ip_output.c revision 162084
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 * 4. 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: head/sys/netinet/ip_output.c 162084 2006-09-06 21:51:59Z andre $ 31 */ 32 33#include "opt_ipfw.h" 34#include "opt_ipsec.h" 35#include "opt_mac.h" 36#include "opt_mbuf_stress_test.h" 37 38#include <sys/param.h> 39#include <sys/systm.h> 40#include <sys/kernel.h> 41#include <sys/mac.h> 42#include <sys/malloc.h> 43#include <sys/mbuf.h> 44#include <sys/protosw.h> 45#include <sys/socket.h> 46#include <sys/socketvar.h> 47#include <sys/sysctl.h> 48 49#include <net/if.h> 50#include <net/netisr.h> 51#include <net/pfil.h> 52#include <net/route.h> 53 54#include <netinet/in.h> 55#include <netinet/in_systm.h> 56#include <netinet/ip.h> 57#include <netinet/in_pcb.h> 58#include <netinet/in_var.h> 59#include <netinet/ip_var.h> 60#include <netinet/ip_options.h> 61 62#if defined(IPSEC) || defined(FAST_IPSEC) 63#include <netinet/ip_ipsec.h> 64#ifdef IPSEC 65#include <netinet6/ipsec.h> 66#endif 67#ifdef FAST_IPSEC 68#include <netipsec/ipsec.h> 69#endif 70#endif /*IPSEC*/ 71 72#include <machine/in_cksum.h> 73 74static MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 75 76#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 77 x, (ntohl(a.s_addr)>>24)&0xFF,\ 78 (ntohl(a.s_addr)>>16)&0xFF,\ 79 (ntohl(a.s_addr)>>8)&0xFF,\ 80 (ntohl(a.s_addr))&0xFF, y); 81 82u_short ip_id; 83 84#ifdef MBUF_STRESS_TEST 85int mbuf_frag_size = 0; 86SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 87 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 88#endif 89 90static struct ifnet *ip_multicast_if(struct in_addr *, int *); 91static void ip_mloopback 92 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 93static int ip_getmoptions(struct inpcb *, struct sockopt *); 94static int ip_setmoptions(struct inpcb *, struct sockopt *); 95 96 97extern struct protosw inetsw[]; 98 99/* 100 * IP output. The packet in mbuf chain m contains a skeletal IP 101 * header (with len, off, ttl, proto, tos, src, dst). 102 * The mbuf chain containing the packet will be freed. 103 * The mbuf opt, if present, will not be freed. 104 * In the IP forwarding case, the packet will arrive with options already 105 * inserted, so must have a NULL opt pointer. 106 */ 107int 108ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, 109 int flags, struct ip_moptions *imo, struct inpcb *inp) 110{ 111 struct ip *ip; 112 struct ifnet *ifp = NULL; /* keep compiler happy */ 113 struct mbuf *m0; 114 int hlen = sizeof (struct ip); 115 int len, error = 0; 116 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 117 struct in_ifaddr *ia = NULL; 118 int isbroadcast, sw_csum; 119 struct route iproute; 120 struct in_addr odst; 121#ifdef IPFIREWALL_FORWARD 122 struct m_tag *fwd_tag = NULL; 123#endif 124 M_ASSERTPKTHDR(m); 125 126 if (ro == NULL) { 127 ro = &iproute; 128 bzero(ro, sizeof (*ro)); 129 } 130 131 if (inp != NULL) 132 INP_LOCK_ASSERT(inp); 133 134 if (opt) { 135 len = 0; 136 m = ip_insertoptions(m, opt, &len); 137 if (len != 0) 138 hlen = len; 139 } 140 ip = mtod(m, struct ip *); 141 142 /* 143 * Fill in IP header. If we are not allowing fragmentation, 144 * then the ip_id field is meaningless, but we don't set it 145 * to zero. Doing so causes various problems when devices along 146 * the path (routers, load balancers, firewalls, etc.) illegally 147 * disable DF on our packet. Note that a 16-bit counter 148 * will wrap around in less than 10 seconds at 100 Mbit/s on a 149 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 150 * for Counting NATted Hosts", Proc. IMW'02, available at 151 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 152 */ 153 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 154 ip->ip_v = IPVERSION; 155 ip->ip_hl = hlen >> 2; 156 ip->ip_id = ip_newid(); 157 ipstat.ips_localout++; 158 } else { 159 hlen = ip->ip_hl << 2; 160 } 161 162 dst = (struct sockaddr_in *)&ro->ro_dst; 163again: 164 /* 165 * If there is a cached route, 166 * check that it is to the same destination 167 * and is still up. If not, free it and try again. 168 * The address family should also be checked in case of sharing the 169 * cache with IPv6. 170 */ 171 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 172 dst->sin_family != AF_INET || 173 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 174 RTFREE(ro->ro_rt); 175 ro->ro_rt = (struct rtentry *)0; 176 } 177#ifdef IPFIREWALL_FORWARD 178 if (ro->ro_rt == NULL && fwd_tag == NULL) { 179#else 180 if (ro->ro_rt == NULL) { 181#endif 182 bzero(dst, sizeof(*dst)); 183 dst->sin_family = AF_INET; 184 dst->sin_len = sizeof(*dst); 185 dst->sin_addr = ip->ip_dst; 186 } 187 /* 188 * If routing to interface only, 189 * short circuit routing lookup. 190 */ 191 if (flags & IP_ROUTETOIF) { 192 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 193 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) { 194 ipstat.ips_noroute++; 195 error = ENETUNREACH; 196 goto bad; 197 } 198 ifp = ia->ia_ifp; 199 ip->ip_ttl = 1; 200 isbroadcast = in_broadcast(dst->sin_addr, ifp); 201 } else if (flags & IP_SENDONES) { 202 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL) { 203 ipstat.ips_noroute++; 204 error = ENETUNREACH; 205 goto bad; 206 } 207 ifp = ia->ia_ifp; 208 ip->ip_dst.s_addr = INADDR_BROADCAST; 209 dst->sin_addr = ip->ip_dst; 210 ip->ip_ttl = 1; 211 isbroadcast = 1; 212 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 213 imo != NULL && imo->imo_multicast_ifp != NULL) { 214 /* 215 * Bypass the normal routing lookup for multicast 216 * packets if the interface is specified. 217 */ 218 ifp = imo->imo_multicast_ifp; 219 IFP_TO_IA(ifp, ia); 220 isbroadcast = 0; /* fool gcc */ 221 } else { 222 /* 223 * We want to do any cloning requested by the link layer, 224 * as this is probably required in all cases for correct 225 * operation (as it is for ARP). 226 */ 227 if (ro->ro_rt == NULL) 228 rtalloc_ign(ro, 0); 229 if (ro->ro_rt == NULL) { 230 ipstat.ips_noroute++; 231 error = EHOSTUNREACH; 232 goto bad; 233 } 234 ia = ifatoia(ro->ro_rt->rt_ifa); 235 ifp = ro->ro_rt->rt_ifp; 236 ro->ro_rt->rt_rmx.rmx_pksent++; 237 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 238 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 239 if (ro->ro_rt->rt_flags & RTF_HOST) 240 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 241 else 242 isbroadcast = in_broadcast(dst->sin_addr, ifp); 243 } 244 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 245 struct in_multi *inm; 246 247 m->m_flags |= M_MCAST; 248 /* 249 * IP destination address is multicast. Make sure "dst" 250 * still points to the address in "ro". (It may have been 251 * changed to point to a gateway address, above.) 252 */ 253 dst = (struct sockaddr_in *)&ro->ro_dst; 254 /* 255 * See if the caller provided any multicast options 256 */ 257 if (imo != NULL) { 258 ip->ip_ttl = imo->imo_multicast_ttl; 259 if (imo->imo_multicast_vif != -1) 260 ip->ip_src.s_addr = 261 ip_mcast_src ? 262 ip_mcast_src(imo->imo_multicast_vif) : 263 INADDR_ANY; 264 } else 265 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 266 /* 267 * Confirm that the outgoing interface supports multicast. 268 */ 269 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 270 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 271 ipstat.ips_noroute++; 272 error = ENETUNREACH; 273 goto bad; 274 } 275 } 276 /* 277 * If source address not specified yet, use address 278 * of outgoing interface. 279 */ 280 if (ip->ip_src.s_addr == INADDR_ANY) { 281 /* Interface may have no addresses. */ 282 if (ia != NULL) 283 ip->ip_src = IA_SIN(ia)->sin_addr; 284 } 285 286 IN_MULTI_LOCK(); 287 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 288 if (inm != NULL && 289 (imo == NULL || imo->imo_multicast_loop)) { 290 IN_MULTI_UNLOCK(); 291 /* 292 * If we belong to the destination multicast group 293 * on the outgoing interface, and the caller did not 294 * forbid loopback, loop back a copy. 295 */ 296 ip_mloopback(ifp, m, dst, hlen); 297 } 298 else { 299 IN_MULTI_UNLOCK(); 300 /* 301 * If we are acting as a multicast router, perform 302 * multicast forwarding as if the packet had just 303 * arrived on the interface to which we are about 304 * to send. The multicast forwarding function 305 * recursively calls this function, using the 306 * IP_FORWARDING flag to prevent infinite recursion. 307 * 308 * Multicasts that are looped back by ip_mloopback(), 309 * above, will be forwarded by the ip_input() routine, 310 * if necessary. 311 */ 312 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 313 /* 314 * If rsvp daemon is not running, do not 315 * set ip_moptions. This ensures that the packet 316 * is multicast and not just sent down one link 317 * as prescribed by rsvpd. 318 */ 319 if (!rsvp_on) 320 imo = NULL; 321 if (ip_mforward && 322 ip_mforward(ip, ifp, m, imo) != 0) { 323 m_freem(m); 324 goto done; 325 } 326 } 327 } 328 329 /* 330 * Multicasts with a time-to-live of zero may be looped- 331 * back, above, but must not be transmitted on a network. 332 * Also, multicasts addressed to the loopback interface 333 * are not sent -- the above call to ip_mloopback() will 334 * loop back a copy if this host actually belongs to the 335 * destination group on the loopback interface. 336 */ 337 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 338 m_freem(m); 339 goto done; 340 } 341 342 goto sendit; 343 } 344#ifndef notdef 345 /* 346 * If the source address is not specified yet, use the address 347 * of the outoing interface. 348 */ 349 if (ip->ip_src.s_addr == INADDR_ANY) { 350 /* Interface may have no addresses. */ 351 if (ia != NULL) { 352 ip->ip_src = IA_SIN(ia)->sin_addr; 353 } 354 } 355#endif /* notdef */ 356 /* 357 * Verify that we have any chance at all of being able to queue the 358 * packet or packet fragments, unless ALTQ is enabled on the given 359 * interface in which case packetdrop should be done by queueing. 360 */ 361#ifdef ALTQ 362 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) && 363 ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 364 ifp->if_snd.ifq_maxlen)) 365#else 366 if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 367 ifp->if_snd.ifq_maxlen) 368#endif /* ALTQ */ 369 { 370 error = ENOBUFS; 371 ipstat.ips_odropped++; 372 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1); 373 goto bad; 374 } 375 376 /* 377 * Look for broadcast address and 378 * verify user is allowed to send 379 * such a packet. 380 */ 381 if (isbroadcast) { 382 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 383 error = EADDRNOTAVAIL; 384 goto bad; 385 } 386 if ((flags & IP_ALLOWBROADCAST) == 0) { 387 error = EACCES; 388 goto bad; 389 } 390 /* don't allow broadcast messages to be fragmented */ 391 if (ip->ip_len > ifp->if_mtu) { 392 error = EMSGSIZE; 393 goto bad; 394 } 395 m->m_flags |= M_BCAST; 396 } else { 397 m->m_flags &= ~M_BCAST; 398 } 399 400sendit: 401#if defined(IPSEC) || defined(FAST_IPSEC) 402 switch(ip_ipsec_output(&m, inp, &flags, &error, &ro, &iproute, &dst, &ia, &ifp)) { 403 case 1: 404 goto bad; 405 case -1: 406 goto done; 407 case 0: 408 default: 409 break; /* Continue with packet processing. */ 410 } 411 /* Update variables that are affected by ipsec4_output(). */ 412 ip = mtod(m, struct ip *); 413 hlen = ip->ip_hl << 2; 414#endif /* IPSEC */ 415 416 /* Jump over all PFIL processing if hooks are not active. */ 417 if (!PFIL_HOOKED(&inet_pfil_hook)) 418 goto passout; 419 420 /* Run through list of hooks for output packets. */ 421 odst.s_addr = ip->ip_dst.s_addr; 422 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 423 if (error != 0 || m == NULL) 424 goto done; 425 426 ip = mtod(m, struct ip *); 427 428 /* See if destination IP address was changed by packet filter. */ 429 if (odst.s_addr != ip->ip_dst.s_addr) { 430 m->m_flags |= M_SKIP_FIREWALL; 431 /* If destination is now ourself drop to ip_input(). */ 432 if (in_localip(ip->ip_dst)) { 433 m->m_flags |= M_FASTFWD_OURS; 434 if (m->m_pkthdr.rcvif == NULL) 435 m->m_pkthdr.rcvif = loif; 436 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 437 m->m_pkthdr.csum_flags |= 438 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 439 m->m_pkthdr.csum_data = 0xffff; 440 } 441 m->m_pkthdr.csum_flags |= 442 CSUM_IP_CHECKED | CSUM_IP_VALID; 443 444 error = netisr_queue(NETISR_IP, m); 445 goto done; 446 } else 447 goto again; /* Redo the routing table lookup. */ 448 } 449 450#ifdef IPFIREWALL_FORWARD 451 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 452 if (m->m_flags & M_FASTFWD_OURS) { 453 if (m->m_pkthdr.rcvif == NULL) 454 m->m_pkthdr.rcvif = loif; 455 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 456 m->m_pkthdr.csum_flags |= 457 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 458 m->m_pkthdr.csum_data = 0xffff; 459 } 460 m->m_pkthdr.csum_flags |= 461 CSUM_IP_CHECKED | CSUM_IP_VALID; 462 463 error = netisr_queue(NETISR_IP, m); 464 goto done; 465 } 466 /* Or forward to some other address? */ 467 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 468 if (fwd_tag) { 469 dst = (struct sockaddr_in *)&ro->ro_dst; 470 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 471 m->m_flags |= M_SKIP_FIREWALL; 472 m_tag_delete(m, fwd_tag); 473 goto again; 474 } 475#endif /* IPFIREWALL_FORWARD */ 476 477passout: 478 /* 127/8 must not appear on wire - RFC1122. */ 479 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 480 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 481 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 482 ipstat.ips_badaddr++; 483 error = EADDRNOTAVAIL; 484 goto bad; 485 } 486 } 487 488 m->m_pkthdr.csum_flags |= CSUM_IP; 489 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 490 if (sw_csum & CSUM_DELAY_DATA) { 491 in_delayed_cksum(m); 492 sw_csum &= ~CSUM_DELAY_DATA; 493 } 494 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 495 496 /* 497 * If small enough for interface, or the interface will take 498 * care of the fragmentation for us, we can just send directly. 499 */ 500 if (ip->ip_len <= ifp->if_mtu || 501 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 502 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 503 ip->ip_len = htons(ip->ip_len); 504 ip->ip_off = htons(ip->ip_off); 505 ip->ip_sum = 0; 506 if (sw_csum & CSUM_DELAY_IP) 507 ip->ip_sum = in_cksum(m, hlen); 508 509 /* 510 * Record statistics for this interface address. 511 * With CSUM_TSO the byte/packet count will be slightly 512 * incorrect because we count the IP+TCP headers only 513 * once instead of for every generated packet. 514 */ 515 if (!(flags & IP_FORWARDING) && ia) { 516 if (m->m_pkthdr.csum_flags & CSUM_TSO) 517 ia->ia_ifa.if_opackets += 518 m->m_pkthdr.len / m->m_pkthdr.tso_segsz; 519 else 520 ia->ia_ifa.if_opackets++; 521 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 522 } 523#ifdef IPSEC 524 /* clean ipsec history once it goes out of the node */ 525 ipsec_delaux(m); 526#endif 527#ifdef MBUF_STRESS_TEST 528 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 529 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 530#endif 531 /* 532 * Reset layer specific mbuf flags 533 * to avoid confusing lower layers. 534 */ 535 m->m_flags &= ~(M_PROTOFLAGS); 536 537 error = (*ifp->if_output)(ifp, m, 538 (struct sockaddr *)dst, ro->ro_rt); 539 goto done; 540 } 541 542 /* Balk when DF bit is set or the interface didn't support TSO. */ 543 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 544 error = EMSGSIZE; 545 /* 546 * This case can happen if the user changed the MTU 547 * of an interface after enabling IP on it. Because 548 * most netifs don't keep track of routes pointing to 549 * them, there is no way for one to update all its 550 * routes when the MTU is changed. 551 */ 552 if (ro != NULL && 553 (ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) && 554 (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 555 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 556 } 557 ipstat.ips_cantfrag++; 558 goto bad; 559 } 560 561 /* 562 * Too large for interface; fragment if possible. If successful, 563 * on return, m will point to a list of packets to be sent. 564 */ 565 error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist, sw_csum); 566 if (error) 567 goto bad; 568 for (; m; m = m0) { 569 m0 = m->m_nextpkt; 570 m->m_nextpkt = 0; 571#ifdef IPSEC 572 /* clean ipsec history once it goes out of the node */ 573 ipsec_delaux(m); 574#endif 575 if (error == 0) { 576 /* Record statistics for this interface address. */ 577 if (ia != NULL) { 578 ia->ia_ifa.if_opackets++; 579 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 580 } 581 /* 582 * Reset layer specific mbuf flags 583 * to avoid confusing upper layers. 584 */ 585 m->m_flags &= ~(M_PROTOFLAGS); 586 587 error = (*ifp->if_output)(ifp, m, 588 (struct sockaddr *)dst, ro->ro_rt); 589 } else 590 m_freem(m); 591 } 592 593 if (error == 0) 594 ipstat.ips_fragmented++; 595 596done: 597 if (ro == &iproute && ro->ro_rt) { 598 RTFREE(ro->ro_rt); 599 } 600 return (error); 601bad: 602 m_freem(m); 603 goto done; 604} 605 606/* 607 * Create a chain of fragments which fit the given mtu. m_frag points to the 608 * mbuf to be fragmented; on return it points to the chain with the fragments. 609 * Return 0 if no error. If error, m_frag may contain a partially built 610 * chain of fragments that should be freed by the caller. 611 * 612 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 613 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 614 */ 615int 616ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 617 u_long if_hwassist_flags, int sw_csum) 618{ 619 int error = 0; 620 int hlen = ip->ip_hl << 2; 621 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 622 int off; 623 struct mbuf *m0 = *m_frag; /* the original packet */ 624 int firstlen; 625 struct mbuf **mnext; 626 int nfrags; 627 628 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 629 ipstat.ips_cantfrag++; 630 return EMSGSIZE; 631 } 632 633 /* 634 * Must be able to put at least 8 bytes per fragment. 635 */ 636 if (len < 8) 637 return EMSGSIZE; 638 639 /* 640 * If the interface will not calculate checksums on 641 * fragmented packets, then do it here. 642 */ 643 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 644 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 645 in_delayed_cksum(m0); 646 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 647 } 648 649 if (len > PAGE_SIZE) { 650 /* 651 * Fragment large datagrams such that each segment 652 * contains a multiple of PAGE_SIZE amount of data, 653 * plus headers. This enables a receiver to perform 654 * page-flipping zero-copy optimizations. 655 * 656 * XXX When does this help given that sender and receiver 657 * could have different page sizes, and also mtu could 658 * be less than the receiver's page size ? 659 */ 660 int newlen; 661 struct mbuf *m; 662 663 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 664 off += m->m_len; 665 666 /* 667 * firstlen (off - hlen) must be aligned on an 668 * 8-byte boundary 669 */ 670 if (off < hlen) 671 goto smart_frag_failure; 672 off = ((off - hlen) & ~7) + hlen; 673 newlen = (~PAGE_MASK) & mtu; 674 if ((newlen + sizeof (struct ip)) > mtu) { 675 /* we failed, go back the default */ 676smart_frag_failure: 677 newlen = len; 678 off = hlen + len; 679 } 680 len = newlen; 681 682 } else { 683 off = hlen + len; 684 } 685 686 firstlen = off - hlen; 687 mnext = &m0->m_nextpkt; /* pointer to next packet */ 688 689 /* 690 * Loop through length of segment after first fragment, 691 * make new header and copy data of each part and link onto chain. 692 * Here, m0 is the original packet, m is the fragment being created. 693 * The fragments are linked off the m_nextpkt of the original 694 * packet, which after processing serves as the first fragment. 695 */ 696 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 697 struct ip *mhip; /* ip header on the fragment */ 698 struct mbuf *m; 699 int mhlen = sizeof (struct ip); 700 701 MGETHDR(m, M_DONTWAIT, MT_DATA); 702 if (m == NULL) { 703 error = ENOBUFS; 704 ipstat.ips_odropped++; 705 goto done; 706 } 707 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 708 /* 709 * In the first mbuf, leave room for the link header, then 710 * copy the original IP header including options. The payload 711 * goes into an additional mbuf chain returned by m_copy(). 712 */ 713 m->m_data += max_linkhdr; 714 mhip = mtod(m, struct ip *); 715 *mhip = *ip; 716 if (hlen > sizeof (struct ip)) { 717 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 718 mhip->ip_v = IPVERSION; 719 mhip->ip_hl = mhlen >> 2; 720 } 721 m->m_len = mhlen; 722 /* XXX do we need to add ip->ip_off below ? */ 723 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 724 if (off + len >= ip->ip_len) { /* last fragment */ 725 len = ip->ip_len - off; 726 m->m_flags |= M_LASTFRAG; 727 } else 728 mhip->ip_off |= IP_MF; 729 mhip->ip_len = htons((u_short)(len + mhlen)); 730 m->m_next = m_copy(m0, off, len); 731 if (m->m_next == NULL) { /* copy failed */ 732 m_free(m); 733 error = ENOBUFS; /* ??? */ 734 ipstat.ips_odropped++; 735 goto done; 736 } 737 m->m_pkthdr.len = mhlen + len; 738 m->m_pkthdr.rcvif = NULL; 739#ifdef MAC 740 mac_create_fragment(m0, m); 741#endif 742 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 743 mhip->ip_off = htons(mhip->ip_off); 744 mhip->ip_sum = 0; 745 if (sw_csum & CSUM_DELAY_IP) 746 mhip->ip_sum = in_cksum(m, mhlen); 747 *mnext = m; 748 mnext = &m->m_nextpkt; 749 } 750 ipstat.ips_ofragments += nfrags; 751 752 /* set first marker for fragment chain */ 753 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 754 m0->m_pkthdr.csum_data = nfrags; 755 756 /* 757 * Update first fragment by trimming what's been copied out 758 * and updating header. 759 */ 760 m_adj(m0, hlen + firstlen - ip->ip_len); 761 m0->m_pkthdr.len = hlen + firstlen; 762 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 763 ip->ip_off |= IP_MF; 764 ip->ip_off = htons(ip->ip_off); 765 ip->ip_sum = 0; 766 if (sw_csum & CSUM_DELAY_IP) 767 ip->ip_sum = in_cksum(m0, hlen); 768 769done: 770 *m_frag = m0; 771 return error; 772} 773 774void 775in_delayed_cksum(struct mbuf *m) 776{ 777 struct ip *ip; 778 u_short csum, offset; 779 780 ip = mtod(m, struct ip *); 781 offset = ip->ip_hl << 2 ; 782 csum = in_cksum_skip(m, ip->ip_len, offset); 783 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 784 csum = 0xffff; 785 offset += m->m_pkthdr.csum_data; /* checksum offset */ 786 787 if (offset + sizeof(u_short) > m->m_len) { 788 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 789 m->m_len, offset, ip->ip_p); 790 /* 791 * XXX 792 * this shouldn't happen, but if it does, the 793 * correct behavior may be to insert the checksum 794 * in the appropriate next mbuf in the chain. 795 */ 796 return; 797 } 798 *(u_short *)(m->m_data + offset) = csum; 799} 800 801/* 802 * IP socket option processing. 803 */ 804int 805ip_ctloutput(so, sopt) 806 struct socket *so; 807 struct sockopt *sopt; 808{ 809 struct inpcb *inp = sotoinpcb(so); 810 int error, optval; 811 812 error = optval = 0; 813 if (sopt->sopt_level != IPPROTO_IP) { 814 return (EINVAL); 815 } 816 817 switch (sopt->sopt_dir) { 818 case SOPT_SET: 819 switch (sopt->sopt_name) { 820 case IP_OPTIONS: 821#ifdef notyet 822 case IP_RETOPTS: 823#endif 824 { 825 struct mbuf *m; 826 if (sopt->sopt_valsize > MLEN) { 827 error = EMSGSIZE; 828 break; 829 } 830 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 831 if (m == NULL) { 832 error = ENOBUFS; 833 break; 834 } 835 m->m_len = sopt->sopt_valsize; 836 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 837 m->m_len); 838 if (error) { 839 m_free(m); 840 break; 841 } 842 INP_LOCK(inp); 843 error = ip_pcbopts(inp, sopt->sopt_name, m); 844 INP_UNLOCK(inp); 845 return (error); 846 } 847 848 case IP_TOS: 849 case IP_TTL: 850 case IP_MINTTL: 851 case IP_RECVOPTS: 852 case IP_RECVRETOPTS: 853 case IP_RECVDSTADDR: 854 case IP_RECVTTL: 855 case IP_RECVIF: 856 case IP_FAITH: 857 case IP_ONESBCAST: 858 case IP_DONTFRAG: 859 error = sooptcopyin(sopt, &optval, sizeof optval, 860 sizeof optval); 861 if (error) 862 break; 863 864 switch (sopt->sopt_name) { 865 case IP_TOS: 866 inp->inp_ip_tos = optval; 867 break; 868 869 case IP_TTL: 870 inp->inp_ip_ttl = optval; 871 break; 872 873 case IP_MINTTL: 874 if (optval > 0 && optval <= MAXTTL) 875 inp->inp_ip_minttl = optval; 876 else 877 error = EINVAL; 878 break; 879 880#define OPTSET(bit) do { \ 881 INP_LOCK(inp); \ 882 if (optval) \ 883 inp->inp_flags |= bit; \ 884 else \ 885 inp->inp_flags &= ~bit; \ 886 INP_UNLOCK(inp); \ 887} while (0) 888 889 case IP_RECVOPTS: 890 OPTSET(INP_RECVOPTS); 891 break; 892 893 case IP_RECVRETOPTS: 894 OPTSET(INP_RECVRETOPTS); 895 break; 896 897 case IP_RECVDSTADDR: 898 OPTSET(INP_RECVDSTADDR); 899 break; 900 901 case IP_RECVTTL: 902 OPTSET(INP_RECVTTL); 903 break; 904 905 case IP_RECVIF: 906 OPTSET(INP_RECVIF); 907 break; 908 909 case IP_FAITH: 910 OPTSET(INP_FAITH); 911 break; 912 913 case IP_ONESBCAST: 914 OPTSET(INP_ONESBCAST); 915 break; 916 case IP_DONTFRAG: 917 OPTSET(INP_DONTFRAG); 918 break; 919 } 920 break; 921#undef OPTSET 922 923 case IP_MULTICAST_IF: 924 case IP_MULTICAST_VIF: 925 case IP_MULTICAST_TTL: 926 case IP_MULTICAST_LOOP: 927 case IP_ADD_MEMBERSHIP: 928 case IP_DROP_MEMBERSHIP: 929 error = ip_setmoptions(inp, sopt); 930 break; 931 932 case IP_PORTRANGE: 933 error = sooptcopyin(sopt, &optval, sizeof optval, 934 sizeof optval); 935 if (error) 936 break; 937 938 INP_LOCK(inp); 939 switch (optval) { 940 case IP_PORTRANGE_DEFAULT: 941 inp->inp_flags &= ~(INP_LOWPORT); 942 inp->inp_flags &= ~(INP_HIGHPORT); 943 break; 944 945 case IP_PORTRANGE_HIGH: 946 inp->inp_flags &= ~(INP_LOWPORT); 947 inp->inp_flags |= INP_HIGHPORT; 948 break; 949 950 case IP_PORTRANGE_LOW: 951 inp->inp_flags &= ~(INP_HIGHPORT); 952 inp->inp_flags |= INP_LOWPORT; 953 break; 954 955 default: 956 error = EINVAL; 957 break; 958 } 959 INP_UNLOCK(inp); 960 break; 961 962#if defined(IPSEC) || defined(FAST_IPSEC) 963 case IP_IPSEC_POLICY: 964 { 965 caddr_t req; 966 size_t len = 0; 967 int priv; 968 struct mbuf *m; 969 int optname; 970 971 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 972 break; 973 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 974 break; 975 priv = (sopt->sopt_td != NULL && 976 suser(sopt->sopt_td) != 0) ? 0 : 1; 977 req = mtod(m, caddr_t); 978 len = m->m_len; 979 optname = sopt->sopt_name; 980 error = ipsec4_set_policy(inp, optname, req, len, priv); 981 m_freem(m); 982 break; 983 } 984#endif /*IPSEC*/ 985 986 default: 987 error = ENOPROTOOPT; 988 break; 989 } 990 break; 991 992 case SOPT_GET: 993 switch (sopt->sopt_name) { 994 case IP_OPTIONS: 995 case IP_RETOPTS: 996 if (inp->inp_options) 997 error = sooptcopyout(sopt, 998 mtod(inp->inp_options, 999 char *), 1000 inp->inp_options->m_len); 1001 else 1002 sopt->sopt_valsize = 0; 1003 break; 1004 1005 case IP_TOS: 1006 case IP_TTL: 1007 case IP_MINTTL: 1008 case IP_RECVOPTS: 1009 case IP_RECVRETOPTS: 1010 case IP_RECVDSTADDR: 1011 case IP_RECVTTL: 1012 case IP_RECVIF: 1013 case IP_PORTRANGE: 1014 case IP_FAITH: 1015 case IP_ONESBCAST: 1016 case IP_DONTFRAG: 1017 switch (sopt->sopt_name) { 1018 1019 case IP_TOS: 1020 optval = inp->inp_ip_tos; 1021 break; 1022 1023 case IP_TTL: 1024 optval = inp->inp_ip_ttl; 1025 break; 1026 1027 case IP_MINTTL: 1028 optval = inp->inp_ip_minttl; 1029 break; 1030 1031#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1032 1033 case IP_RECVOPTS: 1034 optval = OPTBIT(INP_RECVOPTS); 1035 break; 1036 1037 case IP_RECVRETOPTS: 1038 optval = OPTBIT(INP_RECVRETOPTS); 1039 break; 1040 1041 case IP_RECVDSTADDR: 1042 optval = OPTBIT(INP_RECVDSTADDR); 1043 break; 1044 1045 case IP_RECVTTL: 1046 optval = OPTBIT(INP_RECVTTL); 1047 break; 1048 1049 case IP_RECVIF: 1050 optval = OPTBIT(INP_RECVIF); 1051 break; 1052 1053 case IP_PORTRANGE: 1054 if (inp->inp_flags & INP_HIGHPORT) 1055 optval = IP_PORTRANGE_HIGH; 1056 else if (inp->inp_flags & INP_LOWPORT) 1057 optval = IP_PORTRANGE_LOW; 1058 else 1059 optval = 0; 1060 break; 1061 1062 case IP_FAITH: 1063 optval = OPTBIT(INP_FAITH); 1064 break; 1065 1066 case IP_ONESBCAST: 1067 optval = OPTBIT(INP_ONESBCAST); 1068 break; 1069 case IP_DONTFRAG: 1070 optval = OPTBIT(INP_DONTFRAG); 1071 break; 1072 } 1073 error = sooptcopyout(sopt, &optval, sizeof optval); 1074 break; 1075 1076 case IP_MULTICAST_IF: 1077 case IP_MULTICAST_VIF: 1078 case IP_MULTICAST_TTL: 1079 case IP_MULTICAST_LOOP: 1080 case IP_ADD_MEMBERSHIP: 1081 case IP_DROP_MEMBERSHIP: 1082 error = ip_getmoptions(inp, sopt); 1083 break; 1084 1085#if defined(IPSEC) || defined(FAST_IPSEC) 1086 case IP_IPSEC_POLICY: 1087 { 1088 struct mbuf *m = NULL; 1089 caddr_t req = NULL; 1090 size_t len = 0; 1091 1092 if (m != 0) { 1093 req = mtod(m, caddr_t); 1094 len = m->m_len; 1095 } 1096 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1097 if (error == 0) 1098 error = soopt_mcopyout(sopt, m); /* XXX */ 1099 if (error == 0) 1100 m_freem(m); 1101 break; 1102 } 1103#endif /*IPSEC*/ 1104 1105 default: 1106 error = ENOPROTOOPT; 1107 break; 1108 } 1109 break; 1110 } 1111 return (error); 1112} 1113 1114/* 1115 * XXX 1116 * The whole multicast option thing needs to be re-thought. 1117 * Several of these options are equally applicable to non-multicast 1118 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1119 * standard option (IP_TTL). 1120 */ 1121 1122/* 1123 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1124 */ 1125static struct ifnet * 1126ip_multicast_if(a, ifindexp) 1127 struct in_addr *a; 1128 int *ifindexp; 1129{ 1130 int ifindex; 1131 struct ifnet *ifp; 1132 1133 if (ifindexp) 1134 *ifindexp = 0; 1135 if (ntohl(a->s_addr) >> 24 == 0) { 1136 ifindex = ntohl(a->s_addr) & 0xffffff; 1137 if (ifindex < 0 || if_index < ifindex) 1138 return NULL; 1139 ifp = ifnet_byindex(ifindex); 1140 if (ifindexp) 1141 *ifindexp = ifindex; 1142 } else { 1143 INADDR_TO_IFP(*a, ifp); 1144 } 1145 return ifp; 1146} 1147 1148/* 1149 * Given an inpcb, return its multicast options structure pointer. Accepts 1150 * an unlocked inpcb pointer, but will return it locked. May sleep. 1151 */ 1152static struct ip_moptions * 1153ip_findmoptions(struct inpcb *inp) 1154{ 1155 struct ip_moptions *imo; 1156 struct in_multi **immp; 1157 1158 INP_LOCK(inp); 1159 if (inp->inp_moptions != NULL) 1160 return (inp->inp_moptions); 1161 1162 INP_UNLOCK(inp); 1163 1164 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, M_WAITOK); 1165 immp = (struct in_multi **)malloc((sizeof(*immp) * IP_MIN_MEMBERSHIPS), 1166 M_IPMOPTS, M_WAITOK); 1167 1168 imo->imo_multicast_ifp = NULL; 1169 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1170 imo->imo_multicast_vif = -1; 1171 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1172 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1173 imo->imo_num_memberships = 0; 1174 imo->imo_max_memberships = IP_MIN_MEMBERSHIPS; 1175 imo->imo_membership = immp; 1176 1177 INP_LOCK(inp); 1178 if (inp->inp_moptions != NULL) { 1179 free(immp, M_IPMOPTS); 1180 free(imo, M_IPMOPTS); 1181 return (inp->inp_moptions); 1182 } 1183 inp->inp_moptions = imo; 1184 return (imo); 1185} 1186 1187/* 1188 * Set the IP multicast options in response to user setsockopt(). 1189 */ 1190static int 1191ip_setmoptions(struct inpcb *inp, struct sockopt *sopt) 1192{ 1193 int error = 0; 1194 int i; 1195 struct in_addr addr; 1196 struct ip_mreq mreq; 1197 struct ifnet *ifp; 1198 struct ip_moptions *imo; 1199 struct route ro; 1200 struct sockaddr_in *dst; 1201 int ifindex; 1202 int s; 1203 1204 switch (sopt->sopt_name) { 1205 /* store an index number for the vif you wanna use in the send */ 1206 case IP_MULTICAST_VIF: 1207 if (legal_vif_num == 0) { 1208 error = EOPNOTSUPP; 1209 break; 1210 } 1211 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1212 if (error) 1213 break; 1214 if (!legal_vif_num(i) && (i != -1)) { 1215 error = EINVAL; 1216 break; 1217 } 1218 imo = ip_findmoptions(inp); 1219 imo->imo_multicast_vif = i; 1220 INP_UNLOCK(inp); 1221 break; 1222 1223 case IP_MULTICAST_IF: 1224 /* 1225 * Select the interface for outgoing multicast packets. 1226 */ 1227 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1228 if (error) 1229 break; 1230 /* 1231 * INADDR_ANY is used to remove a previous selection. 1232 * When no interface is selected, a default one is 1233 * chosen every time a multicast packet is sent. 1234 */ 1235 imo = ip_findmoptions(inp); 1236 if (addr.s_addr == INADDR_ANY) { 1237 imo->imo_multicast_ifp = NULL; 1238 INP_UNLOCK(inp); 1239 break; 1240 } 1241 /* 1242 * The selected interface is identified by its local 1243 * IP address. Find the interface and confirm that 1244 * it supports multicasting. 1245 */ 1246 s = splimp(); 1247 ifp = ip_multicast_if(&addr, &ifindex); 1248 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1249 INP_UNLOCK(inp); 1250 splx(s); 1251 error = EADDRNOTAVAIL; 1252 break; 1253 } 1254 imo->imo_multicast_ifp = ifp; 1255 if (ifindex) 1256 imo->imo_multicast_addr = addr; 1257 else 1258 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1259 INP_UNLOCK(inp); 1260 splx(s); 1261 break; 1262 1263 case IP_MULTICAST_TTL: 1264 /* 1265 * Set the IP time-to-live for outgoing multicast packets. 1266 * The original multicast API required a char argument, 1267 * which is inconsistent with the rest of the socket API. 1268 * We allow either a char or an int. 1269 */ 1270 if (sopt->sopt_valsize == 1) { 1271 u_char ttl; 1272 error = sooptcopyin(sopt, &ttl, 1, 1); 1273 if (error) 1274 break; 1275 imo = ip_findmoptions(inp); 1276 imo->imo_multicast_ttl = ttl; 1277 INP_UNLOCK(inp); 1278 } else { 1279 u_int ttl; 1280 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1281 sizeof ttl); 1282 if (error) 1283 break; 1284 if (ttl > 255) 1285 error = EINVAL; 1286 else { 1287 imo = ip_findmoptions(inp); 1288 imo->imo_multicast_ttl = ttl; 1289 INP_UNLOCK(inp); 1290 } 1291 } 1292 break; 1293 1294 case IP_MULTICAST_LOOP: 1295 /* 1296 * Set the loopback flag for outgoing multicast packets. 1297 * Must be zero or one. The original multicast API required a 1298 * char argument, which is inconsistent with the rest 1299 * of the socket API. We allow either a char or an int. 1300 */ 1301 if (sopt->sopt_valsize == 1) { 1302 u_char loop; 1303 error = sooptcopyin(sopt, &loop, 1, 1); 1304 if (error) 1305 break; 1306 imo = ip_findmoptions(inp); 1307 imo->imo_multicast_loop = !!loop; 1308 INP_UNLOCK(inp); 1309 } else { 1310 u_int loop; 1311 error = sooptcopyin(sopt, &loop, sizeof loop, 1312 sizeof loop); 1313 if (error) 1314 break; 1315 imo = ip_findmoptions(inp); 1316 imo->imo_multicast_loop = !!loop; 1317 INP_UNLOCK(inp); 1318 } 1319 break; 1320 1321 case IP_ADD_MEMBERSHIP: 1322 /* 1323 * Add a multicast group membership. 1324 * Group must be a valid IP multicast address. 1325 */ 1326 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1327 if (error) 1328 break; 1329 1330 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1331 error = EINVAL; 1332 break; 1333 } 1334 s = splimp(); 1335 /* 1336 * If no interface address was provided, use the interface of 1337 * the route to the given multicast address. 1338 */ 1339 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1340 bzero((caddr_t)&ro, sizeof(ro)); 1341 dst = (struct sockaddr_in *)&ro.ro_dst; 1342 dst->sin_len = sizeof(*dst); 1343 dst->sin_family = AF_INET; 1344 dst->sin_addr = mreq.imr_multiaddr; 1345 rtalloc_ign(&ro, RTF_CLONING); 1346 if (ro.ro_rt == NULL) { 1347 error = EADDRNOTAVAIL; 1348 splx(s); 1349 break; 1350 } 1351 ifp = ro.ro_rt->rt_ifp; 1352 RTFREE(ro.ro_rt); 1353 } 1354 else { 1355 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1356 } 1357 1358 /* 1359 * See if we found an interface, and confirm that it 1360 * supports multicast. 1361 */ 1362 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1363 error = EADDRNOTAVAIL; 1364 splx(s); 1365 break; 1366 } 1367 /* 1368 * See if the membership already exists or if all the 1369 * membership slots are full. 1370 */ 1371 imo = ip_findmoptions(inp); 1372 for (i = 0; i < imo->imo_num_memberships; ++i) { 1373 if (imo->imo_membership[i]->inm_ifp == ifp && 1374 imo->imo_membership[i]->inm_addr.s_addr 1375 == mreq.imr_multiaddr.s_addr) 1376 break; 1377 } 1378 if (i < imo->imo_num_memberships) { 1379 INP_UNLOCK(inp); 1380 error = EADDRINUSE; 1381 splx(s); 1382 break; 1383 } 1384 if (imo->imo_num_memberships == imo->imo_max_memberships) { 1385 struct in_multi **nmships, **omships; 1386 size_t newmax; 1387 /* 1388 * Resize the vector to next power-of-two minus 1. If the 1389 * size would exceed the maximum then we know we've really 1390 * run out of entries. Otherwise, we realloc() the vector 1391 * with the INP lock held to avoid introducing a race. 1392 */ 1393 nmships = NULL; 1394 omships = imo->imo_membership; 1395 newmax = ((imo->imo_max_memberships + 1) * 2) - 1; 1396 if (newmax <= IP_MAX_MEMBERSHIPS) { 1397 nmships = (struct in_multi **)realloc(omships, 1398sizeof(*nmships) * newmax, M_IPMOPTS, M_NOWAIT); 1399 if (nmships != NULL) { 1400 imo->imo_membership = nmships; 1401 imo->imo_max_memberships = newmax; 1402 } 1403 } 1404 if (nmships == NULL) { 1405 INP_UNLOCK(inp); 1406 error = ETOOMANYREFS; 1407 splx(s); 1408 break; 1409 } 1410 } 1411 /* 1412 * Everything looks good; add a new record to the multicast 1413 * address list for the given interface. 1414 */ 1415 if ((imo->imo_membership[i] = 1416 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1417 INP_UNLOCK(inp); 1418 error = ENOBUFS; 1419 splx(s); 1420 break; 1421 } 1422 ++imo->imo_num_memberships; 1423 INP_UNLOCK(inp); 1424 splx(s); 1425 break; 1426 1427 case IP_DROP_MEMBERSHIP: 1428 /* 1429 * Drop a multicast group membership. 1430 * Group must be a valid IP multicast address. 1431 */ 1432 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1433 if (error) 1434 break; 1435 1436 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1437 error = EINVAL; 1438 break; 1439 } 1440 1441 s = splimp(); 1442 /* 1443 * If an interface address was specified, get a pointer 1444 * to its ifnet structure. 1445 */ 1446 if (mreq.imr_interface.s_addr == INADDR_ANY) 1447 ifp = NULL; 1448 else { 1449 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1450 if (ifp == NULL) { 1451 error = EADDRNOTAVAIL; 1452 splx(s); 1453 break; 1454 } 1455 } 1456 /* 1457 * Find the membership in the membership array. 1458 */ 1459 imo = ip_findmoptions(inp); 1460 for (i = 0; i < imo->imo_num_memberships; ++i) { 1461 if ((ifp == NULL || 1462 imo->imo_membership[i]->inm_ifp == ifp) && 1463 imo->imo_membership[i]->inm_addr.s_addr == 1464 mreq.imr_multiaddr.s_addr) 1465 break; 1466 } 1467 if (i == imo->imo_num_memberships) { 1468 INP_UNLOCK(inp); 1469 error = EADDRNOTAVAIL; 1470 splx(s); 1471 break; 1472 } 1473 /* 1474 * Give up the multicast address record to which the 1475 * membership points. 1476 */ 1477 in_delmulti(imo->imo_membership[i]); 1478 /* 1479 * Remove the gap in the membership array. 1480 */ 1481 for (++i; i < imo->imo_num_memberships; ++i) 1482 imo->imo_membership[i-1] = imo->imo_membership[i]; 1483 --imo->imo_num_memberships; 1484 INP_UNLOCK(inp); 1485 splx(s); 1486 break; 1487 1488 default: 1489 error = EOPNOTSUPP; 1490 break; 1491 } 1492 1493 return (error); 1494} 1495 1496/* 1497 * Return the IP multicast options in response to user getsockopt(). 1498 */ 1499static int 1500ip_getmoptions(struct inpcb *inp, struct sockopt *sopt) 1501{ 1502 struct ip_moptions *imo; 1503 struct in_addr addr; 1504 struct in_ifaddr *ia; 1505 int error, optval; 1506 u_char coptval; 1507 1508 INP_LOCK(inp); 1509 imo = inp->inp_moptions; 1510 1511 error = 0; 1512 switch (sopt->sopt_name) { 1513 case IP_MULTICAST_VIF: 1514 if (imo != NULL) 1515 optval = imo->imo_multicast_vif; 1516 else 1517 optval = -1; 1518 INP_UNLOCK(inp); 1519 error = sooptcopyout(sopt, &optval, sizeof optval); 1520 break; 1521 1522 case IP_MULTICAST_IF: 1523 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1524 addr.s_addr = INADDR_ANY; 1525 else if (imo->imo_multicast_addr.s_addr) { 1526 /* return the value user has set */ 1527 addr = imo->imo_multicast_addr; 1528 } else { 1529 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1530 addr.s_addr = (ia == NULL) ? INADDR_ANY 1531 : IA_SIN(ia)->sin_addr.s_addr; 1532 } 1533 INP_UNLOCK(inp); 1534 error = sooptcopyout(sopt, &addr, sizeof addr); 1535 break; 1536 1537 case IP_MULTICAST_TTL: 1538 if (imo == 0) 1539 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1540 else 1541 optval = coptval = imo->imo_multicast_ttl; 1542 INP_UNLOCK(inp); 1543 if (sopt->sopt_valsize == 1) 1544 error = sooptcopyout(sopt, &coptval, 1); 1545 else 1546 error = sooptcopyout(sopt, &optval, sizeof optval); 1547 break; 1548 1549 case IP_MULTICAST_LOOP: 1550 if (imo == 0) 1551 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1552 else 1553 optval = coptval = imo->imo_multicast_loop; 1554 INP_UNLOCK(inp); 1555 if (sopt->sopt_valsize == 1) 1556 error = sooptcopyout(sopt, &coptval, 1); 1557 else 1558 error = sooptcopyout(sopt, &optval, sizeof optval); 1559 break; 1560 1561 default: 1562 INP_UNLOCK(inp); 1563 error = ENOPROTOOPT; 1564 break; 1565 } 1566 INP_UNLOCK_ASSERT(inp); 1567 1568 return (error); 1569} 1570 1571/* 1572 * Discard the IP multicast options. 1573 */ 1574void 1575ip_freemoptions(imo) 1576 register struct ip_moptions *imo; 1577{ 1578 register int i; 1579 1580 if (imo != NULL) { 1581 for (i = 0; i < imo->imo_num_memberships; ++i) 1582 in_delmulti(imo->imo_membership[i]); 1583 free(imo->imo_membership, M_IPMOPTS); 1584 free(imo, M_IPMOPTS); 1585 } 1586} 1587 1588/* 1589 * Routine called from ip_output() to loop back a copy of an IP multicast 1590 * packet to the input queue of a specified interface. Note that this 1591 * calls the output routine of the loopback "driver", but with an interface 1592 * pointer that might NOT be a loopback interface -- evil, but easier than 1593 * replicating that code here. 1594 */ 1595static void 1596ip_mloopback(ifp, m, dst, hlen) 1597 struct ifnet *ifp; 1598 register struct mbuf *m; 1599 register struct sockaddr_in *dst; 1600 int hlen; 1601{ 1602 register struct ip *ip; 1603 struct mbuf *copym; 1604 1605 copym = m_copy(m, 0, M_COPYALL); 1606 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1607 copym = m_pullup(copym, hlen); 1608 if (copym != NULL) { 1609 /* If needed, compute the checksum and mark it as valid. */ 1610 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1611 in_delayed_cksum(copym); 1612 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1613 copym->m_pkthdr.csum_flags |= 1614 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1615 copym->m_pkthdr.csum_data = 0xffff; 1616 } 1617 /* 1618 * We don't bother to fragment if the IP length is greater 1619 * than the interface's MTU. Can this possibly matter? 1620 */ 1621 ip = mtod(copym, struct ip *); 1622 ip->ip_len = htons(ip->ip_len); 1623 ip->ip_off = htons(ip->ip_off); 1624 ip->ip_sum = 0; 1625 ip->ip_sum = in_cksum(copym, hlen); 1626 /* 1627 * NB: 1628 * It's not clear whether there are any lingering 1629 * reentrancy problems in other areas which might 1630 * be exposed by using ip_input directly (in 1631 * particular, everything which modifies the packet 1632 * in-place). Yet another option is using the 1633 * protosw directly to deliver the looped back 1634 * packet. For the moment, we'll err on the side 1635 * of safety by using if_simloop(). 1636 */ 1637#if 1 /* XXX */ 1638 if (dst->sin_family != AF_INET) { 1639 printf("ip_mloopback: bad address family %d\n", 1640 dst->sin_family); 1641 dst->sin_family = AF_INET; 1642 } 1643#endif 1644 1645#ifdef notdef 1646 copym->m_pkthdr.rcvif = ifp; 1647 ip_input(copym); 1648#else 1649 if_simloop(ifp, copym, dst->sin_family, 0); 1650#endif 1651 } 1652} 1653