159 160 /* 161 * Make sure that the IP header is in one mbuf, 162 * required by ip_input 163 */ 164 if (m->m_len < hlen) { 165 m = m_pullup(m, hlen); 166 if (m == NULL) { 167 /* The packet was freed; we are done */ 168 return 1; 169 } 170 } 171 ip = mtod(m, struct ip *); 172 173 ip->ip_len = htons(ip->ip_len); 174 ip->ip_off = htons(ip->ip_off); 175 ip_input(m); 176 177 return 1; /* The packet gets forwarded locally */ 178 } 179 return 0; 180 } 181 182 /* 183 * IP output. The packet in mbuf chain m contains a skeletal IP 184 * header (with len, off, ttl, proto, tos, src, dst). 185 * The mbuf chain containing the packet will be freed. 186 * The mbuf opt, if present, will not be freed. 187 */ 188 int 189 ip_output(struct mbuf *m0, struct mbuf *opt, struct route *ro, 190 int flags, struct ip_moptions *imo, struct inpcb *inp) 191 { 192 struct ip *ip; 193 struct ifnet *ifp = NULL; /* keep compiler happy */ 194 struct mbuf *m; 195 int hlen = sizeof(struct ip); 196 int len, error = 0; 197 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 198 struct in_ifaddr *ia = NULL; 199 int isbroadcast, sw_csum; 200 struct in_addr pkt_dst; 201 struct route iproute; 202 struct m_tag *mtag; 203 #ifdef IPSEC 204 struct secpolicy *sp = NULL; 205 struct socket *so = inp ? inp->inp_socket : NULL; 206 #endif 207 #ifdef FAST_IPSEC 208 struct secpolicy *sp = NULL; 209 struct tdb_ident *tdbi; 210 #endif /* FAST_IPSEC */ 211 struct sockaddr_in *next_hop = NULL; 212 int src_was_INADDR_ANY = 0; /* as the name says... */ 213 214 m = m0; 215 M_ASSERTPKTHDR(m); 216 217 if (ro == NULL) { 218 ro = &iproute; 219 bzero(ro, sizeof *ro); 220 } else if (ro->ro_rt != NULL && ro->ro_rt->rt_cpuid != mycpuid) { 221 if (flags & IP_DEBUGROUTE) { 222 if (route_assert_owner_access) { 223 panic("ip_output: " 224 "rt rt_cpuid %d accessed on cpu %d\n", 225 ro->ro_rt->rt_cpuid, mycpuid); 226 } else { 227 kprintf("ip_output: " 228 "rt rt_cpuid %d accessed on cpu %d\n", 229 ro->ro_rt->rt_cpuid, mycpuid); 230 print_backtrace(-1); 231 } 232 } 233 234 /* 235 * XXX 236 * If the cached rtentry's owner CPU is not the current CPU, 237 * then don't touch the cached rtentry (remote free is too 238 * expensive in this context); just relocate the route. 239 */ 240 ro = &iproute; 241 bzero(ro, sizeof *ro); 242 } 243 244 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 245 /* Next hop */ 246 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 247 KKASSERT(mtag != NULL); 248 next_hop = m_tag_data(mtag); 249 } 250 251 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 252 struct dn_pkt *dn_pkt; 253 254 /* Extract info from dummynet tag */ 255 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 256 KKASSERT(mtag != NULL); 257 dn_pkt = m_tag_data(mtag); 258 259 /* 260 * The packet was already tagged, so part of the 261 * processing was already done, and we need to go down. 262 * Get the calculated parameters from the tag. 263 */ 264 ifp = dn_pkt->ifp; 265 266 KKASSERT(ro == &iproute); 267 *ro = dn_pkt->ro; /* structure copy */ 268 KKASSERT(ro->ro_rt == NULL || ro->ro_rt->rt_cpuid == mycpuid); 269 270 dst = dn_pkt->dn_dst; 271 if (dst == (struct sockaddr_in *)&(dn_pkt->ro.ro_dst)) { 272 /* If 'dst' points into dummynet tag, adjust it */ 273 dst = (struct sockaddr_in *)&(ro->ro_dst); 274 } 275 276 ip = mtod(m, struct ip *); 277 hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 278 if (ro->ro_rt) 279 ia = ifatoia(ro->ro_rt->rt_ifa); 280 goto sendit; 281 } 282 283 if (opt) { 284 len = 0; 285 m = ip_insertoptions(m, opt, &len); 286 if (len != 0) 287 hlen = len; 288 } 289 ip = mtod(m, struct ip *); 290 291 /* 292 * Fill in IP header. 293 */ 294 if (!(flags & (IP_FORWARDING|IP_RAWOUTPUT))) { 295 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 296 ip->ip_off &= IP_DF; 297 ip->ip_id = ip_newid(); 298 ipstat.ips_localout++; 299 } else { 300 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 301 } 302 303 reroute: 304 pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst; 305 306 dst = (struct sockaddr_in *)&ro->ro_dst; 307 /* 308 * If there is a cached route, 309 * check that it is to the same destination 310 * and is still up. If not, free it and try again. 311 * The address family should also be checked in case of sharing the 312 * cache with IPv6. 313 */ 314 if (ro->ro_rt && 315 (!(ro->ro_rt->rt_flags & RTF_UP) || 316 dst->sin_family != AF_INET || 317 dst->sin_addr.s_addr != pkt_dst.s_addr)) { 318 rtfree(ro->ro_rt); 319 ro->ro_rt = NULL; 320 } 321 if (ro->ro_rt == NULL) { 322 bzero(dst, sizeof *dst); 323 dst->sin_family = AF_INET; 324 dst->sin_len = sizeof *dst; 325 dst->sin_addr = pkt_dst; 326 } 327 /* 328 * If routing to interface only, 329 * short circuit routing lookup. 330 */ 331 if (flags & IP_ROUTETOIF) { 332 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 333 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) { 334 ipstat.ips_noroute++; 335 error = ENETUNREACH; 336 goto bad; 337 } 338 ifp = ia->ia_ifp; 339 ip->ip_ttl = 1; 340 isbroadcast = in_broadcast(dst->sin_addr, ifp); 341 } else if (IN_MULTICAST(ntohl(pkt_dst.s_addr)) && 342 imo != NULL && imo->imo_multicast_ifp != NULL) { 343 /* 344 * Bypass the normal routing lookup for multicast 345 * packets if the interface is specified. 346 */ 347 ifp = imo->imo_multicast_ifp; 348 ia = IFP_TO_IA(ifp); 349 isbroadcast = 0; /* fool gcc */ 350 } else { 351 /* 352 * If this is the case, we probably don't want to allocate 353 * a protocol-cloned route since we didn't get one from the 354 * ULP. This lets TCP do its thing, while not burdening 355 * forwarding or ICMP with the overhead of cloning a route. 356 * Of course, we still want to do any cloning requested by 357 * the link layer, as this is probably required in all cases 358 * for correct operation (as it is for ARP). 359 */ 360 if (ro->ro_rt == NULL) 361 rtalloc_ign(ro, RTF_PRCLONING); 362 if (ro->ro_rt == NULL) { 363 ipstat.ips_noroute++; 364 error = EHOSTUNREACH; 365 goto bad; 366 } 367 ia = ifatoia(ro->ro_rt->rt_ifa); 368 ifp = ro->ro_rt->rt_ifp; 369 ro->ro_rt->rt_use++; 370 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 371 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 372 if (ro->ro_rt->rt_flags & RTF_HOST) 373 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 374 else 375 isbroadcast = in_broadcast(dst->sin_addr, ifp); 376 } 377 if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 378 struct in_multi *inm; 379 380 m->m_flags |= M_MCAST; 381 /* 382 * IP destination address is multicast. Make sure "dst" 383 * still points to the address in "ro". (It may have been 384 * changed to point to a gateway address, above.) 385 */ 386 dst = (struct sockaddr_in *)&ro->ro_dst; 387 /* 388 * See if the caller provided any multicast options 389 */ 390 if (imo != NULL) { 391 ip->ip_ttl = imo->imo_multicast_ttl; 392 if (imo->imo_multicast_vif != -1) { 393 ip->ip_src.s_addr = 394 ip_mcast_src ? 395 ip_mcast_src(imo->imo_multicast_vif) : 396 INADDR_ANY; 397 } 398 } else { 399 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 400 } 401 /* 402 * Confirm that the outgoing interface supports multicast. 403 */ 404 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 405 if (!(ifp->if_flags & IFF_MULTICAST)) { 406 ipstat.ips_noroute++; 407 error = ENETUNREACH; 408 goto bad; 409 } 410 } 411 /* 412 * If source address not specified yet, use address 413 * of outgoing interface. 414 */ 415 if (ip->ip_src.s_addr == INADDR_ANY) { 416 /* Interface may have no addresses. */ 417 if (ia != NULL) 418 ip->ip_src = IA_SIN(ia)->sin_addr; 419 } 420 421 IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 422 if (inm != NULL && 423 (imo == NULL || imo->imo_multicast_loop)) { 424 /* 425 * If we belong to the destination multicast group 426 * on the outgoing interface, and the caller did not 427 * forbid loopback, loop back a copy. 428 */ 429 ip_mloopback(ifp, m, dst, hlen); 430 } else { 431 /* 432 * If we are acting as a multicast router, perform 433 * multicast forwarding as if the packet had just 434 * arrived on the interface to which we are about 435 * to send. The multicast forwarding function 436 * recursively calls this function, using the 437 * IP_FORWARDING flag to prevent infinite recursion. 438 * 439 * Multicasts that are looped back by ip_mloopback(), 440 * above, will be forwarded by the ip_input() routine, 441 * if necessary. 442 */ 443 if (ip_mrouter && !(flags & IP_FORWARDING)) { 444 /* 445 * If rsvp daemon is not running, do not 446 * set ip_moptions. This ensures that the packet 447 * is multicast and not just sent down one link 448 * as prescribed by rsvpd. 449 */ 450 if (!rsvp_on) 451 imo = NULL; 452 if (ip_mforward) { 453 get_mplock(); 454 if (ip_mforward(ip, ifp, m, imo) != 0) { 455 m_freem(m); 456 rel_mplock(); 457 goto done; 458 } 459 rel_mplock(); 460 } 461 } 462 } 463 464 /* 465 * Multicasts with a time-to-live of zero may be looped- 466 * back, above, but must not be transmitted on a network. 467 * Also, multicasts addressed to the loopback interface 468 * are not sent -- the above call to ip_mloopback() will 469 * loop back a copy if this host actually belongs to the 470 * destination group on the loopback interface. 471 */ 472 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 473 m_freem(m); 474 goto done; 475 } 476 477 goto sendit; 478 } else { 479 m->m_flags &= ~M_MCAST; 480 } 481 482 /* 483 * If the source address is not specified yet, use the address 484 * of the outoing interface. In case, keep note we did that, so 485 * if the the firewall changes the next-hop causing the output 486 * interface to change, we can fix that. 487 */ 488 if (ip->ip_src.s_addr == INADDR_ANY || src_was_INADDR_ANY) { 489 /* Interface may have no addresses. */ 490 if (ia != NULL) { 491 ip->ip_src = IA_SIN(ia)->sin_addr; 492 src_was_INADDR_ANY = 1; 493 } 494 } 495 496 /* 497 * Look for broadcast address and 498 * verify user is allowed to send 499 * such a packet. 500 */ 501 if (isbroadcast) { 502 if (!(ifp->if_flags & IFF_BROADCAST)) { 503 error = EADDRNOTAVAIL; 504 goto bad; 505 } 506 if (!(flags & IP_ALLOWBROADCAST)) { 507 error = EACCES; 508 goto bad; 509 } 510 /* don't allow broadcast messages to be fragmented */ 511 if (ip->ip_len > ifp->if_mtu) { 512 error = EMSGSIZE; 513 goto bad; 514 } 515 m->m_flags |= M_BCAST; 516 } else { 517 m->m_flags &= ~M_BCAST; 518 } 519 520 sendit: 521 #ifdef IPSEC 522 /* get SP for this packet */ 523 if (so == NULL) 524 sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 525 else 526 sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 527 528 if (sp == NULL) { 529 ipsecstat.out_inval++; 530 goto bad; 531 } 532 533 error = 0; 534 535 /* check policy */ 536 switch (sp->policy) { 537 case IPSEC_POLICY_DISCARD: 538 /* 539 * This packet is just discarded. 540 */ 541 ipsecstat.out_polvio++; 542 goto bad; 543 544 case IPSEC_POLICY_BYPASS: 545 case IPSEC_POLICY_NONE: 546 case IPSEC_POLICY_TCP: 547 /* no need to do IPsec. */ 548 goto skip_ipsec; 549 550 case IPSEC_POLICY_IPSEC: 551 if (sp->req == NULL) { 552 /* acquire a policy */ 553 error = key_spdacquire(sp); 554 goto bad; 555 } 556 break; 557 558 case IPSEC_POLICY_ENTRUST: 559 default: 560 kprintf("ip_output: Invalid policy found. %d\n", sp->policy); 561 } 562 { 563 struct ipsec_output_state state; 564 bzero(&state, sizeof state); 565 state.m = m; 566 if (flags & IP_ROUTETOIF) { 567 state.ro = &iproute; 568 bzero(&iproute, sizeof iproute); 569 } else 570 state.ro = ro; 571 state.dst = (struct sockaddr *)dst; 572 573 ip->ip_sum = 0; 574 575 /* 576 * XXX 577 * delayed checksums are not currently compatible with IPsec 578 */ 579 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 580 in_delayed_cksum(m); 581 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 582 } 583 584 ip->ip_len = htons(ip->ip_len); 585 ip->ip_off = htons(ip->ip_off); 586 587 error = ipsec4_output(&state, sp, flags); 588 589 m = state.m; 590 if (flags & IP_ROUTETOIF) { 591 /* 592 * if we have tunnel mode SA, we may need to ignore 593 * IP_ROUTETOIF. 594 */ 595 if (state.ro != &iproute || state.ro->ro_rt != NULL) { 596 flags &= ~IP_ROUTETOIF; 597 ro = state.ro; 598 } 599 } else 600 ro = state.ro; 601 dst = (struct sockaddr_in *)state.dst; 602 if (error) { 603 /* mbuf is already reclaimed in ipsec4_output. */ 604 m0 = NULL; 605 switch (error) { 606 case EHOSTUNREACH: 607 case ENETUNREACH: 608 case EMSGSIZE: 609 case ENOBUFS: 610 case ENOMEM: 611 break; 612 default: 613 kprintf("ip4_output (ipsec): error code %d\n", error); 614 /*fall through*/ 615 case ENOENT: 616 /* don't show these error codes to the user */ 617 error = 0; 618 break; 619 } 620 goto bad; 621 } 622 } 623 624 /* be sure to update variables that are affected by ipsec4_output() */ 625 ip = mtod(m, struct ip *); 626 #ifdef _IP_VHL 627 hlen = IP_VHL_HL(ip->ip_vhl) << 2; 628 #else 629 hlen = ip->ip_hl << 2; 630 #endif 631 if (ro->ro_rt == NULL) { 632 if (!(flags & IP_ROUTETOIF)) { 633 kprintf("ip_output: " 634 "can't update route after IPsec processing\n"); 635 error = EHOSTUNREACH; /*XXX*/ 636 goto bad; 637 } 638 } else { 639 ia = ifatoia(ro->ro_rt->rt_ifa); 640 ifp = ro->ro_rt->rt_ifp; 641 } 642 643 /* make it flipped, again. */ 644 ip->ip_len = ntohs(ip->ip_len); 645 ip->ip_off = ntohs(ip->ip_off); 646 skip_ipsec: 647 #endif /*IPSEC*/ 648 #ifdef FAST_IPSEC 649 /* 650 * Check the security policy (SP) for the packet and, if 651 * required, do IPsec-related processing. There are two 652 * cases here; the first time a packet is sent through 653 * it will be untagged and handled by ipsec4_checkpolicy. 654 * If the packet is resubmitted to ip_output (e.g. after 655 * AH, ESP, etc. processing), there will be a tag to bypass 656 * the lookup and related policy checking. 657 */ 658 mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 659 crit_enter(); 660 if (mtag != NULL) { 661 tdbi = (struct tdb_ident *)m_tag_data(mtag); 662 sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 663 if (sp == NULL) 664 error = -EINVAL; /* force silent drop */ 665 m_tag_delete(m, mtag); 666 } else { 667 sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 668 &error, inp); 669 } 670 /* 671 * There are four return cases: 672 * sp != NULL apply IPsec policy 673 * sp == NULL, error == 0 no IPsec handling needed 674 * sp == NULL, error == -EINVAL discard packet w/o error 675 * sp == NULL, error != 0 discard packet, report error 676 */ 677 if (sp != NULL) { 678 /* Loop detection, check if ipsec processing already done */ 679 KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 680 for (mtag = m_tag_first(m); mtag != NULL; 681 mtag = m_tag_next(m, mtag)) { 682 if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 683 continue; 684 if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 685 mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 686 continue; 687 /* 688 * Check if policy has an SA associated with it. 689 * This can happen when an SP has yet to acquire 690 * an SA; e.g. on first reference. If it occurs, 691 * then we let ipsec4_process_packet do its thing. 692 */ 693 if (sp->req->sav == NULL) 694 break; 695 tdbi = (struct tdb_ident *)m_tag_data(mtag); 696 if (tdbi->spi == sp->req->sav->spi && 697 tdbi->proto == sp->req->sav->sah->saidx.proto && 698 bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 699 sizeof(union sockaddr_union)) == 0) { 700 /* 701 * No IPsec processing is needed, free 702 * reference to SP. 703 * 704 * NB: null pointer to avoid free at 705 * done: below. 706 */ 707 KEY_FREESP(&sp), sp = NULL; 708 crit_exit(); 709 goto spd_done; 710 } 711 } 712 713 /* 714 * Do delayed checksums now because we send before 715 * this is done in the normal processing path. 716 */ 717 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 718 in_delayed_cksum(m); 719 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 720 } 721 722 ip->ip_len = htons(ip->ip_len); 723 ip->ip_off = htons(ip->ip_off); 724 725 /* NB: callee frees mbuf */ 726 error = ipsec4_process_packet(m, sp->req, flags, 0); 727 /* 728 * Preserve KAME behaviour: ENOENT can be returned 729 * when an SA acquire is in progress. Don't propagate 730 * this to user-level; it confuses applications. 731 * 732 * XXX this will go away when the SADB is redone. 733 */ 734 if (error == ENOENT) 735 error = 0; 736 crit_exit(); 737 goto done; 738 } else { 739 crit_exit(); 740 741 if (error != 0) { 742 /* 743 * Hack: -EINVAL is used to signal that a packet 744 * should be silently discarded. This is typically 745 * because we asked key management for an SA and 746 * it was delayed (e.g. kicked up to IKE). 747 */ 748 if (error == -EINVAL) 749 error = 0; 750 goto bad; 751 } else { 752 /* No IPsec processing for this packet. */ 753 } 754 #ifdef notyet 755 /* 756 * If deferred crypto processing is needed, check that 757 * the interface supports it. 758 */ 759 mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 760 if (mtag != NULL && !(ifp->if_capenable & IFCAP_IPSEC)) { 761 /* notify IPsec to do its own crypto */ 762 ipsp_skipcrypto_unmark((struct tdb_ident *)m_tag_data(mtag)); 763 error = EHOSTUNREACH; 764 goto bad; 765 } 766 #endif 767 } 768 spd_done: 769 #endif /* FAST_IPSEC */ 770 771 /* We are already being fwd'd from a firewall. */ 772 if (next_hop != NULL) 773 goto pass; 774 775 /* No pfil hooks */ 776 if (!pfil_has_hooks(&inet_pfil_hook)) { 777 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 778 /* 779 * Strip dummynet tags from stranded packets 780 */ 781 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 782 KKASSERT(mtag != NULL); 783 m_tag_delete(m, mtag); 784 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED; 785 } 786 goto pass; 787 } 788 789 /* 790 * IpHack's section. 791 * - Xlate: translate packet's addr/port (NAT). 792 * - Firewall: deny/allow/etc. 793 * - Wrap: fake packet's addr/port <unimpl.> 794 * - Encapsulate: put it in another IP and send out. <unimp.> 795 */ 796 797 /* 798 * Run through list of hooks for output packets. 799 */ 800 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT); 801 if (error != 0 || m == NULL) 802 goto done; 803 ip = mtod(m, struct ip *); 804 805 if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) { 806 /* 807 * Check dst to make sure it is directly reachable on the 808 * interface we previously thought it was. 809 * If it isn't (which may be likely in some situations) we have 810 * to re-route it (ie, find a route for the next-hop and the 811 * associated interface) and set them here. This is nested 812 * forwarding which in most cases is undesirable, except where 813 * such control is nigh impossible. So we do it here. 814 * And I'm babbling. 815 */ 816 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 817 KKASSERT(mtag != NULL); 818 next_hop = m_tag_data(mtag); 819 820 /* 821 * Try local forwarding first 822 */ 823 if (ip_localforward(m, next_hop, hlen)) 824 goto done; 825 826 /* 827 * Relocate the route based on next_hop. 828 * If the current route is inp's cache, keep it untouched. 829 */ 830 if (ro == &iproute && ro->ro_rt != NULL) { 831 RTFREE(ro->ro_rt); 832 ro->ro_rt = NULL; 833 } 834 ro = &iproute; 835 bzero(ro, sizeof *ro); 836 837 /* 838 * Forwarding to broadcast address is not allowed. 839 * XXX Should we follow IP_ROUTETOIF? 840 */ 841 flags &= ~(IP_ALLOWBROADCAST | IP_ROUTETOIF); 842 843 /* We are doing forwarding now */ 844 flags |= IP_FORWARDING; 845 846 goto reroute; 847 } 848 849 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) { 850 struct dn_pkt *dn_pkt; 851 852 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL); 853 KKASSERT(mtag != NULL); 854 dn_pkt = m_tag_data(mtag); 855 856 /* 857 * Under certain cases it is not possible to recalculate 858 * 'ro' and 'dst', let alone 'flags', so just save them in 859 * dummynet tag and avoid the possible wrong reculcalation 860 * when we come back to ip_output() again. 861 * 862 * All other parameters have been already used and so they 863 * are not needed anymore. 864 * XXX if the ifp is deleted while a pkt is in dummynet, 865 * we are in trouble! (TODO use ifnet_detach_event) 866 * 867 * We need to copy *ro because for ICMP pkts (and maybe 868 * others) the caller passed a pointer into the stack; 869 * dst might also be a pointer into *ro so it needs to 870 * be updated. 871 */ 872 dn_pkt->ro = *ro; 873 if (ro->ro_rt) 874 ro->ro_rt->rt_refcnt++; 875 if (dst == (struct sockaddr_in *)&ro->ro_dst) { 876 /* 'dst' points into 'ro' */ 877 dst = (struct sockaddr_in *)&(dn_pkt->ro.ro_dst); 878 } 879 dn_pkt->dn_dst = dst; 880 dn_pkt->flags = flags; 881 882 ip_dn_queue(m); 883 goto done; 884 } 885 pass: 886 /* 127/8 must not appear on wire - RFC1122. */ 887 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 888 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 889 if (!(ifp->if_flags & IFF_LOOPBACK)) { 890 ipstat.ips_badaddr++; 891 error = EADDRNOTAVAIL; 892 goto bad; 893 } 894 } 895 896 if ((m->m_pkthdr.csum_flags & CSUM_TSO) == 0) { 897 m->m_pkthdr.csum_flags |= CSUM_IP; 898 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 899 if (sw_csum & CSUM_DELAY_DATA) { 900 in_delayed_cksum(m); 901 sw_csum &= ~CSUM_DELAY_DATA; 902 } 903 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 904 } else { 905 sw_csum = 0; 906 } 907 m->m_pkthdr.csum_iphlen = hlen; 908 909 /* 910 * If small enough for interface, or the interface will take 911 * care of the fragmentation or segmentation for us, can just 912 * send directly. 913 */ 914 if (ip->ip_len <= ifp->if_mtu || 915 ((ifp->if_hwassist & CSUM_FRAGMENT) && !(ip->ip_off & IP_DF)) || 916 (m->m_pkthdr.csum_flags & CSUM_TSO)) { 917 ip->ip_len = htons(ip->ip_len); 918 ip->ip_off = htons(ip->ip_off); 919 ip->ip_sum = 0; 920 if (sw_csum & CSUM_DELAY_IP) { 921 if (ip->ip_vhl == IP_VHL_BORING) 922 ip->ip_sum = in_cksum_hdr(ip); 923 else 924 ip->ip_sum = in_cksum(m, hlen); 925 } 926 927 /* Record statistics for this interface address. */ 928 if (!(flags & IP_FORWARDING) && ia) { 929 ia->ia_ifa.if_opackets++; 930 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 931 } 932 933 #ifdef IPSEC 934 /* clean ipsec history once it goes out of the node */ 935 ipsec_delaux(m); 936 #endif 937 938 #ifdef MBUF_STRESS_TEST 939 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) { 940 struct mbuf *m1, *m2; 941 int length, tmp; 942 943 tmp = length = m->m_pkthdr.len; 944 945 while ((length -= mbuf_frag_size) >= 1) { 946 m1 = m_split(m, length, MB_DONTWAIT); 947 if (m1 == NULL) 948 break; 949 m2 = m; 950 while (m2->m_next != NULL) 951 m2 = m2->m_next; 952 m2->m_next = m1; 953 } 954 m->m_pkthdr.len = tmp; 955 } 956 #endif 957 958 #ifdef MPLS 959 if (!mpls_output_process(m, ro->ro_rt)) 960 goto done; 961 #endif 962 error = ifp->if_output(ifp, m, (struct sockaddr *)dst, 963 ro->ro_rt); 964 goto done; 965 } 966 967 if (ip->ip_off & IP_DF) { 968 error = EMSGSIZE; 969 /* 970 * This case can happen if the user changed the MTU 971 * of an interface after enabling IP on it. Because 972 * most netifs don't keep track of routes pointing to 973 * them, there is no way for one to update all its 974 * routes when the MTU is changed. 975 */ 976 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 977 !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) && 978 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 979 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 980 } 981 ipstat.ips_cantfrag++; 982 goto bad; 983 } 984 985 /* 986 * Too large for interface; fragment if possible. If successful, 987 * on return, m will point to a list of packets to be sent. 988 */ 989 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 990 if (error) 991 goto bad; 992 for (; m; m = m0) { 993 m0 = m->m_nextpkt; 994 m->m_nextpkt = NULL; 995 #ifdef IPSEC 996 /* clean ipsec history once it goes out of the node */ 997 ipsec_delaux(m); 998 #endif 999 if (error == 0) { 1000 /* Record statistics for this interface address. */ 1001 if (ia != NULL) { 1002 ia->ia_ifa.if_opackets++; 1003 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1004 } 1005 #ifdef MPLS 1006 if (!mpls_output_process(m, ro->ro_rt)) 1007 continue; 1008 #endif 1009 error = ifp->if_output(ifp, m, (struct sockaddr *)dst, 1010 ro->ro_rt); 1011 } else { 1012 m_freem(m); 1013 } 1014 } 1015 1016 if (error == 0) 1017 ipstat.ips_fragmented++; 1018 1019 done: 1020 if (ro == &iproute && ro->ro_rt != NULL) { 1021 RTFREE(ro->ro_rt); 1022 ro->ro_rt = NULL; 1023 } 1024 #ifdef IPSEC 1025 if (sp != NULL) { 1026 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1027 kprintf("DP ip_output call free SP:%p\n", sp)); 1028 key_freesp(sp); 1029 } 1030 #endif 1031 #ifdef FAST_IPSEC 1032 if (sp != NULL) 1033 KEY_FREESP(&sp); 1034 #endif 1035 return (error); 1036 bad: 1037 m_freem(m); 1038 goto done; 1039 } 1040 1041 /* 1042 * Create a chain of fragments which fit the given mtu. m_frag points to the 1043 * mbuf to be fragmented; on return it points to the chain with the fragments. 1044 * Return 0 if no error. If error, m_frag may contain a partially built 1045 * chain of fragments that should be freed by the caller. 1046 * 1047 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 1048 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 1049 */ 1050 int 1051 ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 1052 u_long if_hwassist_flags, int sw_csum) 1053 { 1054 int error = 0; 1055 int hlen = IP_VHL_HL(ip->ip_vhl) << 2; 1056 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 1057 int off; 1058 struct mbuf *m0 = *m_frag; /* the original packet */ 1059 int firstlen; 1060 struct mbuf **mnext; 1061 int nfrags; 1062 1063 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 1064 ipstat.ips_cantfrag++; 1065 return EMSGSIZE; 1066 } 1067 1068 /* 1069 * Must be able to put at least 8 bytes per fragment. 1070 */ 1071 if (len < 8) 1072 return EMSGSIZE; 1073 1074 /* 1075 * If the interface will not calculate checksums on 1076 * fragmented packets, then do it here. 1077 */ 1078 if ((m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) && 1079 !(if_hwassist_flags & CSUM_IP_FRAGS)) { 1080 in_delayed_cksum(m0); 1081 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1082 } 1083 1084 if (len > PAGE_SIZE) { 1085 /* 1086 * Fragment large datagrams such that each segment 1087 * contains a multiple of PAGE_SIZE amount of data, 1088 * plus headers. This enables a receiver to perform 1089 * page-flipping zero-copy optimizations. 1090 * 1091 * XXX When does this help given that sender and receiver 1092 * could have different page sizes, and also mtu could 1093 * be less than the receiver's page size ? 1094 */ 1095 int newlen; 1096 struct mbuf *m; 1097 1098 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 1099 off += m->m_len; 1100 1101 /* 1102 * firstlen (off - hlen) must be aligned on an 1103 * 8-byte boundary 1104 */ 1105 if (off < hlen) 1106 goto smart_frag_failure; 1107 off = ((off - hlen) & ~7) + hlen; 1108 newlen = (~PAGE_MASK) & mtu; 1109 if ((newlen + sizeof(struct ip)) > mtu) { 1110 /* we failed, go back the default */ 1111 smart_frag_failure: 1112 newlen = len; 1113 off = hlen + len; 1114 } 1115 len = newlen; 1116 1117 } else { 1118 off = hlen + len; 1119 } 1120 1121 firstlen = off - hlen; 1122 mnext = &m0->m_nextpkt; /* pointer to next packet */ 1123 1124 /* 1125 * Loop through length of segment after first fragment, 1126 * make new header and copy data of each part and link onto chain. 1127 * Here, m0 is the original packet, m is the fragment being created. 1128 * The fragments are linked off the m_nextpkt of the original 1129 * packet, which after processing serves as the first fragment. 1130 */ 1131 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 1132 struct ip *mhip; /* ip header on the fragment */ 1133 struct mbuf *m; 1134 int mhlen = sizeof(struct ip); 1135 1136 MGETHDR(m, MB_DONTWAIT, MT_HEADER); 1137 if (m == NULL) { 1138 error = ENOBUFS; 1139 ipstat.ips_odropped++; 1140 goto done; 1141 } 1142 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1143 /* 1144 * In the first mbuf, leave room for the link header, then 1145 * copy the original IP header including options. The payload 1146 * goes into an additional mbuf chain returned by m_copy(). 1147 */ 1148 m->m_data += max_linkhdr; 1149 mhip = mtod(m, struct ip *); 1150 *mhip = *ip; 1151 if (hlen > sizeof(struct ip)) { 1152 mhlen = ip_optcopy(ip, mhip) + sizeof(struct ip); 1153 mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 1154 } 1155 m->m_len = mhlen; 1156 /* XXX do we need to add ip->ip_off below ? */ 1157 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1158 if (off + len >= ip->ip_len) { /* last fragment */ 1159 len = ip->ip_len - off; 1160 m->m_flags |= M_LASTFRAG; 1161 } else 1162 mhip->ip_off |= IP_MF; 1163 mhip->ip_len = htons((u_short)(len + mhlen)); 1164 m->m_next = m_copy(m0, off, len); 1165 if (m->m_next == NULL) { /* copy failed */ 1166 m_free(m); 1167 error = ENOBUFS; /* ??? */ 1168 ipstat.ips_odropped++; 1169 goto done; 1170 } 1171 m->m_pkthdr.len = mhlen + len; 1172 m->m_pkthdr.rcvif = NULL; 1173 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1174 m->m_pkthdr.csum_iphlen = mhlen; 1175 mhip->ip_off = htons(mhip->ip_off); 1176 mhip->ip_sum = 0; 1177 if (sw_csum & CSUM_DELAY_IP) 1178 mhip->ip_sum = in_cksum(m, mhlen); 1179 *mnext = m; 1180 mnext = &m->m_nextpkt; 1181 } 1182 ipstat.ips_ofragments += nfrags; 1183 1184 /* set first marker for fragment chain */ 1185 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1186 m0->m_pkthdr.csum_data = nfrags; 1187 1188 /* 1189 * Update first fragment by trimming what's been copied out 1190 * and updating header. 1191 */ 1192 m_adj(m0, hlen + firstlen - ip->ip_len); 1193 m0->m_pkthdr.len = hlen + firstlen; 1194 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 1195 ip->ip_off |= IP_MF; 1196 ip->ip_off = htons(ip->ip_off); 1197 ip->ip_sum = 0; 1198 if (sw_csum & CSUM_DELAY_IP) 1199 ip->ip_sum = in_cksum(m0, hlen); 1200 1201 done: 1202 *m_frag = m0; 1203 return error; 1204 } 1205 1206 void 1207 in_delayed_cksum(struct mbuf *m) 1208 { 1209 struct ip *ip; 1210 u_short csum, offset; 1211 1212 ip = mtod(m, struct ip *); 1213 offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 1214 csum = in_cksum_skip(m, ip->ip_len, offset); 1215 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1216 csum = 0xffff; 1217 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1218 1219 if (offset + sizeof(u_short) > m->m_len) { 1220 kprintf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1221 m->m_len, offset, ip->ip_p); 1222 /* 1223 * XXX 1224 * this shouldn't happen, but if it does, the 1225 * correct behavior may be to insert the checksum 1226 * in the existing chain instead of rearranging it. 1227 */ 1228 m = m_pullup(m, offset + sizeof(u_short)); 1229 } 1230 *(u_short *)(m->m_data + offset) = csum; 1231 } 1232 1233 /* 1234 * Insert IP options into preformed packet. 1235 * Adjust IP destination as required for IP source routing, 1236 * as indicated by a non-zero in_addr at the start of the options. 1237 * 1238 * XXX This routine assumes that the packet has no options in place. 1239 */ 1240 static struct mbuf * 1241 ip_insertoptions(struct mbuf *m, struct mbuf *opt, int *phlen) 1242 { 1243 struct ipoption *p = mtod(opt, struct ipoption *); 1244 struct mbuf *n; 1245 struct ip *ip = mtod(m, struct ip *); 1246 unsigned optlen; 1247 1248 optlen = opt->m_len - sizeof p->ipopt_dst; 1249 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) { 1250 *phlen = 0; 1251 return (m); /* XXX should fail */ 1252 } 1253 if (p->ipopt_dst.s_addr) 1254 ip->ip_dst = p->ipopt_dst; 1255 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1256 MGETHDR(n, MB_DONTWAIT, MT_HEADER); 1257 if (n == NULL) { 1258 *phlen = 0; 1259 return (m); 1260 } 1261 n->m_pkthdr.rcvif = NULL; 1262 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1263 m->m_len -= sizeof(struct ip); 1264 m->m_data += sizeof(struct ip); 1265 n->m_next = m; 1266 m = n; 1267 m->m_len = optlen + sizeof(struct ip); 1268 m->m_data += max_linkhdr; 1269 memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1270 } else { 1271 m->m_data -= optlen; 1272 m->m_len += optlen; 1273 m->m_pkthdr.len += optlen; 1274 ovbcopy(ip, mtod(m, caddr_t), sizeof(struct ip)); 1275 } 1276 ip = mtod(m, struct ip *); 1277 bcopy(p->ipopt_list, ip + 1, optlen); 1278 *phlen = sizeof(struct ip) + optlen; 1279 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1280 ip->ip_len += optlen; 1281 return (m); 1282 } 1283 1284 /* 1285 * Copy options from ip to jp, 1286 * omitting those not copied during fragmentation. 1287 */ 1288 int 1289 ip_optcopy(struct ip *ip, struct ip *jp) 1290 { 1291 u_char *cp, *dp; 1292 int opt, optlen, cnt; 1293 1294 cp = (u_char *)(ip + 1); 1295 dp = (u_char *)(jp + 1); 1296 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip); 1297 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1298 opt = cp[0]; 1299 if (opt == IPOPT_EOL) 1300 break; 1301 if (opt == IPOPT_NOP) { 1302 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1303 *dp++ = IPOPT_NOP; 1304 optlen = 1; 1305 continue; 1306 } 1307 1308 KASSERT(cnt >= IPOPT_OLEN + sizeof *cp, 1309 ("ip_optcopy: malformed ipv4 option")); 1310 optlen = cp[IPOPT_OLEN]; 1311 KASSERT(optlen >= IPOPT_OLEN + sizeof *cp && optlen <= cnt, 1312 ("ip_optcopy: malformed ipv4 option")); 1313 1314 /* bogus lengths should have been caught by ip_dooptions */ 1315 if (optlen > cnt) 1316 optlen = cnt; 1317 if (IPOPT_COPIED(opt)) { 1318 bcopy(cp, dp, optlen); 1319 dp += optlen; 1320 } 1321 } 1322 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1323 *dp++ = IPOPT_EOL; 1324 return (optlen); 1325 } 1326 1327 /* 1328 * IP socket option processing. 1329 */ 1330 void 1331 ip_ctloutput(netmsg_t msg) 1332 { 1333 struct socket *so = msg->base.nm_so; 1334 struct sockopt *sopt = msg->ctloutput.nm_sopt; 1335 struct inpcb *inp = so->so_pcb; 1336 int error, optval; 1337 1338 error = optval = 0; 1339 if (sopt->sopt_level != IPPROTO_IP) { 1340 error = EINVAL; 1341 goto done; 1342 } 1343 1344 switch (sopt->sopt_dir) { 1345 case SOPT_SET: 1346 switch (sopt->sopt_name) { 1347 case IP_OPTIONS: 1348 #ifdef notyet 1349 case IP_RETOPTS: 1350 #endif 1351 { 1352 struct mbuf *m; 1353 if (sopt->sopt_valsize > MLEN) { 1354 error = EMSGSIZE; 1355 break; 1356 } 1357 MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER); 1358 if (m == NULL) { 1359 error = ENOBUFS; 1360 break; 1361 } 1362 m->m_len = sopt->sopt_valsize; 1363 error = soopt_to_kbuf(sopt, mtod(m, void *), m->m_len, 1364 m->m_len); 1365 error = ip_pcbopts(sopt->sopt_name, 1366 &inp->inp_options, m); 1367 goto done; 1368 } 1369 1370 case IP_TOS: 1371 case IP_TTL: 1372 case IP_MINTTL: 1373 case IP_RECVOPTS: 1374 case IP_RECVRETOPTS: 1375 case IP_RECVDSTADDR: 1376 case IP_RECVIF: 1377 case IP_RECVTTL: 1378 case IP_FAITH: 1379 error = soopt_to_kbuf(sopt, &optval, sizeof optval, 1380 sizeof optval); 1381 if (error) 1382 break; 1383 switch (sopt->sopt_name) { 1384 case IP_TOS: 1385 inp->inp_ip_tos = optval; 1386 break; 1387 1388 case IP_TTL: 1389 inp->inp_ip_ttl = optval; 1390 break; 1391 case IP_MINTTL: 1392 if (optval >= 0 && optval <= MAXTTL) 1393 inp->inp_ip_minttl = optval; 1394 else 1395 error = EINVAL; 1396 break; 1397 #define OPTSET(bit) \ 1398 if (optval) \ 1399 inp->inp_flags |= bit; \ 1400 else \ 1401 inp->inp_flags &= ~bit; 1402 1403 case IP_RECVOPTS: 1404 OPTSET(INP_RECVOPTS); 1405 break; 1406 1407 case IP_RECVRETOPTS: 1408 OPTSET(INP_RECVRETOPTS); 1409 break; 1410 1411 case IP_RECVDSTADDR: 1412 OPTSET(INP_RECVDSTADDR); 1413 break; 1414 1415 case IP_RECVIF: 1416 OPTSET(INP_RECVIF); 1417 break; 1418 1419 case IP_RECVTTL: 1420 OPTSET(INP_RECVTTL); 1421 break; 1422 1423 case IP_FAITH: 1424 OPTSET(INP_FAITH); 1425 break; 1426 } 1427 break; 1428 #undef OPTSET 1429 1430 case IP_MULTICAST_IF: 1431 case IP_MULTICAST_VIF: 1432 case IP_MULTICAST_TTL: 1433 case IP_MULTICAST_LOOP: 1434 case IP_ADD_MEMBERSHIP: 1435 case IP_DROP_MEMBERSHIP: 1436 error = ip_setmoptions(sopt, &inp->inp_moptions); 1437 break; 1438 1439 case IP_PORTRANGE: 1440 error = soopt_to_kbuf(sopt, &optval, sizeof optval, 1441 sizeof optval); 1442 if (error) 1443 break; 1444 1445 switch (optval) { 1446 case IP_PORTRANGE_DEFAULT: 1447 inp->inp_flags &= ~(INP_LOWPORT); 1448 inp->inp_flags &= ~(INP_HIGHPORT); 1449 break; 1450 1451 case IP_PORTRANGE_HIGH: 1452 inp->inp_flags &= ~(INP_LOWPORT); 1453 inp->inp_flags |= INP_HIGHPORT; 1454 break; 1455 1456 case IP_PORTRANGE_LOW: 1457 inp->inp_flags &= ~(INP_HIGHPORT); 1458 inp->inp_flags |= INP_LOWPORT; 1459 break; 1460 1461 default: 1462 error = EINVAL; 1463 break; 1464 } 1465 break; 1466 1467 #if defined(IPSEC) || defined(FAST_IPSEC) 1468 case IP_IPSEC_POLICY: 1469 { 1470 caddr_t req; 1471 size_t len = 0; 1472 int priv; 1473 struct mbuf *m; 1474 int optname; 1475 1476 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1477 break; 1478 soopt_to_mbuf(sopt, m); 1479 priv = (sopt->sopt_td != NULL && 1480 priv_check(sopt->sopt_td, PRIV_ROOT) != 0) ? 0 : 1; 1481 req = mtod(m, caddr_t); 1482 len = m->m_len; 1483 optname = sopt->sopt_name; 1484 error = ipsec4_set_policy(inp, optname, req, len, priv); 1485 m_freem(m); 1486 break; 1487 } 1488 #endif /*IPSEC*/ 1489 1490 default: 1491 error = ENOPROTOOPT; 1492 break; 1493 } 1494 break; 1495 1496 case SOPT_GET: 1497 switch (sopt->sopt_name) { 1498 case IP_OPTIONS: 1499 case IP_RETOPTS: 1500 if (inp->inp_options) 1501 soopt_from_kbuf(sopt, mtod(inp->inp_options, 1502 char *), 1503 inp->inp_options->m_len); 1504 else 1505 sopt->sopt_valsize = 0; 1506 break; 1507 1508 case IP_TOS: 1509 case IP_TTL: 1510 case IP_MINTTL: 1511 case IP_RECVOPTS: 1512 case IP_RECVRETOPTS: 1513 case IP_RECVDSTADDR: 1514 case IP_RECVTTL: 1515 case IP_RECVIF: 1516 case IP_PORTRANGE: 1517 case IP_FAITH: 1518 switch (sopt->sopt_name) { 1519 1520 case IP_TOS: 1521 optval = inp->inp_ip_tos; 1522 break; 1523 1524 case IP_TTL: 1525 optval = inp->inp_ip_ttl; 1526 break; 1527 case IP_MINTTL: 1528 optval = inp->inp_ip_minttl; 1529 break; 1530 1531 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1532 1533 case IP_RECVOPTS: 1534 optval = OPTBIT(INP_RECVOPTS); 1535 break; 1536 1537 case IP_RECVRETOPTS: 1538 optval = OPTBIT(INP_RECVRETOPTS); 1539 break; 1540 1541 case IP_RECVDSTADDR: 1542 optval = OPTBIT(INP_RECVDSTADDR); 1543 break; 1544 1545 case IP_RECVTTL: 1546 optval = OPTBIT(INP_RECVTTL); 1547 break; 1548 1549 case IP_RECVIF: 1550 optval = OPTBIT(INP_RECVIF); 1551 break; 1552 1553 case IP_PORTRANGE: 1554 if (inp->inp_flags & INP_HIGHPORT) 1555 optval = IP_PORTRANGE_HIGH; 1556 else if (inp->inp_flags & INP_LOWPORT) 1557 optval = IP_PORTRANGE_LOW; 1558 else 1559 optval = 0; 1560 break; 1561 1562 case IP_FAITH: 1563 optval = OPTBIT(INP_FAITH); 1564 break; 1565 } 1566 soopt_from_kbuf(sopt, &optval, sizeof optval); 1567 break; 1568 1569 case IP_MULTICAST_IF: 1570 case IP_MULTICAST_VIF: 1571 case IP_MULTICAST_TTL: 1572 case IP_MULTICAST_LOOP: 1573 case IP_ADD_MEMBERSHIP: 1574 case IP_DROP_MEMBERSHIP: 1575 error = ip_getmoptions(sopt, inp->inp_moptions); 1576 break; 1577 1578 #if defined(IPSEC) || defined(FAST_IPSEC) 1579 case IP_IPSEC_POLICY: 1580 { 1581 struct mbuf *m = NULL; 1582 caddr_t req = NULL; 1583 size_t len = 0; 1584 1585 if (m != NULL) { 1586 req = mtod(m, caddr_t); 1587 len = m->m_len; 1588 } 1589 error = ipsec4_get_policy(so->so_pcb, req, len, &m); 1590 if (error == 0) 1591 error = soopt_from_mbuf(sopt, m); /* XXX */ 1592 if (error == 0) 1593 m_freem(m); 1594 break; 1595 } 1596 #endif /*IPSEC*/ 1597 1598 default: 1599 error = ENOPROTOOPT; 1600 break; 1601 } 1602 break; 1603 } 1604 done: 1605 lwkt_replymsg(&msg->lmsg, error); 1606 } 1607 1608 /* 1609 * Set up IP options in pcb for insertion in output packets. 1610 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1611 * with destination address if source routed. 1612 */ 1613 static int 1614 ip_pcbopts(int optname, struct mbuf **pcbopt, struct mbuf *m) 1615 { 1616 int cnt, optlen; 1617 u_char *cp; 1618 u_char opt; 1619 1620 /* turn off any old options */ 1621 if (*pcbopt) 1622 m_free(*pcbopt); 1623 *pcbopt = NULL; 1624 if (m == NULL || m->m_len == 0) { 1625 /* 1626 * Only turning off any previous options. 1627 */ 1628 if (m != NULL) 1629 m_free(m); 1630 return (0); 1631 } 1632 1633 if (m->m_len % sizeof(int32_t)) 1634 goto bad; 1635 /* 1636 * IP first-hop destination address will be stored before 1637 * actual options; move other options back 1638 * and clear it when none present. 1639 */ 1640 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1641 goto bad; 1642 cnt = m->m_len; 1643 m->m_len += sizeof(struct in_addr); 1644 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1645 ovbcopy(mtod(m, caddr_t), cp, cnt); 1646 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1647 1648 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1649 opt = cp[IPOPT_OPTVAL]; 1650 if (opt == IPOPT_EOL) 1651 break; 1652 if (opt == IPOPT_NOP) 1653 optlen = 1; 1654 else { 1655 if (cnt < IPOPT_OLEN + sizeof *cp) 1656 goto bad; 1657 optlen = cp[IPOPT_OLEN]; 1658 if (optlen < IPOPT_OLEN + sizeof *cp || optlen > cnt) 1659 goto bad; 1660 } 1661 switch (opt) { 1662 1663 default: 1664 break; 1665 1666 case IPOPT_LSRR: 1667 case IPOPT_SSRR: 1668 /* 1669 * user process specifies route as: 1670 * ->A->B->C->D 1671 * D must be our final destination (but we can't 1672 * check that since we may not have connected yet). 1673 * A is first hop destination, which doesn't appear in 1674 * actual IP option, but is stored before the options. 1675 */ 1676 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1677 goto bad; 1678 m->m_len -= sizeof(struct in_addr); 1679 cnt -= sizeof(struct in_addr); 1680 optlen -= sizeof(struct in_addr); 1681 cp[IPOPT_OLEN] = optlen; 1682 /* 1683 * Move first hop before start of options. 1684 */ 1685 bcopy(&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1686 sizeof(struct in_addr)); 1687 /* 1688 * Then copy rest of options back 1689 * to close up the deleted entry. 1690 */ 1691 ovbcopy(&cp[IPOPT_OFFSET+1] + sizeof(struct in_addr), 1692 &cp[IPOPT_OFFSET+1], 1693 cnt - (IPOPT_MINOFF - 1)); 1694 break; 1695 } 1696 } 1697 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1698 goto bad; 1699 *pcbopt = m; 1700 return (0); 1701 1702 bad: 1703 m_free(m); 1704 return (EINVAL); 1705 } 1706 1707 /* 1708 * XXX 1709 * The whole multicast option thing needs to be re-thought. 1710 * Several of these options are equally applicable to non-multicast 1711 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1712 * standard option (IP_TTL). 1713 */ 1714 1715 /* 1716 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1717 */ 1718 static struct ifnet * 1719 ip_multicast_if(struct in_addr *a, int *ifindexp) 1720 { 1721 int ifindex; 1722 struct ifnet *ifp; 1723 1724 if (ifindexp) 1725 *ifindexp = 0; 1726 if (ntohl(a->s_addr) >> 24 == 0) { 1727 ifindex = ntohl(a->s_addr) & 0xffffff; 1728 if (ifindex < 0 || if_index < ifindex) 1729 return NULL; 1730 ifp = ifindex2ifnet[ifindex]; 1731 if (ifindexp) 1732 *ifindexp = ifindex; 1733 } else { 1734 ifp = INADDR_TO_IFP(a); 1735 } 1736 return ifp; 1737 } 1738 1739 /* 1740 * Set the IP multicast options in response to user setsockopt(). 1741 */ 1742 static int 1743 ip_setmoptions(struct sockopt *sopt, struct ip_moptions **imop) 1744 { 1745 int error = 0; 1746 int i; 1747 struct in_addr addr; 1748 struct ip_mreq mreq; 1749 struct ifnet *ifp; 1750 struct ip_moptions *imo = *imop; 1751 int ifindex; 1752 1753 if (imo == NULL) { 1754 /* 1755 * No multicast option buffer attached to the pcb; 1756 * allocate one and initialize to default values. 1757 */ 1758 imo = kmalloc(sizeof *imo, M_IPMOPTS, M_WAITOK); 1759 1760 *imop = imo; 1761 imo->imo_multicast_ifp = NULL; 1762 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1763 imo->imo_multicast_vif = -1; 1764 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1765 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1766 imo->imo_num_memberships = 0; 1767 } 1768 switch (sopt->sopt_name) { 1769 /* store an index number for the vif you wanna use in the send */ 1770 case IP_MULTICAST_VIF: 1771 if (legal_vif_num == 0) { 1772 error = EOPNOTSUPP; 1773 break; 1774 } 1775 error = soopt_to_kbuf(sopt, &i, sizeof i, sizeof i); 1776 if (error) 1777 break; 1778 if (!legal_vif_num(i) && (i != -1)) { 1779 error = EINVAL; 1780 break; 1781 } 1782 imo->imo_multicast_vif = i; 1783 break; 1784 1785 case IP_MULTICAST_IF: 1786 /* 1787 * Select the interface for outgoing multicast packets. 1788 */ 1789 error = soopt_to_kbuf(sopt, &addr, sizeof addr, sizeof addr); 1790 if (error) 1791 break; 1792 1793 /* 1794 * INADDR_ANY is used to remove a previous selection. 1795 * When no interface is selected, a default one is 1796 * chosen every time a multicast packet is sent. 1797 */ 1798 if (addr.s_addr == INADDR_ANY) { 1799 imo->imo_multicast_ifp = NULL; 1800 break; 1801 } 1802 /* 1803 * The selected interface is identified by its local 1804 * IP address. Find the interface and confirm that 1805 * it supports multicasting. 1806 */ 1807 crit_enter(); 1808 ifp = ip_multicast_if(&addr, &ifindex); 1809 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) { 1810 crit_exit(); 1811 error = EADDRNOTAVAIL; 1812 break; 1813 } 1814 imo->imo_multicast_ifp = ifp; 1815 if (ifindex) 1816 imo->imo_multicast_addr = addr; 1817 else 1818 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1819 crit_exit(); 1820 break; 1821 1822 case IP_MULTICAST_TTL: 1823 /* 1824 * Set the IP time-to-live for outgoing multicast packets. 1825 * The original multicast API required a char argument, 1826 * which is inconsistent with the rest of the socket API. 1827 * We allow either a char or an int. 1828 */ 1829 if (sopt->sopt_valsize == 1) { 1830 u_char ttl; 1831 error = soopt_to_kbuf(sopt, &ttl, 1, 1); 1832 if (error) 1833 break; 1834 imo->imo_multicast_ttl = ttl; 1835 } else { 1836 u_int ttl; 1837 error = soopt_to_kbuf(sopt, &ttl, sizeof ttl, sizeof ttl); 1838 if (error) 1839 break; 1840 if (ttl > 255) 1841 error = EINVAL; 1842 else 1843 imo->imo_multicast_ttl = ttl; 1844 } 1845 break; 1846 1847 case IP_MULTICAST_LOOP: 1848 /* 1849 * Set the loopback flag for outgoing multicast packets. 1850 * Must be zero or one. The original multicast API required a 1851 * char argument, which is inconsistent with the rest 1852 * of the socket API. We allow either a char or an int. 1853 */ 1854 if (sopt->sopt_valsize == 1) { 1855 u_char loop; 1856 1857 error = soopt_to_kbuf(sopt, &loop, 1, 1); 1858 if (error) 1859 break; 1860 imo->imo_multicast_loop = !!loop; 1861 } else { 1862 u_int loop; 1863 1864 error = soopt_to_kbuf(sopt, &loop, sizeof loop, 1865 sizeof loop); 1866 if (error) 1867 break; 1868 imo->imo_multicast_loop = !!loop; 1869 } 1870 break; 1871 1872 case IP_ADD_MEMBERSHIP: 1873 /* 1874 * Add a multicast group membership. 1875 * Group must be a valid IP multicast address. 1876 */ 1877 error = soopt_to_kbuf(sopt, &mreq, sizeof mreq, sizeof mreq); 1878 if (error) 1879 break; 1880 1881 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1882 error = EINVAL; 1883 break; 1884 } 1885 crit_enter(); 1886 /* 1887 * If no interface address was provided, use the interface of 1888 * the route to the given multicast address. 1889 */ 1890 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1891 struct sockaddr_in dst; 1892 struct rtentry *rt; 1893 1894 bzero(&dst, sizeof(struct sockaddr_in)); 1895 dst.sin_len = sizeof(struct sockaddr_in); 1896 dst.sin_family = AF_INET; 1897 dst.sin_addr = mreq.imr_multiaddr; 1898 rt = rtlookup((struct sockaddr *)&dst); 1899 if (rt == NULL) { 1900 error = EADDRNOTAVAIL; 1901 crit_exit(); 1902 break; 1903 } 1904 --rt->rt_refcnt; 1905 ifp = rt->rt_ifp; 1906 } else { 1907 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1908 } 1909 1910 /* 1911 * See if we found an interface, and confirm that it 1912 * supports multicast. 1913 */ 1914 if (ifp == NULL || !(ifp->if_flags & IFF_MULTICAST)) { 1915 error = EADDRNOTAVAIL; 1916 crit_exit(); 1917 break; 1918 } 1919 /* 1920 * See if the membership already exists or if all the 1921 * membership slots are full. 1922 */ 1923 for (i = 0; i < imo->imo_num_memberships; ++i) { 1924 if (imo->imo_membership[i]->inm_ifp == ifp && 1925 imo->imo_membership[i]->inm_addr.s_addr 1926 == mreq.imr_multiaddr.s_addr) 1927 break; 1928 } 1929 if (i < imo->imo_num_memberships) { 1930 error = EADDRINUSE; 1931 crit_exit(); 1932 break; 1933 } 1934 if (i == IP_MAX_MEMBERSHIPS) { 1935 error = ETOOMANYREFS; 1936 crit_exit(); 1937 break; 1938 } 1939 /* 1940 * Everything looks good; add a new record to the multicast 1941 * address list for the given interface. 1942 */ 1943 if ((imo->imo_membership[i] = 1944 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1945 error = ENOBUFS; 1946 crit_exit(); 1947 break; 1948 } 1949 ++imo->imo_num_memberships; 1950 crit_exit(); 1951 break; 1952 1953 case IP_DROP_MEMBERSHIP: 1954 /* 1955 * Drop a multicast group membership. 1956 * Group must be a valid IP multicast address. 1957 */ 1958 error = soopt_to_kbuf(sopt, &mreq, sizeof mreq, sizeof mreq); 1959 if (error) 1960 break; 1961 1962 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1963 error = EINVAL; 1964 break; 1965 } 1966 1967 crit_enter(); 1968 /* 1969 * If an interface address was specified, get a pointer 1970 * to its ifnet structure. 1971 */ 1972 if (mreq.imr_interface.s_addr == INADDR_ANY) 1973 ifp = NULL; 1974 else { 1975 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1976 if (ifp == NULL) { 1977 error = EADDRNOTAVAIL; 1978 crit_exit(); 1979 break; 1980 } 1981 } 1982 /* 1983 * Find the membership in the membership array. 1984 */ 1985 for (i = 0; i < imo->imo_num_memberships; ++i) { 1986 if ((ifp == NULL || 1987 imo->imo_membership[i]->inm_ifp == ifp) && 1988 imo->imo_membership[i]->inm_addr.s_addr == 1989 mreq.imr_multiaddr.s_addr) 1990 break; 1991 } 1992 if (i == imo->imo_num_memberships) { 1993 error = EADDRNOTAVAIL; 1994 crit_exit(); 1995 break; 1996 } 1997 /* 1998 * Give up the multicast address record to which the 1999 * membership points. 2000 */ 2001 in_delmulti(imo->imo_membership[i]); 2002 /* 2003 * Remove the gap in the membership array. 2004 */ 2005 for (++i; i < imo->imo_num_memberships; ++i) 2006 imo->imo_membership[i-1] = imo->imo_membership[i]; 2007 --imo->imo_num_memberships; 2008 crit_exit(); 2009 break; 2010 2011 default: 2012 error = EOPNOTSUPP; 2013 break; 2014 } 2015 2016 /* 2017 * If all options have default values, no need to keep the mbuf. 2018 */ 2019 if (imo->imo_multicast_ifp == NULL && 2020 imo->imo_multicast_vif == -1 && 2021 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2022 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2023 imo->imo_num_memberships == 0) { 2024 kfree(*imop, M_IPMOPTS); 2025 *imop = NULL; 2026 } 2027 2028 return (error); 2029 } 2030 2031 /* 2032 * Return the IP multicast options in response to user getsockopt(). 2033 */ 2034 static int 2035 ip_getmoptions(struct sockopt *sopt, struct ip_moptions *imo) 2036 { 2037 struct in_addr addr; 2038 struct in_ifaddr *ia; 2039 int error, optval; 2040 u_char coptval; 2041 2042 error = 0; 2043 switch (sopt->sopt_name) { 2044 case IP_MULTICAST_VIF: 2045 if (imo != NULL) 2046 optval = imo->imo_multicast_vif; 2047 else 2048 optval = -1; 2049 soopt_from_kbuf(sopt, &optval, sizeof optval); 2050 break; 2051 2052 case IP_MULTICAST_IF: 2053 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2054 addr.s_addr = INADDR_ANY; 2055 else if (imo->imo_multicast_addr.s_addr) { 2056 /* return the value user has set */ 2057 addr = imo->imo_multicast_addr; 2058 } else { 2059 ia = IFP_TO_IA(imo->imo_multicast_ifp); 2060 addr.s_addr = (ia == NULL) ? INADDR_ANY 2061 : IA_SIN(ia)->sin_addr.s_addr; 2062 } 2063 soopt_from_kbuf(sopt, &addr, sizeof addr); 2064 break; 2065 2066 case IP_MULTICAST_TTL: 2067 if (imo == NULL) 2068 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2069 else 2070 optval = coptval = imo->imo_multicast_ttl; 2071 if (sopt->sopt_valsize == 1) 2072 soopt_from_kbuf(sopt, &coptval, 1); 2073 else 2074 soopt_from_kbuf(sopt, &optval, sizeof optval); 2075 break; 2076 2077 case IP_MULTICAST_LOOP: 2078 if (imo == NULL) 2079 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2080 else 2081 optval = coptval = imo->imo_multicast_loop; 2082 if (sopt->sopt_valsize == 1) 2083 soopt_from_kbuf(sopt, &coptval, 1); 2084 else 2085 soopt_from_kbuf(sopt, &optval, sizeof optval); 2086 break; 2087 2088 default: 2089 error = ENOPROTOOPT; 2090 break; 2091 } 2092 return (error); 2093 } 2094 2095 /* 2096 * Discard the IP multicast options. 2097 */ 2098 void 2099 ip_freemoptions(struct ip_moptions *imo) 2100 { 2101 int i; 2102 2103 if (imo != NULL) { 2104 for (i = 0; i < imo->imo_num_memberships; ++i) 2105 in_delmulti(imo->imo_membership[i]); 2106 kfree(imo, M_IPMOPTS); 2107 } 2108 } 2109 2110 /* 2111 * Routine called from ip_output() to loop back a copy of an IP multicast 2112 * packet to the input queue of a specified interface. Note that this 2113 * calls the output routine of the loopback "driver", but with an interface 2114 * pointer that might NOT be a loopback interface -- evil, but easier than 2115 * replicating that code here. 2116 */ 2117 static void 2118 ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst, 2119 int hlen) 2120 { 2121 struct ip *ip; 2122 struct mbuf *copym; 2123 2124 copym = m_copypacket(m, MB_DONTWAIT); 2125 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2126 copym = m_pullup(copym, hlen); 2127 if (copym != NULL) { 2128 /* 2129 * if the checksum hasn't been computed, mark it as valid 2130 */ 2131 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2132 in_delayed_cksum(copym); 2133 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2134 copym->m_pkthdr.csum_flags |= 2135 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2136 copym->m_pkthdr.csum_data = 0xffff; 2137 } 2138 /* 2139 * We don't bother to fragment if the IP length is greater 2140 * than the interface's MTU. Can this possibly matter? 2141 */ 2142 ip = mtod(copym, struct ip *); 2143 ip->ip_len = htons(ip->ip_len); 2144 ip->ip_off = htons(ip->ip_off); 2145 ip->ip_sum = 0; 2146 if (ip->ip_vhl == IP_VHL_BORING) { 2147 ip->ip_sum = in_cksum_hdr(ip); 2148 } else { 2149 ip->ip_sum = in_cksum(copym, hlen); 2150 } 2151 /* 2152 * NB: 2153 * It's not clear whether there are any lingering 2154 * reentrancy problems in other areas which might 2155 * be exposed by using ip_input directly (in 2156 * particular, everything which modifies the packet 2157 * in-place). Yet another option is using the 2158 * protosw directly to deliver the looped back 2159 * packet. For the moment, we'll err on the side 2160 * of safety by using if_simloop(). 2161 */ 2162 #if 1 /* XXX */ 2163 if (dst->sin_family != AF_INET) { 2164 kprintf("ip_mloopback: bad address family %d\n", 2165 dst->sin_family); 2166 dst->sin_family = AF_INET; 2167 } 2168 #endif 2169 get_mplock(); /* is if_simloop() mpsafe yet? */ 2170 if_simloop(ifp, copym, dst->sin_family, 0); 2171 rel_mplock(); 2172 } 2173 } 2174