ip_output.c revision 192528
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 192528 2009-05-21 09:45:47Z rwatson $"); 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 (inp != NULL) { 154 INP_LOCK_ASSERT(inp); 155 M_SETFIB(m, inp->inp_inc.inc_fibnum); 156 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) { 157 m->m_pkthdr.flowid = inp->inp_flowid; 158 m->m_flags |= M_FLOWID; 159 } 160 } 161 162 if (ro == NULL) { 163 ro = &iproute; 164 bzero(ro, sizeof (*ro)); 165 166 /* 167 * The flow table returns route entries valid for up to 30 168 * seconds; we rely on the remainder of ip_output() taking no 169 * longer than that long for the stability of ro_rt. The 170 * flow ID assignment must have happened before this point. 171 */ 172 if (flowtable_lookup(ip_ft, m, ro) == 0) 173 nortfree = 1; 174 } 175 176 if (opt) { 177 len = 0; 178 m = ip_insertoptions(m, opt, &len); 179 if (len != 0) 180 hlen = len; 181 } 182 ip = mtod(m, struct ip *); 183 184 /* 185 * Fill in IP header. If we are not allowing fragmentation, 186 * then the ip_id field is meaningless, but we don't set it 187 * to zero. Doing so causes various problems when devices along 188 * the path (routers, load balancers, firewalls, etc.) illegally 189 * disable DF on our packet. Note that a 16-bit counter 190 * will wrap around in less than 10 seconds at 100 Mbit/s on a 191 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 192 * for Counting NATted Hosts", Proc. IMW'02, available at 193 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 194 */ 195 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 196 ip->ip_v = IPVERSION; 197 ip->ip_hl = hlen >> 2; 198 ip->ip_id = ip_newid(); 199 IPSTAT_INC(ips_localout); 200 } else { 201 hlen = ip->ip_hl << 2; 202 } 203 204 dst = (struct sockaddr_in *)&ro->ro_dst; 205again: 206 /* 207 * If there is a cached route, 208 * check that it is to the same destination 209 * and is still up. If not, free it and try again. 210 * The address family should also be checked in case of sharing the 211 * cache with IPv6. 212 */ 213 if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 214 dst->sin_family != AF_INET || 215 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 216 if (!nortfree) 217 RTFREE(ro->ro_rt); 218 ro->ro_rt = (struct rtentry *)NULL; 219 } 220#ifdef IPFIREWALL_FORWARD 221 if (ro->ro_rt == NULL && fwd_tag == NULL) { 222#else 223 if (ro->ro_rt == NULL) { 224#endif 225 bzero(dst, sizeof(*dst)); 226 dst->sin_family = AF_INET; 227 dst->sin_len = sizeof(*dst); 228 dst->sin_addr = ip->ip_dst; 229 } 230 /* 231 * If routing to interface only, short circuit routing lookup. 232 * The use of an all-ones broadcast address implies this; an 233 * interface is specified by the broadcast address of an interface, 234 * or the destination address of a ptp interface. 235 */ 236 if (flags & IP_SENDONES) { 237 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL && 238 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) { 239 IPSTAT_INC(ips_noroute); 240 error = ENETUNREACH; 241 goto bad; 242 } 243 ip->ip_dst.s_addr = INADDR_BROADCAST; 244 dst->sin_addr = ip->ip_dst; 245 ifp = ia->ia_ifp; 246 ip->ip_ttl = 1; 247 isbroadcast = 1; 248 } else if (flags & IP_ROUTETOIF) { 249 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 250 (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == NULL) { 251 IPSTAT_INC(ips_noroute); 252 error = ENETUNREACH; 253 goto bad; 254 } 255 ifp = ia->ia_ifp; 256 ip->ip_ttl = 1; 257 isbroadcast = in_broadcast(dst->sin_addr, ifp); 258 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 259 imo != NULL && imo->imo_multicast_ifp != NULL) { 260 /* 261 * Bypass the normal routing lookup for multicast 262 * packets if the interface is specified. 263 */ 264 ifp = imo->imo_multicast_ifp; 265 IFP_TO_IA(ifp, ia); 266 isbroadcast = 0; /* fool gcc */ 267 } else { 268 /* 269 * We want to do any cloning requested by the link layer, 270 * as this is probably required in all cases for correct 271 * operation (as it is for ARP). 272 */ 273 if (ro->ro_rt == NULL) 274#ifdef RADIX_MPATH 275 rtalloc_mpath_fib(ro, 276 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 277 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 278#else 279 in_rtalloc_ign(ro, 0, 280 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 281#endif 282 if (ro->ro_rt == NULL) { 283#ifdef IPSEC 284 /* 285 * There is no route for this packet, but it is 286 * possible that a matching SPD entry exists. 287 */ 288 no_route_but_check_spd = 1; 289 mtu = 0; /* Silence GCC warning. */ 290 goto sendit; 291#endif 292 IPSTAT_INC(ips_noroute); 293 error = EHOSTUNREACH; 294 goto bad; 295 } 296 ia = ifatoia(ro->ro_rt->rt_ifa); 297 ifp = ro->ro_rt->rt_ifp; 298 ro->ro_rt->rt_rmx.rmx_pksent++; 299 if (ro->ro_rt->rt_flags & RTF_GATEWAY) 300 dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 301 if (ro->ro_rt->rt_flags & RTF_HOST) 302 isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 303 else 304 isbroadcast = in_broadcast(dst->sin_addr, ifp); 305 } 306 /* 307 * Calculate MTU. If we have a route that is up, use that, 308 * otherwise use the interface's MTU. 309 */ 310 if (ro->ro_rt != NULL && (ro->ro_rt->rt_flags & (RTF_UP|RTF_HOST))) { 311 /* 312 * This case can happen if the user changed the MTU 313 * of an interface after enabling IP on it. Because 314 * most netifs don't keep track of routes pointing to 315 * them, there is no way for one to update all its 316 * routes when the MTU is changed. 317 */ 318 if (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu) 319 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 320 mtu = ro->ro_rt->rt_rmx.rmx_mtu; 321 } else { 322 mtu = ifp->if_mtu; 323 } 324 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 325 m->m_flags |= M_MCAST; 326 /* 327 * IP destination address is multicast. Make sure "dst" 328 * still points to the address in "ro". (It may have been 329 * changed to point to a gateway address, above.) 330 */ 331 dst = (struct sockaddr_in *)&ro->ro_dst; 332 /* 333 * See if the caller provided any multicast options 334 */ 335 if (imo != NULL) { 336 ip->ip_ttl = imo->imo_multicast_ttl; 337 if (imo->imo_multicast_vif != -1) 338 ip->ip_src.s_addr = 339 ip_mcast_src ? 340 ip_mcast_src(imo->imo_multicast_vif) : 341 INADDR_ANY; 342 } else 343 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 344 /* 345 * Confirm that the outgoing interface supports multicast. 346 */ 347 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 348 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 349 IPSTAT_INC(ips_noroute); 350 error = ENETUNREACH; 351 goto bad; 352 } 353 } 354 /* 355 * If source address not specified yet, use address 356 * of outgoing interface. 357 */ 358 if (ip->ip_src.s_addr == INADDR_ANY) { 359 /* Interface may have no addresses. */ 360 if (ia != NULL) 361 ip->ip_src = IA_SIN(ia)->sin_addr; 362 } 363 364 if ((imo == NULL && in_mcast_loop) || 365 (imo && imo->imo_multicast_loop)) { 366 /* 367 * Loop back multicast datagram if not expressly 368 * forbidden to do so, even if we are not a member 369 * of the group; ip_input() will filter it later, 370 * thus deferring a hash lookup and mutex acquisition 371 * at the expense of a cheap copy using m_copym(). 372 */ 373 ip_mloopback(ifp, m, dst, hlen); 374 } else { 375 /* 376 * If we are acting as a multicast router, perform 377 * multicast forwarding as if the packet had just 378 * arrived on the interface to which we are about 379 * to send. The multicast forwarding function 380 * recursively calls this function, using the 381 * IP_FORWARDING flag to prevent infinite recursion. 382 * 383 * Multicasts that are looped back by ip_mloopback(), 384 * above, will be forwarded by the ip_input() routine, 385 * if necessary. 386 */ 387 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { 388 /* 389 * If rsvp daemon is not running, do not 390 * set ip_moptions. This ensures that the packet 391 * is multicast and not just sent down one link 392 * as prescribed by rsvpd. 393 */ 394 if (!V_rsvp_on) 395 imo = NULL; 396 if (ip_mforward && 397 ip_mforward(ip, ifp, m, imo) != 0) { 398 m_freem(m); 399 goto done; 400 } 401 } 402 } 403 404 /* 405 * Multicasts with a time-to-live of zero may be looped- 406 * back, above, but must not be transmitted on a network. 407 * Also, multicasts addressed to the loopback interface 408 * are not sent -- the above call to ip_mloopback() will 409 * loop back a copy. ip_input() will drop the copy if 410 * this host does not belong to the destination group on 411 * the loopback interface. 412 */ 413 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 414 m_freem(m); 415 goto done; 416 } 417 418 goto sendit; 419 } 420 421 /* 422 * If the source address is not specified yet, use the address 423 * of the outoing interface. 424 */ 425 if (ip->ip_src.s_addr == INADDR_ANY) { 426 /* Interface may have no addresses. */ 427 if (ia != NULL) { 428 ip->ip_src = IA_SIN(ia)->sin_addr; 429 } 430 } 431 432 /* 433 * Verify that we have any chance at all of being able to queue the 434 * packet or packet fragments, unless ALTQ is enabled on the given 435 * interface in which case packetdrop should be done by queueing. 436 */ 437#ifdef ALTQ 438 if ((!ALTQ_IS_ENABLED(&ifp->if_snd)) && 439 ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >= 440 ifp->if_snd.ifq_maxlen)) 441#else 442 if ((ifp->if_snd.ifq_len + ip->ip_len / mtu + 1) >= 443 ifp->if_snd.ifq_maxlen) 444#endif /* ALTQ */ 445 { 446 error = ENOBUFS; 447 IPSTAT_INC(ips_odropped); 448 ifp->if_snd.ifq_drops += (ip->ip_len / ifp->if_mtu + 1); 449 goto bad; 450 } 451 452 /* 453 * Look for broadcast address and 454 * verify user is allowed to send 455 * such a packet. 456 */ 457 if (isbroadcast) { 458 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 459 error = EADDRNOTAVAIL; 460 goto bad; 461 } 462 if ((flags & IP_ALLOWBROADCAST) == 0) { 463 error = EACCES; 464 goto bad; 465 } 466 /* don't allow broadcast messages to be fragmented */ 467 if (ip->ip_len > mtu) { 468 error = EMSGSIZE; 469 goto bad; 470 } 471 m->m_flags |= M_BCAST; 472 } else { 473 m->m_flags &= ~M_BCAST; 474 } 475 476sendit: 477#ifdef IPSEC 478 switch(ip_ipsec_output(&m, inp, &flags, &error, &ro, &iproute, &dst, &ia, &ifp)) { 479 case 1: 480 goto bad; 481 case -1: 482 goto done; 483 case 0: 484 default: 485 break; /* Continue with packet processing. */ 486 } 487 /* 488 * Check if there was a route for this packet; return error if not. 489 */ 490 if (no_route_but_check_spd) { 491 IPSTAT_INC(ips_noroute); 492 error = EHOSTUNREACH; 493 goto bad; 494 } 495 /* Update variables that are affected by ipsec4_output(). */ 496 ip = mtod(m, struct ip *); 497 hlen = ip->ip_hl << 2; 498#endif /* IPSEC */ 499 500 /* Jump over all PFIL processing if hooks are not active. */ 501 if (!PFIL_HOOKED(&inet_pfil_hook)) 502 goto passout; 503 504 /* Run through list of hooks for output packets. */ 505 odst.s_addr = ip->ip_dst.s_addr; 506 error = pfil_run_hooks(&inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 507 if (error != 0 || m == NULL) 508 goto done; 509 510 ip = mtod(m, struct ip *); 511 512 /* See if destination IP address was changed by packet filter. */ 513 if (odst.s_addr != ip->ip_dst.s_addr) { 514 m->m_flags |= M_SKIP_FIREWALL; 515 /* If destination is now ourself drop to ip_input(). */ 516 if (in_localip(ip->ip_dst)) { 517 m->m_flags |= M_FASTFWD_OURS; 518 if (m->m_pkthdr.rcvif == NULL) 519 m->m_pkthdr.rcvif = V_loif; 520 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 521 m->m_pkthdr.csum_flags |= 522 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 523 m->m_pkthdr.csum_data = 0xffff; 524 } 525 m->m_pkthdr.csum_flags |= 526 CSUM_IP_CHECKED | CSUM_IP_VALID; 527#ifdef SCTP 528 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 529 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 530#endif 531 error = netisr_queue(NETISR_IP, m); 532 goto done; 533 } else 534 goto again; /* Redo the routing table lookup. */ 535 } 536 537#ifdef IPFIREWALL_FORWARD 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 fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL); 559 if (fwd_tag) { 560 dst = (struct sockaddr_in *)&ro->ro_dst; 561 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 562 m->m_flags |= M_SKIP_FIREWALL; 563 m_tag_delete(m, fwd_tag); 564 goto again; 565 } 566#endif /* IPFIREWALL_FORWARD */ 567 568passout: 569 /* 127/8 must not appear on wire - RFC1122. */ 570 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 571 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 572 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 573 IPSTAT_INC(ips_badaddr); 574 error = EADDRNOTAVAIL; 575 goto bad; 576 } 577 } 578 579 m->m_pkthdr.csum_flags |= CSUM_IP; 580 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 581 if (sw_csum & CSUM_DELAY_DATA) { 582 in_delayed_cksum(m); 583 sw_csum &= ~CSUM_DELAY_DATA; 584 } 585#ifdef SCTP 586 if (sw_csum & CSUM_SCTP) { 587 sctp_delayed_cksum(m); 588 sw_csum &= ~CSUM_SCTP; 589 } 590#endif 591 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 592 593 /* 594 * If small enough for interface, or the interface will take 595 * care of the fragmentation for us, we can just send directly. 596 */ 597 if (ip->ip_len <= mtu || 598 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 599 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 600 ip->ip_len = htons(ip->ip_len); 601 ip->ip_off = htons(ip->ip_off); 602 ip->ip_sum = 0; 603 if (sw_csum & CSUM_DELAY_IP) 604 ip->ip_sum = in_cksum(m, hlen); 605 606 /* 607 * Record statistics for this interface address. 608 * With CSUM_TSO the byte/packet count will be slightly 609 * incorrect because we count the IP+TCP headers only 610 * once instead of for every generated packet. 611 */ 612 if (!(flags & IP_FORWARDING) && ia) { 613 if (m->m_pkthdr.csum_flags & CSUM_TSO) 614 ia->ia_ifa.if_opackets += 615 m->m_pkthdr.len / m->m_pkthdr.tso_segsz; 616 else 617 ia->ia_ifa.if_opackets++; 618 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 619 } 620#ifdef MBUF_STRESS_TEST 621 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 622 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 623#endif 624 /* 625 * Reset layer specific mbuf flags 626 * to avoid confusing lower layers. 627 */ 628 m->m_flags &= ~(M_PROTOFLAGS); 629 error = (*ifp->if_output)(ifp, m, 630 (struct sockaddr *)dst, ro); 631 goto done; 632 } 633 634 /* Balk when DF bit is set or the interface didn't support TSO. */ 635 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 636 error = EMSGSIZE; 637 IPSTAT_INC(ips_cantfrag); 638 goto bad; 639 } 640 641 /* 642 * Too large for interface; fragment if possible. If successful, 643 * on return, m will point to a list of packets to be sent. 644 */ 645 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum); 646 if (error) 647 goto bad; 648 for (; m; m = m0) { 649 m0 = m->m_nextpkt; 650 m->m_nextpkt = 0; 651 if (error == 0) { 652 /* Record statistics for this interface address. */ 653 if (ia != NULL) { 654 ia->ia_ifa.if_opackets++; 655 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 656 } 657 /* 658 * Reset layer specific mbuf flags 659 * to avoid confusing upper layers. 660 */ 661 m->m_flags &= ~(M_PROTOFLAGS); 662 663 error = (*ifp->if_output)(ifp, m, 664 (struct sockaddr *)dst, ro); 665 } else 666 m_freem(m); 667 } 668 669 if (error == 0) 670 IPSTAT_INC(ips_fragmented); 671 672done: 673 if (ro == &iproute && ro->ro_rt && !nortfree) { 674 RTFREE(ro->ro_rt); 675 } 676 return (error); 677bad: 678 m_freem(m); 679 goto done; 680} 681 682/* 683 * Create a chain of fragments which fit the given mtu. m_frag points to the 684 * mbuf to be fragmented; on return it points to the chain with the fragments. 685 * Return 0 if no error. If error, m_frag may contain a partially built 686 * chain of fragments that should be freed by the caller. 687 * 688 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 689 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 690 */ 691int 692ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 693 u_long if_hwassist_flags, int sw_csum) 694{ 695 INIT_VNET_INET(curvnet); 696 int error = 0; 697 int hlen = ip->ip_hl << 2; 698 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 699 int off; 700 struct mbuf *m0 = *m_frag; /* the original packet */ 701 int firstlen; 702 struct mbuf **mnext; 703 int nfrags; 704 705 if (ip->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 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 722 in_delayed_cksum(m0); 723 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 724 } 725#ifdef SCTP 726 if (m0->m_pkthdr.csum_flags & CSUM_SCTP && 727 (if_hwassist_flags & CSUM_IP_FRAGS) == 0) { 728 sctp_delayed_cksum(m0); 729 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; 730 } 731#endif 732 if (len > PAGE_SIZE) { 733 /* 734 * Fragment large datagrams such that each segment 735 * contains a multiple of PAGE_SIZE amount of data, 736 * plus headers. This enables a receiver to perform 737 * page-flipping zero-copy optimizations. 738 * 739 * XXX When does this help given that sender and receiver 740 * could have different page sizes, and also mtu could 741 * be less than the receiver's page size ? 742 */ 743 int newlen; 744 struct mbuf *m; 745 746 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 747 off += m->m_len; 748 749 /* 750 * firstlen (off - hlen) must be aligned on an 751 * 8-byte boundary 752 */ 753 if (off < hlen) 754 goto smart_frag_failure; 755 off = ((off - hlen) & ~7) + hlen; 756 newlen = (~PAGE_MASK) & mtu; 757 if ((newlen + sizeof (struct ip)) > mtu) { 758 /* we failed, go back the default */ 759smart_frag_failure: 760 newlen = len; 761 off = hlen + len; 762 } 763 len = newlen; 764 765 } else { 766 off = hlen + len; 767 } 768 769 firstlen = off - hlen; 770 mnext = &m0->m_nextpkt; /* pointer to next packet */ 771 772 /* 773 * Loop through length of segment after first fragment, 774 * make new header and copy data of each part and link onto chain. 775 * Here, m0 is the original packet, m is the fragment being created. 776 * The fragments are linked off the m_nextpkt of the original 777 * packet, which after processing serves as the first fragment. 778 */ 779 for (nfrags = 1; off < ip->ip_len; off += len, nfrags++) { 780 struct ip *mhip; /* ip header on the fragment */ 781 struct mbuf *m; 782 int mhlen = sizeof (struct ip); 783 784 MGETHDR(m, M_DONTWAIT, MT_DATA); 785 if (m == NULL) { 786 error = ENOBUFS; 787 IPSTAT_INC(ips_odropped); 788 goto done; 789 } 790 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 791 /* 792 * In the first mbuf, leave room for the link header, then 793 * copy the original IP header including options. The payload 794 * goes into an additional mbuf chain returned by m_copym(). 795 */ 796 m->m_data += max_linkhdr; 797 mhip = mtod(m, struct ip *); 798 *mhip = *ip; 799 if (hlen > sizeof (struct ip)) { 800 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 801 mhip->ip_v = IPVERSION; 802 mhip->ip_hl = mhlen >> 2; 803 } 804 m->m_len = mhlen; 805 /* XXX do we need to add ip->ip_off below ? */ 806 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 807 if (off + len >= ip->ip_len) { /* last fragment */ 808 len = ip->ip_len - off; 809 m->m_flags |= M_LASTFRAG; 810 } else 811 mhip->ip_off |= IP_MF; 812 mhip->ip_len = htons((u_short)(len + mhlen)); 813 m->m_next = m_copym(m0, off, len, M_DONTWAIT); 814 if (m->m_next == NULL) { /* copy failed */ 815 m_free(m); 816 error = ENOBUFS; /* ??? */ 817 IPSTAT_INC(ips_odropped); 818 goto done; 819 } 820 m->m_pkthdr.len = mhlen + len; 821 m->m_pkthdr.rcvif = NULL; 822#ifdef MAC 823 mac_netinet_fragment(m0, m); 824#endif 825 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 826 mhip->ip_off = htons(mhip->ip_off); 827 mhip->ip_sum = 0; 828 if (sw_csum & CSUM_DELAY_IP) 829 mhip->ip_sum = in_cksum(m, mhlen); 830 *mnext = m; 831 mnext = &m->m_nextpkt; 832 } 833 IPSTAT_ADD(ips_ofragments, nfrags); 834 835 /* set first marker for fragment chain */ 836 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 837 m0->m_pkthdr.csum_data = nfrags; 838 839 /* 840 * Update first fragment by trimming what's been copied out 841 * and updating header. 842 */ 843 m_adj(m0, hlen + firstlen - ip->ip_len); 844 m0->m_pkthdr.len = hlen + firstlen; 845 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 846 ip->ip_off |= IP_MF; 847 ip->ip_off = htons(ip->ip_off); 848 ip->ip_sum = 0; 849 if (sw_csum & CSUM_DELAY_IP) 850 ip->ip_sum = in_cksum(m0, hlen); 851 852done: 853 *m_frag = m0; 854 return error; 855} 856 857void 858in_delayed_cksum(struct mbuf *m) 859{ 860 struct ip *ip; 861 u_short csum, offset; 862 863 ip = mtod(m, struct ip *); 864 offset = ip->ip_hl << 2 ; 865 csum = in_cksum_skip(m, ip->ip_len, offset); 866 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 867 csum = 0xffff; 868 offset += m->m_pkthdr.csum_data; /* checksum offset */ 869 870 if (offset + sizeof(u_short) > m->m_len) { 871 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 872 m->m_len, offset, ip->ip_p); 873 /* 874 * XXX 875 * this shouldn't happen, but if it does, the 876 * correct behavior may be to insert the checksum 877 * in the appropriate next mbuf in the chain. 878 */ 879 return; 880 } 881 *(u_short *)(m->m_data + offset) = csum; 882} 883 884/* 885 * IP socket option processing. 886 */ 887int 888ip_ctloutput(struct socket *so, struct sockopt *sopt) 889{ 890 struct inpcb *inp = sotoinpcb(so); 891 int error, optval; 892 893 error = optval = 0; 894 if (sopt->sopt_level != IPPROTO_IP) { 895 if ((sopt->sopt_level == SOL_SOCKET) && 896 (sopt->sopt_name == SO_SETFIB)) { 897 inp->inp_inc.inc_fibnum = so->so_fibnum; 898 return (0); 899 } 900 return (EINVAL); 901 } 902 903 switch (sopt->sopt_dir) { 904 case SOPT_SET: 905 switch (sopt->sopt_name) { 906 case IP_OPTIONS: 907#ifdef notyet 908 case IP_RETOPTS: 909#endif 910 { 911 struct mbuf *m; 912 if (sopt->sopt_valsize > MLEN) { 913 error = EMSGSIZE; 914 break; 915 } 916 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA); 917 if (m == NULL) { 918 error = ENOBUFS; 919 break; 920 } 921 m->m_len = sopt->sopt_valsize; 922 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 923 m->m_len); 924 if (error) { 925 m_free(m); 926 break; 927 } 928 INP_WLOCK(inp); 929 error = ip_pcbopts(inp, sopt->sopt_name, m); 930 INP_WUNLOCK(inp); 931 return (error); 932 } 933 934#if defined(IP_NONLOCALBIND) 935 case IP_NONLOCALOK: 936 if (! ip_nonlocalok) { 937 error = ENOPROTOOPT; 938 break; 939 } 940 /* FALLTHROUGH */ 941#endif 942 case IP_TOS: 943 case IP_TTL: 944 case IP_MINTTL: 945 case IP_RECVOPTS: 946 case IP_RECVRETOPTS: 947 case IP_RECVDSTADDR: 948 case IP_RECVTTL: 949 case IP_RECVIF: 950 case IP_FAITH: 951 case IP_ONESBCAST: 952 case IP_DONTFRAG: 953 error = sooptcopyin(sopt, &optval, sizeof optval, 954 sizeof optval); 955 if (error) 956 break; 957 958 switch (sopt->sopt_name) { 959 case IP_TOS: 960 inp->inp_ip_tos = optval; 961 break; 962 963 case IP_TTL: 964 inp->inp_ip_ttl = optval; 965 break; 966 967 case IP_MINTTL: 968 if (optval >= 0 && optval <= MAXTTL) 969 inp->inp_ip_minttl = optval; 970 else 971 error = EINVAL; 972 break; 973 974#define OPTSET(bit) do { \ 975 INP_WLOCK(inp); \ 976 if (optval) \ 977 inp->inp_flags |= bit; \ 978 else \ 979 inp->inp_flags &= ~bit; \ 980 INP_WUNLOCK(inp); \ 981} while (0) 982 983 case IP_RECVOPTS: 984 OPTSET(INP_RECVOPTS); 985 break; 986 987 case IP_RECVRETOPTS: 988 OPTSET(INP_RECVRETOPTS); 989 break; 990 991 case IP_RECVDSTADDR: 992 OPTSET(INP_RECVDSTADDR); 993 break; 994 995 case IP_RECVTTL: 996 OPTSET(INP_RECVTTL); 997 break; 998 999 case IP_RECVIF: 1000 OPTSET(INP_RECVIF); 1001 break; 1002 1003 case IP_FAITH: 1004 OPTSET(INP_FAITH); 1005 break; 1006 1007 case IP_ONESBCAST: 1008 OPTSET(INP_ONESBCAST); 1009 break; 1010 case IP_DONTFRAG: 1011 OPTSET(INP_DONTFRAG); 1012 break; 1013#if defined(IP_NONLOCALBIND) 1014 case IP_NONLOCALOK: 1015 OPTSET(INP_NONLOCALOK); 1016 break; 1017#endif 1018 } 1019 break; 1020#undef OPTSET 1021 1022 /* 1023 * Multicast socket options are processed by the in_mcast 1024 * module. 1025 */ 1026 case IP_MULTICAST_IF: 1027 case IP_MULTICAST_VIF: 1028 case IP_MULTICAST_TTL: 1029 case IP_MULTICAST_LOOP: 1030 case IP_ADD_MEMBERSHIP: 1031 case IP_DROP_MEMBERSHIP: 1032 case IP_ADD_SOURCE_MEMBERSHIP: 1033 case IP_DROP_SOURCE_MEMBERSHIP: 1034 case IP_BLOCK_SOURCE: 1035 case IP_UNBLOCK_SOURCE: 1036 case IP_MSFILTER: 1037 case MCAST_JOIN_GROUP: 1038 case MCAST_LEAVE_GROUP: 1039 case MCAST_JOIN_SOURCE_GROUP: 1040 case MCAST_LEAVE_SOURCE_GROUP: 1041 case MCAST_BLOCK_SOURCE: 1042 case MCAST_UNBLOCK_SOURCE: 1043 error = inp_setmoptions(inp, sopt); 1044 break; 1045 1046 case IP_PORTRANGE: 1047 error = sooptcopyin(sopt, &optval, sizeof optval, 1048 sizeof optval); 1049 if (error) 1050 break; 1051 1052 INP_WLOCK(inp); 1053 switch (optval) { 1054 case IP_PORTRANGE_DEFAULT: 1055 inp->inp_flags &= ~(INP_LOWPORT); 1056 inp->inp_flags &= ~(INP_HIGHPORT); 1057 break; 1058 1059 case IP_PORTRANGE_HIGH: 1060 inp->inp_flags &= ~(INP_LOWPORT); 1061 inp->inp_flags |= INP_HIGHPORT; 1062 break; 1063 1064 case IP_PORTRANGE_LOW: 1065 inp->inp_flags &= ~(INP_HIGHPORT); 1066 inp->inp_flags |= INP_LOWPORT; 1067 break; 1068 1069 default: 1070 error = EINVAL; 1071 break; 1072 } 1073 INP_WUNLOCK(inp); 1074 break; 1075 1076#ifdef IPSEC 1077 case IP_IPSEC_POLICY: 1078 { 1079 caddr_t req; 1080 struct mbuf *m; 1081 1082 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1083 break; 1084 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1085 break; 1086 req = mtod(m, caddr_t); 1087 error = ipsec_set_policy(inp, sopt->sopt_name, req, 1088 m->m_len, (sopt->sopt_td != NULL) ? 1089 sopt->sopt_td->td_ucred : NULL); 1090 m_freem(m); 1091 break; 1092 } 1093#endif /* IPSEC */ 1094 1095 default: 1096 error = ENOPROTOOPT; 1097 break; 1098 } 1099 break; 1100 1101 case SOPT_GET: 1102 switch (sopt->sopt_name) { 1103 case IP_OPTIONS: 1104 case IP_RETOPTS: 1105 if (inp->inp_options) 1106 error = sooptcopyout(sopt, 1107 mtod(inp->inp_options, 1108 char *), 1109 inp->inp_options->m_len); 1110 else 1111 sopt->sopt_valsize = 0; 1112 break; 1113 1114 case IP_TOS: 1115 case IP_TTL: 1116 case IP_MINTTL: 1117 case IP_RECVOPTS: 1118 case IP_RECVRETOPTS: 1119 case IP_RECVDSTADDR: 1120 case IP_RECVTTL: 1121 case IP_RECVIF: 1122 case IP_PORTRANGE: 1123 case IP_FAITH: 1124 case IP_ONESBCAST: 1125 case IP_DONTFRAG: 1126 switch (sopt->sopt_name) { 1127 1128 case IP_TOS: 1129 optval = inp->inp_ip_tos; 1130 break; 1131 1132 case IP_TTL: 1133 optval = inp->inp_ip_ttl; 1134 break; 1135 1136 case IP_MINTTL: 1137 optval = inp->inp_ip_minttl; 1138 break; 1139 1140#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1141 1142 case IP_RECVOPTS: 1143 optval = OPTBIT(INP_RECVOPTS); 1144 break; 1145 1146 case IP_RECVRETOPTS: 1147 optval = OPTBIT(INP_RECVRETOPTS); 1148 break; 1149 1150 case IP_RECVDSTADDR: 1151 optval = OPTBIT(INP_RECVDSTADDR); 1152 break; 1153 1154 case IP_RECVTTL: 1155 optval = OPTBIT(INP_RECVTTL); 1156 break; 1157 1158 case IP_RECVIF: 1159 optval = OPTBIT(INP_RECVIF); 1160 break; 1161 1162 case IP_PORTRANGE: 1163 if (inp->inp_flags & INP_HIGHPORT) 1164 optval = IP_PORTRANGE_HIGH; 1165 else if (inp->inp_flags & INP_LOWPORT) 1166 optval = IP_PORTRANGE_LOW; 1167 else 1168 optval = 0; 1169 break; 1170 1171 case IP_FAITH: 1172 optval = OPTBIT(INP_FAITH); 1173 break; 1174 1175 case IP_ONESBCAST: 1176 optval = OPTBIT(INP_ONESBCAST); 1177 break; 1178 case IP_DONTFRAG: 1179 optval = OPTBIT(INP_DONTFRAG); 1180 break; 1181 } 1182 error = sooptcopyout(sopt, &optval, sizeof optval); 1183 break; 1184 1185 /* 1186 * Multicast socket options are processed by the in_mcast 1187 * module. 1188 */ 1189 case IP_MULTICAST_IF: 1190 case IP_MULTICAST_VIF: 1191 case IP_MULTICAST_TTL: 1192 case IP_MULTICAST_LOOP: 1193 case IP_MSFILTER: 1194 error = inp_getmoptions(inp, sopt); 1195 break; 1196 1197#ifdef IPSEC 1198 case IP_IPSEC_POLICY: 1199 { 1200 struct mbuf *m = NULL; 1201 caddr_t req = NULL; 1202 size_t len = 0; 1203 1204 if (m != 0) { 1205 req = mtod(m, caddr_t); 1206 len = m->m_len; 1207 } 1208 error = ipsec_get_policy(sotoinpcb(so), req, len, &m); 1209 if (error == 0) 1210 error = soopt_mcopyout(sopt, m); /* XXX */ 1211 if (error == 0) 1212 m_freem(m); 1213 break; 1214 } 1215#endif /* IPSEC */ 1216 1217 default: 1218 error = ENOPROTOOPT; 1219 break; 1220 } 1221 break; 1222 } 1223 return (error); 1224} 1225 1226/* 1227 * Routine called from ip_output() to loop back a copy of an IP multicast 1228 * packet to the input queue of a specified interface. Note that this 1229 * calls the output routine of the loopback "driver", but with an interface 1230 * pointer that might NOT be a loopback interface -- evil, but easier than 1231 * replicating that code here. 1232 */ 1233static void 1234ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst, 1235 int hlen) 1236{ 1237 register struct ip *ip; 1238 struct mbuf *copym; 1239 1240 /* 1241 * Make a deep copy of the packet because we're going to 1242 * modify the pack in order to generate checksums. 1243 */ 1244 copym = m_dup(m, M_DONTWAIT); 1245 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1246 copym = m_pullup(copym, hlen); 1247 if (copym != NULL) { 1248 /* If needed, compute the checksum and mark it as valid. */ 1249 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1250 in_delayed_cksum(copym); 1251 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1252 copym->m_pkthdr.csum_flags |= 1253 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1254 copym->m_pkthdr.csum_data = 0xffff; 1255 } 1256 /* 1257 * We don't bother to fragment if the IP length is greater 1258 * than the interface's MTU. Can this possibly matter? 1259 */ 1260 ip = mtod(copym, struct ip *); 1261 ip->ip_len = htons(ip->ip_len); 1262 ip->ip_off = htons(ip->ip_off); 1263 ip->ip_sum = 0; 1264 ip->ip_sum = in_cksum(copym, hlen); 1265#if 1 /* XXX */ 1266 if (dst->sin_family != AF_INET) { 1267 printf("ip_mloopback: bad address family %d\n", 1268 dst->sin_family); 1269 dst->sin_family = AF_INET; 1270 } 1271#endif 1272 if_simloop(ifp, copym, dst->sin_family, 0); 1273 } 1274} 1275