ip_output.c revision 266192
1/*- 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 30 */ 31 32#include <sys/cdefs.h> 33__FBSDID("$FreeBSD: stable/9/sys/netinet/ip_output.c 266192 2014-05-15 20:27:57Z tuexen $"); 34 35#include "opt_ipfw.h" 36#include "opt_ipsec.h" 37#include "opt_route.h" 38#include "opt_mbuf_stress_test.h" 39#include "opt_mpath.h" 40#include "opt_sctp.h" 41 42#include <sys/param.h> 43#include <sys/systm.h> 44#include <sys/kernel.h> 45#include <sys/malloc.h> 46#include <sys/mbuf.h> 47#include <sys/priv.h> 48#include <sys/proc.h> 49#include <sys/protosw.h> 50#include <sys/socket.h> 51#include <sys/socketvar.h> 52#include <sys/sysctl.h> 53#include <sys/ucred.h> 54 55#include <net/if.h> 56#include <net/if_llatbl.h> 57#include <net/netisr.h> 58#include <net/pfil.h> 59#include <net/route.h> 60#include <net/flowtable.h> 61#ifdef RADIX_MPATH 62#include <net/radix_mpath.h> 63#endif 64#include <net/vnet.h> 65 66#include <netinet/in.h> 67#include <netinet/in_systm.h> 68#include <netinet/ip.h> 69#include <netinet/in_pcb.h> 70#include <netinet/in_var.h> 71#include <netinet/ip_var.h> 72#include <netinet/ip_options.h> 73#ifdef SCTP 74#include <netinet/sctp.h> 75#include <netinet/sctp_crc32.h> 76#endif 77 78#ifdef IPSEC 79#include <netinet/ip_ipsec.h> 80#include <netipsec/ipsec.h> 81#endif /* IPSEC*/ 82 83#include <machine/in_cksum.h> 84 85#include <security/mac/mac_framework.h> 86 87VNET_DEFINE(u_short, ip_id); 88 89#ifdef MBUF_STRESS_TEST 90static int mbuf_frag_size = 0; 91SYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 92 &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 93#endif 94 95static void ip_mloopback 96 (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 97 98 99extern int in_mcast_loop; 100extern struct protosw inetsw[]; 101 102/* 103 * IP output. The packet in mbuf chain m contains a skeletal IP 104 * header (with len, off, ttl, proto, tos, src, dst). 105 * ip_len and ip_off are in host format. 106 * The mbuf chain containing the packet will be freed. 107 * The mbuf opt, if present, will not be freed. 108 * If route ro is present and has ro_rt initialized, route lookup would be 109 * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL, 110 * then result of route lookup is stored in ro->ro_rt. 111 * 112 * In the IP forwarding case, the packet will arrive with options already 113 * inserted, so must have a NULL opt pointer. 114 */ 115int 116ip_output(struct mbuf *m, struct mbuf *opt, struct route *ro, int flags, 117 struct ip_moptions *imo, struct inpcb *inp) 118{ 119 struct ip *ip; 120 struct ifnet *ifp = NULL; /* keep compiler happy */ 121 struct mbuf *m0; 122 int hlen = sizeof (struct ip); 123 int mtu; 124 int n; /* scratchpad */ 125 int error = 0; 126 struct sockaddr_in *dst; 127 struct in_ifaddr *ia; 128 int isbroadcast, sw_csum; 129 struct route iproute; 130 struct rtentry *rte; /* cache for ro->ro_rt */ 131 struct in_addr odst; 132 struct m_tag *fwd_tag = NULL; 133#ifdef IPSEC 134 int no_route_but_check_spd = 0; 135#endif 136 M_ASSERTPKTHDR(m); 137 138 if (inp != NULL) { 139 INP_LOCK_ASSERT(inp); 140 M_SETFIB(m, inp->inp_inc.inc_fibnum); 141 if (inp->inp_flags & (INP_HW_FLOWID|INP_SW_FLOWID)) { 142 m->m_pkthdr.flowid = inp->inp_flowid; 143 m->m_flags |= M_FLOWID; 144 } 145 } 146 147 if (ro == NULL) { 148 ro = &iproute; 149 bzero(ro, sizeof (*ro)); 150 } 151 152#ifdef FLOWTABLE 153 if (ro->ro_rt == NULL) { 154 struct flentry *fle; 155 156 /* 157 * The flow table returns route entries valid for up to 30 158 * seconds; we rely on the remainder of ip_output() taking no 159 * longer than that long for the stability of ro_rt. The 160 * flow ID assignment must have happened before this point. 161 */ 162 fle = flowtable_lookup_mbuf(V_ip_ft, m, AF_INET); 163 if (fle != NULL) 164 flow_to_route(fle, ro); 165 } 166#endif 167 168 if (opt) { 169 int len = 0; 170 m = ip_insertoptions(m, opt, &len); 171 if (len != 0) 172 hlen = len; /* ip->ip_hl is updated above */ 173 } 174 ip = mtod(m, struct ip *); 175 176 /* 177 * Fill in IP header. If we are not allowing fragmentation, 178 * then the ip_id field is meaningless, but we don't set it 179 * to zero. Doing so causes various problems when devices along 180 * the path (routers, load balancers, firewalls, etc.) illegally 181 * disable DF on our packet. Note that a 16-bit counter 182 * will wrap around in less than 10 seconds at 100 Mbit/s on a 183 * medium with MTU 1500. See Steven M. Bellovin, "A Technique 184 * for Counting NATted Hosts", Proc. IMW'02, available at 185 * <http://www.cs.columbia.edu/~smb/papers/fnat.pdf>. 186 */ 187 if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 188 ip->ip_v = IPVERSION; 189 ip->ip_hl = hlen >> 2; 190 ip->ip_id = ip_newid(); 191 IPSTAT_INC(ips_localout); 192 } else { 193 /* Header already set, fetch hlen from there */ 194 hlen = ip->ip_hl << 2; 195 } 196 197again: 198 dst = (struct sockaddr_in *)&ro->ro_dst; 199 ia = NULL; 200 /* 201 * If there is a cached route, 202 * check that it is to the same destination 203 * and is still up. If not, free it and try again. 204 * The address family should also be checked in case of sharing the 205 * cache with IPv6. 206 */ 207 rte = ro->ro_rt; 208 if (rte && ((rte->rt_flags & RTF_UP) == 0 || 209 rte->rt_ifp == NULL || 210 !RT_LINK_IS_UP(rte->rt_ifp) || 211 dst->sin_family != AF_INET || 212 dst->sin_addr.s_addr != ip->ip_dst.s_addr)) { 213 RO_RTFREE(ro); 214 ro->ro_lle = NULL; 215 rte = NULL; 216 } 217 if (rte == NULL && fwd_tag == NULL) { 218 bzero(dst, sizeof(*dst)); 219 dst->sin_family = AF_INET; 220 dst->sin_len = sizeof(*dst); 221 dst->sin_addr = ip->ip_dst; 222 } 223 /* 224 * If routing to interface only, short circuit routing lookup. 225 * The use of an all-ones broadcast address implies this; an 226 * interface is specified by the broadcast address of an interface, 227 * or the destination address of a ptp interface. 228 */ 229 if (flags & IP_SENDONES) { 230 if ((ia = ifatoia(ifa_ifwithbroadaddr(sintosa(dst)))) == NULL && 231 (ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL) { 232 IPSTAT_INC(ips_noroute); 233 error = ENETUNREACH; 234 goto bad; 235 } 236 ip->ip_dst.s_addr = INADDR_BROADCAST; 237 dst->sin_addr = ip->ip_dst; 238 ifp = ia->ia_ifp; 239 ip->ip_ttl = 1; 240 isbroadcast = 1; 241 } else if (flags & IP_ROUTETOIF) { 242 if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == NULL && 243 (ia = ifatoia(ifa_ifwithnet(sintosa(dst), 0))) == NULL) { 244 IPSTAT_INC(ips_noroute); 245 error = ENETUNREACH; 246 goto bad; 247 } 248 ifp = ia->ia_ifp; 249 ip->ip_ttl = 1; 250 isbroadcast = in_broadcast(dst->sin_addr, ifp); 251 } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 252 imo != NULL && imo->imo_multicast_ifp != NULL) { 253 /* 254 * Bypass the normal routing lookup for multicast 255 * packets if the interface is specified. 256 */ 257 ifp = imo->imo_multicast_ifp; 258 IFP_TO_IA(ifp, ia); 259 isbroadcast = 0; /* fool gcc */ 260 } else { 261 /* 262 * We want to do any cloning requested by the link layer, 263 * as this is probably required in all cases for correct 264 * operation (as it is for ARP). 265 */ 266 if (rte == NULL) { 267#ifdef RADIX_MPATH 268 rtalloc_mpath_fib(ro, 269 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 270 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 271#else 272 in_rtalloc_ign(ro, 0, 273 inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m)); 274#endif 275 rte = ro->ro_rt; 276 } 277 if (rte == NULL || 278 rte->rt_ifp == NULL || 279 !RT_LINK_IS_UP(rte->rt_ifp)) { 280#ifdef IPSEC 281 /* 282 * There is no route for this packet, but it is 283 * possible that a matching SPD entry exists. 284 */ 285 no_route_but_check_spd = 1; 286 mtu = 0; /* Silence GCC warning. */ 287 goto sendit; 288#endif 289 IPSTAT_INC(ips_noroute); 290 error = EHOSTUNREACH; 291 goto bad; 292 } 293 ia = ifatoia(rte->rt_ifa); 294 ifa_ref(&ia->ia_ifa); 295 ifp = rte->rt_ifp; 296 rte->rt_rmx.rmx_pksent++; 297 if (rte->rt_flags & RTF_GATEWAY) 298 dst = (struct sockaddr_in *)rte->rt_gateway; 299 if (rte->rt_flags & RTF_HOST) 300 isbroadcast = (rte->rt_flags & RTF_BROADCAST); 301 else 302 isbroadcast = in_broadcast(dst->sin_addr, ifp); 303 } 304 /* 305 * Calculate MTU. If we have a route that is up, use that, 306 * otherwise use the interface's MTU. 307 */ 308 if (rte != NULL && (rte->rt_flags & (RTF_UP|RTF_HOST))) { 309 /* 310 * This case can happen if the user changed the MTU 311 * of an interface after enabling IP on it. Because 312 * most netifs don't keep track of routes pointing to 313 * them, there is no way for one to update all its 314 * routes when the MTU is changed. 315 */ 316 if (rte->rt_rmx.rmx_mtu > ifp->if_mtu) 317 rte->rt_rmx.rmx_mtu = ifp->if_mtu; 318 mtu = rte->rt_rmx.rmx_mtu; 319 } else { 320 mtu = ifp->if_mtu; 321 } 322 /* Catch a possible divide by zero later. */ 323 KASSERT(mtu > 0, ("%s: mtu %d <= 0, rte=%p (rt_flags=0x%08x) ifp=%p", 324 __func__, mtu, rte, (rte != NULL) ? rte->rt_flags : 0, ifp)); 325 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 326 m->m_flags |= M_MCAST; 327 /* 328 * IP destination address is multicast. Make sure "dst" 329 * still points to the address in "ro". (It may have been 330 * changed to point to a gateway address, above.) 331 */ 332 dst = (struct sockaddr_in *)&ro->ro_dst; 333 /* 334 * See if the caller provided any multicast options 335 */ 336 if (imo != NULL) { 337 ip->ip_ttl = imo->imo_multicast_ttl; 338 if (imo->imo_multicast_vif != -1) 339 ip->ip_src.s_addr = 340 ip_mcast_src ? 341 ip_mcast_src(imo->imo_multicast_vif) : 342 INADDR_ANY; 343 } else 344 ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 345 /* 346 * Confirm that the outgoing interface supports multicast. 347 */ 348 if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 349 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 350 IPSTAT_INC(ips_noroute); 351 error = ENETUNREACH; 352 goto bad; 353 } 354 } 355 /* 356 * If source address not specified yet, use address 357 * of outgoing interface. 358 */ 359 if (ip->ip_src.s_addr == INADDR_ANY) { 360 /* Interface may have no addresses. */ 361 if (ia != NULL) 362 ip->ip_src = IA_SIN(ia)->sin_addr; 363 } 364 365 if ((imo == NULL && in_mcast_loop) || 366 (imo && imo->imo_multicast_loop)) { 367 /* 368 * Loop back multicast datagram if not expressly 369 * forbidden to do so, even if we are not a member 370 * of the group; ip_input() will filter it later, 371 * thus deferring a hash lookup and mutex acquisition 372 * at the expense of a cheap copy using m_copym(). 373 */ 374 ip_mloopback(ifp, m, dst, hlen); 375 } else { 376 /* 377 * If we are acting as a multicast router, perform 378 * multicast forwarding as if the packet had just 379 * arrived on the interface to which we are about 380 * to send. The multicast forwarding function 381 * recursively calls this function, using the 382 * IP_FORWARDING flag to prevent infinite recursion. 383 * 384 * Multicasts that are looped back by ip_mloopback(), 385 * above, will be forwarded by the ip_input() routine, 386 * if necessary. 387 */ 388 if (V_ip_mrouter && (flags & IP_FORWARDING) == 0) { 389 /* 390 * If rsvp daemon is not running, do not 391 * set ip_moptions. This ensures that the packet 392 * is multicast and not just sent down one link 393 * as prescribed by rsvpd. 394 */ 395 if (!V_rsvp_on) 396 imo = NULL; 397 if (ip_mforward && 398 ip_mforward(ip, ifp, m, imo) != 0) { 399 m_freem(m); 400 goto done; 401 } 402 } 403 } 404 405 /* 406 * Multicasts with a time-to-live of zero may be looped- 407 * back, above, but must not be transmitted on a network. 408 * Also, multicasts addressed to the loopback interface 409 * are not sent -- the above call to ip_mloopback() will 410 * loop back a copy. ip_input() will drop the copy if 411 * this host does not belong to the destination group on 412 * the loopback interface. 413 */ 414 if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 415 m_freem(m); 416 goto done; 417 } 418 419 goto sendit; 420 } 421 422 /* 423 * If the source address is not specified yet, use the address 424 * of the outoing interface. 425 */ 426 if (ip->ip_src.s_addr == INADDR_ANY) { 427 /* Interface may have no addresses. */ 428 if (ia != NULL) { 429 ip->ip_src = IA_SIN(ia)->sin_addr; 430 } 431 } 432 433 /* 434 * Verify that we have any chance at all of being able to queue the 435 * packet or packet fragments, unless ALTQ is enabled on the given 436 * interface in which case packetdrop should be done by queueing. 437 */ 438 n = ip->ip_len / mtu + 1; /* how many fragments ? */ 439 if ( 440#ifdef ALTQ 441 (!ALTQ_IS_ENABLED(&ifp->if_snd)) && 442#endif /* ALTQ */ 443 (ifp->if_snd.ifq_len + n) >= ifp->if_snd.ifq_maxlen ) { 444 error = ENOBUFS; 445 IPSTAT_INC(ips_odropped); 446 ifp->if_snd.ifq_drops += n; 447 goto bad; 448 } 449 450 /* 451 * Look for broadcast address and 452 * verify user is allowed to send 453 * such a packet. 454 */ 455 if (isbroadcast) { 456 if ((ifp->if_flags & IFF_BROADCAST) == 0) { 457 error = EADDRNOTAVAIL; 458 goto bad; 459 } 460 if ((flags & IP_ALLOWBROADCAST) == 0) { 461 error = EACCES; 462 goto bad; 463 } 464 /* don't allow broadcast messages to be fragmented */ 465 if (ip->ip_len > mtu) { 466 error = EMSGSIZE; 467 goto bad; 468 } 469 m->m_flags |= M_BCAST; 470 } else { 471 m->m_flags &= ~M_BCAST; 472 } 473 474sendit: 475#ifdef IPSEC 476 switch(ip_ipsec_output(&m, inp, &flags, &error)) { 477 case 1: 478 goto bad; 479 case -1: 480 goto done; 481 case 0: 482 default: 483 break; /* Continue with packet processing. */ 484 } 485 /* 486 * Check if there was a route for this packet; return error if not. 487 */ 488 if (no_route_but_check_spd) { 489 IPSTAT_INC(ips_noroute); 490 error = EHOSTUNREACH; 491 goto bad; 492 } 493 /* Update variables that are affected by ipsec4_output(). */ 494 ip = mtod(m, struct ip *); 495 hlen = ip->ip_hl << 2; 496#endif /* IPSEC */ 497 498 /* Jump over all PFIL processing if hooks are not active. */ 499 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 500 goto passout; 501 502 /* Run through list of hooks for output packets. */ 503 odst.s_addr = ip->ip_dst.s_addr; 504 error = pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_OUT, inp); 505 if (error != 0 || m == NULL) 506 goto done; 507 508 ip = mtod(m, struct ip *); 509 510 /* See if destination IP address was changed by packet filter. */ 511 if (odst.s_addr != ip->ip_dst.s_addr) { 512 m->m_flags |= M_SKIP_FIREWALL; 513 /* If destination is now ourself drop to ip_input(). */ 514 if (in_localip(ip->ip_dst)) { 515 m->m_flags |= M_FASTFWD_OURS; 516 if (m->m_pkthdr.rcvif == NULL) 517 m->m_pkthdr.rcvif = V_loif; 518 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 519 m->m_pkthdr.csum_flags |= 520 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 521 m->m_pkthdr.csum_data = 0xffff; 522 } 523 m->m_pkthdr.csum_flags |= 524 CSUM_IP_CHECKED | CSUM_IP_VALID; 525#ifdef SCTP 526 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 527 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 528#endif 529 error = netisr_queue(NETISR_IP, m); 530 goto done; 531 } else { 532 if (ia != NULL) 533 ifa_free(&ia->ia_ifa); 534 goto again; /* Redo the routing table lookup. */ 535 } 536 } 537 538 /* See if local, if yes, send it to netisr with IP_FASTFWD_OURS. */ 539 if (m->m_flags & M_FASTFWD_OURS) { 540 if (m->m_pkthdr.rcvif == NULL) 541 m->m_pkthdr.rcvif = V_loif; 542 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 543 m->m_pkthdr.csum_flags |= 544 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 545 m->m_pkthdr.csum_data = 0xffff; 546 } 547#ifdef SCTP 548 if (m->m_pkthdr.csum_flags & CSUM_SCTP) 549 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 550#endif 551 m->m_pkthdr.csum_flags |= 552 CSUM_IP_CHECKED | CSUM_IP_VALID; 553 554 error = netisr_queue(NETISR_IP, m); 555 goto done; 556 } 557 /* Or forward to some other address? */ 558 if ((m->m_flags & M_IP_NEXTHOP) && 559 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 560 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->m_flags &= ~M_IP_NEXTHOP; 564 m_tag_delete(m, fwd_tag); 565 if (ia != NULL) 566 ifa_free(&ia->ia_ifa); 567 goto again; 568 } 569 570passout: 571 /* 127/8 must not appear on wire - RFC1122. */ 572 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 573 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 574 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 575 IPSTAT_INC(ips_badaddr); 576 error = EADDRNOTAVAIL; 577 goto bad; 578 } 579 } 580 581 m->m_pkthdr.csum_flags |= CSUM_IP; 582 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 583 if (sw_csum & CSUM_DELAY_DATA) { 584 in_delayed_cksum(m); 585 sw_csum &= ~CSUM_DELAY_DATA; 586 } 587#ifdef SCTP 588 if (sw_csum & CSUM_SCTP) { 589 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); 590 sw_csum &= ~CSUM_SCTP; 591 } 592#endif 593 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 594 595 /* 596 * If small enough for interface, or the interface will take 597 * care of the fragmentation for us, we can just send directly. 598 */ 599 if (ip->ip_len <= mtu || 600 (m->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0 || 601 ((ip->ip_off & IP_DF) == 0 && (ifp->if_hwassist & CSUM_FRAGMENT))) { 602 ip->ip_len = htons(ip->ip_len); 603 ip->ip_off = htons(ip->ip_off); 604 ip->ip_sum = 0; 605 if (sw_csum & CSUM_DELAY_IP) 606 ip->ip_sum = in_cksum(m, hlen); 607 608 /* 609 * Record statistics for this interface address. 610 * With CSUM_TSO the byte/packet count will be slightly 611 * incorrect because we count the IP+TCP headers only 612 * once instead of for every generated packet. 613 */ 614 if (!(flags & IP_FORWARDING) && ia) { 615 if (m->m_pkthdr.csum_flags & CSUM_TSO) 616 ia->ia_ifa.if_opackets += 617 m->m_pkthdr.len / m->m_pkthdr.tso_segsz; 618 else 619 ia->ia_ifa.if_opackets++; 620 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 621 } 622#ifdef MBUF_STRESS_TEST 623 if (mbuf_frag_size && m->m_pkthdr.len > mbuf_frag_size) 624 m = m_fragment(m, M_DONTWAIT, mbuf_frag_size); 625#endif 626 /* 627 * Reset layer specific mbuf flags 628 * to avoid confusing lower layers. 629 */ 630 m->m_flags &= ~(M_PROTOFLAGS); 631 error = (*ifp->if_output)(ifp, m, 632 (struct sockaddr *)dst, ro); 633 goto done; 634 } 635 636 /* Balk when DF bit is set or the interface didn't support TSO. */ 637 if ((ip->ip_off & IP_DF) || (m->m_pkthdr.csum_flags & CSUM_TSO)) { 638 error = EMSGSIZE; 639 IPSTAT_INC(ips_cantfrag); 640 goto bad; 641 } 642 643 /* 644 * Too large for interface; fragment if possible. If successful, 645 * on return, m will point to a list of packets to be sent. 646 */ 647 error = ip_fragment(ip, &m, mtu, ifp->if_hwassist, sw_csum); 648 if (error) 649 goto bad; 650 for (; m; m = m0) { 651 m0 = m->m_nextpkt; 652 m->m_nextpkt = 0; 653 if (error == 0) { 654 /* Record statistics for this interface address. */ 655 if (ia != NULL) { 656 ia->ia_ifa.if_opackets++; 657 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 658 } 659 /* 660 * Reset layer specific mbuf flags 661 * to avoid confusing upper layers. 662 */ 663 m->m_flags &= ~(M_PROTOFLAGS); 664 665 error = (*ifp->if_output)(ifp, m, 666 (struct sockaddr *)dst, ro); 667 } else 668 m_freem(m); 669 } 670 671 if (error == 0) 672 IPSTAT_INC(ips_fragmented); 673 674done: 675 if (ro == &iproute) 676 RO_RTFREE(ro); 677 if (ia != NULL) 678 ifa_free(&ia->ia_ifa); 679 return (error); 680bad: 681 m_freem(m); 682 goto done; 683} 684 685/* 686 * Create a chain of fragments which fit the given mtu. m_frag points to the 687 * mbuf to be fragmented; on return it points to the chain with the fragments. 688 * Return 0 if no error. If error, m_frag may contain a partially built 689 * chain of fragments that should be freed by the caller. 690 * 691 * if_hwassist_flags is the hw offload capabilities (see if_data.ifi_hwassist) 692 * sw_csum contains the delayed checksums flags (e.g., CSUM_DELAY_IP). 693 */ 694int 695ip_fragment(struct ip *ip, struct mbuf **m_frag, int mtu, 696 u_long if_hwassist_flags, int sw_csum) 697{ 698 int error = 0; 699 int hlen = ip->ip_hl << 2; 700 int len = (mtu - hlen) & ~7; /* size of payload in each fragment */ 701 int off; 702 struct mbuf *m0 = *m_frag; /* the original packet */ 703 int firstlen; 704 struct mbuf **mnext; 705 int nfrags; 706 707 if (ip->ip_off & IP_DF) { /* Fragmentation not allowed */ 708 IPSTAT_INC(ips_cantfrag); 709 return EMSGSIZE; 710 } 711 712 /* 713 * Must be able to put at least 8 bytes per fragment. 714 */ 715 if (len < 8) 716 return EMSGSIZE; 717 718 /* 719 * If the interface will not calculate checksums on 720 * fragmented packets, then do it here. 721 */ 722 if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 723 in_delayed_cksum(m0); 724 m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 725 } 726#ifdef SCTP 727 if (m0->m_pkthdr.csum_flags & CSUM_SCTP) { 728 sctp_delayed_cksum(m0, hlen); 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 /* find the mbuf in the chain where the checksum starts*/ 871 while ((m != NULL) && (offset >= m->m_len)) { 872 offset -= m->m_len; 873 m = m->m_next; 874 } 875 KASSERT(m != NULL, ("in_delayed_cksum: checksum outside mbuf chain.")); 876 KASSERT(offset + sizeof(u_short) <= m->m_len, ("in_delayed_cksum: checksum split between mbufs.")); 877 *(u_short *)(m->m_data + offset) = csum; 878} 879 880/* 881 * IP socket option processing. 882 */ 883int 884ip_ctloutput(struct socket *so, struct sockopt *sopt) 885{ 886 struct inpcb *inp = sotoinpcb(so); 887 int error, optval; 888 889 error = optval = 0; 890 if (sopt->sopt_level != IPPROTO_IP) { 891 error = EINVAL; 892 893 if (sopt->sopt_level == SOL_SOCKET && 894 sopt->sopt_dir == SOPT_SET) { 895 switch (sopt->sopt_name) { 896 case SO_REUSEADDR: 897 INP_WLOCK(inp); 898 if ((so->so_options & SO_REUSEADDR) != 0) 899 inp->inp_flags2 |= INP_REUSEADDR; 900 else 901 inp->inp_flags2 &= ~INP_REUSEADDR; 902 INP_WUNLOCK(inp); 903 error = 0; 904 break; 905 case SO_REUSEPORT: 906 INP_WLOCK(inp); 907 if ((so->so_options & SO_REUSEPORT) != 0) 908 inp->inp_flags2 |= INP_REUSEPORT; 909 else 910 inp->inp_flags2 &= ~INP_REUSEPORT; 911 INP_WUNLOCK(inp); 912 error = 0; 913 break; 914 case SO_SETFIB: 915 INP_WLOCK(inp); 916 inp->inp_inc.inc_fibnum = so->so_fibnum; 917 INP_WUNLOCK(inp); 918 error = 0; 919 break; 920 default: 921 break; 922 } 923 } 924 return (error); 925 } 926 927 switch (sopt->sopt_dir) { 928 case SOPT_SET: 929 switch (sopt->sopt_name) { 930 case IP_OPTIONS: 931#ifdef notyet 932 case IP_RETOPTS: 933#endif 934 { 935 struct mbuf *m; 936 if (sopt->sopt_valsize > MLEN) { 937 error = EMSGSIZE; 938 break; 939 } 940 MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA); 941 if (m == NULL) { 942 error = ENOBUFS; 943 break; 944 } 945 m->m_len = sopt->sopt_valsize; 946 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 947 m->m_len); 948 if (error) { 949 m_free(m); 950 break; 951 } 952 INP_WLOCK(inp); 953 error = ip_pcbopts(inp, sopt->sopt_name, m); 954 INP_WUNLOCK(inp); 955 return (error); 956 } 957 958 case IP_BINDANY: 959 if (sopt->sopt_td != NULL) { 960 error = priv_check(sopt->sopt_td, 961 PRIV_NETINET_BINDANY); 962 if (error) 963 break; 964 } 965 /* FALLTHROUGH */ 966 case IP_TOS: 967 case IP_TTL: 968 case IP_MINTTL: 969 case IP_RECVOPTS: 970 case IP_RECVRETOPTS: 971 case IP_RECVDSTADDR: 972 case IP_RECVTTL: 973 case IP_RECVIF: 974 case IP_FAITH: 975 case IP_ONESBCAST: 976 case IP_DONTFRAG: 977 case IP_RECVTOS: 978 error = sooptcopyin(sopt, &optval, sizeof optval, 979 sizeof optval); 980 if (error) 981 break; 982 983 switch (sopt->sopt_name) { 984 case IP_TOS: 985 inp->inp_ip_tos = optval; 986 break; 987 988 case IP_TTL: 989 inp->inp_ip_ttl = optval; 990 break; 991 992 case IP_MINTTL: 993 if (optval >= 0 && optval <= MAXTTL) 994 inp->inp_ip_minttl = optval; 995 else 996 error = EINVAL; 997 break; 998 999#define OPTSET(bit) do { \ 1000 INP_WLOCK(inp); \ 1001 if (optval) \ 1002 inp->inp_flags |= bit; \ 1003 else \ 1004 inp->inp_flags &= ~bit; \ 1005 INP_WUNLOCK(inp); \ 1006} while (0) 1007 1008 case IP_RECVOPTS: 1009 OPTSET(INP_RECVOPTS); 1010 break; 1011 1012 case IP_RECVRETOPTS: 1013 OPTSET(INP_RECVRETOPTS); 1014 break; 1015 1016 case IP_RECVDSTADDR: 1017 OPTSET(INP_RECVDSTADDR); 1018 break; 1019 1020 case IP_RECVTTL: 1021 OPTSET(INP_RECVTTL); 1022 break; 1023 1024 case IP_RECVIF: 1025 OPTSET(INP_RECVIF); 1026 break; 1027 1028 case IP_FAITH: 1029 OPTSET(INP_FAITH); 1030 break; 1031 1032 case IP_ONESBCAST: 1033 OPTSET(INP_ONESBCAST); 1034 break; 1035 case IP_DONTFRAG: 1036 OPTSET(INP_DONTFRAG); 1037 break; 1038 case IP_BINDANY: 1039 OPTSET(INP_BINDANY); 1040 break; 1041 case IP_RECVTOS: 1042 OPTSET(INP_RECVTOS); 1043 break; 1044 } 1045 break; 1046#undef OPTSET 1047 1048 /* 1049 * Multicast socket options are processed by the in_mcast 1050 * module. 1051 */ 1052 case IP_MULTICAST_IF: 1053 case IP_MULTICAST_VIF: 1054 case IP_MULTICAST_TTL: 1055 case IP_MULTICAST_LOOP: 1056 case IP_ADD_MEMBERSHIP: 1057 case IP_DROP_MEMBERSHIP: 1058 case IP_ADD_SOURCE_MEMBERSHIP: 1059 case IP_DROP_SOURCE_MEMBERSHIP: 1060 case IP_BLOCK_SOURCE: 1061 case IP_UNBLOCK_SOURCE: 1062 case IP_MSFILTER: 1063 case MCAST_JOIN_GROUP: 1064 case MCAST_LEAVE_GROUP: 1065 case MCAST_JOIN_SOURCE_GROUP: 1066 case MCAST_LEAVE_SOURCE_GROUP: 1067 case MCAST_BLOCK_SOURCE: 1068 case MCAST_UNBLOCK_SOURCE: 1069 error = inp_setmoptions(inp, sopt); 1070 break; 1071 1072 case IP_PORTRANGE: 1073 error = sooptcopyin(sopt, &optval, sizeof optval, 1074 sizeof optval); 1075 if (error) 1076 break; 1077 1078 INP_WLOCK(inp); 1079 switch (optval) { 1080 case IP_PORTRANGE_DEFAULT: 1081 inp->inp_flags &= ~(INP_LOWPORT); 1082 inp->inp_flags &= ~(INP_HIGHPORT); 1083 break; 1084 1085 case IP_PORTRANGE_HIGH: 1086 inp->inp_flags &= ~(INP_LOWPORT); 1087 inp->inp_flags |= INP_HIGHPORT; 1088 break; 1089 1090 case IP_PORTRANGE_LOW: 1091 inp->inp_flags &= ~(INP_HIGHPORT); 1092 inp->inp_flags |= INP_LOWPORT; 1093 break; 1094 1095 default: 1096 error = EINVAL; 1097 break; 1098 } 1099 INP_WUNLOCK(inp); 1100 break; 1101 1102#ifdef IPSEC 1103 case IP_IPSEC_POLICY: 1104 { 1105 caddr_t req; 1106 struct mbuf *m; 1107 1108 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1109 break; 1110 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1111 break; 1112 req = mtod(m, caddr_t); 1113 error = ipsec_set_policy(inp, sopt->sopt_name, req, 1114 m->m_len, (sopt->sopt_td != NULL) ? 1115 sopt->sopt_td->td_ucred : NULL); 1116 m_freem(m); 1117 break; 1118 } 1119#endif /* IPSEC */ 1120 1121 default: 1122 error = ENOPROTOOPT; 1123 break; 1124 } 1125 break; 1126 1127 case SOPT_GET: 1128 switch (sopt->sopt_name) { 1129 case IP_OPTIONS: 1130 case IP_RETOPTS: 1131 if (inp->inp_options) 1132 error = sooptcopyout(sopt, 1133 mtod(inp->inp_options, 1134 char *), 1135 inp->inp_options->m_len); 1136 else 1137 sopt->sopt_valsize = 0; 1138 break; 1139 1140 case IP_TOS: 1141 case IP_TTL: 1142 case IP_MINTTL: 1143 case IP_RECVOPTS: 1144 case IP_RECVRETOPTS: 1145 case IP_RECVDSTADDR: 1146 case IP_RECVTTL: 1147 case IP_RECVIF: 1148 case IP_PORTRANGE: 1149 case IP_FAITH: 1150 case IP_ONESBCAST: 1151 case IP_DONTFRAG: 1152 case IP_BINDANY: 1153 case IP_RECVTOS: 1154 switch (sopt->sopt_name) { 1155 1156 case IP_TOS: 1157 optval = inp->inp_ip_tos; 1158 break; 1159 1160 case IP_TTL: 1161 optval = inp->inp_ip_ttl; 1162 break; 1163 1164 case IP_MINTTL: 1165 optval = inp->inp_ip_minttl; 1166 break; 1167 1168#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1169 1170 case IP_RECVOPTS: 1171 optval = OPTBIT(INP_RECVOPTS); 1172 break; 1173 1174 case IP_RECVRETOPTS: 1175 optval = OPTBIT(INP_RECVRETOPTS); 1176 break; 1177 1178 case IP_RECVDSTADDR: 1179 optval = OPTBIT(INP_RECVDSTADDR); 1180 break; 1181 1182 case IP_RECVTTL: 1183 optval = OPTBIT(INP_RECVTTL); 1184 break; 1185 1186 case IP_RECVIF: 1187 optval = OPTBIT(INP_RECVIF); 1188 break; 1189 1190 case IP_PORTRANGE: 1191 if (inp->inp_flags & INP_HIGHPORT) 1192 optval = IP_PORTRANGE_HIGH; 1193 else if (inp->inp_flags & INP_LOWPORT) 1194 optval = IP_PORTRANGE_LOW; 1195 else 1196 optval = 0; 1197 break; 1198 1199 case IP_FAITH: 1200 optval = OPTBIT(INP_FAITH); 1201 break; 1202 1203 case IP_ONESBCAST: 1204 optval = OPTBIT(INP_ONESBCAST); 1205 break; 1206 case IP_DONTFRAG: 1207 optval = OPTBIT(INP_DONTFRAG); 1208 break; 1209 case IP_BINDANY: 1210 optval = OPTBIT(INP_BINDANY); 1211 break; 1212 case IP_RECVTOS: 1213 optval = OPTBIT(INP_RECVTOS); 1214 break; 1215 } 1216 error = sooptcopyout(sopt, &optval, sizeof optval); 1217 break; 1218 1219 /* 1220 * Multicast socket options are processed by the in_mcast 1221 * module. 1222 */ 1223 case IP_MULTICAST_IF: 1224 case IP_MULTICAST_VIF: 1225 case IP_MULTICAST_TTL: 1226 case IP_MULTICAST_LOOP: 1227 case IP_MSFILTER: 1228 error = inp_getmoptions(inp, sopt); 1229 break; 1230 1231#ifdef IPSEC 1232 case IP_IPSEC_POLICY: 1233 { 1234 struct mbuf *m = NULL; 1235 caddr_t req = NULL; 1236 size_t len = 0; 1237 1238 if (m != 0) { 1239 req = mtod(m, caddr_t); 1240 len = m->m_len; 1241 } 1242 error = ipsec_get_policy(sotoinpcb(so), req, len, &m); 1243 if (error == 0) 1244 error = soopt_mcopyout(sopt, m); /* XXX */ 1245 if (error == 0) 1246 m_freem(m); 1247 break; 1248 } 1249#endif /* IPSEC */ 1250 1251 default: 1252 error = ENOPROTOOPT; 1253 break; 1254 } 1255 break; 1256 } 1257 return (error); 1258} 1259 1260/* 1261 * Routine called from ip_output() to loop back a copy of an IP multicast 1262 * packet to the input queue of a specified interface. Note that this 1263 * calls the output routine of the loopback "driver", but with an interface 1264 * pointer that might NOT be a loopback interface -- evil, but easier than 1265 * replicating that code here. 1266 */ 1267static void 1268ip_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in *dst, 1269 int hlen) 1270{ 1271 register struct ip *ip; 1272 struct mbuf *copym; 1273 1274 /* 1275 * Make a deep copy of the packet because we're going to 1276 * modify the pack in order to generate checksums. 1277 */ 1278 copym = m_dup(m, M_DONTWAIT); 1279 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 1280 copym = m_pullup(copym, hlen); 1281 if (copym != NULL) { 1282 /* If needed, compute the checksum and mark it as valid. */ 1283 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1284 in_delayed_cksum(copym); 1285 copym->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1286 copym->m_pkthdr.csum_flags |= 1287 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 1288 copym->m_pkthdr.csum_data = 0xffff; 1289 } 1290 /* 1291 * We don't bother to fragment if the IP length is greater 1292 * than the interface's MTU. Can this possibly matter? 1293 */ 1294 ip = mtod(copym, struct ip *); 1295 ip->ip_len = htons(ip->ip_len); 1296 ip->ip_off = htons(ip->ip_off); 1297 ip->ip_sum = 0; 1298 ip->ip_sum = in_cksum(copym, hlen); 1299#if 1 /* XXX */ 1300 if (dst->sin_family != AF_INET) { 1301 printf("ip_mloopback: bad address family %d\n", 1302 dst->sin_family); 1303 dst->sin_family = AF_INET; 1304 } 1305#endif 1306 if_simloop(ifp, copym, dst->sin_family, 0); 1307 } 1308} 1309