ip_output.c revision 260378
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: releng/10.0/sys/netinet/ip_output.c 260378 2014-01-06 19:14:46Z glebius $"); 34 35#include "opt_ipfw.h" 36#include "opt_ipsec.h" 37#include "opt_kdtrace.h" 38#include "opt_mbuf_stress_test.h" 39#include "opt_mpath.h" 40#include "opt_route.h" 41#include "opt_sctp.h" 42 43#include <sys/param.h> 44#include <sys/systm.h> 45#include <sys/kernel.h> 46#include <sys/malloc.h> 47#include <sys/mbuf.h> 48#include <sys/priv.h> 49#include <sys/proc.h> 50#include <sys/protosw.h> 51#include <sys/sdt.h> 52#include <sys/socket.h> 53#include <sys/socketvar.h> 54#include <sys/sysctl.h> 55#include <sys/ucred.h> 56 57#include <net/if.h> 58#include <net/if_llatbl.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_kdtrace.h> 70#include <netinet/in_systm.h> 71#include <netinet/ip.h> 72#include <netinet/in_pcb.h> 73#include <netinet/in_var.h> 74#include <netinet/ip_var.h> 75#include <netinet/ip_options.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 90VNET_DEFINE(u_short, ip_id); 91 92#ifdef MBUF_STRESS_TEST 93static int mbuf_frag_size = 0; 94SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 95 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 96#endif 97 98static void ip_mloopback 99 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 100 101 102extern int in_mcast_loop; 103extern struct protosw inetsw[]; 104 105/* 106 * IP output. The packet in mbuf chain m contains a skeletal IP 107 * header (with len, off, ttl, proto, tos, src, dst). 108 * The mbuf chain containing the packet will be freed. 109 * The mbuf opt, if present, will not be freed. 110 * If route ro is present and has ro_rt initialized, route lookup would be 111 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, 112 * then result of route lookup is stored in ro->ro_rt. 113 * 114 * In the IP forwarding case, the packet will arrive with options already 115 * inserted, so must have a NULL opt pointer. 116 */ 117int 118ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags, 119 struct ip_moptions *imo, struct inpcb *inp) 120{ 121 struct ip *ip; 122 struct ifnet *ifp = NULL; /* keep compiler happy */ 123 struct mbuf *m0; 124 int hlen = sizeof (struct ip); 125 int mtu; 126 int n; /* scratchpad */ 127 int error = 0; 128 struct sockaddr_in *dst; 129 const struct sockaddr_in *gw; 130 struct in_ifaddr *ia; 131 int isbroadcast; 132 uint16_t ip_len, ip_off; 133 struct route iproute; 134 struct rtentry *rte; /* cache for ro->ro_rt */ 135 struct in_addr odst; 136 struct m_tag *fwd_tag = NULL; 137#ifdef IPSEC 138 int no_route_but_check_spd = 0; 139#endif 140 M_ASSERTPKTHDR(m); 141 142 if (inp != NULL) { 143 INP_LOCK_ASSERT(inp); 144 M_SETFIB(m, inp->inp_inc.inc_fibnum); 145 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) { 146 m->m_pkthdr.flowid = inp->inp_flowid; 147 m->m_flags |= M_FLOWID; 148 } 149 } 150 151 if (ro == NULL) { 152 ro = &iproute; 153 bzero(ro, sizeof (*ro)); 154 } 155 156#ifdef FLOWTABLE 157 if (ro->ro_rt == NULL) { 158 struct flentry *fle; 159 160 /* 161 * The flow table returns route entries valid for up to 30 162 * seconds; we rely on the remainder of ip_output() taking no 163 * longer than that long for the stability of ro_rt. The 164 * flow ID assignment must have happened before this point. 165 */ 166 fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET); 167 if (fle != NULL) 168 flow_to_route(fle, ro); 169 } 170#endif 171 172 if (opt) { 173 int len = 0; 174 m = ip_insertoptions(m, opt, &len); 175 if (len != 0) 176 hlen = len; /* ip->ip_hl is updated above */ 177 } 178 ip = mtod(m, struct ip *); 179 ip_len = ntohs(ip->ip_len); 180 ip_off = ntohs(ip->ip_off); 181 182 /* 183 * Fill in IP header. If we are not allowing fragmentation, 184 * then the ip_id field is meaningless, but we don't set it 185 * to zero. Doing so causes various problems when devices along 186 * the path (routers, load balancers, firewalls, etc.) illegally 187 * disable DF on our packet. Note that a 16-bit counter 188 * will wrap around in less than 10 seconds at 100 Mbit/s on a 189 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 190 * for Counting NATted Hosts", Proc. IMW'02, available at 191 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 192 */ 193 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 194 ip->ip_v = IPVERSION; 195 ip->ip_hl = hlen >> 2; 196 ip->ip_id = ip_newid(); 197 IPSTAT_INC(ips_localout); 198 } else { 199 /* Header already set, fetch hlen from there */ 200 hlen = ip->ip_hl << 2; 201 } 202 203 gw = dst = (struct sockaddr_in *)&ro->ro_dst; 204again: 205 ia = NULL; 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 rte = ro->ro_rt; 214 if (rte && ((rte->rt_flags & RTF_UP) == 0 || 215 rte->rt_ifp == NULL || 216 !RT_LINK_IS_UP(rte->rt_ifp) || 217 dst->sin_family != AF_INET || 218 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 219 RO_RTFREE(ro); 220 ro->ro_lle = NULL; 221 rte = NULL; 222 } 223 if (rte == NULL && fwd_tag == NULL) { 224 bzero(dst, sizeof(*dst)); 225 dst->sin_family = AF_INET; 226 dst->sin_len = sizeof(*dst); 227 dst->sin_addr = ip->ip_dst; 228 } 229 /* 230 * If routing to interface only, short circuit routing lookup. 231 * The use of an all-ones broadcast address implies this; an 232 * interface is specified by the broadcast address of an interface, 233 * or the destination address of a ptp interface. 234 */ 235 if (flags & IP_SENDONES) { 236 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL && 237 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) { 238 IPSTAT_INC(ips_noroute); 239 error = ENETUNREACH; 240 goto bad; 241 } 242 ip->ip_dst.s_addr = INADDR_BROADCAST; 243 dst->sin_addr = ip->ip_dst; 244 ifp = ia->ia_ifp; 245 ip->ip_ttl = 1; 246 isbroadcast = 1; 247 } else if (flags & IP_ROUTETOIF) { 248 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 249 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) { 250 IPSTAT_INC(ips_noroute); 251 error = ENETUNREACH; 252 goto bad; 253 } 254 ifp = ia->ia_ifp; 255 ip->ip_ttl = 1; 256 isbroadcast = in_broadcast(dst->sin_addr, ifp); 257 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 258 imo != NULL && imo->imo_multicast_ifp != NULL) { 259 /* 260 * Bypass the normal routing lookup for multicast 261 * packets if the interface is specified. 262 */ 263 ifp = imo->imo_multicast_ifp; 264 IFP_TO_IA(ifp, ia); 265 isbroadcast = 0; /* fool gcc */ 266 } else { 267 /* 268 * We want to do any cloning requested by the link layer, 269 * as this is probably required in all cases for correct 270 * operation (as it is for ARP). 271 */ 272 if (rte == NULL) { 273#ifdef RADIX_MPATH 274 rtalloc_mpath_fib(ro, 275 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 276 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 277#else 278 in_rtalloc_ign(ro, 0, 279 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 280#endif 281 rte = ro->ro_rt; 282 } 283 if (rte == NULL || 284 rte->rt_ifp == NULL || 285 !RT_LINK_IS_UP(rte->rt_ifp)) { 286#ifdef IPSEC 287 /* 288 * There is no route for this packet, but it is 289 * possible that a matching SPD entry exists. 290 */ 291 no_route_but_check_spd = 1; 292 mtu = 0; /* Silence GCC warning. */ 293 goto sendit; 294#endif 295 IPSTAT_INC(ips_noroute); 296 error = EHOSTUNREACH; 297 goto bad; 298 } 299 ia = ifatoia(rte->rt_ifa); 300 ifa_ref(&ia->ia_ifa); 301 ifp = rte->rt_ifp; 302 rte->rt_rmx.rmx_pksent++; 303 if (rte->rt_flags & RTF_GATEWAY) 304 gw = (struct sockaddr_in *)rte->rt_gateway; 305 if (rte->rt_flags & RTF_HOST) 306 isbroadcast = (rte->rt_flags & RTF_BROADCAST); 307 else 308 isbroadcast = in_broadcast(gw->sin_addr, ifp); 309 } 310 /* 311 * Calculate MTU. If we have a route that is up, use that, 312 * otherwise use the interface's MTU. 313 */ 314 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) { 315 /* 316 * This case can happen if the user changed the MTU 317 * of an interface after enabling IP on it. Because 318 * most netifs don't keep track of routes pointing to 319 * them, there is no way for one to update all its 320 * routes when the MTU is changed. 321 */ 322 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu) 323 rte->rt_rmx.rmx_mtu = ifp->if_mtu; 324 mtu = rte->rt_rmx.rmx_mtu; 325 } else { 326 mtu = ifp->if_mtu; 327 } 328 /* Catch a possible divide by zero later. */ 329 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p", 330 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp)); 331 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 332 m->m_flags |= M_MCAST; 333 /* 334 * IP destination address is multicast. Make sure "gw" 335 * still points to the address in "ro". (It may have been 336 * changed to point to a gateway address, above.) 337 */ 338 gw = dst; 339 /* 340 * See if the caller provided any multicast options 341 */ 342 if (imo != NULL) { 343 ip->ip_ttl = imo->imo_multicast_ttl; 344 if (imo->imo_multicast_vif != -1) 345 ip->ip_src.s_addr = 346 ip_mcast_src ? 347 ip_mcast_src(imo->imo_multicast_vif) : 348 INADDR_ANY; 349 } else 350 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 351 /* 352 * Confirm that the outgoing interface supports multicast. 353 */ 354 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 355 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 356 IPSTAT_INC(ips_noroute); 357 error = ENETUNREACH; 358 goto bad; 359 } 360 } 361 /* 362 * If source address not specified yet, use address 363 * of outgoing interface. 364 */ 365 if (ip->ip_src.s_addr == INADDR_ANY) { 366 /* Interface may have no addresses. */ 367 if (ia != NULL) 368 ip->ip_src = IA_SIN(ia)->sin_addr; 369 } 370 371 if ((imo == NULL && in_mcast_loop) || 372 (imo && imo->imo_multicast_loop)) { 373 /* 374 * Loop back multicast datagram if not expressly 375 * forbidden to do so, even if we are not a member 376 * of the group; ip_input() will filter it later, 377 * thus deferring a hash lookup and mutex acquisition 378 * at the expense of a cheap copy using m_copym(). 379 */ 380 ip_mloopback(ifp, m, dst, hlen); 381 } else { 382 /* 383 * If we are acting as a multicast router, perform 384 * multicast forwarding as if the packet had just 385 * arrived on the interface to which we are about 386 * to send. The multicast forwarding function 387 * recursively calls this function, using the 388 * IP_FORWARDING flag to prevent infinite recursion. 389 * 390 * Multicasts that are looped back by ip_mloopback(), 391 * above, will be forwarded by the ip_input() routine, 392 * if necessary. 393 */ 394 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { 395 /* 396 * If rsvp daemon is not running, do not 397 * set ip_moptions. This ensures that the packet 398 * is multicast and not just sent down one link 399 * as prescribed by rsvpd. 400 */ 401 if (!V_rsvp_on) 402 imo = NULL; 403 if (ip_mforward && 404 ip_mforward(ip, ifp, m, imo) != 0) { 405 m_freem(m); 406 goto done; 407 } 408 } 409 } 410 411 /* 412 * Multicasts with a time-to-live of zero may be looped- 413 * back, above, but must not be transmitted on a network. 414 * Also, multicasts addressed to the loopback interface 415 * are not sent -- the above call to ip_mloopback() will 416 * loop back a copy. ip_input() will drop the copy if 417 * this host does not belong to the destination group on 418 * the loopback interface. 419 */ 420 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 421 m_freem(m); 422 goto done; 423 } 424 425 goto sendit; 426 } 427 428 /* 429 * If the source address is not specified yet, use the address 430 * of the outoing interface. 431 */ 432 if (ip->ip_src.s_addr == INADDR_ANY) { 433 /* Interface may have no addresses. */ 434 if (ia != NULL) { 435 ip->ip_src = IA_SIN(ia)->sin_addr; 436 } 437 } 438 439 /* 440 * Verify that we have any chance at all of being able to queue the 441 * packet or packet fragments, unless ALTQ is enabled on the given 442 * interface in which case packetdrop should be done by queueing. 443 */ 444 n = ip_len / mtu + 1; /* how many fragments ? */ 445 if ( 446#ifdef ALTQ 447 (!ALTQ_IS_ENABLED(&ifp->if_snd)) && 448#endif /* ALTQ */ 449 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) { 450 error = ENOBUFS; 451 IPSTAT_INC(ips_odropped); 452 ifp->if_snd.ifq_drops += n; 453 goto bad; 454 } 455 456 /* 457 * Look for broadcast address and 458 * verify user is allowed to send 459 * such a packet. 460 */ 461 if (isbroadcast) { 462 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 463 error = EADDRNOTAVAIL; 464 goto bad; 465 } 466 if ((flags & IP_ALLOWBROADCAST) == 0) { 467 error = EACCES; 468 goto bad; 469 } 470 /* don't allow broadcast messages to be fragmented */ 471 if (ip_len > mtu) { 472 error = EMSGSIZE; 473 goto bad; 474 } 475 m->m_flags |= M_BCAST; 476 } else { 477 m->m_flags &= ~M_BCAST; 478 } 479 480sendit: 481#ifdef IPSEC 482 switch(ip_ipsec_output(&m, inp, &flags, &error)) { 483 case 1: 484 goto bad; 485 case -1: 486 goto done; 487 case 0: 488 default: 489 break; /* Continue with packet processing. */ 490 } 491 /* 492 * Check if there was a route for this packet; return error if not. 493 */ 494 if (no_route_but_check_spd) { 495 IPSTAT_INC(ips_noroute); 496 error = EHOSTUNREACH; 497 goto bad; 498 } 499 /* Update variables that are affected by ipsec4_output(). */ 500 ip = mtod(m, struct ip *); 501 hlen = ip->ip_hl << 2; 502#endif /* IPSEC */ 503 504 /* Jump over all PFIL processing if hooks are not active. */ 505 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 506 goto passout; 507 508 /* Run through list of hooks for output packets. */ 509 odst.s_addr = ip->ip_dst.s_addr; 510 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 511 if (error != 0 || m == NULL) 512 goto done; 513 514 ip = mtod(m, struct ip *); 515 516 /* See if destination IP address was changed by packet filter. */ 517 if (odst.s_addr != ip->ip_dst.s_addr) { 518 m->m_flags |= M_SKIP_FIREWALL; 519 /* If destination is now ourself drop to ip_input(). */ 520 if (in_localip(ip->ip_dst)) { 521 m->m_flags |= M_FASTFWD_OURS; 522 if (m->m_pkthdr.rcvif == NULL) 523 m->m_pkthdr.rcvif = V_loif; 524 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 525 m->m_pkthdr.csum_flags |= 526 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 527 m->m_pkthdr.csum_data = 0xffff; 528 } 529 m->m_pkthdr.csum_flags |= 530 CSUM_IP_CHECKED | CSUM_IP_VALID; 531#ifdef SCTP 532 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 533 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 534#endif 535 error = netisr_queue(NETISR_IP, m); 536 goto done; 537 } else { 538 if (ia != NULL) 539 ifa_free(&ia->ia_ifa); 540 goto again; /* Redo the routing table lookup. */ 541 } 542 } 543 544 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 545 if (m->m_flags & M_FASTFWD_OURS) { 546 if (m->m_pkthdr.rcvif == NULL) 547 m->m_pkthdr.rcvif = V_loif; 548 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 549 m->m_pkthdr.csum_flags |= 550 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 551 m->m_pkthdr.csum_data = 0xffff; 552 } 553#ifdef SCTP 554 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 555 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 556#endif 557 m->m_pkthdr.csum_flags |= 558 CSUM_IP_CHECKED | CSUM_IP_VALID; 559 560 error = netisr_queue(NETISR_IP, m); 561 goto done; 562 } 563 /* Or forward to some other address? */ 564 if ((m->m_flags & M_IP_NEXTHOP) && 565 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 566 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in)); 567 m->m_flags |= M_SKIP_FIREWALL; 568 m->m_flags &= ~M_IP_NEXTHOP; 569 m_tag_delete(m, fwd_tag); 570 if (ia != NULL) 571 ifa_free(&ia->ia_ifa); 572 goto again; 573 } 574 575passout: 576 /* 127/8 must not appear on wire - RFC1122. */ 577 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 578 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 579 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 580 IPSTAT_INC(ips_badaddr); 581 error = EADDRNOTAVAIL; 582 goto bad; 583 } 584 } 585 586 m->m_pkthdr.csum_flags |= CSUM_IP; 587 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) { 588 in_delayed_cksum(m); 589 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 590 } 591#ifdef SCTP 592 if (m->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) { 593 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); 594 m->m_pkthdr.csum_flags &= ~CSUM_SCTP; 595 } 596#endif 597 598 /* 599 * If small enough for interface, or the interface will take 600 * care of the fragmentation for us, we can just send directly. 601 */ 602 if (ip_len <= mtu || 603 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 604 ((ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 605 ip->ip_sum = 0; 606 if (m->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) { 607 ip->ip_sum = in_cksum(m, hlen); 608 m->m_pkthdr.csum_flags &= ~CSUM_IP; 609 } 610 611 /* 612 * Record statistics for this interface address. 613 * With CSUM_TSO the byte/packet count will be slightly 614 * incorrect because we count the IP+TCP headers only 615 * once instead of for every generated packet. 616 */ 617 if (!(flags & IP_FORWARDING) && ia) { 618 if (m->m_pkthdr.csum_flags & CSUM_TSO) 619 ia->ia_ifa.if_opackets += 620 m->m_pkthdr.len / m->m_pkthdr.tso_segsz; 621 else 622 ia->ia_ifa.if_opackets++; 623 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 624 } 625#ifdef MBUF_STRESS_TEST 626 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 627 m = m_fragment(m, M_NOWAIT, mbuf_frag_size); 628#endif 629 /* 630 * Reset layer specific mbuf flags 631 * to avoid confusing lower layers. 632 */ 633 m_clrprotoflags(m); 634 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 635 error = (*ifp->if_output)(ifp, m, 636 (const struct sockaddr *)gw, ro); 637 goto done; 638 } 639 640 /* Balk when DF bit is set or the interface didn't support TSO. */ 641 if ((ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 642 error = EMSGSIZE; 643 IPSTAT_INC(ips_cantfrag); 644 goto bad; 645 } 646 647 /* 648 * Too large for interface; fragment if possible. If successful, 649 * on return, m will point to a list of packets to be sent. 650 */ 651 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist); 652 if (error) 653 goto bad; 654 for (; m; m = m0) { 655 m0 = m->m_nextpkt; 656 m->m_nextpkt = 0; 657 if (error == 0) { 658 /* Record statistics for this interface address. */ 659 if (ia != NULL) { 660 ia->ia_ifa.if_opackets++; 661 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 662 } 663 /* 664 * Reset layer specific mbuf flags 665 * to avoid confusing upper layers. 666 */ 667 m_clrprotoflags(m); 668 669 IP_PROBE(send, NULL, NULL, ip, ifp, ip, NULL); 670 error = (*ifp->if_output)(ifp, m, 671 (const struct sockaddr *)gw, ro); 672 } else 673 m_freem(m); 674 } 675 676 if (error == 0) 677 IPSTAT_INC(ips_fragmented); 678 679done: 680 if (ro == &iproute) 681 RO_RTFREE(ro); 682 if (ia != NULL) 683 ifa_free(&ia->ia_ifa); 684 return (error); 685bad: 686 m_freem(m); 687 goto done; 688} 689 690/* 691 * Create a chain of fragments which fit the given mtu. m_frag points to the 692 * mbuf to be fragmented; on return it points to the chain with the fragments. 693 * Return 0 if no error. If error, m_frag may contain a partially built 694 * chain of fragments that should be freed by the caller. 695 * 696 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 697 */ 698int 699ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 700 u_long if_hwassist_flags) 701{ 702 int error = 0; 703 int hlen = ip->ip_hl << 2; 704 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 705 int off; 706 struct mbuf *m0 = *m_frag; /* the original packet */ 707 int firstlen; 708 struct mbuf **mnext; 709 int nfrags; 710 uint16_t ip_len, ip_off; 711 712 ip_len = ntohs(ip->ip_len); 713 ip_off = ntohs(ip->ip_off); 714 715 if (ip_off & IP_DF) { /* Fragmentation not allowed */ 716 IPSTAT_INC(ips_cantfrag); 717 return EMSGSIZE; 718 } 719 720 /* 721 * Must be able to put at least 8 bytes per fragment. 722 */ 723 if (len < 8) 724 return EMSGSIZE; 725 726 /* 727 * If the interface will not calculate checksums on 728 * fragmented packets, then do it here. 729 */ 730 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 731 in_delayed_cksum(m0); 732 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 733 } 734#ifdef SCTP 735 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) { 736 sctp_delayed_cksum(m0, hlen); 737 m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; 738 } 739#endif 740 if (len > PAGE_SIZE) { 741 /* 742 * Fragment large datagrams such that each segment 743 * contains a multiple of PAGE_SIZE amount of data, 744 * plus headers. This enables a receiver to perform 745 * page-flipping zero-copy optimizations. 746 * 747 * XXX When does this help given that sender and receiver 748 * could have different page sizes, and also mtu could 749 * be less than the receiver's page size ? 750 */ 751 int newlen; 752 struct mbuf *m; 753 754 for (m = m0, off = 0; m && (off+m->m_len) <= mtu; m = m->m_next) 755 off += m->m_len; 756 757 /* 758 * firstlen (off - hlen) must be aligned on an 759 * 8-byte boundary 760 */ 761 if (off < hlen) 762 goto smart_frag_failure; 763 off = ((off - hlen) & ~7) + hlen; 764 newlen = (~PAGE_MASK) & mtu; 765 if ((newlen + sizeof (struct ip)) > mtu) { 766 /* we failed, go back the default */ 767smart_frag_failure: 768 newlen = len; 769 off = hlen + len; 770 } 771 len = newlen; 772 773 } else { 774 off = hlen + len; 775 } 776 777 firstlen = off - hlen; 778 mnext = &m0->m_nextpkt; /* pointer to next packet */ 779 780 /* 781 * Loop through length of segment after first fragment, 782 * make new header and copy data of each part and link onto chain. 783 * Here, m0 is the original packet, m is the fragment being created. 784 * The fragments are linked off the m_nextpkt of the original 785 * packet, which after processing serves as the first fragment. 786 */ 787 for (nfrags = 1; off < ip_len; off += len, nfrags++) { 788 struct ip *mhip; /* ip header on the fragment */ 789 struct mbuf *m; 790 int mhlen = sizeof (struct ip); 791 792 m = m_gethdr(M_NOWAIT, MT_DATA); 793 if (m == NULL) { 794 error = ENOBUFS; 795 IPSTAT_INC(ips_odropped); 796 goto done; 797 } 798 m->m_flags |= (m0->m_flags & M_MCAST); 799 /* 800 * In the first mbuf, leave room for the link header, then 801 * copy the original IP header including options. The payload 802 * goes into an additional mbuf chain returned by m_copym(). 803 */ 804 m->m_data += max_linkhdr; 805 mhip = mtod(m, struct ip *); 806 *mhip = *ip; 807 if (hlen > sizeof (struct ip)) { 808 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 809 mhip->ip_v = IPVERSION; 810 mhip->ip_hl = mhlen >> 2; 811 } 812 m->m_len = mhlen; 813 /* XXX do we need to add ip_off below ? */ 814 mhip->ip_off = ((off - hlen) >> 3) + ip_off; 815 if (off + len >= ip_len) 816 len = ip_len - off; 817 else 818 mhip->ip_off |= IP_MF; 819 mhip->ip_len = htons((u_short)(len + mhlen)); 820 m->m_next = m_copym(m0, off, len, M_NOWAIT); 821 if (m->m_next == NULL) { /* copy failed */ 822 m_free(m); 823 error = ENOBUFS; /* ??? */ 824 IPSTAT_INC(ips_odropped); 825 goto done; 826 } 827 m->m_pkthdr.len = mhlen + len; 828 m->m_pkthdr.rcvif = NULL; 829#ifdef MAC 830 mac_netinet_fragment(m0, m); 831#endif 832 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 833 mhip->ip_off = htons(mhip->ip_off); 834 mhip->ip_sum = 0; 835 if (m->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 836 mhip->ip_sum = in_cksum(m, mhlen); 837 m->m_pkthdr.csum_flags &= ~CSUM_IP; 838 } 839 *mnext = m; 840 mnext = &m->m_nextpkt; 841 } 842 IPSTAT_ADD(ips_ofragments, nfrags); 843 844 /* 845 * Update first fragment by trimming what's been copied out 846 * and updating header. 847 */ 848 m_adj(m0, hlen + firstlen - ip_len); 849 m0->m_pkthdr.len = hlen + firstlen; 850 ip->ip_len = htons((u_short)m0->m_pkthdr.len); 851 ip->ip_off = htons(ip_off | IP_MF); 852 ip->ip_sum = 0; 853 if (m0->m_pkthdr.csum_flags & CSUM_IP & ~if_hwassist_flags) { 854 ip->ip_sum = in_cksum(m0, hlen); 855 m0->m_pkthdr.csum_flags &= ~CSUM_IP; 856 } 857 858done: 859 *m_frag = m0; 860 return error; 861} 862 863void 864in_delayed_cksum(struct mbuf *m) 865{ 866 struct ip *ip; 867 uint16_t csum, offset, ip_len; 868 869 ip = mtod(m, struct ip *); 870 offset = ip->ip_hl << 2 ; 871 ip_len = ntohs(ip->ip_len); 872 csum = in_cksum_skip(m, ip_len, offset); 873 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 874 csum = 0xffff; 875 offset += m->m_pkthdr.csum_data; /* checksum offset */ 876 877 if (offset + sizeof(u_short) > m->m_len) { 878 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 879 m->m_len, offset, ip->ip_p); 880 /* 881 * XXX 882 * this shouldn't happen, but if it does, the 883 * correct behavior may be to insert the checksum 884 * in the appropriate next mbuf in the chain. 885 */ 886 return; 887 } 888 *(u_short *)(m->m_data + offset) = csum; 889} 890 891/* 892 * IP socket option processing. 893 */ 894int 895ip_ctloutput(struct socket *so, struct sockopt *sopt) 896{ 897 struct inpcb *inp = sotoinpcb(so); 898 int error, optval; 899 900 error = optval = 0; 901 if (sopt->sopt_level != IPPROTO_IP) { 902 error = EINVAL; 903 904 if (sopt->sopt_level == SOL_SOCKET && 905 sopt->sopt_dir == SOPT_SET) { 906 switch (sopt->sopt_name) { 907 case SO_REUSEADDR: 908 INP_WLOCK(inp); 909 if ((so->so_options & SO_REUSEADDR) != 0) 910 inp->inp_flags2 |= INP_REUSEADDR; 911 else 912 inp->inp_flags2 &= ~INP_REUSEADDR; 913 INP_WUNLOCK(inp); 914 error = 0; 915 break; 916 case SO_REUSEPORT: 917 INP_WLOCK(inp); 918 if ((so->so_options & SO_REUSEPORT) != 0) 919 inp->inp_flags2 |= INP_REUSEPORT; 920 else 921 inp->inp_flags2 &= ~INP_REUSEPORT; 922 INP_WUNLOCK(inp); 923 error = 0; 924 break; 925 case SO_SETFIB: 926 INP_WLOCK(inp); 927 inp->inp_inc.inc_fibnum = so->so_fibnum; 928 INP_WUNLOCK(inp); 929 error = 0; 930 break; 931 default: 932 break; 933 } 934 } 935 return (error); 936 } 937 938 switch (sopt->sopt_dir) { 939 case SOPT_SET: 940 switch (sopt->sopt_name) { 941 case IP_OPTIONS: 942#ifdef notyet 943 case IP_RETOPTS: 944#endif 945 { 946 struct mbuf *m; 947 if (sopt->sopt_valsize > MLEN) { 948 error = EMSGSIZE; 949 break; 950 } 951 m = m_get(sopt->sopt_td ? M_WAITOK : M_NOWAIT, MT_DATA); 952 if (m == NULL) { 953 error = ENOBUFS; 954 break; 955 } 956 m->m_len = sopt->sopt_valsize; 957 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 958 m->m_len); 959 if (error) { 960 m_free(m); 961 break; 962 } 963 INP_WLOCK(inp); 964 error = ip_pcbopts(inp, sopt->sopt_name, m); 965 INP_WUNLOCK(inp); 966 return (error); 967 } 968 969 case IP_BINDANY: 970 if (sopt->sopt_td != NULL) { 971 error = priv_check(sopt->sopt_td, 972 PRIV_NETINET_BINDANY); 973 if (error) 974 break; 975 } 976 /* FALLTHROUGH */ 977 case IP_TOS: 978 case IP_TTL: 979 case IP_MINTTL: 980 case IP_RECVOPTS: 981 case IP_RECVRETOPTS: 982 case IP_RECVDSTADDR: 983 case IP_RECVTTL: 984 case IP_RECVIF: 985 case IP_FAITH: 986 case IP_ONESBCAST: 987 case IP_DONTFRAG: 988 case IP_RECVTOS: 989 error = sooptcopyin(sopt, &optval, sizeof optval, 990 sizeof optval); 991 if (error) 992 break; 993 994 switch (sopt->sopt_name) { 995 case IP_TOS: 996 inp->inp_ip_tos = optval; 997 break; 998 999 case IP_TTL: 1000 inp->inp_ip_ttl = optval; 1001 break; 1002 1003 case IP_MINTTL: 1004 if (optval >= 0 && optval <= MAXTTL) 1005 inp->inp_ip_minttl = optval; 1006 else 1007 error = EINVAL; 1008 break; 1009 1010#define OPTSET(bit) do { \ 1011 INP_WLOCK(inp); \ 1012 if (optval) \ 1013 inp->inp_flags |= bit; \ 1014 else \ 1015 inp->inp_flags &= ~bit; \ 1016 INP_WUNLOCK(inp); \ 1017} while (0) 1018 1019 case IP_RECVOPTS: 1020 OPTSET(INP_RECVOPTS); 1021 break; 1022 1023 case IP_RECVRETOPTS: 1024 OPTSET(INP_RECVRETOPTS); 1025 break; 1026 1027 case IP_RECVDSTADDR: 1028 OPTSET(INP_RECVDSTADDR); 1029 break; 1030 1031 case IP_RECVTTL: 1032 OPTSET(INP_RECVTTL); 1033 break; 1034 1035 case IP_RECVIF: 1036 OPTSET(INP_RECVIF); 1037 break; 1038 1039 case IP_FAITH: 1040 OPTSET(INP_FAITH); 1041 break; 1042 1043 case IP_ONESBCAST: 1044 OPTSET(INP_ONESBCAST); 1045 break; 1046 case IP_DONTFRAG: 1047 OPTSET(INP_DONTFRAG); 1048 break; 1049 case IP_BINDANY: 1050 OPTSET(INP_BINDANY); 1051 break; 1052 case IP_RECVTOS: 1053 OPTSET(INP_RECVTOS); 1054 break; 1055 } 1056 break; 1057#undef OPTSET 1058 1059 /* 1060 * Multicast socket options are processed by the in_mcast 1061 * module. 1062 */ 1063 case IP_MULTICAST_IF: 1064 case IP_MULTICAST_VIF: 1065 case IP_MULTICAST_TTL: 1066 case IP_MULTICAST_LOOP: 1067 case IP_ADD_MEMBERSHIP: 1068 case IP_DROP_MEMBERSHIP: 1069 case IP_ADD_SOURCE_MEMBERSHIP: 1070 case IP_DROP_SOURCE_MEMBERSHIP: 1071 case IP_BLOCK_SOURCE: 1072 case IP_UNBLOCK_SOURCE: 1073 case IP_MSFILTER: 1074 case MCAST_JOIN_GROUP: 1075 case MCAST_LEAVE_GROUP: 1076 case MCAST_JOIN_SOURCE_GROUP: 1077 case MCAST_LEAVE_SOURCE_GROUP: 1078 case MCAST_BLOCK_SOURCE: 1079 case MCAST_UNBLOCK_SOURCE: 1080 error = inp_setmoptions(inp, sopt); 1081 break; 1082 1083 case IP_PORTRANGE: 1084 error = sooptcopyin(sopt, &optval, sizeof optval, 1085 sizeof optval); 1086 if (error) 1087 break; 1088 1089 INP_WLOCK(inp); 1090 switch (optval) { 1091 case IP_PORTRANGE_DEFAULT: 1092 inp->inp_flags &= ~(INP_LOWPORT); 1093 inp->inp_flags &= ~(INP_HIGHPORT); 1094 break; 1095 1096 case IP_PORTRANGE_HIGH: 1097 inp->inp_flags &= ~(INP_LOWPORT); 1098 inp->inp_flags |= INP_HIGHPORT; 1099 break; 1100 1101 case IP_PORTRANGE_LOW: 1102 inp->inp_flags &= ~(INP_HIGHPORT); 1103 inp->inp_flags |= INP_LOWPORT; 1104 break; 1105 1106 default: 1107 error = EINVAL; 1108 break; 1109 } 1110 INP_WUNLOCK(inp); 1111 break; 1112 1113#ifdef IPSEC 1114 case IP_IPSEC_POLICY: 1115 { 1116 caddr_t req; 1117 struct mbuf *m; 1118 1119 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1120 break; 1121 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1122 break; 1123 req = mtod(m, caddr_t); 1124 error = ipsec_set_policy(inp, sopt->sopt_name, req, 1125 m->m_len, (sopt->sopt_td != NULL) ? 1126 sopt->sopt_td->td_ucred : NULL); 1127 m_freem(m); 1128 break; 1129 } 1130#endif /* IPSEC */ 1131 1132 default: 1133 error = ENOPROTOOPT; 1134 break; 1135 } 1136 break; 1137 1138 case SOPT_GET: 1139 switch (sopt->sopt_name) { 1140 case IP_OPTIONS: 1141 case IP_RETOPTS: 1142 if (inp->inp_options) 1143 error = sooptcopyout(sopt, 1144 mtod(inp->inp_options, 1145 char *), 1146 inp->inp_options->m_len); 1147 else 1148 sopt->sopt_valsize = 0; 1149 break; 1150 1151 case IP_TOS: 1152 case IP_TTL: 1153 case IP_MINTTL: 1154 case IP_RECVOPTS: 1155 case IP_RECVRETOPTS: 1156 case IP_RECVDSTADDR: 1157 case IP_RECVTTL: 1158 case IP_RECVIF: 1159 case IP_PORTRANGE: 1160 case IP_FAITH: 1161 case IP_ONESBCAST: 1162 case IP_DONTFRAG: 1163 case IP_BINDANY: 1164 case IP_RECVTOS: 1165 switch (sopt->sopt_name) { 1166 1167 case IP_TOS: 1168 optval = inp->inp_ip_tos; 1169 break; 1170 1171 case IP_TTL: 1172 optval = inp->inp_ip_ttl; 1173 break; 1174 1175 case IP_MINTTL: 1176 optval = inp->inp_ip_minttl; 1177 break; 1178 1179#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1180 1181 case IP_RECVOPTS: 1182 optval = OPTBIT(INP_RECVOPTS); 1183 break; 1184 1185 case IP_RECVRETOPTS: 1186 optval = OPTBIT(INP_RECVRETOPTS); 1187 break; 1188 1189 case IP_RECVDSTADDR: 1190 optval = OPTBIT(INP_RECVDSTADDR); 1191 break; 1192 1193 case IP_RECVTTL: 1194 optval = OPTBIT(INP_RECVTTL); 1195 break; 1196 1197 case IP_RECVIF: 1198 optval = OPTBIT(INP_RECVIF); 1199 break; 1200 1201 case IP_PORTRANGE: 1202 if (inp->inp_flags & INP_HIGHPORT) 1203 optval = IP_PORTRANGE_HIGH; 1204 else if (inp->inp_flags & INP_LOWPORT) 1205 optval = IP_PORTRANGE_LOW; 1206 else 1207 optval = 0; 1208 break; 1209 1210 case IP_FAITH: 1211 optval = OPTBIT(INP_FAITH); 1212 break; 1213 1214 case IP_ONESBCAST: 1215 optval = OPTBIT(INP_ONESBCAST); 1216 break; 1217 case IP_DONTFRAG: 1218 optval = OPTBIT(INP_DONTFRAG); 1219 break; 1220 case IP_BINDANY: 1221 optval = OPTBIT(INP_BINDANY); 1222 break; 1223 case IP_RECVTOS: 1224 optval = OPTBIT(INP_RECVTOS); 1225 break; 1226 } 1227 error = sooptcopyout(sopt, &optval, sizeof optval); 1228 break; 1229 1230 /* 1231 * Multicast socket options are processed by the in_mcast 1232 * module. 1233 */ 1234 case IP_MULTICAST_IF: 1235 case IP_MULTICAST_VIF: 1236 case IP_MULTICAST_TTL: 1237 case IP_MULTICAST_LOOP: 1238 case IP_MSFILTER: 1239 error = inp_getmoptions(inp, sopt); 1240 break; 1241 1242#ifdef IPSEC 1243 case IP_IPSEC_POLICY: 1244 { 1245 struct mbuf *m = NULL; 1246 caddr_t req = NULL; 1247 size_t len = 0; 1248 1249 if (m != 0) { 1250 req = mtod(m, caddr_t); 1251 len = m->m_len; 1252 } 1253 error = ipsec_get_policy(sotoinpcb(so), req, len, &m); 1254 if (error == 0) 1255 error = soopt_mcopyout(sopt, m); /* XXX */ 1256 if (error == 0) 1257 m_freem(m); 1258 break; 1259 } 1260#endif /* IPSEC */ 1261 1262 default: 1263 error = ENOPROTOOPT; 1264 break; 1265 } 1266 break; 1267 } 1268 return (error); 1269} 1270 1271/* 1272 * Routine called from ip_output() to loop back a copy of an IP multicast 1273 * packet to the input queue of a specified interface. Note that this 1274 * calls the output routine of the loopback "driver", but with an interface 1275 * pointer that might NOT be a loopback interface -- evil, but easier than 1276 * replicating that code here. 1277 */ 1278static void 1279ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst, 1280 int hlen) 1281{ 1282 register struct ip *ip; 1283 struct mbuf *copym; 1284 1285 /* 1286 * Make a deep copy of the packet because we're going to 1287 * modify the pack in order to generate checksums. 1288 */ 1289 copym = m_dup(m, M_NOWAIT); 1290 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1291 copym = m_pullup(copym, hlen); 1292 if (copym != NULL) { 1293 /* If needed, compute the checksum and mark it as valid. */ 1294 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1295 in_delayed_cksum(copym); 1296 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1297 copym->m_pkthdr.csum_flags |= 1298 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1299 copym->m_pkthdr.csum_data = 0xffff; 1300 } 1301 /* 1302 * We don't bother to fragment if the IP length is greater 1303 * than the interface's MTU. Can this possibly matter? 1304 */ 1305 ip = mtod(copym, struct ip *); 1306 ip->ip_sum = 0; 1307 ip->ip_sum = in_cksum(copym, hlen); 1308#if 1 /* XXX */ 1309 if (dst->sin_family != AF_INET) { 1310 printf("ip_mloopback: bad address family %d\n", 1311 dst->sin_family); 1312 dst->sin_family = AF_INET; 1313 } 1314#endif 1315 if_simloop(ifp, copym, dst->sin_family, 0); 1316 } 1317} 1318