ip_output.c revision 267736
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 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: stable/10/sys/netinet/ip_output.c 267736 2014-06-22 16:54:59Z tuexen $"); 34 35#include "opt_inet.h" 36#include "opt_ipfw.h" 37#include "opt_ipsec.h" 38#include "opt_kdtrace.h" 39#include "opt_mbuf_stress_test.h" 40#include "opt_mpath.h" 41#include "opt_route.h" 42#include "opt_sctp.h" 43 44#include <sys/param.h> 45#include <sys/systm.h> 46#include <sys/kernel.h> 47#include <sys/malloc.h> 48#include <sys/mbuf.h> 49#include <sys/priv.h> 50#include <sys/proc.h> 51#include <sys/protosw.h> 52#include <sys/sdt.h> 53#include <sys/socket.h> 54#include <sys/socketvar.h> 55#include <sys/sysctl.h> 56#include <sys/ucred.h> 57 58#include <net/if.h> 59#include <net/if_llatbl.h> 60#include <net/netisr.h> 61#include <net/pfil.h> 62#include <net/route.h> 63#include <net/flowtable.h> 64#ifdef RADIX_MPATH 65#include <net/radix_mpath.h> 66#endif 67#include <net/vnet.h> 68 69#include <netinet/in.h> 70#include <netinet/in_kdtrace.h> 71#include <netinet/in_systm.h> 72#include <netinet/ip.h> 73#include <netinet/in_pcb.h> 74#include <netinet/in_var.h> 75#include <netinet/ip_var.h> 76#include <netinet/ip_options.h> 77#ifdef SCTP 78#include <netinet/sctp.h> 79#include <netinet/sctp_crc32.h> 80#endif 81 82#ifdef IPSEC 83#include <netinet/ip_ipsec.h> 84#include <netipsec/ipsec.h> 85#endif /* IPSEC*/ 86 87#include <machine/in_cksum.h> 88 89#include <security/mac/mac_framework.h> 90 91VNET_DEFINE(u_short, ip_id); 92 93#ifdef MBUF_STRESS_TEST 94static int mbuf_frag_size = 0; 95SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 96 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 97#endif 98 99static void ip_mloopback 100 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 101 102 103extern int in_mcast_loop; 104extern struct protosw inetsw[]; 105 106/* 107 * IP output. The packet in mbuf chain m contains a skeletal IP 108 * header (with len, off, ttl, proto, tos, src, dst). 109 * The mbuf chain containing the packet will be freed. 110 * The mbuf opt, if present, will not be freed. 111 * If route ro is present and has ro_rt initialized, route lookup would be 112 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, 113 * then result of route lookup is stored in ro->ro_rt. 114 * 115 * In the IP forwarding case, the packet will arrive with options already 116 * inserted, so must have a NULL opt pointer. 117 */ 118int 119ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags, 120 struct ip_moptions *imo, struct inpcb *inp) 121{ 122 struct ip *ip; 123 struct ifnet *ifp = NULL; /* keep compiler happy */ 124 struct mbuf *m0; 125 int hlen = sizeof (struct ip); 126 int mtu; 127 int n; /* scratchpad */ 128 int error = 0; 129 struct sockaddr_in *dst; 130 const struct sockaddr_in *gw; 131 struct in_ifaddr *ia; 132 int isbroadcast; 133 uint16_t ip_len, ip_off; 134 struct route iproute; 135 struct rtentry *rte; /* cache for ro->ro_rt */ 136 struct in_addr odst; 137 struct m_tag *fwd_tag = NULL; 138#ifdef IPSEC 139 int no_route_but_check_spd = 0; 140#endif 141 M_ASSERTPKTHDR(m); 142 143 if (inp != NULL) { 144 INP_LOCK_ASSERT(inp); 145 M_SETFIB(m, inp->inp_inc.inc_fibnum); 146 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) { 147 m->m_pkthdr.flowid = inp->inp_flowid; 148 m->m_flags |= M_FLOWID; 149 } 150 } 151 152 if (ro == NULL) { 153 ro = &iproute; 154 bzero(ro, sizeof (*ro)); 155 } 156 157#ifdef FLOWTABLE 158 if (ro->ro_rt == NULL) 159 (void )flowtable_lookup(AF_INET, m, ro); 160#endif 161 162 if (opt) { 163 int len = 0; 164 m = ip_insertoptions(m, opt, &len); 165 if (len != 0) 166 hlen = len; /* ip->ip_hl is updated above */ 167 } 168 ip = mtod(m, struct ip *); 169 ip_len = ntohs(ip->ip_len); 170 ip_off = ntohs(ip->ip_off); 171 172 /* 173 * Fill in IP header. If we are not allowing fragmentation, 174 * then the ip_id field is meaningless, but we don't set it 175 * to zero. Doing so causes various problems when devices along 176 * the path (routers, load balancers, firewalls, etc.) illegally 177 * disable DF on our packet. Note that a 16-bit counter 178 * will wrap around in less than 10 seconds at 100 Mbit/s on a 179 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 180 * for Counting NATted Hosts", Proc. IMW'02, available at 181 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 182 */ 183 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 184 ip->ip_v = IPVERSION; 185 ip->ip_hl = hlen >> 2; 186 ip->ip_id = ip_newid(); 187 IPSTAT_INC(ips_localout); 188 } else { 189 /* Header already set, fetch hlen from there */ 190 hlen = ip->ip_hl << 2; 191 } 192 193 /* 194 * dst/gw handling: 195 * 196 * dst can be rewritten but always point to &ro->ro_dst 197 * gw is readonly but can be pointed either to dst OR rt_gatewy 198 * therefore we need restore GW if we're re-doing lookup 199 */ 200 gw = dst = (struct sockaddr_in *)&ro->ro_dst; 201again: 202 ia = NULL; 203 /* 204 * If there is a cached route, 205 * check that it is to the same destination 206 * and is still up. If not, free it and try again. 207 * The address family should also be checked in case of sharing the 208 * cache with IPv6. 209 */ 210 rte = ro->ro_rt; 211 if (rte && ((rte->rt_flags & RTF_UP) == 0 || 212 rte->rt_ifp == NULL || 213 !RT_LINK_IS_UP(rte->rt_ifp) || 214 dst->sin_family != AF_INET || 215 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 216 RO_RTFREE(ro); 217 ro->ro_lle = NULL; 218 rte = NULL; 219 gw = dst; 220 } 221 if (rte == NULL && fwd_tag == NULL) { 222 bzero(dst, sizeof(*dst)); 223 dst->sin_family = AF_INET; 224 dst->sin_len = sizeof(*dst); 225 dst->sin_addr = ip->ip_dst; 226 } 227 /* 228 * If routing to interface only, short circuit routing lookup. 229 * The use of an all-ones broadcast address implies this; an 230 * interface is specified by the broadcast address of an interface, 231 * or the destination address of a ptp interface. 232 */ 233 if (flags & IP_SENDONES) { 234 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL && 235 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) { 236 IPSTAT_INC(ips_noroute); 237 error = ENETUNREACH; 238 goto bad; 239 } 240 ip->ip_dst.s_addr = INADDR_BROADCAST; 241 dst->sin_addr = ip->ip_dst; 242 ifp = ia->ia_ifp; 243 ip->ip_ttl = 1; 244 isbroadcast = 1; 245 } else if (flags & IP_ROUTETOIF) { 246 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 247 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) { 248 IPSTAT_INC(ips_noroute); 249 error = ENETUNREACH; 250 goto bad; 251 } 252 ifp = ia->ia_ifp; 253 ip->ip_ttl = 1; 254 isbroadcast = in_broadcast(dst->sin_addr, ifp); 255 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 256 imo != NULL && imo->imo_multicast_ifp != NULL) { 257 /* 258 * Bypass the normal routing lookup for multicast 259 * packets if the interface is specified. 260 */ 261 ifp = imo->imo_multicast_ifp; 262 IFP_TO_IA(ifp, ia); 263 isbroadcast = 0; /* fool gcc */ 264 } else { 265 /* 266 * We want to do any cloning requested by the link layer, 267 * as this is probably required in all cases for correct 268 * operation (as it is for ARP). 269 */ 270 if (rte == NULL) { 271#ifdef RADIX_MPATH 272 rtalloc_mpath_fib(ro, 273 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 274 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 275#else 276 in_rtalloc_ign(ro, 0, 277 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 278#endif 279 rte = ro->ro_rt; 280 } 281 if (rte == NULL || 282 rte->rt_ifp == NULL || 283 !RT_LINK_IS_UP(rte->rt_ifp)) { 284#ifdef IPSEC 285 /* 286 * There is no route for this packet, but it is 287 * possible that a matching SPD entry exists. 288 */ 289 no_route_but_check_spd = 1; 290 mtu = 0; /* Silence GCC warning. */ 291 goto sendit; 292#endif 293 IPSTAT_INC(ips_noroute); 294 error = EHOSTUNREACH; 295 goto bad; 296 } 297 ia = ifatoia(rte->rt_ifa); 298 ifp = rte->rt_ifp; 299 counter_u64_add(rte->rt_pksent, 1); 300 if (rte->rt_flags & RTF_GATEWAY) 301 gw = (struct sockaddr_in *)rte->rt_gateway; 302 if (rte->rt_flags & RTF_HOST) 303 isbroadcast = (rte->rt_flags & RTF_BROADCAST); 304 else 305 isbroadcast = in_broadcast(gw->sin_addr, ifp); 306 } 307 /* 308 * Calculate MTU. If we have a route that is up, use that, 309 * otherwise use the interface's MTU. 310 */ 311 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) { 312 /* 313 * This case can happen if the user changed the MTU 314 * of an interface after enabling IP on it. Because 315 * most netifs don't keep track of routes pointing to 316 * them, there is no way for one to update all its 317 * routes when the MTU is changed. 318 */ 319 if (rte->rt_mtu > ifp->if_mtu) 320 rte->rt_mtu = ifp->if_mtu; 321 mtu = rte->rt_mtu; 322 } else { 323 mtu = ifp->if_mtu; 324 } 325 /* Catch a possible divide by zero later. */ 326 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p", 327 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp)); 328 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 329 m->m_flags |= M_MCAST; 330 /* 331 * IP destination address is multicast. Make sure "gw" 332 * still points to the address in "ro". (It may have been 333 * changed to point to a gateway address, above.) 334 */ 335 gw = dst; 336 /* 337 * See if the caller provided any multicast options 338 */ 339 if (imo != NULL) { 340 ip->ip_ttl = imo->imo_multicast_ttl; 341 if (imo->imo_multicast_vif != -1) 342 ip->ip_src.s_addr = 343 ip_mcast_src ? 344 ip_mcast_src(imo->imo_multicast_vif) : 345 INADDR_ANY; 346 } else 347 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 348 /* 349 * Confirm that the outgoing interface supports multicast. 350 */ 351 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 352 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 353 IPSTAT_INC(ips_noroute); 354 error = ENETUNREACH; 355 goto bad; 356 } 357 } 358 /* 359 * If source address not specified yet, use address 360 * of outgoing interface. 361 */ 362 if (ip->ip_src.s_addr == INADDR_ANY) { 363 /* Interface may have no addresses. */ 364 if (ia != NULL) 365 ip->ip_src = IA_SIN(ia)->sin_addr; 366 } 367 368 if ((imo == NULL && in_mcast_loop) || 369 (imo && imo->imo_multicast_loop)) { 370 /* 371 * Loop back multicast datagram if not expressly 372 * forbidden to do so, even if we are not a member 373 * of the group; ip_input() will filter it later, 374 * thus deferring a hash lookup and mutex acquisition 375 * at the expense of a cheap copy using m_copym(). 376 */ 377 ip_mloopback(ifp, m, dst, hlen); 378 } else { 379 /* 380 * If we are acting as a multicast router, perform 381 * multicast forwarding as if the packet had just 382 * arrived on the interface to which we are about 383 * to send. The multicast forwarding function 384 * recursively calls this function, using the 385 * IP_FORWARDING flag to prevent infinite recursion. 386 * 387 * Multicasts that are looped back by ip_mloopback(), 388 * above, will be forwarded by the ip_input() routine, 389 * if necessary. 390 */ 391 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { 392 /* 393 * If rsvp daemon is not running, do not 394 * set ip_moptions. This ensures that the packet 395 * is multicast and not just sent down one link 396 * as prescribed by rsvpd. 397 */ 398 if (!V_rsvp_on) 399 imo = NULL; 400 if (ip_mforward && 401 ip_mforward(ip, ifp, m, imo) != 0) { 402 m_freem(m); 403 goto done; 404 } 405 } 406 } 407 408 /* 409 * Multicasts with a time-to-live of zero may be looped- 410 * back, above, but must not be transmitted on a network. 411 * Also, multicasts addressed to the loopback interface 412 * are not sent -- the above call to ip_mloopback() will 413 * loop back a copy. ip_input() will drop the copy if 414 * this host does not belong to the destination group on 415 * the loopback interface. 416 */ 417 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 418 m_freem(m); 419 goto done; 420 } 421 422 goto sendit; 423 } 424 425 /* 426 * If the source address is not specified yet, use the address 427 * of the outoing interface. 428 */ 429 if (ip->ip_src.s_addr == INADDR_ANY) { 430 /* Interface may have no addresses. */ 431 if (ia != NULL) { 432 ip->ip_src = IA_SIN(ia)->sin_addr; 433 } 434 } 435 436 /* 437 * Verify that we have any chance at all of being able to queue the 438 * packet or packet fragments, unless ALTQ is enabled on the given 439 * interface in which case packetdrop should be done by queueing. 440 */ 441 n = ip_len / mtu + 1; /* how many fragments ? */ 442 if ( 443#ifdef ALTQ 444 (!ALTQ_IS_ENABLED(&ifp->if_snd)) && 445#endif /* ALTQ */ 446 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) { 447 error = ENOBUFS; 448 IPSTAT_INC(ips_odropped); 449 ifp->if_snd.ifq_drops += n; 450 goto bad; 451 } 452 453 /* 454 * Look for broadcast address and 455 * verify user is allowed to send 456 * such a packet. 457 */ 458 if (isbroadcast) { 459 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 460 error = EADDRNOTAVAIL; 461 goto bad; 462 } 463 if ((flags & IP_ALLOWBROADCAST) == 0) { 464 error = EACCES; 465 goto bad; 466 } 467 /* don't allow broadcast messages to be fragmented */ 468 if (ip_len > mtu) { 469 error = EMSGSIZE; 470 goto bad; 471 } 472 m->m_flags |= M_BCAST; 473 } else { 474 m->m_flags &= ~M_BCAST; 475 } 476 477sendit: 478#ifdef IPSEC 479 switch(ip_ipsec_output(&m, inp, &flags, &error)) { 480 case 1: 481 goto bad; 482 case -1: 483 goto done; 484 case 0: 485 default: 486 break; /* Continue with packet processing. */ 487 } 488 /* 489 * Check if there was a route for this packet; return error if not. 490 */ 491 if (no_route_but_check_spd) { 492 IPSTAT_INC(ips_noroute); 493 error = EHOSTUNREACH; 494 goto bad; 495 } 496 /* Update variables that are affected by ipsec4_output(). */ 497 ip = mtod(m, struct ip *); 498 hlen = ip->ip_hl << 2; 499#endif /* IPSEC */ 500 501 /* Jump over all PFIL processing if hooks are not active. */ 502 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 503 goto passout; 504 505 /* Run through list of hooks for output packets. */ 506 odst.s_addr = ip->ip_dst.s_addr; 507 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 508 if (error != 0 || m == NULL) 509 goto done; 510 511 ip = mtod(m, struct ip *); 512 513 /* See if destination IP address was changed by packet filter. */ 514 if (odst.s_addr != ip->ip_dst.s_addr) { 515 m->m_flags |= M_SKIP_FIREWALL; 516 /* If destination is now ourself drop to ip_input(). */ 517 if (in_localip(ip->ip_dst)) { 518 m->m_flags |= M_FASTFWD_OURS; 519 if (m->m_pkthdr.rcvif == NULL) 520 m->m_pkthdr.rcvif = V_loif; 521 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 522 m->m_pkthdr.csum_flags |= 523 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 524 m->m_pkthdr.csum_data = 0xffff; 525 } 526 m->m_pkthdr.csum_flags |= 527 CSUM_IP_CHECKED | CSUM_IP_VALID; 528#ifdef SCTP 529 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 530 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 531#endif 532 error = netisr_queue(NETISR_IP, m); 533 goto done; 534 } else 535 goto again; /* Redo the routing table lookup. */ 536 } 537 538 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 539 if (m->m_flags & M_FASTFWD_OURS) { 540 if (m->m_pkthdr.rcvif == NULL) 541 m->m_pkthdr.rcvif = V_loif; 542 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 543 m->m_pkthdr.csum_flags |= 544 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 545 m->m_pkthdr.csum_data = 0xffff; 546 } 547#ifdef SCTP 548 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 549 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 550#endif 551 m->m_pkthdr.csum_flags |= 552 CSUM_IP_CHECKED | CSUM_IP_VALID; 553 554 error = netisr_queue(NETISR_IP, m); 555 goto done; 556 } 557 /* Or forward to some other address? */ 558 if ((m->m_flags & M_IP_NEXTHOP) && 559 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 560 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 561 m->m_flags |= M_SKIP_FIREWALL; 562 m->m_flags &= ~M_IP_NEXTHOP; 563 m_tag_delete(m, fwd_tag); 564 goto again; 565 } 566 567passout: 568 /* 127/8 must not appear on wire - RFC1122. */ 569 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 570 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 571 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 572 IPSTAT_INC(ips_badaddr); 573 error = EADDRNOTAVAIL; 574 goto bad; 575 } 576 } 577 578 m->m_pkthdr.csum_flags |= CSUM_IP; 579 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) { 580 in_delayed_cksum(m); 581 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 582 } 583#ifdef SCTP 584 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) { 585 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); 586 m->m_pkthdr.csum_flags &= ~CSUM_SCTP; 587 } 588#endif 589 590 /* 591 * If small enough for interface, or the interface will take 592 * care of the fragmentation for us, we can just send directly. 593 */ 594 if (ip_len <= mtu || 595 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 596 ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 597 ip->ip_sum = 0; 598 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) { 599 ip->ip_sum = in_cksum(m, hlen); 600 m->m_pkthdr.csum_flags &= ~CSUM_IP; 601 } 602 603 /* 604 * Record statistics for this interface address. 605 * With CSUM_TSO the byte/packet count will be slightly 606 * incorrect because we count the IP+TCP headers only 607 * once instead of for every generated packet. 608 */ 609 if (!(flags & IP_FORWARDING) && ia) { 610 if (m->m_pkthdr.csum_flags & CSUM_TSO) 611 ia->ia_ifa.if_opackets += 612 m->m_pkthdr.len / m->m_pkthdr.tso_segsz; 613 else 614 ia->ia_ifa.if_opackets++; 615 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 616 } 617#ifdef MBUF_STRESS_TEST 618 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 619 m = m_fragment(m, M_NOWAIT, mbuf_frag_size); 620#endif 621 /* 622 * Reset layer specific mbuf flags 623 * to avoid confusing lower layers. 624 */ 625 m_clrprotoflags(m); 626 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 627 error = (*ifp->if_output)(ifp, m, 628 (const struct sockaddr *)gw, ro); 629 goto done; 630 } 631 632 /* Balk when DF bit is set or the interface didn't support TSO. */ 633 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 634 error = EMSGSIZE; 635 IPSTAT_INC(ips_cantfrag); 636 goto bad; 637 } 638 639 /* 640 * Too large for interface; fragment if possible. If successful, 641 * on return, m will point to a list of packets to be sent. 642 */ 643 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist); 644 if (error) 645 goto bad; 646 for (; m; m = m0) { 647 m0 = m->m_nextpkt; 648 m->m_nextpkt = 0; 649 if (error == 0) { 650 /* Record statistics for this interface address. */ 651 if (ia != NULL) { 652 ia->ia_ifa.if_opackets++; 653 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 654 } 655 /* 656 * Reset layer specific mbuf flags 657 * to avoid confusing upper layers. 658 */ 659 m_clrprotoflags(m); 660 661 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 662 error = (*ifp->if_output)(ifp, m, 663 (const struct sockaddr *)gw, ro); 664 } else 665 m_freem(m); 666 } 667 668 if (error == 0) 669 IPSTAT_INC(ips_fragmented); 670 671done: 672 if (ro == &iproute) 673 RO_RTFREE(ro); 674 return (error); 675bad: 676 m_freem(m); 677 goto done; 678} 679 680/* 681 * Create a chain of fragments which fit the given mtu. m_frag points to the 682 * mbuf to be fragmented; on return it points to the chain with the fragments. 683 * Return 0 if no error. If error, m_frag may contain a partially built 684 * chain of fragments that should be freed by the caller. 685 * 686 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 687 */ 688int 689ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 690 u_long if_hwassist_flags) 691{ 692 int error = 0; 693 int hlen = ip->ip_hl << 2; 694 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 695 int off; 696 struct mbuf *m0 = *m_frag; /* the original packet */ 697 int firstlen; 698 struct mbuf **mnext; 699 int nfrags; 700 uint16_t ip_len, ip_off; 701 702 ip_len = ntohs(ip->ip_len); 703 ip_off = ntohs(ip->ip_off); 704 705 if (ip_off & IP_DF) { /* Fragmentation not allowed */ 706 IPSTAT_INC(ips_cantfrag); 707 return EMSGSIZE; 708 } 709 710 /* 711 * Must be able to put at least 8 bytes per fragment. 712 */ 713 if (len < 8) 714 return EMSGSIZE; 715 716 /* 717 * If the interface will not calculate checksums on 718 * fragmented packets, then do it here. 719 */ 720 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 721 in_delayed_cksum(m0); 722 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 723 } 724#ifdef SCTP 725 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) { 726 sctp_delayed_cksum(m0, hlen); 727 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; 728 } 729#endif 730 if (len > PAGE_SIZE) { 731 /* 732 * Fragment large datagrams such that each segment 733 * contains a multiple of PAGE_SIZE amount of data, 734 * plus headers. This enables a receiver to perform 735 * page-flipping zero-copy optimizations. 736 * 737 * XXX When does this help given that sender and receiver 738 * could have different page sizes, and also mtu could 739 * be less than the receiver's page size ? 740 */ 741 int newlen; 742 struct mbuf *m; 743 744 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 745 off += m->m_len; 746 747 /* 748 * firstlen (off - hlen) must be aligned on an 749 * 8-byte boundary 750 */ 751 if (off < hlen) 752 goto smart_frag_failure; 753 off = ((off - hlen) & ~7) + hlen; 754 newlen = (~PAGE_MASK) & mtu; 755 if ((newlen + sizeof (struct ip)) > mtu) { 756 /* we failed, go back the default */ 757smart_frag_failure: 758 newlen = len; 759 off = hlen + len; 760 } 761 len = newlen; 762 763 } else { 764 off = hlen + len; 765 } 766 767 firstlen = off - hlen; 768 mnext = &m0->m_nextpkt; /* pointer to next packet */ 769 770 /* 771 * Loop through length of segment after first fragment, 772 * make new header and copy data of each part and link onto chain. 773 * Here, m0 is the original packet, m is the fragment being created. 774 * The fragments are linked off the m_nextpkt of the original 775 * packet, which after processing serves as the first fragment. 776 */ 777 for (nfrags = 1; off < ip_len; off += len, nfrags++) { 778 struct ip *mhip; /* ip header on the fragment */ 779 struct mbuf *m; 780 int mhlen = sizeof (struct ip); 781 782 m = m_gethdr(M_NOWAIT, MT_DATA); 783 if (m == NULL) { 784 error = ENOBUFS; 785 IPSTAT_INC(ips_odropped); 786 goto done; 787 } 788 m->m_flags |= (m0->m_flags & M_MCAST); 789 /* 790 * In the first mbuf, leave room for the link header, then 791 * copy the original IP header including options. The payload 792 * goes into an additional mbuf chain returned by m_copym(). 793 */ 794 m->m_data += max_linkhdr; 795 mhip = mtod(m, struct ip *); 796 *mhip = *ip; 797 if (hlen > sizeof (struct ip)) { 798 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 799 mhip->ip_v = IPVERSION; 800 mhip->ip_hl = mhlen >> 2; 801 } 802 m->m_len = mhlen; 803 /* XXX do we need to add ip_off below ? */ 804 mhip->ip_off = ((off - hlen) >> 3) + ip_off; 805 if (off + len >= ip_len) 806 len = ip_len - off; 807 else 808 mhip->ip_off |= IP_MF; 809 mhip->ip_len = htons((u_short)(len + mhlen)); 810 m->m_next = m_copym(m0, off, len, M_NOWAIT); 811 if (m->m_next == NULL) { /* copy failed */ 812 m_free(m); 813 error = ENOBUFS; /* ??? */ 814 IPSTAT_INC(ips_odropped); 815 goto done; 816 } 817 m->m_pkthdr.len = mhlen + len; 818 m->m_pkthdr.rcvif = NULL; 819#ifdef MAC 820 mac_netinet_fragment(m0, m); 821#endif 822 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 823 mhip->ip_off = htons(mhip->ip_off); 824 mhip->ip_sum = 0; 825 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 826 mhip->ip_sum = in_cksum(m, mhlen); 827 m->m_pkthdr.csum_flags &= ~CSUM_IP; 828 } 829 *mnext = m; 830 mnext = &m->m_nextpkt; 831 } 832 IPSTAT_ADD(ips_ofragments, nfrags); 833 834 /* 835 * Update first fragment by trimming what's been copied out 836 * and updating header. 837 */ 838 m_adj(m0, hlen + firstlen - ip_len); 839 m0->m_pkthdr.len = hlen + firstlen; 840 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 841 ip->ip_off = htons(ip_off | IP_MF); 842 ip->ip_sum = 0; 843 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 844 ip->ip_sum = in_cksum(m0, hlen); 845 m0->m_pkthdr.csum_flags &= ~CSUM_IP; 846 } 847 848done: 849 *m_frag = m0; 850 return error; 851} 852 853void 854in_delayed_cksum(struct mbuf *m) 855{ 856 struct ip *ip; 857 uint16_t csum, offset, ip_len; 858 859 ip = mtod(m, struct ip *); 860 offset = ip->ip_hl << 2 ; 861 ip_len = ntohs(ip->ip_len); 862 csum = in_cksum_skip(m, ip_len, offset); 863 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 864 csum = 0xffff; 865 offset += m->m_pkthdr.csum_data; /* checksum offset */ 866 867 /* find the mbuf in the chain where the checksum starts*/ 868 while ((m != NULL) && (offset >= m->m_len)) { 869 offset -= m->m_len; 870 m = m->m_next; 871 } 872 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain.")); 873 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs.")); 874 *(u_short *)(m->m_data + offset) = csum; 875} 876 877/* 878 * IP socket option processing. 879 */ 880int 881ip_ctloutput(struct socket *so, struct sockopt *sopt) 882{ 883 struct inpcb *inp = sotoinpcb(so); 884 int error, optval; 885 886 error = optval = 0; 887 if (sopt->sopt_level != IPPROTO_IP) { 888 error = EINVAL; 889 890 if (sopt->sopt_level == SOL_SOCKET && 891 sopt->sopt_dir == SOPT_SET) { 892 switch (sopt->sopt_name) { 893 case SO_REUSEADDR: 894 INP_WLOCK(inp); 895 if ((so->so_options & SO_REUSEADDR) != 0) 896 inp->inp_flags2 |= INP_REUSEADDR; 897 else 898 inp->inp_flags2 &= ~INP_REUSEADDR; 899 INP_WUNLOCK(inp); 900 error = 0; 901 break; 902 case SO_REUSEPORT: 903 INP_WLOCK(inp); 904 if ((so->so_options & SO_REUSEPORT) != 0) 905 inp->inp_flags2 |= INP_REUSEPORT; 906 else 907 inp->inp_flags2 &= ~INP_REUSEPORT; 908 INP_WUNLOCK(inp); 909 error = 0; 910 break; 911 case SO_SETFIB: 912 INP_WLOCK(inp); 913 inp->inp_inc.inc_fibnum = so->so_fibnum; 914 INP_WUNLOCK(inp); 915 error = 0; 916 break; 917 default: 918 break; 919 } 920 } 921 return (error); 922 } 923 924 switch (sopt->sopt_dir) { 925 case SOPT_SET: 926 switch (sopt->sopt_name) { 927 case IP_OPTIONS: 928#ifdef notyet 929 case IP_RETOPTS: 930#endif 931 { 932 struct mbuf *m; 933 if (sopt->sopt_valsize > MLEN) { 934 error = EMSGSIZE; 935 break; 936 } 937 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA); 938 if (m == NULL) { 939 error = ENOBUFS; 940 break; 941 } 942 m->m_len = sopt->sopt_valsize; 943 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 944 m->m_len); 945 if (error) { 946 m_free(m); 947 break; 948 } 949 INP_WLOCK(inp); 950 error = ip_pcbopts(inp, sopt->sopt_name, m); 951 INP_WUNLOCK(inp); 952 return (error); 953 } 954 955 case IP_BINDANY: 956 if (sopt->sopt_td != NULL) { 957 error = priv_check(sopt->sopt_td, 958 PRIV_NETINET_BINDANY); 959 if (error) 960 break; 961 } 962 /* FALLTHROUGH */ 963 case IP_TOS: 964 case IP_TTL: 965 case IP_MINTTL: 966 case IP_RECVOPTS: 967 case IP_RECVRETOPTS: 968 case IP_RECVDSTADDR: 969 case IP_RECVTTL: 970 case IP_RECVIF: 971 case IP_FAITH: 972 case IP_ONESBCAST: 973 case IP_DONTFRAG: 974 case IP_RECVTOS: 975 error = sooptcopyin(sopt, &optval, sizeof optval, 976 sizeof optval); 977 if (error) 978 break; 979 980 switch (sopt->sopt_name) { 981 case IP_TOS: 982 inp->inp_ip_tos = optval; 983 break; 984 985 case IP_TTL: 986 inp->inp_ip_ttl = optval; 987 break; 988 989 case IP_MINTTL: 990 if (optval >= 0 && optval <= MAXTTL) 991 inp->inp_ip_minttl = optval; 992 else 993 error = EINVAL; 994 break; 995 996#define OPTSET(bit) do { \ 997 INP_WLOCK(inp); \ 998 if (optval) \ 999 inp->inp_flags |= bit; \ 1000 else \ 1001 inp->inp_flags &= ~bit; \ 1002 INP_WUNLOCK(inp); \ 1003} while (0) 1004 1005 case IP_RECVOPTS: 1006 OPTSET(INP_RECVOPTS); 1007 break; 1008 1009 case IP_RECVRETOPTS: 1010 OPTSET(INP_RECVRETOPTS); 1011 break; 1012 1013 case IP_RECVDSTADDR: 1014 OPTSET(INP_RECVDSTADDR); 1015 break; 1016 1017 case IP_RECVTTL: 1018 OPTSET(INP_RECVTTL); 1019 break; 1020 1021 case IP_RECVIF: 1022 OPTSET(INP_RECVIF); 1023 break; 1024 1025 case IP_FAITH: 1026 OPTSET(INP_FAITH); 1027 break; 1028 1029 case IP_ONESBCAST: 1030 OPTSET(INP_ONESBCAST); 1031 break; 1032 case IP_DONTFRAG: 1033 OPTSET(INP_DONTFRAG); 1034 break; 1035 case IP_BINDANY: 1036 OPTSET(INP_BINDANY); 1037 break; 1038 case IP_RECVTOS: 1039 OPTSET(INP_RECVTOS); 1040 break; 1041 } 1042 break; 1043#undef OPTSET 1044 1045 /* 1046 * Multicast socket options are processed by the in_mcast 1047 * module. 1048 */ 1049 case IP_MULTICAST_IF: 1050 case IP_MULTICAST_VIF: 1051 case IP_MULTICAST_TTL: 1052 case IP_MULTICAST_LOOP: 1053 case IP_ADD_MEMBERSHIP: 1054 case IP_DROP_MEMBERSHIP: 1055 case IP_ADD_SOURCE_MEMBERSHIP: 1056 case IP_DROP_SOURCE_MEMBERSHIP: 1057 case IP_BLOCK_SOURCE: 1058 case IP_UNBLOCK_SOURCE: 1059 case IP_MSFILTER: 1060 case MCAST_JOIN_GROUP: 1061 case MCAST_LEAVE_GROUP: 1062 case MCAST_JOIN_SOURCE_GROUP: 1063 case MCAST_LEAVE_SOURCE_GROUP: 1064 case MCAST_BLOCK_SOURCE: 1065 case MCAST_UNBLOCK_SOURCE: 1066 error = inp_setmoptions(inp, sopt); 1067 break; 1068 1069 case IP_PORTRANGE: 1070 error = sooptcopyin(sopt, &optval, sizeof optval, 1071 sizeof optval); 1072 if (error) 1073 break; 1074 1075 INP_WLOCK(inp); 1076 switch (optval) { 1077 case IP_PORTRANGE_DEFAULT: 1078 inp->inp_flags &= ~(INP_LOWPORT); 1079 inp->inp_flags &= ~(INP_HIGHPORT); 1080 break; 1081 1082 case IP_PORTRANGE_HIGH: 1083 inp->inp_flags &= ~(INP_LOWPORT); 1084 inp->inp_flags |= INP_HIGHPORT; 1085 break; 1086 1087 case IP_PORTRANGE_LOW: 1088 inp->inp_flags &= ~(INP_HIGHPORT); 1089 inp->inp_flags |= INP_LOWPORT; 1090 break; 1091 1092 default: 1093 error = EINVAL; 1094 break; 1095 } 1096 INP_WUNLOCK(inp); 1097 break; 1098 1099#ifdef IPSEC 1100 case IP_IPSEC_POLICY: 1101 { 1102 caddr_t req; 1103 struct mbuf *m; 1104 1105 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1106 break; 1107 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1108 break; 1109 req = mtod(m, caddr_t); 1110 error = ipsec_set_policy(inp, sopt->sopt_name, req, 1111 m->m_len, (sopt->sopt_td != NULL) ? 1112 sopt->sopt_td->td_ucred : NULL); 1113 m_freem(m); 1114 break; 1115 } 1116#endif /* IPSEC */ 1117 1118 default: 1119 error = ENOPROTOOPT; 1120 break; 1121 } 1122 break; 1123 1124 case SOPT_GET: 1125 switch (sopt->sopt_name) { 1126 case IP_OPTIONS: 1127 case IP_RETOPTS: 1128 if (inp->inp_options) 1129 error = sooptcopyout(sopt, 1130 mtod(inp->inp_options, 1131 char *), 1132 inp->inp_options->m_len); 1133 else 1134 sopt->sopt_valsize = 0; 1135 break; 1136 1137 case IP_TOS: 1138 case IP_TTL: 1139 case IP_MINTTL: 1140 case IP_RECVOPTS: 1141 case IP_RECVRETOPTS: 1142 case IP_RECVDSTADDR: 1143 case IP_RECVTTL: 1144 case IP_RECVIF: 1145 case IP_PORTRANGE: 1146 case IP_FAITH: 1147 case IP_ONESBCAST: 1148 case IP_DONTFRAG: 1149 case IP_BINDANY: 1150 case IP_RECVTOS: 1151 switch (sopt->sopt_name) { 1152 1153 case IP_TOS: 1154 optval = inp->inp_ip_tos; 1155 break; 1156 1157 case IP_TTL: 1158 optval = inp->inp_ip_ttl; 1159 break; 1160 1161 case IP_MINTTL: 1162 optval = inp->inp_ip_minttl; 1163 break; 1164 1165#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1166 1167 case IP_RECVOPTS: 1168 optval = OPTBIT(INP_RECVOPTS); 1169 break; 1170 1171 case IP_RECVRETOPTS: 1172 optval = OPTBIT(INP_RECVRETOPTS); 1173 break; 1174 1175 case IP_RECVDSTADDR: 1176 optval = OPTBIT(INP_RECVDSTADDR); 1177 break; 1178 1179 case IP_RECVTTL: 1180 optval = OPTBIT(INP_RECVTTL); 1181 break; 1182 1183 case IP_RECVIF: 1184 optval = OPTBIT(INP_RECVIF); 1185 break; 1186 1187 case IP_PORTRANGE: 1188 if (inp->inp_flags & INP_HIGHPORT) 1189 optval = IP_PORTRANGE_HIGH; 1190 else if (inp->inp_flags & INP_LOWPORT) 1191 optval = IP_PORTRANGE_LOW; 1192 else 1193 optval = 0; 1194 break; 1195 1196 case IP_FAITH: 1197 optval = OPTBIT(INP_FAITH); 1198 break; 1199 1200 case IP_ONESBCAST: 1201 optval = OPTBIT(INP_ONESBCAST); 1202 break; 1203 case IP_DONTFRAG: 1204 optval = OPTBIT(INP_DONTFRAG); 1205 break; 1206 case IP_BINDANY: 1207 optval = OPTBIT(INP_BINDANY); 1208 break; 1209 case IP_RECVTOS: 1210 optval = OPTBIT(INP_RECVTOS); 1211 break; 1212 } 1213 error = sooptcopyout(sopt, &optval, sizeof optval); 1214 break; 1215 1216 /* 1217 * Multicast socket options are processed by the in_mcast 1218 * module. 1219 */ 1220 case IP_MULTICAST_IF: 1221 case IP_MULTICAST_VIF: 1222 case IP_MULTICAST_TTL: 1223 case IP_MULTICAST_LOOP: 1224 case IP_MSFILTER: 1225 error = inp_getmoptions(inp, sopt); 1226 break; 1227 1228#ifdef IPSEC 1229 case IP_IPSEC_POLICY: 1230 { 1231 struct mbuf *m = NULL; 1232 caddr_t req = NULL; 1233 size_t len = 0; 1234 1235 if (m != 0) { 1236 req = mtod(m, caddr_t); 1237 len = m->m_len; 1238 } 1239 error = ipsec_get_policy(sotoinpcb(so), req, len, &m); 1240 if (error == 0) 1241 error = soopt_mcopyout(sopt, m); /* XXX */ 1242 if (error == 0) 1243 m_freem(m); 1244 break; 1245 } 1246#endif /* IPSEC */ 1247 1248 default: 1249 error = ENOPROTOOPT; 1250 break; 1251 } 1252 break; 1253 } 1254 return (error); 1255} 1256 1257/* 1258 * Routine called from ip_output() to loop back a copy of an IP multicast 1259 * packet to the input queue of a specified interface. Note that this 1260 * calls the output routine of the loopback "driver", but with an interface 1261 * pointer that might NOT be a loopback interface -- evil, but easier than 1262 * replicating that code here. 1263 */ 1264static void 1265ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst, 1266 int hlen) 1267{ 1268 register struct ip *ip; 1269 struct mbuf *copym; 1270 1271 /* 1272 * Make a deep copy of the packet because we're going to 1273 * modify the pack in order to generate checksums. 1274 */ 1275 copym = m_dup(m, M_NOWAIT); 1276 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1277 copym = m_pullup(copym, hlen); 1278 if (copym != NULL) { 1279 /* If needed, compute the checksum and mark it as valid. */ 1280 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1281 in_delayed_cksum(copym); 1282 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1283 copym->m_pkthdr.csum_flags |= 1284 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1285 copym->m_pkthdr.csum_data = 0xffff; 1286 } 1287 /* 1288 * We don't bother to fragment if the IP length is greater 1289 * than the interface's MTU. Can this possibly matter? 1290 */ 1291 ip = mtod(copym, struct ip *); 1292 ip->ip_sum = 0; 1293 ip->ip_sum = in_cksum(copym, hlen); 1294#if 1 /* XXX */ 1295 if (dst->sin_family != AF_INET) { 1296 printf("ip_mloopback: bad address family %d\n", 1297 dst->sin_family); 1298 dst->sin_family = AF_INET; 1299 } 1300#endif 1301 if_simloop(ifp, copym, dst->sin_family, 0); 1302 } 1303} 1304