ip_output.c revision 162231
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 162231 2006-09-11 19:56:10Z 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 mtu; 116 int len, error = 0; 117 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 118 struct in_ifaddr *ia = NULL; 119 int isbroadcast, sw_csum; 120 struct route iproute; 121 struct in_addr odst; 122#ifdef IPFIREWALL_FORWARD 123 struct m_tag *fwd_tag = NULL; 124#endif 125 M_ASSERTPKTHDR(m); 126 127 if (ro == NULL) { 128 ro = &iproute; 129 bzero(ro, sizeof (*ro)); 130 } 131 132 if (inp != NULL) 133 INP_LOCK_ASSERT(inp); 134 135 if (opt) { 136 len = 0; 137 m = ip_insertoptions(m, opt, &len); 138 if (len != 0) 139 hlen = len; 140 } 141 ip = mtod(m, struct ip *); 142 143 /* 144 * Fill in IP header. If we are not allowing fragmentation, 145 * then the ip_id field is meaningless, but we don't set it 146 * to zero. Doing so causes various problems when devices along 147 * the path (routers, load balancers, firewalls, etc.) illegally 148 * disable DF on our packet. Note that a 16-bit counter 149 * will wrap around in less than 10 seconds at 100 Mbit/s on a 150 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 151 * for Counting NATted Hosts", Proc. IMW'02, available at 152 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 153 */ 154 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 155 ip->ip_v = IPVERSION; 156 ip->ip_hl = hlen >> 2; 157 ip->ip_id = ip_newid(); 158 ipstat.ips_localout++; 159 } else { 160 hlen = ip->ip_hl << 2; 161 } 162 163 dst = (struct sockaddr_in *)&ro->ro_dst; 164again: 165 /* 166 * If there is a cached route, 167 * check that it is to the same destination 168 * and is still up. If not, free it and try again. 169 * The address family should also be checked in case of sharing the 170 * cache with IPv6. 171 */ 172 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 173 dst->sin_family != AF_INET || 174 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 175 RTFREE(ro->ro_rt); 176 ro->ro_rt = (struct rtentry *)NULL; 177 } 178#ifdef IPFIREWALL_FORWARD 179 if (ro->ro_rt == NULL && fwd_tag == NULL) { 180#else 181 if (ro->ro_rt == NULL) { 182#endif 183 bzero(dst, sizeof(*dst)); 184 dst->sin_family = AF_INET; 185 dst->sin_len = sizeof(*dst); 186 dst->sin_addr = ip->ip_dst; 187 } 188 /* 189 * If routing to interface only, 190 * short circuit routing lookup. 191 */ 192 if (flags & IP_ROUTETOIF) { 193 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 194 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) { 195 ipstat.ips_noroute++; 196 error = ENETUNREACH; 197 goto bad; 198 } 199 ifp = ia->ia_ifp; 200 ip->ip_ttl = 1; 201 isbroadcast = in_broadcast(dst->sin_addr, ifp); 202 } else if (flags & IP_SENDONES) { 203 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL) { 204 ipstat.ips_noroute++; 205 error = ENETUNREACH; 206 goto bad; 207 } 208 ifp = ia->ia_ifp; 209 ip->ip_dst.s_addr = INADDR_BROADCAST; 210 dst->sin_addr = ip->ip_dst; 211 ip->ip_ttl = 1; 212 isbroadcast = 1; 213 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 214 imo != NULL && imo->imo_multicast_ifp != NULL) { 215 /* 216 * Bypass the normal routing lookup for multicast 217 * packets if the interface is specified. 218 */ 219 ifp = imo->imo_multicast_ifp; 220 IFP_TO_IA(ifp, ia); 221 isbroadcast = 0; /* fool gcc */ 222 } else { 223 /* 224 * We want to do any cloning requested by the link layer, 225 * as this is probably required in all cases for correct 226 * operation (as it is for ARP). 227 */ 228 if (ro->ro_rt == NULL) 229 rtalloc_ign(ro, 0); 230 if (ro->ro_rt == NULL) { 231 ipstat.ips_noroute++; 232 error = EHOSTUNREACH; 233 goto bad; 234 } 235 ia = ifatoia(ro->ro_rt->rt_ifa); 236 ifp = ro->ro_rt->rt_ifp; 237 ro->ro_rt->rt_rmx.rmx_pksent++; 238 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 239 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 240 if (ro->ro_rt->rt_flags & RTF_HOST) 241 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 242 else 243 isbroadcast = in_broadcast(dst->sin_addr, ifp); 244 } 245 /* 246 * Calculate MTU. If we have a route that is up, use that, 247 * otherwise use the interface's MTU. 248 */ 249 if (ro->ro_rt != NULL && (ro->ro_rt->rt_flags & (RTF_UP|RTF_HOST))) { 250 /* 251 * This case can happen if the user changed the MTU 252 * of an interface after enabling IP on it. Because 253 * most netifs don't keep track of routes pointing to 254 * them, there is no way for one to update all its 255 * routes when the MTU is changed. 256 */ 257 if (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu) 258 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 259 mtu = ro->ro_rt->rt_rmx.rmx_mtu; 260 } else { 261 mtu = ifp->if_mtu; 262 } 263 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 264 struct in_multi *inm; 265 266 m->m_flags |= M_MCAST; 267 /* 268 * IP destination address is multicast. Make sure "dst" 269 * still points to the address in "ro". (It may have been 270 * changed to point to a gateway address, above.) 271 */ 272 dst = (struct sockaddr_in *)&ro->ro_dst; 273 /* 274 * See if the caller provided any multicast options 275 */ 276 if (imo != NULL) { 277 ip->ip_ttl = imo->imo_multicast_ttl; 278 if (imo->imo_multicast_vif != -1) 279 ip->ip_src.s_addr = 280 ip_mcast_src ? 281 ip_mcast_src(imo->imo_multicast_vif) : 282 INADDR_ANY; 283 } else 284 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 285 /* 286 * Confirm that the outgoing interface supports multicast. 287 */ 288 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 289 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 290 ipstat.ips_noroute++; 291 error = ENETUNREACH; 292 goto bad; 293 } 294 } 295 /* 296 * If source address not specified yet, use address 297 * of outgoing interface. 298 */ 299 if (ip->ip_src.s_addr == INADDR_ANY) { 300 /* Interface may have no addresses. */ 301 if (ia != NULL) 302 ip->ip_src = IA_SIN(ia)->sin_addr; 303 } 304 305 IN_MULTI_LOCK(); 306 IN_LOOKUP_MULTI(ip->ip_dst, ifp, inm); 307 if (inm != NULL && 308 (imo == NULL || imo->imo_multicast_loop)) { 309 IN_MULTI_UNLOCK(); 310 /* 311 * If we belong to the destination multicast group 312 * on the outgoing interface, and the caller did not 313 * forbid loopback, loop back a copy. 314 */ 315 ip_mloopback(ifp, m, dst, hlen); 316 } 317 else { 318 IN_MULTI_UNLOCK(); 319 /* 320 * If we are acting as a multicast router, perform 321 * multicast forwarding as if the packet had just 322 * arrived on the interface to which we are about 323 * to send. The multicast forwarding function 324 * recursively calls this function, using the 325 * IP_FORWARDING flag to prevent infinite recursion. 326 * 327 * Multicasts that are looped back by ip_mloopback(), 328 * above, will be forwarded by the ip_input() routine, 329 * if necessary. 330 */ 331 if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 332 /* 333 * If rsvp daemon is not running, do not 334 * set ip_moptions. This ensures that the packet 335 * is multicast and not just sent down one link 336 * as prescribed by rsvpd. 337 */ 338 if (!rsvp_on) 339 imo = NULL; 340 if (ip_mforward && 341 ip_mforward(ip, ifp, m, imo) != 0) { 342 m_freem(m); 343 goto done; 344 } 345 } 346 } 347 348 /* 349 * Multicasts with a time-to-live of zero may be looped- 350 * back, above, but must not be transmitted on a network. 351 * Also, multicasts addressed to the loopback interface 352 * are not sent -- the above call to ip_mloopback() will 353 * loop back a copy if this host actually belongs to the 354 * destination group on the loopback interface. 355 */ 356 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 357 m_freem(m); 358 goto done; 359 } 360 361 goto sendit; 362 } 363#ifndef notdef 364 /* 365 * If the source address is not specified yet, use the address 366 * of the outoing interface. 367 */ 368 if (ip->ip_src.s_addr == INADDR_ANY) { 369 /* Interface may have no addresses. */ 370 if (ia != NULL) { 371 ip->ip_src = IA_SIN(ia)->sin_addr; 372 } 373 } 374#endif /* notdef */ 375 /* 376 * Verify that we have any chance at all of being able to queue the 377 * packet or packet fragments, unless ALTQ is enabled on the given 378 * interface in which case packetdrop should be done by queueing. 379 */ 380#ifdef ALTQ 381 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) && 382 ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >= 383 ifp->if_snd.ifq_maxlen)) 384#else 385 if ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >= 386 ifp->if_snd.ifq_maxlen) 387#endif /* ALTQ */ 388 { 389 error = ENOBUFS; 390 ipstat.ips_odropped++; 391 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1); 392 goto bad; 393 } 394 395 /* 396 * Look for broadcast address and 397 * verify user is allowed to send 398 * such a packet. 399 */ 400 if (isbroadcast) { 401 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 402 error = EADDRNOTAVAIL; 403 goto bad; 404 } 405 if ((flags & IP_ALLOWBROADCAST) == 0) { 406 error = EACCES; 407 goto bad; 408 } 409 /* don't allow broadcast messages to be fragmented */ 410 if (ip->ip_len > mtu) { 411 error = EMSGSIZE; 412 goto bad; 413 } 414 m->m_flags |= M_BCAST; 415 } else { 416 m->m_flags &= ~M_BCAST; 417 } 418 419sendit: 420#if defined(IPSEC) || defined(FAST_IPSEC) 421 switch(ip_ipsec_output(&m, inp, &flags, &error, &ro, &iproute, &dst, &ia, &ifp)) { 422 case 1: 423 goto bad; 424 case -1: 425 goto done; 426 case 0: 427 default: 428 break; /* Continue with packet processing. */ 429 } 430 /* Update variables that are affected by ipsec4_output(). */ 431 ip = mtod(m, struct ip *); 432 hlen = ip->ip_hl << 2; 433#endif /* IPSEC */ 434 435 /* Jump over all PFIL processing if hooks are not active. */ 436 if (!PFIL_HOOKED(&inet_pfil_hook)) 437 goto passout; 438 439 /* Run through list of hooks for output packets. */ 440 odst.s_addr = ip->ip_dst.s_addr; 441 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 442 if (error != 0 || m == NULL) 443 goto done; 444 445 ip = mtod(m, struct ip *); 446 447 /* See if destination IP address was changed by packet filter. */ 448 if (odst.s_addr != ip->ip_dst.s_addr) { 449 m->m_flags |= M_SKIP_FIREWALL; 450 /* If destination is now ourself drop to ip_input(). */ 451 if (in_localip(ip->ip_dst)) { 452 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 } else 466 goto again; /* Redo the routing table lookup. */ 467 } 468 469#ifdef IPFIREWALL_FORWARD 470 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 471 if (m->m_flags & M_FASTFWD_OURS) { 472 if (m->m_pkthdr.rcvif == NULL) 473 m->m_pkthdr.rcvif = loif; 474 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 475 m->m_pkthdr.csum_flags |= 476 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 477 m->m_pkthdr.csum_data = 0xffff; 478 } 479 m->m_pkthdr.csum_flags |= 480 CSUM_IP_CHECKED | CSUM_IP_VALID; 481 482 error = netisr_queue(NETISR_IP, m); 483 goto done; 484 } 485 /* Or forward to some other address? */ 486 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 487 if (fwd_tag) { 488 dst = (struct sockaddr_in *)&ro->ro_dst; 489 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 490 m->m_flags |= M_SKIP_FIREWALL; 491 m_tag_delete(m, fwd_tag); 492 goto again; 493 } 494#endif /* IPFIREWALL_FORWARD */ 495 496passout: 497 /* 127/8 must not appear on wire - RFC1122. */ 498 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 499 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 500 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 501 ipstat.ips_badaddr++; 502 error = EADDRNOTAVAIL; 503 goto bad; 504 } 505 } 506 507 m->m_pkthdr.csum_flags |= CSUM_IP; 508 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 509 if (sw_csum & CSUM_DELAY_DATA) { 510 in_delayed_cksum(m); 511 sw_csum &= ~CSUM_DELAY_DATA; 512 } 513 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 514 515 /* 516 * If small enough for interface, or the interface will take 517 * care of the fragmentation for us, we can just send directly. 518 */ 519 if (ip->ip_len <= mtu || 520 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 521 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 522 ip->ip_len = htons(ip->ip_len); 523 ip->ip_off = htons(ip->ip_off); 524 ip->ip_sum = 0; 525 if (sw_csum & CSUM_DELAY_IP) 526 ip->ip_sum = in_cksum(m, hlen); 527 528 /* 529 * Record statistics for this interface address. 530 * With CSUM_TSO the byte/packet count will be slightly 531 * incorrect because we count the IP+TCP headers only 532 * once instead of for every generated packet. 533 */ 534 if (!(flags & IP_FORWARDING) && ia) { 535 if (m->m_pkthdr.csum_flags & CSUM_TSO) 536 ia->ia_ifa.if_opackets += 537 m->m_pkthdr.len / m->m_pkthdr.tso_segsz; 538 else 539 ia->ia_ifa.if_opackets++; 540 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 541 } 542#ifdef IPSEC 543 /* clean ipsec history once it goes out of the node */ 544 ipsec_delaux(m); 545#endif 546#ifdef MBUF_STRESS_TEST 547 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 548 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 549#endif 550 /* 551 * Reset layer specific mbuf flags 552 * to avoid confusing lower layers. 553 */ 554 m->m_flags &= ~(M_PROTOFLAGS); 555 556 error = (*ifp->if_output)(ifp, m, 557 (struct sockaddr *)dst, ro->ro_rt); 558 goto done; 559 } 560 561 /* Balk when DF bit is set or the interface didn't support TSO. */ 562 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 563 error = EMSGSIZE; 564 ipstat.ips_cantfrag++; 565 goto bad; 566 } 567 568 /* 569 * Too large for interface; fragment if possible. If successful, 570 * on return, m will point to a list of packets to be sent. 571 */ 572 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum); 573 if (error) 574 goto bad; 575 for (; m; m = m0) { 576 m0 = m->m_nextpkt; 577 m->m_nextpkt = 0; 578#ifdef IPSEC 579 /* clean ipsec history once it goes out of the node */ 580 ipsec_delaux(m); 581#endif 582 if (error == 0) { 583 /* Record statistics for this interface address. */ 584 if (ia != NULL) { 585 ia->ia_ifa.if_opackets++; 586 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 587 } 588 /* 589 * Reset layer specific mbuf flags 590 * to avoid confusing upper layers. 591 */ 592 m->m_flags &= ~(M_PROTOFLAGS); 593 594 error = (*ifp->if_output)(ifp, m, 595 (struct sockaddr *)dst, ro->ro_rt); 596 } else 597 m_freem(m); 598 } 599 600 if (error == 0) 601 ipstat.ips_fragmented++; 602 603done: 604 if (ro == &iproute && ro->ro_rt) { 605 RTFREE(ro->ro_rt); 606 } 607 return (error); 608bad: 609 m_freem(m); 610 goto done; 611} 612 613/* 614 * Create a chain of fragments which fit the given mtu. m_frag points to the 615 * mbuf to be fragmented; on return it points to the chain with the fragments. 616 * Return 0 if no error. If error, m_frag may contain a partially built 617 * chain of fragments that should be freed by the caller. 618 * 619 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 620 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 621 */ 622int 623ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 624 u_long if_hwassist_flags, int sw_csum) 625{ 626 int error = 0; 627 int hlen = ip->ip_hl << 2; 628 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 629 int off; 630 struct mbuf *m0 = *m_frag; /* the original packet */ 631 int firstlen; 632 struct mbuf **mnext; 633 int nfrags; 634 635 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 636 ipstat.ips_cantfrag++; 637 return EMSGSIZE; 638 } 639 640 /* 641 * Must be able to put at least 8 bytes per fragment. 642 */ 643 if (len < 8) 644 return EMSGSIZE; 645 646 /* 647 * If the interface will not calculate checksums on 648 * fragmented packets, then do it here. 649 */ 650 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 651 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 652 in_delayed_cksum(m0); 653 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 654 } 655 656 if (len > PAGE_SIZE) { 657 /* 658 * Fragment large datagrams such that each segment 659 * contains a multiple of PAGE_SIZE amount of data, 660 * plus headers. This enables a receiver to perform 661 * page-flipping zero-copy optimizations. 662 * 663 * XXX When does this help given that sender and receiver 664 * could have different page sizes, and also mtu could 665 * be less than the receiver's page size ? 666 */ 667 int newlen; 668 struct mbuf *m; 669 670 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 671 off += m->m_len; 672 673 /* 674 * firstlen (off - hlen) must be aligned on an 675 * 8-byte boundary 676 */ 677 if (off < hlen) 678 goto smart_frag_failure; 679 off = ((off - hlen) & ~7) + hlen; 680 newlen = (~PAGE_MASK) & mtu; 681 if ((newlen + sizeof (struct ip)) > mtu) { 682 /* we failed, go back the default */ 683smart_frag_failure: 684 newlen = len; 685 off = hlen + len; 686 } 687 len = newlen; 688 689 } else { 690 off = hlen + len; 691 } 692 693 firstlen = off - hlen; 694 mnext = &m0->m_nextpkt; /* pointer to next packet */ 695 696 /* 697 * Loop through length of segment after first fragment, 698 * make new header and copy data of each part and link onto chain. 699 * Here, m0 is the original packet, m is the fragment being created. 700 * The fragments are linked off the m_nextpkt of the original 701 * packet, which after processing serves as the first fragment. 702 */ 703 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 704 struct ip *mhip; /* ip header on the fragment */ 705 struct mbuf *m; 706 int mhlen = sizeof (struct ip); 707 708 MGETHDR(m, M_DONTWAIT, MT_DATA); 709 if (m == NULL) { 710 error = ENOBUFS; 711 ipstat.ips_odropped++; 712 goto done; 713 } 714 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 715 /* 716 * In the first mbuf, leave room for the link header, then 717 * copy the original IP header including options. The payload 718 * goes into an additional mbuf chain returned by m_copy(). 719 */ 720 m->m_data += max_linkhdr; 721 mhip = mtod(m, struct ip *); 722 *mhip = *ip; 723 if (hlen > sizeof (struct ip)) { 724 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 725 mhip->ip_v = IPVERSION; 726 mhip->ip_hl = mhlen >> 2; 727 } 728 m->m_len = mhlen; 729 /* XXX do we need to add ip->ip_off below ? */ 730 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 731 if (off + len >= ip->ip_len) { /* last fragment */ 732 len = ip->ip_len - off; 733 m->m_flags |= M_LASTFRAG; 734 } else 735 mhip->ip_off |= IP_MF; 736 mhip->ip_len = htons((u_short)(len + mhlen)); 737 m->m_next = m_copy(m0, off, len); 738 if (m->m_next == NULL) { /* copy failed */ 739 m_free(m); 740 error = ENOBUFS; /* ??? */ 741 ipstat.ips_odropped++; 742 goto done; 743 } 744 m->m_pkthdr.len = mhlen + len; 745 m->m_pkthdr.rcvif = NULL; 746#ifdef MAC 747 mac_create_fragment(m0, m); 748#endif 749 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 750 mhip->ip_off = htons(mhip->ip_off); 751 mhip->ip_sum = 0; 752 if (sw_csum & CSUM_DELAY_IP) 753 mhip->ip_sum = in_cksum(m, mhlen); 754 *mnext = m; 755 mnext = &m->m_nextpkt; 756 } 757 ipstat.ips_ofragments += nfrags; 758 759 /* set first marker for fragment chain */ 760 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 761 m0->m_pkthdr.csum_data = nfrags; 762 763 /* 764 * Update first fragment by trimming what's been copied out 765 * and updating header. 766 */ 767 m_adj(m0, hlen + firstlen - ip->ip_len); 768 m0->m_pkthdr.len = hlen + firstlen; 769 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 770 ip->ip_off |= IP_MF; 771 ip->ip_off = htons(ip->ip_off); 772 ip->ip_sum = 0; 773 if (sw_csum & CSUM_DELAY_IP) 774 ip->ip_sum = in_cksum(m0, hlen); 775 776done: 777 *m_frag = m0; 778 return error; 779} 780 781void 782in_delayed_cksum(struct mbuf *m) 783{ 784 struct ip *ip; 785 u_short csum, offset; 786 787 ip = mtod(m, struct ip *); 788 offset = ip->ip_hl << 2 ; 789 csum = in_cksum_skip(m, ip->ip_len, offset); 790 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 791 csum = 0xffff; 792 offset += m->m_pkthdr.csum_data; /* checksum offset */ 793 794 if (offset + sizeof(u_short) > m->m_len) { 795 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 796 m->m_len, offset, ip->ip_p); 797 /* 798 * XXX 799 * this shouldn't happen, but if it does, the 800 * correct behavior may be to insert the checksum 801 * in the appropriate next mbuf in the chain. 802 */ 803 return; 804 } 805 *(u_short *)(m->m_data + offset) = csum; 806} 807 808/* 809 * IP socket option processing. 810 */ 811int 812ip_ctloutput(so, sopt) 813 struct socket *so; 814 struct sockopt *sopt; 815{ 816 struct inpcb *inp = sotoinpcb(so); 817 int error, optval; 818 819 error = optval = 0; 820 if (sopt->sopt_level != IPPROTO_IP) { 821 return (EINVAL); 822 } 823 824 switch (sopt->sopt_dir) { 825 case SOPT_SET: 826 switch (sopt->sopt_name) { 827 case IP_OPTIONS: 828#ifdef notyet 829 case IP_RETOPTS: 830#endif 831 { 832 struct mbuf *m; 833 if (sopt->sopt_valsize > MLEN) { 834 error = EMSGSIZE; 835 break; 836 } 837 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_DATA); 838 if (m == NULL) { 839 error = ENOBUFS; 840 break; 841 } 842 m->m_len = sopt->sopt_valsize; 843 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 844 m->m_len); 845 if (error) { 846 m_free(m); 847 break; 848 } 849 INP_LOCK(inp); 850 error = ip_pcbopts(inp, sopt->sopt_name, m); 851 INP_UNLOCK(inp); 852 return (error); 853 } 854 855 case IP_TOS: 856 case IP_TTL: 857 case IP_MINTTL: 858 case IP_RECVOPTS: 859 case IP_RECVRETOPTS: 860 case IP_RECVDSTADDR: 861 case IP_RECVTTL: 862 case IP_RECVIF: 863 case IP_FAITH: 864 case IP_ONESBCAST: 865 case IP_DONTFRAG: 866 error = sooptcopyin(sopt, &optval, sizeof optval, 867 sizeof optval); 868 if (error) 869 break; 870 871 switch (sopt->sopt_name) { 872 case IP_TOS: 873 inp->inp_ip_tos = optval; 874 break; 875 876 case IP_TTL: 877 inp->inp_ip_ttl = optval; 878 break; 879 880 case IP_MINTTL: 881 if (optval > 0 && optval <= MAXTTL) 882 inp->inp_ip_minttl = optval; 883 else 884 error = EINVAL; 885 break; 886 887#define OPTSET(bit) do { \ 888 INP_LOCK(inp); \ 889 if (optval) \ 890 inp->inp_flags |= bit; \ 891 else \ 892 inp->inp_flags &= ~bit; \ 893 INP_UNLOCK(inp); \ 894} while (0) 895 896 case IP_RECVOPTS: 897 OPTSET(INP_RECVOPTS); 898 break; 899 900 case IP_RECVRETOPTS: 901 OPTSET(INP_RECVRETOPTS); 902 break; 903 904 case IP_RECVDSTADDR: 905 OPTSET(INP_RECVDSTADDR); 906 break; 907 908 case IP_RECVTTL: 909 OPTSET(INP_RECVTTL); 910 break; 911 912 case IP_RECVIF: 913 OPTSET(INP_RECVIF); 914 break; 915 916 case IP_FAITH: 917 OPTSET(INP_FAITH); 918 break; 919 920 case IP_ONESBCAST: 921 OPTSET(INP_ONESBCAST); 922 break; 923 case IP_DONTFRAG: 924 OPTSET(INP_DONTFRAG); 925 break; 926 } 927 break; 928#undef OPTSET 929 930 case IP_MULTICAST_IF: 931 case IP_MULTICAST_VIF: 932 case IP_MULTICAST_TTL: 933 case IP_MULTICAST_LOOP: 934 case IP_ADD_MEMBERSHIP: 935 case IP_DROP_MEMBERSHIP: 936 error = ip_setmoptions(inp, sopt); 937 break; 938 939 case IP_PORTRANGE: 940 error = sooptcopyin(sopt, &optval, sizeof optval, 941 sizeof optval); 942 if (error) 943 break; 944 945 INP_LOCK(inp); 946 switch (optval) { 947 case IP_PORTRANGE_DEFAULT: 948 inp->inp_flags &= ~(INP_LOWPORT); 949 inp->inp_flags &= ~(INP_HIGHPORT); 950 break; 951 952 case IP_PORTRANGE_HIGH: 953 inp->inp_flags &= ~(INP_LOWPORT); 954 inp->inp_flags |= INP_HIGHPORT; 955 break; 956 957 case IP_PORTRANGE_LOW: 958 inp->inp_flags &= ~(INP_HIGHPORT); 959 inp->inp_flags |= INP_LOWPORT; 960 break; 961 962 default: 963 error = EINVAL; 964 break; 965 } 966 INP_UNLOCK(inp); 967 break; 968 969#if defined(IPSEC) || defined(FAST_IPSEC) 970 case IP_IPSEC_POLICY: 971 { 972 caddr_t req; 973 size_t len = 0; 974 int priv; 975 struct mbuf *m; 976 int optname; 977 978 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 979 break; 980 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 981 break; 982 priv = (sopt->sopt_td != NULL && 983 suser(sopt->sopt_td) != 0) ? 0 : 1; 984 req = mtod(m, caddr_t); 985 len = m->m_len; 986 optname = sopt->sopt_name; 987 error = ipsec4_set_policy(inp, optname, req, len, priv); 988 m_freem(m); 989 break; 990 } 991#endif /*IPSEC*/ 992 993 default: 994 error = ENOPROTOOPT; 995 break; 996 } 997 break; 998 999 case SOPT_GET: 1000 switch (sopt->sopt_name) { 1001 case IP_OPTIONS: 1002 case IP_RETOPTS: 1003 if (inp->inp_options) 1004 error = sooptcopyout(sopt, 1005 mtod(inp->inp_options, 1006 char *), 1007 inp->inp_options->m_len); 1008 else 1009 sopt->sopt_valsize = 0; 1010 break; 1011 1012 case IP_TOS: 1013 case IP_TTL: 1014 case IP_MINTTL: 1015 case IP_RECVOPTS: 1016 case IP_RECVRETOPTS: 1017 case IP_RECVDSTADDR: 1018 case IP_RECVTTL: 1019 case IP_RECVIF: 1020 case IP_PORTRANGE: 1021 case IP_FAITH: 1022 case IP_ONESBCAST: 1023 case IP_DONTFRAG: 1024 switch (sopt->sopt_name) { 1025 1026 case IP_TOS: 1027 optval = inp->inp_ip_tos; 1028 break; 1029 1030 case IP_TTL: 1031 optval = inp->inp_ip_ttl; 1032 break; 1033 1034 case IP_MINTTL: 1035 optval = inp->inp_ip_minttl; 1036 break; 1037 1038#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1039 1040 case IP_RECVOPTS: 1041 optval = OPTBIT(INP_RECVOPTS); 1042 break; 1043 1044 case IP_RECVRETOPTS: 1045 optval = OPTBIT(INP_RECVRETOPTS); 1046 break; 1047 1048 case IP_RECVDSTADDR: 1049 optval = OPTBIT(INP_RECVDSTADDR); 1050 break; 1051 1052 case IP_RECVTTL: 1053 optval = OPTBIT(INP_RECVTTL); 1054 break; 1055 1056 case IP_RECVIF: 1057 optval = OPTBIT(INP_RECVIF); 1058 break; 1059 1060 case IP_PORTRANGE: 1061 if (inp->inp_flags & INP_HIGHPORT) 1062 optval = IP_PORTRANGE_HIGH; 1063 else if (inp->inp_flags & INP_LOWPORT) 1064 optval = IP_PORTRANGE_LOW; 1065 else 1066 optval = 0; 1067 break; 1068 1069 case IP_FAITH: 1070 optval = OPTBIT(INP_FAITH); 1071 break; 1072 1073 case IP_ONESBCAST: 1074 optval = OPTBIT(INP_ONESBCAST); 1075 break; 1076 case IP_DONTFRAG: 1077 optval = OPTBIT(INP_DONTFRAG); 1078 break; 1079 } 1080 error = sooptcopyout(sopt, &optval, sizeof optval); 1081 break; 1082 1083 case IP_MULTICAST_IF: 1084 case IP_MULTICAST_VIF: 1085 case IP_MULTICAST_TTL: 1086 case IP_MULTICAST_LOOP: 1087 case IP_ADD_MEMBERSHIP: 1088 case IP_DROP_MEMBERSHIP: 1089 error = ip_getmoptions(inp, sopt); 1090 break; 1091 1092#if defined(IPSEC) || defined(FAST_IPSEC) 1093 case IP_IPSEC_POLICY: 1094 { 1095 struct mbuf *m = NULL; 1096 caddr_t req = NULL; 1097 size_t len = 0; 1098 1099 if (m != 0) { 1100 req = mtod(m, caddr_t); 1101 len = m->m_len; 1102 } 1103 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1104 if (error == 0) 1105 error = soopt_mcopyout(sopt, m); /* XXX */ 1106 if (error == 0) 1107 m_freem(m); 1108 break; 1109 } 1110#endif /*IPSEC*/ 1111 1112 default: 1113 error = ENOPROTOOPT; 1114 break; 1115 } 1116 break; 1117 } 1118 return (error); 1119} 1120 1121/* 1122 * XXX 1123 * The whole multicast option thing needs to be re-thought. 1124 * Several of these options are equally applicable to non-multicast 1125 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1126 * standard option (IP_TTL). 1127 */ 1128 1129/* 1130 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1131 */ 1132static struct ifnet * 1133ip_multicast_if(a, ifindexp) 1134 struct in_addr *a; 1135 int *ifindexp; 1136{ 1137 int ifindex; 1138 struct ifnet *ifp; 1139 1140 if (ifindexp) 1141 *ifindexp = 0; 1142 if (ntohl(a->s_addr) >> 24 == 0) { 1143 ifindex = ntohl(a->s_addr) & 0xffffff; 1144 if (ifindex < 0 || if_index < ifindex) 1145 return NULL; 1146 ifp = ifnet_byindex(ifindex); 1147 if (ifindexp) 1148 *ifindexp = ifindex; 1149 } else { 1150 INADDR_TO_IFP(*a, ifp); 1151 } 1152 return ifp; 1153} 1154 1155/* 1156 * Given an inpcb, return its multicast options structure pointer. Accepts 1157 * an unlocked inpcb pointer, but will return it locked. May sleep. 1158 */ 1159static struct ip_moptions * 1160ip_findmoptions(struct inpcb *inp) 1161{ 1162 struct ip_moptions *imo; 1163 struct in_multi **immp; 1164 1165 INP_LOCK(inp); 1166 if (inp->inp_moptions != NULL) 1167 return (inp->inp_moptions); 1168 1169 INP_UNLOCK(inp); 1170 1171 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, M_WAITOK); 1172 immp = (struct in_multi **)malloc((sizeof(*immp) * IP_MIN_MEMBERSHIPS), 1173 M_IPMOPTS, M_WAITOK); 1174 1175 imo->imo_multicast_ifp = NULL; 1176 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1177 imo->imo_multicast_vif = -1; 1178 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1179 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1180 imo->imo_num_memberships = 0; 1181 imo->imo_max_memberships = IP_MIN_MEMBERSHIPS; 1182 imo->imo_membership = immp; 1183 1184 INP_LOCK(inp); 1185 if (inp->inp_moptions != NULL) { 1186 free(immp, M_IPMOPTS); 1187 free(imo, M_IPMOPTS); 1188 return (inp->inp_moptions); 1189 } 1190 inp->inp_moptions = imo; 1191 return (imo); 1192} 1193 1194/* 1195 * Set the IP multicast options in response to user setsockopt(). 1196 */ 1197static int 1198ip_setmoptions(struct inpcb *inp, struct sockopt *sopt) 1199{ 1200 int error = 0; 1201 int i; 1202 struct in_addr addr; 1203 struct ip_mreq mreq; 1204 struct ifnet *ifp; 1205 struct ip_moptions *imo; 1206 struct route ro; 1207 struct sockaddr_in *dst; 1208 int ifindex; 1209 int s; 1210 1211 switch (sopt->sopt_name) { 1212 /* store an index number for the vif you wanna use in the send */ 1213 case IP_MULTICAST_VIF: 1214 if (legal_vif_num == 0) { 1215 error = EOPNOTSUPP; 1216 break; 1217 } 1218 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1219 if (error) 1220 break; 1221 if (!legal_vif_num(i) && (i != -1)) { 1222 error = EINVAL; 1223 break; 1224 } 1225 imo = ip_findmoptions(inp); 1226 imo->imo_multicast_vif = i; 1227 INP_UNLOCK(inp); 1228 break; 1229 1230 case IP_MULTICAST_IF: 1231 /* 1232 * Select the interface for outgoing multicast packets. 1233 */ 1234 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1235 if (error) 1236 break; 1237 /* 1238 * INADDR_ANY is used to remove a previous selection. 1239 * When no interface is selected, a default one is 1240 * chosen every time a multicast packet is sent. 1241 */ 1242 imo = ip_findmoptions(inp); 1243 if (addr.s_addr == INADDR_ANY) { 1244 imo->imo_multicast_ifp = NULL; 1245 INP_UNLOCK(inp); 1246 break; 1247 } 1248 /* 1249 * The selected interface is identified by its local 1250 * IP address. Find the interface and confirm that 1251 * it supports multicasting. 1252 */ 1253 s = splimp(); 1254 ifp = ip_multicast_if(&addr, &ifindex); 1255 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1256 INP_UNLOCK(inp); 1257 splx(s); 1258 error = EADDRNOTAVAIL; 1259 break; 1260 } 1261 imo->imo_multicast_ifp = ifp; 1262 if (ifindex) 1263 imo->imo_multicast_addr = addr; 1264 else 1265 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1266 INP_UNLOCK(inp); 1267 splx(s); 1268 break; 1269 1270 case IP_MULTICAST_TTL: 1271 /* 1272 * Set the IP time-to-live for outgoing multicast packets. 1273 * The original multicast API required a char argument, 1274 * which is inconsistent with the rest of the socket API. 1275 * We allow either a char or an int. 1276 */ 1277 if (sopt->sopt_valsize == 1) { 1278 u_char ttl; 1279 error = sooptcopyin(sopt, &ttl, 1, 1); 1280 if (error) 1281 break; 1282 imo = ip_findmoptions(inp); 1283 imo->imo_multicast_ttl = ttl; 1284 INP_UNLOCK(inp); 1285 } else { 1286 u_int ttl; 1287 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1288 sizeof ttl); 1289 if (error) 1290 break; 1291 if (ttl > 255) 1292 error = EINVAL; 1293 else { 1294 imo = ip_findmoptions(inp); 1295 imo->imo_multicast_ttl = ttl; 1296 INP_UNLOCK(inp); 1297 } 1298 } 1299 break; 1300 1301 case IP_MULTICAST_LOOP: 1302 /* 1303 * Set the loopback flag for outgoing multicast packets. 1304 * Must be zero or one. The original multicast API required a 1305 * char argument, which is inconsistent with the rest 1306 * of the socket API. We allow either a char or an int. 1307 */ 1308 if (sopt->sopt_valsize == 1) { 1309 u_char loop; 1310 error = sooptcopyin(sopt, &loop, 1, 1); 1311 if (error) 1312 break; 1313 imo = ip_findmoptions(inp); 1314 imo->imo_multicast_loop = !!loop; 1315 INP_UNLOCK(inp); 1316 } else { 1317 u_int loop; 1318 error = sooptcopyin(sopt, &loop, sizeof loop, 1319 sizeof loop); 1320 if (error) 1321 break; 1322 imo = ip_findmoptions(inp); 1323 imo->imo_multicast_loop = !!loop; 1324 INP_UNLOCK(inp); 1325 } 1326 break; 1327 1328 case IP_ADD_MEMBERSHIP: 1329 /* 1330 * Add a multicast group membership. 1331 * Group must be a valid IP multicast address. 1332 */ 1333 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1334 if (error) 1335 break; 1336 1337 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1338 error = EINVAL; 1339 break; 1340 } 1341 s = splimp(); 1342 /* 1343 * If no interface address was provided, use the interface of 1344 * the route to the given multicast address. 1345 */ 1346 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1347 bzero((caddr_t)&ro, sizeof(ro)); 1348 dst = (struct sockaddr_in *)&ro.ro_dst; 1349 dst->sin_len = sizeof(*dst); 1350 dst->sin_family = AF_INET; 1351 dst->sin_addr = mreq.imr_multiaddr; 1352 rtalloc_ign(&ro, RTF_CLONING); 1353 if (ro.ro_rt == NULL) { 1354 error = EADDRNOTAVAIL; 1355 splx(s); 1356 break; 1357 } 1358 ifp = ro.ro_rt->rt_ifp; 1359 RTFREE(ro.ro_rt); 1360 } 1361 else { 1362 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1363 } 1364 1365 /* 1366 * See if we found an interface, and confirm that it 1367 * supports multicast. 1368 */ 1369 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1370 error = EADDRNOTAVAIL; 1371 splx(s); 1372 break; 1373 } 1374 /* 1375 * See if the membership already exists or if all the 1376 * membership slots are full. 1377 */ 1378 imo = ip_findmoptions(inp); 1379 for (i = 0; i < imo->imo_num_memberships; ++i) { 1380 if (imo->imo_membership[i]->inm_ifp == ifp && 1381 imo->imo_membership[i]->inm_addr.s_addr 1382 == mreq.imr_multiaddr.s_addr) 1383 break; 1384 } 1385 if (i < imo->imo_num_memberships) { 1386 INP_UNLOCK(inp); 1387 error = EADDRINUSE; 1388 splx(s); 1389 break; 1390 } 1391 if (imo->imo_num_memberships == imo->imo_max_memberships) { 1392 struct in_multi **nmships, **omships; 1393 size_t newmax; 1394 /* 1395 * Resize the vector to next power-of-two minus 1. If the 1396 * size would exceed the maximum then we know we've really 1397 * run out of entries. Otherwise, we realloc() the vector 1398 * with the INP lock held to avoid introducing a race. 1399 */ 1400 nmships = NULL; 1401 omships = imo->imo_membership; 1402 newmax = ((imo->imo_max_memberships + 1) * 2) - 1; 1403 if (newmax <= IP_MAX_MEMBERSHIPS) { 1404 nmships = (struct in_multi **)realloc(omships, 1405sizeof(*nmships) * newmax, M_IPMOPTS, M_NOWAIT); 1406 if (nmships != NULL) { 1407 imo->imo_membership = nmships; 1408 imo->imo_max_memberships = newmax; 1409 } 1410 } 1411 if (nmships == NULL) { 1412 INP_UNLOCK(inp); 1413 error = ETOOMANYREFS; 1414 splx(s); 1415 break; 1416 } 1417 } 1418 /* 1419 * Everything looks good; add a new record to the multicast 1420 * address list for the given interface. 1421 */ 1422 if ((imo->imo_membership[i] = 1423 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1424 INP_UNLOCK(inp); 1425 error = ENOBUFS; 1426 splx(s); 1427 break; 1428 } 1429 ++imo->imo_num_memberships; 1430 INP_UNLOCK(inp); 1431 splx(s); 1432 break; 1433 1434 case IP_DROP_MEMBERSHIP: 1435 /* 1436 * Drop a multicast group membership. 1437 * Group must be a valid IP multicast address. 1438 */ 1439 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1440 if (error) 1441 break; 1442 1443 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1444 error = EINVAL; 1445 break; 1446 } 1447 1448 s = splimp(); 1449 /* 1450 * If an interface address was specified, get a pointer 1451 * to its ifnet structure. 1452 */ 1453 if (mreq.imr_interface.s_addr == INADDR_ANY) 1454 ifp = NULL; 1455 else { 1456 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1457 if (ifp == NULL) { 1458 error = EADDRNOTAVAIL; 1459 splx(s); 1460 break; 1461 } 1462 } 1463 /* 1464 * Find the membership in the membership array. 1465 */ 1466 imo = ip_findmoptions(inp); 1467 for (i = 0; i < imo->imo_num_memberships; ++i) { 1468 if ((ifp == NULL || 1469 imo->imo_membership[i]->inm_ifp == ifp) && 1470 imo->imo_membership[i]->inm_addr.s_addr == 1471 mreq.imr_multiaddr.s_addr) 1472 break; 1473 } 1474 if (i == imo->imo_num_memberships) { 1475 INP_UNLOCK(inp); 1476 error = EADDRNOTAVAIL; 1477 splx(s); 1478 break; 1479 } 1480 /* 1481 * Give up the multicast address record to which the 1482 * membership points. 1483 */ 1484 in_delmulti(imo->imo_membership[i]); 1485 /* 1486 * Remove the gap in the membership array. 1487 */ 1488 for (++i; i < imo->imo_num_memberships; ++i) 1489 imo->imo_membership[i-1] = imo->imo_membership[i]; 1490 --imo->imo_num_memberships; 1491 INP_UNLOCK(inp); 1492 splx(s); 1493 break; 1494 1495 default: 1496 error = EOPNOTSUPP; 1497 break; 1498 } 1499 1500 return (error); 1501} 1502 1503/* 1504 * Return the IP multicast options in response to user getsockopt(). 1505 */ 1506static int 1507ip_getmoptions(struct inpcb *inp, struct sockopt *sopt) 1508{ 1509 struct ip_moptions *imo; 1510 struct in_addr addr; 1511 struct in_ifaddr *ia; 1512 int error, optval; 1513 u_char coptval; 1514 1515 INP_LOCK(inp); 1516 imo = inp->inp_moptions; 1517 1518 error = 0; 1519 switch (sopt->sopt_name) { 1520 case IP_MULTICAST_VIF: 1521 if (imo != NULL) 1522 optval = imo->imo_multicast_vif; 1523 else 1524 optval = -1; 1525 INP_UNLOCK(inp); 1526 error = sooptcopyout(sopt, &optval, sizeof optval); 1527 break; 1528 1529 case IP_MULTICAST_IF: 1530 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1531 addr.s_addr = INADDR_ANY; 1532 else if (imo->imo_multicast_addr.s_addr) { 1533 /* return the value user has set */ 1534 addr = imo->imo_multicast_addr; 1535 } else { 1536 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1537 addr.s_addr = (ia == NULL) ? INADDR_ANY 1538 : IA_SIN(ia)->sin_addr.s_addr; 1539 } 1540 INP_UNLOCK(inp); 1541 error = sooptcopyout(sopt, &addr, sizeof addr); 1542 break; 1543 1544 case IP_MULTICAST_TTL: 1545 if (imo == 0) 1546 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1547 else 1548 optval = coptval = imo->imo_multicast_ttl; 1549 INP_UNLOCK(inp); 1550 if (sopt->sopt_valsize == 1) 1551 error = sooptcopyout(sopt, &coptval, 1); 1552 else 1553 error = sooptcopyout(sopt, &optval, sizeof optval); 1554 break; 1555 1556 case IP_MULTICAST_LOOP: 1557 if (imo == 0) 1558 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1559 else 1560 optval = coptval = imo->imo_multicast_loop; 1561 INP_UNLOCK(inp); 1562 if (sopt->sopt_valsize == 1) 1563 error = sooptcopyout(sopt, &coptval, 1); 1564 else 1565 error = sooptcopyout(sopt, &optval, sizeof optval); 1566 break; 1567 1568 default: 1569 INP_UNLOCK(inp); 1570 error = ENOPROTOOPT; 1571 break; 1572 } 1573 INP_UNLOCK_ASSERT(inp); 1574 1575 return (error); 1576} 1577 1578/* 1579 * Discard the IP multicast options. 1580 */ 1581void 1582ip_freemoptions(imo) 1583 register struct ip_moptions *imo; 1584{ 1585 register int i; 1586 1587 if (imo != NULL) { 1588 for (i = 0; i < imo->imo_num_memberships; ++i) 1589 in_delmulti(imo->imo_membership[i]); 1590 free(imo->imo_membership, M_IPMOPTS); 1591 free(imo, M_IPMOPTS); 1592 } 1593} 1594 1595/* 1596 * Routine called from ip_output() to loop back a copy of an IP multicast 1597 * packet to the input queue of a specified interface. Note that this 1598 * calls the output routine of the loopback "driver", but with an interface 1599 * pointer that might NOT be a loopback interface -- evil, but easier than 1600 * replicating that code here. 1601 */ 1602static void 1603ip_mloopback(ifp, m, dst, hlen) 1604 struct ifnet *ifp; 1605 register struct mbuf *m; 1606 register struct sockaddr_in *dst; 1607 int hlen; 1608{ 1609 register struct ip *ip; 1610 struct mbuf *copym; 1611 1612 copym = m_copy(m, 0, M_COPYALL); 1613 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1614 copym = m_pullup(copym, hlen); 1615 if (copym != NULL) { 1616 /* If needed, compute the checksum and mark it as valid. */ 1617 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1618 in_delayed_cksum(copym); 1619 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1620 copym->m_pkthdr.csum_flags |= 1621 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1622 copym->m_pkthdr.csum_data = 0xffff; 1623 } 1624 /* 1625 * We don't bother to fragment if the IP length is greater 1626 * than the interface's MTU. Can this possibly matter? 1627 */ 1628 ip = mtod(copym, struct ip *); 1629 ip->ip_len = htons(ip->ip_len); 1630 ip->ip_off = htons(ip->ip_off); 1631 ip->ip_sum = 0; 1632 ip->ip_sum = in_cksum(copym, hlen); 1633 /* 1634 * NB: 1635 * It's not clear whether there are any lingering 1636 * reentrancy problems in other areas which might 1637 * be exposed by using ip_input directly (in 1638 * particular, everything which modifies the packet 1639 * in-place). Yet another option is using the 1640 * protosw directly to deliver the looped back 1641 * packet. For the moment, we'll err on the side 1642 * of safety by using if_simloop(). 1643 */ 1644#if 1 /* XXX */ 1645 if (dst->sin_family != AF_INET) { 1646 printf("ip_mloopback: bad address family %d\n", 1647 dst->sin_family); 1648 dst->sin_family = AF_INET; 1649 } 1650#endif 1651 1652#ifdef notdef 1653 copym->m_pkthdr.rcvif = ifp; 1654 ip_input(copym); 1655#else 1656 if_simloop(ifp, copym, dst->sin_family, 0); 1657#endif 1658 } 1659} 1660