ip_output.c revision 191621
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: head/sys/netinet/ip_output.c 191621 2009-04-28 11:10:33Z trasz $"); 34 35#include "opt_ipfw.h" 36#include "opt_inet.h" 37#include "opt_ipsec.h" 38#include "opt_route.h" 39#include "opt_mac.h" 40#include "opt_mbuf_stress_test.h" 41#include "opt_mpath.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/socket.h> 53#include <sys/socketvar.h> 54#include <sys/sysctl.h> 55#include <sys/ucred.h> 56#include <sys/vimage.h> 57 58#include <net/if.h> 59#include <net/netisr.h> 60#include <net/pfil.h> 61#include <net/route.h> 62#include <net/flowtable.h> 63#ifdef RADIX_MPATH 64#include <net/radix_mpath.h> 65#endif 66#include <net/vnet.h> 67 68#include <netinet/in.h> 69#include <netinet/in_systm.h> 70#include <netinet/ip.h> 71#include <netinet/in_pcb.h> 72#include <netinet/in_var.h> 73#include <netinet/ip_var.h> 74#include <netinet/ip_options.h> 75#include <netinet/vinet.h> 76#ifdef SCTP 77#include <netinet/sctp.h> 78#include <netinet/sctp_crc32.h> 79#endif 80 81#ifdef IPSEC 82#include <netinet/ip_ipsec.h> 83#include <netipsec/ipsec.h> 84#endif /* IPSEC*/ 85 86#include <machine/in_cksum.h> 87 88#include <security/mac/mac_framework.h> 89 90#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 91 x, (ntohl(a.s_addr)>>24)&0xFF,\ 92 (ntohl(a.s_addr)>>16)&0xFF,\ 93 (ntohl(a.s_addr)>>8)&0xFF,\ 94 (ntohl(a.s_addr))&0xFF, y); 95 96#ifdef VIMAGE_GLOBALS 97u_short ip_id; 98#endif 99 100#ifdef MBUF_STRESS_TEST 101int mbuf_frag_size = 0; 102SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 103 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 104#endif 105 106#if defined(IP_NONLOCALBIND) 107static int ip_nonlocalok = 0; 108SYSCTL_INT(_net_inet_ip, OID_AUTO, nonlocalok, 109 CTLFLAG_RW|CTLFLAG_SECURE, &ip_nonlocalok, 0, ""); 110#endif 111 112static void ip_mloopback 113 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 114 115 116extern int in_mcast_loop; 117extern struct protosw inetsw[]; 118 119/* 120 * IP output. The packet in mbuf chain m contains a skeletal IP 121 * header (with len, off, ttl, proto, tos, src, dst). 122 * The mbuf chain containing the packet will be freed. 123 * The mbuf opt, if present, will not be freed. 124 * In the IP forwarding case, the packet will arrive with options already 125 * inserted, so must have a NULL opt pointer. 126 */ 127int 128ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags, 129 struct ip_moptions *imo, struct inpcb *inp) 130{ 131 INIT_VNET_NET(curvnet); 132 INIT_VNET_INET(curvnet); 133 struct ip *ip; 134 struct ifnet *ifp = NULL; /* keep compiler happy */ 135 struct mbuf *m0; 136 int hlen = sizeof (struct ip); 137 int mtu; 138 int len, error = 0; 139 int nortfree = 0; 140 struct sockaddr_in *dst = NULL; /* keep compiler happy */ 141 struct in_ifaddr *ia = NULL; 142 int isbroadcast, sw_csum; 143 struct route iproute; 144 struct in_addr odst; 145#ifdef IPFIREWALL_FORWARD 146 struct m_tag *fwd_tag = NULL; 147#endif 148#ifdef IPSEC 149 int no_route_but_check_spd = 0; 150#endif 151 M_ASSERTPKTHDR(m); 152 153 if (ro == NULL) { 154 ro = &iproute; 155 bzero(ro, sizeof (*ro)); 156 } 157 158 if (inp != NULL) { 159 M_SETFIB(m, inp->inp_inc.inc_fibnum); 160 INP_LOCK_ASSERT(inp); 161 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) { 162 m->m_pkthdr.flowid = inp->inp_flowid; 163 m->m_flags |= M_FLOWID; 164 } 165 } 166 if ((ro == &iproute) && (ro->ro_rt == NULL) && (ro->ro_lle == NULL)) { 167 if (flowtable_lookup(ip_ft, m, ro) == 0) 168 nortfree = 1; 169 } 170 171 if (opt) { 172 len = 0; 173 m = ip_insertoptions(m, opt, &len); 174 if (len != 0) 175 hlen = len; 176 } 177 ip = mtod(m, struct ip *); 178 179 /* 180 * Fill in IP header. If we are not allowing fragmentation, 181 * then the ip_id field is meaningless, but we don't set it 182 * to zero. Doing so causes various problems when devices along 183 * the path (routers, load balancers, firewalls, etc.) illegally 184 * disable DF on our packet. Note that a 16-bit counter 185 * will wrap around in less than 10 seconds at 100 Mbit/s on a 186 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 187 * for Counting NATted Hosts", Proc. IMW'02, available at 188 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 189 */ 190 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 191 ip->ip_v = IPVERSION; 192 ip->ip_hl = hlen >> 2; 193 ip->ip_id = ip_newid(); 194 IPSTAT_INC(ips_localout); 195 } else { 196 hlen = ip->ip_hl << 2; 197 } 198 199 dst = (struct sockaddr_in *)&ro->ro_dst; 200again: 201 /* 202 * If there is a cached route, 203 * check that it is to the same destination 204 * and is still up. If not, free it and try again. 205 * The address family should also be checked in case of sharing the 206 * cache with IPv6. 207 */ 208 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 209 dst->sin_family != AF_INET || 210 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 211 if (!nortfree) 212 RTFREE(ro->ro_rt); 213 ro->ro_rt = (struct rtentry *)NULL; 214 } 215#ifdef IPFIREWALL_FORWARD 216 if (ro->ro_rt == NULL && fwd_tag == NULL) { 217#else 218 if (ro->ro_rt == NULL) { 219#endif 220 bzero(dst, sizeof(*dst)); 221 dst->sin_family = AF_INET; 222 dst->sin_len = sizeof(*dst); 223 dst->sin_addr = ip->ip_dst; 224 } 225 /* 226 * If routing to interface only, short circuit routing lookup. 227 * The use of an all-ones broadcast address implies this; an 228 * interface is specified by the broadcast address of an interface, 229 * or the destination address of a ptp interface. 230 */ 231 if (flags & IP_SENDONES) { 232 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL && 233 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) { 234 IPSTAT_INC(ips_noroute); 235 error = ENETUNREACH; 236 goto bad; 237 } 238 ip->ip_dst.s_addr = INADDR_BROADCAST; 239 dst->sin_addr = ip->ip_dst; 240 ifp = ia->ia_ifp; 241 ip->ip_ttl = 1; 242 isbroadcast = 1; 243 } else if (flags & IP_ROUTETOIF) { 244 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 245 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) { 246 IPSTAT_INC(ips_noroute); 247 error = ENETUNREACH; 248 goto bad; 249 } 250 ifp = ia->ia_ifp; 251 ip->ip_ttl = 1; 252 isbroadcast = in_broadcast(dst->sin_addr, ifp); 253 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 254 imo != NULL && imo->imo_multicast_ifp != NULL) { 255 /* 256 * Bypass the normal routing lookup for multicast 257 * packets if the interface is specified. 258 */ 259 ifp = imo->imo_multicast_ifp; 260 IFP_TO_IA(ifp, ia); 261 isbroadcast = 0; /* fool gcc */ 262 } else { 263 /* 264 * We want to do any cloning requested by the link layer, 265 * as this is probably required in all cases for correct 266 * operation (as it is for ARP). 267 */ 268 if (ro->ro_rt == NULL) 269#ifdef RADIX_MPATH 270 rtalloc_mpath_fib(ro, 271 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 272 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 273#else 274 in_rtalloc_ign(ro, 0, 275 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 276#endif 277 if (ro->ro_rt == NULL) { 278#ifdef IPSEC 279 /* 280 * There is no route for this packet, but it is 281 * possible that a matching SPD entry exists. 282 */ 283 no_route_but_check_spd = 1; 284 mtu = 0; /* Silence GCC warning. */ 285 goto sendit; 286#endif 287 IPSTAT_INC(ips_noroute); 288 error = EHOSTUNREACH; 289 goto bad; 290 } 291 ia = ifatoia(ro->ro_rt->rt_ifa); 292 ifp = ro->ro_rt->rt_ifp; 293 ro->ro_rt->rt_rmx.rmx_pksent++; 294 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 295 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 296 if (ro->ro_rt->rt_flags & RTF_HOST) 297 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 298 else 299 isbroadcast = in_broadcast(dst->sin_addr, ifp); 300 } 301 /* 302 * Calculate MTU. If we have a route that is up, use that, 303 * otherwise use the interface's MTU. 304 */ 305 if (ro->ro_rt != NULL && (ro->ro_rt->rt_flags & (RTF_UP|RTF_HOST))) { 306 /* 307 * This case can happen if the user changed the MTU 308 * of an interface after enabling IP on it. Because 309 * most netifs don't keep track of routes pointing to 310 * them, there is no way for one to update all its 311 * routes when the MTU is changed. 312 */ 313 if (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu) 314 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 315 mtu = ro->ro_rt->rt_rmx.rmx_mtu; 316 } else { 317 mtu = ifp->if_mtu; 318 } 319 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 320 m->m_flags |= M_MCAST; 321 /* 322 * IP destination address is multicast. Make sure "dst" 323 * still points to the address in "ro". (It may have been 324 * changed to point to a gateway address, above.) 325 */ 326 dst = (struct sockaddr_in *)&ro->ro_dst; 327 /* 328 * See if the caller provided any multicast options 329 */ 330 if (imo != NULL) { 331 ip->ip_ttl = imo->imo_multicast_ttl; 332 if (imo->imo_multicast_vif != -1) 333 ip->ip_src.s_addr = 334 ip_mcast_src ? 335 ip_mcast_src(imo->imo_multicast_vif) : 336 INADDR_ANY; 337 } else 338 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 339 /* 340 * Confirm that the outgoing interface supports multicast. 341 */ 342 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 343 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 344 IPSTAT_INC(ips_noroute); 345 error = ENETUNREACH; 346 goto bad; 347 } 348 } 349 /* 350 * If source address not specified yet, use address 351 * of outgoing interface. 352 */ 353 if (ip->ip_src.s_addr == INADDR_ANY) { 354 /* Interface may have no addresses. */ 355 if (ia != NULL) 356 ip->ip_src = IA_SIN(ia)->sin_addr; 357 } 358 359 if ((imo == NULL && in_mcast_loop) || 360 (imo && imo->imo_multicast_loop)) { 361 /* 362 * Loop back multicast datagram if not expressly 363 * forbidden to do so, even if we are not a member 364 * of the group; ip_input() will filter it later, 365 * thus deferring a hash lookup and mutex acquisition 366 * at the expense of a cheap copy using m_copym(). 367 */ 368 ip_mloopback(ifp, m, dst, hlen); 369 } else { 370 /* 371 * If we are acting as a multicast router, perform 372 * multicast forwarding as if the packet had just 373 * arrived on the interface to which we are about 374 * to send. The multicast forwarding function 375 * recursively calls this function, using the 376 * IP_FORWARDING flag to prevent infinite recursion. 377 * 378 * Multicasts that are looped back by ip_mloopback(), 379 * above, will be forwarded by the ip_input() routine, 380 * if necessary. 381 */ 382 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { 383 /* 384 * If rsvp daemon is not running, do not 385 * set ip_moptions. This ensures that the packet 386 * is multicast and not just sent down one link 387 * as prescribed by rsvpd. 388 */ 389 if (!V_rsvp_on) 390 imo = NULL; 391 if (ip_mforward && 392 ip_mforward(ip, ifp, m, imo) != 0) { 393 m_freem(m); 394 goto done; 395 } 396 } 397 } 398 399 /* 400 * Multicasts with a time-to-live of zero may be looped- 401 * back, above, but must not be transmitted on a network. 402 * Also, multicasts addressed to the loopback interface 403 * are not sent -- the above call to ip_mloopback() will 404 * loop back a copy. ip_input() will drop the copy if 405 * this host does not belong to the destination group on 406 * the loopback interface. 407 */ 408 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 409 m_freem(m); 410 goto done; 411 } 412 413 goto sendit; 414 } 415 416 /* 417 * If the source address is not specified yet, use the address 418 * of the outoing interface. 419 */ 420 if (ip->ip_src.s_addr == INADDR_ANY) { 421 /* Interface may have no addresses. */ 422 if (ia != NULL) { 423 ip->ip_src = IA_SIN(ia)->sin_addr; 424 } 425 } 426 427 /* 428 * Verify that we have any chance at all of being able to queue the 429 * packet or packet fragments, unless ALTQ is enabled on the given 430 * interface in which case packetdrop should be done by queueing. 431 */ 432#ifdef ALTQ 433 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) && 434 ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >= 435 ifp->if_snd.ifq_maxlen)) 436#else 437 if ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >= 438 ifp->if_snd.ifq_maxlen) 439#endif /* ALTQ */ 440 { 441 error = ENOBUFS; 442 IPSTAT_INC(ips_odropped); 443 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1); 444 goto bad; 445 } 446 447 /* 448 * Look for broadcast address and 449 * verify user is allowed to send 450 * such a packet. 451 */ 452 if (isbroadcast) { 453 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 454 error = EADDRNOTAVAIL; 455 goto bad; 456 } 457 if ((flags & IP_ALLOWBROADCAST) == 0) { 458 error = EACCES; 459 goto bad; 460 } 461 /* don't allow broadcast messages to be fragmented */ 462 if (ip->ip_len > mtu) { 463 error = EMSGSIZE; 464 goto bad; 465 } 466 m->m_flags |= M_BCAST; 467 } else { 468 m->m_flags &= ~M_BCAST; 469 } 470 471sendit: 472#ifdef IPSEC 473 switch(ip_ipsec_output(&m, inp, &flags, &error, &ro, &iproute, &dst, &ia, &ifp)) { 474 case 1: 475 goto bad; 476 case -1: 477 goto done; 478 case 0: 479 default: 480 break; /* Continue with packet processing. */ 481 } 482 /* 483 * Check if there was a route for this packet; return error if not. 484 */ 485 if (no_route_but_check_spd) { 486 IPSTAT_INC(ips_noroute); 487 error = EHOSTUNREACH; 488 goto bad; 489 } 490 /* Update variables that are affected by ipsec4_output(). */ 491 ip = mtod(m, struct ip *); 492 hlen = ip->ip_hl << 2; 493#endif /* IPSEC */ 494 495 /* Jump over all PFIL processing if hooks are not active. */ 496 if (!PFIL_HOOKED(&inet_pfil_hook)) 497 goto passout; 498 499 /* Run through list of hooks for output packets. */ 500 odst.s_addr = ip->ip_dst.s_addr; 501 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 502 if (error != 0 || m == NULL) 503 goto done; 504 505 ip = mtod(m, struct ip *); 506 507 /* See if destination IP address was changed by packet filter. */ 508 if (odst.s_addr != ip->ip_dst.s_addr) { 509 m->m_flags |= M_SKIP_FIREWALL; 510 /* If destination is now ourself drop to ip_input(). */ 511 if (in_localip(ip->ip_dst)) { 512 m->m_flags |= M_FASTFWD_OURS; 513 if (m->m_pkthdr.rcvif == NULL) 514 m->m_pkthdr.rcvif = V_loif; 515 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 516 m->m_pkthdr.csum_flags |= 517 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 518 m->m_pkthdr.csum_data = 0xffff; 519 } 520 m->m_pkthdr.csum_flags |= 521 CSUM_IP_CHECKED | CSUM_IP_VALID; 522#ifdef SCTP 523 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 524 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 525#endif 526 error = netisr_queue(NETISR_IP, m); 527 goto done; 528 } else 529 goto again; /* Redo the routing table lookup. */ 530 } 531 532#ifdef IPFIREWALL_FORWARD 533 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 534 if (m->m_flags & M_FASTFWD_OURS) { 535 if (m->m_pkthdr.rcvif == NULL) 536 m->m_pkthdr.rcvif = V_loif; 537 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 538 m->m_pkthdr.csum_flags |= 539 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 540 m->m_pkthdr.csum_data = 0xffff; 541 } 542#ifdef SCTP 543 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 544 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 545#endif 546 m->m_pkthdr.csum_flags |= 547 CSUM_IP_CHECKED | CSUM_IP_VALID; 548 549 error = netisr_queue(NETISR_IP, m); 550 goto done; 551 } 552 /* Or forward to some other address? */ 553 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 554 if (fwd_tag) { 555 dst = (struct sockaddr_in *)&ro->ro_dst; 556 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 557 m->m_flags |= M_SKIP_FIREWALL; 558 m_tag_delete(m, fwd_tag); 559 goto again; 560 } 561#endif /* IPFIREWALL_FORWARD */ 562 563passout: 564 /* 127/8 must not appear on wire - RFC1122. */ 565 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 566 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 567 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 568 IPSTAT_INC(ips_badaddr); 569 error = EADDRNOTAVAIL; 570 goto bad; 571 } 572 } 573 574 m->m_pkthdr.csum_flags |= CSUM_IP; 575 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 576 if (sw_csum & CSUM_DELAY_DATA) { 577 in_delayed_cksum(m); 578 sw_csum &= ~CSUM_DELAY_DATA; 579 } 580#ifdef SCTP 581 if (sw_csum & CSUM_SCTP) { 582 sctp_delayed_cksum(m); 583 sw_csum &= ~CSUM_SCTP; 584 } 585#endif 586 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 587 588 /* 589 * If small enough for interface, or the interface will take 590 * care of the fragmentation for us, we can just send directly. 591 */ 592 if (ip->ip_len <= mtu || 593 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 594 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 595 ip->ip_len = htons(ip->ip_len); 596 ip->ip_off = htons(ip->ip_off); 597 ip->ip_sum = 0; 598 if (sw_csum & CSUM_DELAY_IP) 599 ip->ip_sum = in_cksum(m, hlen); 600 601 /* 602 * Record statistics for this interface address. 603 * With CSUM_TSO the byte/packet count will be slightly 604 * incorrect because we count the IP+TCP headers only 605 * once instead of for every generated packet. 606 */ 607 if (!(flags & IP_FORWARDING) && ia) { 608 if (m->m_pkthdr.csum_flags & CSUM_TSO) 609 ia->ia_ifa.if_opackets += 610 m->m_pkthdr.len / m->m_pkthdr.tso_segsz; 611 else 612 ia->ia_ifa.if_opackets++; 613 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 614 } 615#ifdef MBUF_STRESS_TEST 616 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 617 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 618#endif 619 /* 620 * Reset layer specific mbuf flags 621 * to avoid confusing lower layers. 622 */ 623 m->m_flags &= ~(M_PROTOFLAGS); 624 error = (*ifp->if_output)(ifp, m, 625 (struct sockaddr *)dst, ro); 626 goto done; 627 } 628 629 /* Balk when DF bit is set or the interface didn't support TSO. */ 630 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 631 error = EMSGSIZE; 632 IPSTAT_INC(ips_cantfrag); 633 goto bad; 634 } 635 636 /* 637 * Too large for interface; fragment if possible. If successful, 638 * on return, m will point to a list of packets to be sent. 639 */ 640 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum); 641 if (error) 642 goto bad; 643 for (; m; m = m0) { 644 m0 = m->m_nextpkt; 645 m->m_nextpkt = 0; 646 if (error == 0) { 647 /* Record statistics for this interface address. */ 648 if (ia != NULL) { 649 ia->ia_ifa.if_opackets++; 650 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 651 } 652 /* 653 * Reset layer specific mbuf flags 654 * to avoid confusing upper layers. 655 */ 656 m->m_flags &= ~(M_PROTOFLAGS); 657 658 error = (*ifp->if_output)(ifp, m, 659 (struct sockaddr *)dst, ro); 660 } else 661 m_freem(m); 662 } 663 664 if (error == 0) 665 IPSTAT_INC(ips_fragmented); 666 667done: 668 if (ro == &iproute && ro->ro_rt && !nortfree) { 669 RTFREE(ro->ro_rt); 670 } 671 return (error); 672bad: 673 m_freem(m); 674 goto done; 675} 676 677/* 678 * Create a chain of fragments which fit the given mtu. m_frag points to the 679 * mbuf to be fragmented; on return it points to the chain with the fragments. 680 * Return 0 if no error. If error, m_frag may contain a partially built 681 * chain of fragments that should be freed by the caller. 682 * 683 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 684 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 685 */ 686int 687ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 688 u_long if_hwassist_flags, int sw_csum) 689{ 690 INIT_VNET_INET(curvnet); 691 int error = 0; 692 int hlen = ip->ip_hl << 2; 693 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 694 int off; 695 struct mbuf *m0 = *m_frag; /* the original packet */ 696 int firstlen; 697 struct mbuf **mnext; 698 int nfrags; 699 700 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 701 IPSTAT_INC(ips_cantfrag); 702 return EMSGSIZE; 703 } 704 705 /* 706 * Must be able to put at least 8 bytes per fragment. 707 */ 708 if (len < 8) 709 return EMSGSIZE; 710 711 /* 712 * If the interface will not calculate checksums on 713 * fragmented packets, then do it here. 714 */ 715 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 716 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 717 in_delayed_cksum(m0); 718 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 719 } 720#ifdef SCTP 721 if (m0->m_pkthdr.csum_flags & CSUM_SCTP && 722 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 723 sctp_delayed_cksum(m0); 724 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; 725 } 726#endif 727 if (len > PAGE_SIZE) { 728 /* 729 * Fragment large datagrams such that each segment 730 * contains a multiple of PAGE_SIZE amount of data, 731 * plus headers. This enables a receiver to perform 732 * page-flipping zero-copy optimizations. 733 * 734 * XXX When does this help given that sender and receiver 735 * could have different page sizes, and also mtu could 736 * be less than the receiver's page size ? 737 */ 738 int newlen; 739 struct mbuf *m; 740 741 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 742 off += m->m_len; 743 744 /* 745 * firstlen (off - hlen) must be aligned on an 746 * 8-byte boundary 747 */ 748 if (off < hlen) 749 goto smart_frag_failure; 750 off = ((off - hlen) & ~7) + hlen; 751 newlen = (~PAGE_MASK) & mtu; 752 if ((newlen + sizeof (struct ip)) > mtu) { 753 /* we failed, go back the default */ 754smart_frag_failure: 755 newlen = len; 756 off = hlen + len; 757 } 758 len = newlen; 759 760 } else { 761 off = hlen + len; 762 } 763 764 firstlen = off - hlen; 765 mnext = &m0->m_nextpkt; /* pointer to next packet */ 766 767 /* 768 * Loop through length of segment after first fragment, 769 * make new header and copy data of each part and link onto chain. 770 * Here, m0 is the original packet, m is the fragment being created. 771 * The fragments are linked off the m_nextpkt of the original 772 * packet, which after processing serves as the first fragment. 773 */ 774 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 775 struct ip *mhip; /* ip header on the fragment */ 776 struct mbuf *m; 777 int mhlen = sizeof (struct ip); 778 779 MGETHDR(m, M_DONTWAIT, MT_DATA); 780 if (m == NULL) { 781 error = ENOBUFS; 782 IPSTAT_INC(ips_odropped); 783 goto done; 784 } 785 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 786 /* 787 * In the first mbuf, leave room for the link header, then 788 * copy the original IP header including options. The payload 789 * goes into an additional mbuf chain returned by m_copym(). 790 */ 791 m->m_data += max_linkhdr; 792 mhip = mtod(m, struct ip *); 793 *mhip = *ip; 794 if (hlen > sizeof (struct ip)) { 795 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 796 mhip->ip_v = IPVERSION; 797 mhip->ip_hl = mhlen >> 2; 798 } 799 m->m_len = mhlen; 800 /* XXX do we need to add ip->ip_off below ? */ 801 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 802 if (off + len >= ip->ip_len) { /* last fragment */ 803 len = ip->ip_len - off; 804 m->m_flags |= M_LASTFRAG; 805 } else 806 mhip->ip_off |= IP_MF; 807 mhip->ip_len = htons((u_short)(len + mhlen)); 808 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 809 if (m->m_next == NULL) { /* copy failed */ 810 m_free(m); 811 error = ENOBUFS; /* ??? */ 812 IPSTAT_INC(ips_odropped); 813 goto done; 814 } 815 m->m_pkthdr.len = mhlen + len; 816 m->m_pkthdr.rcvif = NULL; 817#ifdef MAC 818 mac_netinet_fragment(m0, m); 819#endif 820 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 821 mhip->ip_off = htons(mhip->ip_off); 822 mhip->ip_sum = 0; 823 if (sw_csum & CSUM_DELAY_IP) 824 mhip->ip_sum = in_cksum(m, mhlen); 825 *mnext = m; 826 mnext = &m->m_nextpkt; 827 } 828 IPSTAT_ADD(ips_ofragments, nfrags); 829 830 /* set first marker for fragment chain */ 831 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 832 m0->m_pkthdr.csum_data = 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->ip_len); 839 m0->m_pkthdr.len = hlen + firstlen; 840 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 841 ip->ip_off |= IP_MF; 842 ip->ip_off = htons(ip->ip_off); 843 ip->ip_sum = 0; 844 if (sw_csum & CSUM_DELAY_IP) 845 ip->ip_sum = in_cksum(m0, hlen); 846 847done: 848 *m_frag = m0; 849 return error; 850} 851 852void 853in_delayed_cksum(struct mbuf *m) 854{ 855 struct ip *ip; 856 u_short csum, offset; 857 858 ip = mtod(m, struct ip *); 859 offset = ip->ip_hl << 2 ; 860 csum = in_cksum_skip(m, ip->ip_len, offset); 861 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 862 csum = 0xffff; 863 offset += m->m_pkthdr.csum_data; /* checksum offset */ 864 865 if (offset + sizeof(u_short) > m->m_len) { 866 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 867 m->m_len, offset, ip->ip_p); 868 /* 869 * XXX 870 * this shouldn't happen, but if it does, the 871 * correct behavior may be to insert the checksum 872 * in the appropriate next mbuf in the chain. 873 */ 874 return; 875 } 876 *(u_short *)(m->m_data + offset) = csum; 877} 878 879/* 880 * IP socket option processing. 881 */ 882int 883ip_ctloutput(struct socket *so, struct sockopt *sopt) 884{ 885 struct inpcb *inp = sotoinpcb(so); 886 int error, optval; 887 888 error = optval = 0; 889 if (sopt->sopt_level != IPPROTO_IP) { 890 if ((sopt->sopt_level == SOL_SOCKET) && 891 (sopt->sopt_name == SO_SETFIB)) { 892 inp->inp_inc.inc_fibnum = so->so_fibnum; 893 return (0); 894 } 895 return (EINVAL); 896 } 897 898 switch (sopt->sopt_dir) { 899 case SOPT_SET: 900 switch (sopt->sopt_name) { 901 case IP_OPTIONS: 902#ifdef notyet 903 case IP_RETOPTS: 904#endif 905 { 906 struct mbuf *m; 907 if (sopt->sopt_valsize > MLEN) { 908 error = EMSGSIZE; 909 break; 910 } 911 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA); 912 if (m == NULL) { 913 error = ENOBUFS; 914 break; 915 } 916 m->m_len = sopt->sopt_valsize; 917 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 918 m->m_len); 919 if (error) { 920 m_free(m); 921 break; 922 } 923 INP_WLOCK(inp); 924 error = ip_pcbopts(inp, sopt->sopt_name, m); 925 INP_WUNLOCK(inp); 926 return (error); 927 } 928 929#if defined(IP_NONLOCALBIND) 930 case IP_NONLOCALOK: 931 if (! ip_nonlocalok) { 932 error = ENOPROTOOPT; 933 break; 934 } 935 /* FALLTHROUGH */ 936#endif 937 case IP_TOS: 938 case IP_TTL: 939 case IP_MINTTL: 940 case IP_RECVOPTS: 941 case IP_RECVRETOPTS: 942 case IP_RECVDSTADDR: 943 case IP_RECVTTL: 944 case IP_RECVIF: 945 case IP_FAITH: 946 case IP_ONESBCAST: 947 case IP_DONTFRAG: 948 error = sooptcopyin(sopt, &optval, sizeof optval, 949 sizeof optval); 950 if (error) 951 break; 952 953 switch (sopt->sopt_name) { 954 case IP_TOS: 955 inp->inp_ip_tos = optval; 956 break; 957 958 case IP_TTL: 959 inp->inp_ip_ttl = optval; 960 break; 961 962 case IP_MINTTL: 963 if (optval >= 0 && optval <= MAXTTL) 964 inp->inp_ip_minttl = optval; 965 else 966 error = EINVAL; 967 break; 968 969#define OPTSET(bit) do { \ 970 INP_WLOCK(inp); \ 971 if (optval) \ 972 inp->inp_flags |= bit; \ 973 else \ 974 inp->inp_flags &= ~bit; \ 975 INP_WUNLOCK(inp); \ 976} while (0) 977 978 case IP_RECVOPTS: 979 OPTSET(INP_RECVOPTS); 980 break; 981 982 case IP_RECVRETOPTS: 983 OPTSET(INP_RECVRETOPTS); 984 break; 985 986 case IP_RECVDSTADDR: 987 OPTSET(INP_RECVDSTADDR); 988 break; 989 990 case IP_RECVTTL: 991 OPTSET(INP_RECVTTL); 992 break; 993 994 case IP_RECVIF: 995 OPTSET(INP_RECVIF); 996 break; 997 998 case IP_FAITH: 999 OPTSET(INP_FAITH); 1000 break; 1001 1002 case IP_ONESBCAST: 1003 OPTSET(INP_ONESBCAST); 1004 break; 1005 case IP_DONTFRAG: 1006 OPTSET(INP_DONTFRAG); 1007 break; 1008#if defined(IP_NONLOCALBIND) 1009 case IP_NONLOCALOK: 1010 OPTSET(INP_NONLOCALOK); 1011 break; 1012#endif 1013 } 1014 break; 1015#undef OPTSET 1016 1017 /* 1018 * Multicast socket options are processed by the in_mcast 1019 * module. 1020 */ 1021 case IP_MULTICAST_IF: 1022 case IP_MULTICAST_VIF: 1023 case IP_MULTICAST_TTL: 1024 case IP_MULTICAST_LOOP: 1025 case IP_ADD_MEMBERSHIP: 1026 case IP_DROP_MEMBERSHIP: 1027 case IP_ADD_SOURCE_MEMBERSHIP: 1028 case IP_DROP_SOURCE_MEMBERSHIP: 1029 case IP_BLOCK_SOURCE: 1030 case IP_UNBLOCK_SOURCE: 1031 case IP_MSFILTER: 1032 case MCAST_JOIN_GROUP: 1033 case MCAST_LEAVE_GROUP: 1034 case MCAST_JOIN_SOURCE_GROUP: 1035 case MCAST_LEAVE_SOURCE_GROUP: 1036 case MCAST_BLOCK_SOURCE: 1037 case MCAST_UNBLOCK_SOURCE: 1038 error = inp_setmoptions(inp, sopt); 1039 break; 1040 1041 case IP_PORTRANGE: 1042 error = sooptcopyin(sopt, &optval, sizeof optval, 1043 sizeof optval); 1044 if (error) 1045 break; 1046 1047 INP_WLOCK(inp); 1048 switch (optval) { 1049 case IP_PORTRANGE_DEFAULT: 1050 inp->inp_flags &= ~(INP_LOWPORT); 1051 inp->inp_flags &= ~(INP_HIGHPORT); 1052 break; 1053 1054 case IP_PORTRANGE_HIGH: 1055 inp->inp_flags &= ~(INP_LOWPORT); 1056 inp->inp_flags |= INP_HIGHPORT; 1057 break; 1058 1059 case IP_PORTRANGE_LOW: 1060 inp->inp_flags &= ~(INP_HIGHPORT); 1061 inp->inp_flags |= INP_LOWPORT; 1062 break; 1063 1064 default: 1065 error = EINVAL; 1066 break; 1067 } 1068 INP_WUNLOCK(inp); 1069 break; 1070 1071#ifdef IPSEC 1072 case IP_IPSEC_POLICY: 1073 { 1074 caddr_t req; 1075 struct mbuf *m; 1076 1077 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1078 break; 1079 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1080 break; 1081 req = mtod(m, caddr_t); 1082 error = ipsec_set_policy(inp, sopt->sopt_name, req, 1083 m->m_len, (sopt->sopt_td != NULL) ? 1084 sopt->sopt_td->td_ucred : NULL); 1085 m_freem(m); 1086 break; 1087 } 1088#endif /* IPSEC */ 1089 1090 default: 1091 error = ENOPROTOOPT; 1092 break; 1093 } 1094 break; 1095 1096 case SOPT_GET: 1097 switch (sopt->sopt_name) { 1098 case IP_OPTIONS: 1099 case IP_RETOPTS: 1100 if (inp->inp_options) 1101 error = sooptcopyout(sopt, 1102 mtod(inp->inp_options, 1103 char *), 1104 inp->inp_options->m_len); 1105 else 1106 sopt->sopt_valsize = 0; 1107 break; 1108 1109 case IP_TOS: 1110 case IP_TTL: 1111 case IP_MINTTL: 1112 case IP_RECVOPTS: 1113 case IP_RECVRETOPTS: 1114 case IP_RECVDSTADDR: 1115 case IP_RECVTTL: 1116 case IP_RECVIF: 1117 case IP_PORTRANGE: 1118 case IP_FAITH: 1119 case IP_ONESBCAST: 1120 case IP_DONTFRAG: 1121 switch (sopt->sopt_name) { 1122 1123 case IP_TOS: 1124 optval = inp->inp_ip_tos; 1125 break; 1126 1127 case IP_TTL: 1128 optval = inp->inp_ip_ttl; 1129 break; 1130 1131 case IP_MINTTL: 1132 optval = inp->inp_ip_minttl; 1133 break; 1134 1135#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1136 1137 case IP_RECVOPTS: 1138 optval = OPTBIT(INP_RECVOPTS); 1139 break; 1140 1141 case IP_RECVRETOPTS: 1142 optval = OPTBIT(INP_RECVRETOPTS); 1143 break; 1144 1145 case IP_RECVDSTADDR: 1146 optval = OPTBIT(INP_RECVDSTADDR); 1147 break; 1148 1149 case IP_RECVTTL: 1150 optval = OPTBIT(INP_RECVTTL); 1151 break; 1152 1153 case IP_RECVIF: 1154 optval = OPTBIT(INP_RECVIF); 1155 break; 1156 1157 case IP_PORTRANGE: 1158 if (inp->inp_flags & INP_HIGHPORT) 1159 optval = IP_PORTRANGE_HIGH; 1160 else if (inp->inp_flags & INP_LOWPORT) 1161 optval = IP_PORTRANGE_LOW; 1162 else 1163 optval = 0; 1164 break; 1165 1166 case IP_FAITH: 1167 optval = OPTBIT(INP_FAITH); 1168 break; 1169 1170 case IP_ONESBCAST: 1171 optval = OPTBIT(INP_ONESBCAST); 1172 break; 1173 case IP_DONTFRAG: 1174 optval = OPTBIT(INP_DONTFRAG); 1175 break; 1176 } 1177 error = sooptcopyout(sopt, &optval, sizeof optval); 1178 break; 1179 1180 /* 1181 * Multicast socket options are processed by the in_mcast 1182 * module. 1183 */ 1184 case IP_MULTICAST_IF: 1185 case IP_MULTICAST_VIF: 1186 case IP_MULTICAST_TTL: 1187 case IP_MULTICAST_LOOP: 1188 case IP_MSFILTER: 1189 error = inp_getmoptions(inp, sopt); 1190 break; 1191 1192#ifdef IPSEC 1193 case IP_IPSEC_POLICY: 1194 { 1195 struct mbuf *m = NULL; 1196 caddr_t req = NULL; 1197 size_t len = 0; 1198 1199 if (m != 0) { 1200 req = mtod(m, caddr_t); 1201 len = m->m_len; 1202 } 1203 error = ipsec_get_policy(sotoinpcb(so), req, len, &m); 1204 if (error == 0) 1205 error = soopt_mcopyout(sopt, m); /* XXX */ 1206 if (error == 0) 1207 m_freem(m); 1208 break; 1209 } 1210#endif /* IPSEC */ 1211 1212 default: 1213 error = ENOPROTOOPT; 1214 break; 1215 } 1216 break; 1217 } 1218 return (error); 1219} 1220 1221/* 1222 * Routine called from ip_output() to loop back a copy of an IP multicast 1223 * packet to the input queue of a specified interface. Note that this 1224 * calls the output routine of the loopback "driver", but with an interface 1225 * pointer that might NOT be a loopback interface -- evil, but easier than 1226 * replicating that code here. 1227 */ 1228static void 1229ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst, 1230 int hlen) 1231{ 1232 register struct ip *ip; 1233 struct mbuf *copym; 1234 1235 /* 1236 * Make a deep copy of the packet because we're going to 1237 * modify the pack in order to generate checksums. 1238 */ 1239 copym = m_dup(m, M_DONTWAIT); 1240 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1241 copym = m_pullup(copym, hlen); 1242 if (copym != NULL) { 1243 /* If needed, compute the checksum and mark it as valid. */ 1244 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1245 in_delayed_cksum(copym); 1246 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1247 copym->m_pkthdr.csum_flags |= 1248 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1249 copym->m_pkthdr.csum_data = 0xffff; 1250 } 1251 /* 1252 * We don't bother to fragment if the IP length is greater 1253 * than the interface's MTU. Can this possibly matter? 1254 */ 1255 ip = mtod(copym, struct ip *); 1256 ip->ip_len = htons(ip->ip_len); 1257 ip->ip_off = htons(ip->ip_off); 1258 ip->ip_sum = 0; 1259 ip->ip_sum = in_cksum(copym, hlen); 1260#if 1 /* XXX */ 1261 if (dst->sin_family != AF_INET) { 1262 printf("ip_mloopback: bad address family %d\n", 1263 dst->sin_family); 1264 dst->sin_family = AF_INET; 1265 } 1266#endif 1267 if_simloop(ifp, copym, dst->sin_family, 0); 1268 } 1269} 1270