ip_output.c revision 112591
1296077Sadrian/* 2296077Sadrian * Copyright (c) 1982, 1986, 1988, 1990, 1993 3296077Sadrian * The Regents of the University of California. All rights reserved. 4296077Sadrian * 5296077Sadrian * Redistribution and use in source and binary forms, with or without 6296077Sadrian * modification, are permitted provided that the following conditions 7296077Sadrian * are met: 8296077Sadrian * 1. Redistributions of source code must retain the above copyright 9296077Sadrian * notice, this list of conditions and the following disclaimer. 10296077Sadrian * 2. Redistributions in binary form must reproduce the above copyright 11296077Sadrian * notice, this list of conditions and the following disclaimer in the 12296077Sadrian * documentation and/or other materials provided with the distribution. 13296077Sadrian * 3. All advertising materials mentioning features or use of this software 14296077Sadrian * must display the following acknowledgement: 15296077Sadrian * This product includes software developed by the University of 16296077Sadrian * California, Berkeley and its contributors. 17296077Sadrian * 4. Neither the name of the University nor the names of its contributors 18296077Sadrian * may be used to endorse or promote products derived from this software 19296077Sadrian * without specific prior written permission. 20296077Sadrian * 21296077Sadrian * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22296077Sadrian * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23296077Sadrian * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24296077Sadrian * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25296077Sadrian * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26296077Sadrian * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27296077Sadrian * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28296077Sadrian * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29296077Sadrian * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30296077Sadrian * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31296077Sadrian * SUCH DAMAGE. 32296077Sadrian * 33296077Sadrian * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 34296077Sadrian * $FreeBSD: head/sys/netinet/ip_output.c 112591 2003-03-25 05:45:05Z silby $ 35296077Sadrian */ 36296077Sadrian 37296077Sadrian#include "opt_ipfw.h" 38296077Sadrian#include "opt_ipdn.h" 39296077Sadrian#include "opt_ipdivert.h" 40296077Sadrian#include "opt_ipfilter.h" 41296077Sadrian#include "opt_ipsec.h" 42296077Sadrian#include "opt_mac.h" 43296077Sadrian#include "opt_pfil_hooks.h" 44296077Sadrian#include "opt_random_ip_id.h" 45296077Sadrian#include "opt_mbuf_frag_test.h" 46296077Sadrian 47296077Sadrian#include <sys/param.h> 48296077Sadrian#include <sys/systm.h> 49296077Sadrian#include <sys/kernel.h> 50296077Sadrian#include <sys/mac.h> 51296077Sadrian#include <sys/malloc.h> 52296077Sadrian#include <sys/mbuf.h> 53296077Sadrian#include <sys/protosw.h> 54296077Sadrian#include <sys/socket.h> 55296077Sadrian#include <sys/socketvar.h> 56296077Sadrian#include <sys/sysctl.h> 57296077Sadrian 58296077Sadrian#include <net/if.h> 59296077Sadrian#include <net/route.h> 60296077Sadrian 61296077Sadrian#include <netinet/in.h> 62296077Sadrian#include <netinet/in_systm.h> 63296077Sadrian#include <netinet/ip.h> 64296077Sadrian#include <netinet/in_pcb.h> 65296077Sadrian#include <netinet/in_var.h> 66296077Sadrian#include <netinet/ip_var.h> 67296077Sadrian 68296077Sadrian#include <machine/in_cksum.h> 69296077Sadrian 70296077Sadrianstatic MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 71296077Sadrian 72296077Sadrian#ifdef IPSEC 73296077Sadrian#include <netinet6/ipsec.h> 74296077Sadrian#include <netkey/key.h> 75296077Sadrian#ifdef IPSEC_DEBUG 76296077Sadrian#include <netkey/key_debug.h> 77296077Sadrian#else 78296077Sadrian#define KEYDEBUG(lev,arg) 79296077Sadrian#endif 80296077Sadrian#endif /*IPSEC*/ 81296077Sadrian 82296077Sadrian#ifdef FAST_IPSEC 83296077Sadrian#include <netipsec/ipsec.h> 84296077Sadrian#include <netipsec/xform.h> 85296077Sadrian#include <netipsec/key.h> 86296077Sadrian#endif /*FAST_IPSEC*/ 87296077Sadrian 88296077Sadrian#include <netinet/ip_fw.h> 89296077Sadrian#include <netinet/ip_dummynet.h> 90296077Sadrian 91296077Sadrian#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 92296077Sadrian x, (ntohl(a.s_addr)>>24)&0xFF,\ 93296077Sadrian (ntohl(a.s_addr)>>16)&0xFF,\ 94296077Sadrian (ntohl(a.s_addr)>>8)&0xFF,\ 95296077Sadrian (ntohl(a.s_addr))&0xFF, y); 96296077Sadrian 97296077Sadrianu_short ip_id; 98296077Sadrian 99296077Sadrian#ifdef MBUF_FRAG_TEST 100296077Sadrianint mbuf_frag_size = 0; 101296077SadrianSYSCTL_INT(_net_inet_ip, OID_AUTO, mbuf_frag_size, CTLFLAG_RW, 102296077Sadrian &mbuf_frag_size, 0, "Fragment outgoing mbufs to this size"); 103296077Sadrian#endif 104296077Sadrian 105296077Sadrianstatic struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 106296077Sadrianstatic struct ifnet *ip_multicast_if(struct in_addr *, int *); 107296077Sadrianstatic void ip_mloopback 108296077Sadrian (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 109296077Sadrianstatic int ip_getmoptions 110296077Sadrian (struct sockopt *, struct ip_moptions *); 111296077Sadrianstatic int ip_pcbopts(int, struct mbuf **, struct mbuf *); 112296077Sadrianstatic int ip_setmoptions 113296077Sadrian (struct sockopt *, struct ip_moptions **); 114296077Sadrian 115296077Sadrianint ip_optcopy(struct ip *, struct ip *); 116296077Sadrian 117296077Sadrian 118296077Sadrianextern struct protosw inetsw[]; 119296077Sadrian 120296077Sadrian/* 121296077Sadrian * IP output. The packet in mbuf chain m contains a skeletal IP 122296077Sadrian * header (with len, off, ttl, proto, tos, src, dst). 123296077Sadrian * The mbuf chain containing the packet will be freed. 124296077Sadrian * The mbuf opt, if present, will not be freed. 125296077Sadrian */ 126296077Sadrianint 127296077Sadrianip_output(m0, opt, ro, flags, imo, inp) 128296077Sadrian struct mbuf *m0; 129296077Sadrian struct mbuf *opt; 130296077Sadrian struct route *ro; 131296077Sadrian int flags; 132296077Sadrian struct ip_moptions *imo; 133296077Sadrian struct inpcb *inp; 134296077Sadrian{ 135296077Sadrian struct ip *ip, *mhip; 136296077Sadrian struct ifnet *ifp = NULL; /* keep compiler happy */ 137296077Sadrian struct mbuf *m; 138296077Sadrian int hlen = sizeof (struct ip); 139296077Sadrian int len, off, error = 0; 140296077Sadrian struct sockaddr_in *dst = NULL; /* keep compiler happy */ 141296077Sadrian struct in_ifaddr *ia = NULL; 142296077Sadrian int isbroadcast, sw_csum; 143296077Sadrian struct in_addr pkt_dst; 144296077Sadrian#ifdef IPSEC 145296077Sadrian struct route iproute; 146296077Sadrian struct secpolicy *sp = NULL; 147296077Sadrian#endif 148296077Sadrian#ifdef FAST_IPSEC 149296077Sadrian struct route iproute; 150296077Sadrian struct m_tag *mtag; 151296077Sadrian struct secpolicy *sp = NULL; 152296077Sadrian struct tdb_ident *tdbi; 153296077Sadrian int s; 154296077Sadrian#endif /* FAST_IPSEC */ 155296077Sadrian struct ip_fw_args args; 156296077Sadrian int src_was_INADDR_ANY = 0; /* as the name says... */ 157296077Sadrian#ifdef PFIL_HOOKS 158296077Sadrian struct packet_filter_hook *pfh; 159296077Sadrian struct mbuf *m1; 160296077Sadrian int rv; 161296077Sadrian#endif /* PFIL_HOOKS */ 162296077Sadrian 163296077Sadrian args.eh = NULL; 164296077Sadrian args.rule = NULL; 165296077Sadrian args.next_hop = NULL; 166296077Sadrian args.divert_rule = 0; /* divert cookie */ 167296077Sadrian 168296077Sadrian /* Grab info from MT_TAG mbufs prepended to the chain. */ 169296077Sadrian for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) { 170296077Sadrian switch(m0->_m_tag_id) { 171296077Sadrian default: 172296077Sadrian printf("ip_output: unrecognised MT_TAG tag %d\n", 173296077Sadrian m0->_m_tag_id); 174296077Sadrian break; 175296077Sadrian 176296077Sadrian case PACKET_TAG_DUMMYNET: 177296077Sadrian /* 178296077Sadrian * the packet was already tagged, so part of the 179296077Sadrian * processing was already done, and we need to go down. 180296077Sadrian * Get parameters from the header. 181296077Sadrian */ 182296077Sadrian args.rule = ((struct dn_pkt *)m0)->rule; 183296077Sadrian opt = NULL ; 184296077Sadrian ro = & ( ((struct dn_pkt *)m0)->ro ) ; 185296077Sadrian imo = NULL ; 186296077Sadrian dst = ((struct dn_pkt *)m0)->dn_dst ; 187296077Sadrian ifp = ((struct dn_pkt *)m0)->ifp ; 188296077Sadrian flags = ((struct dn_pkt *)m0)->flags ; 189296077Sadrian break; 190296077Sadrian 191296077Sadrian case PACKET_TAG_DIVERT: 192296077Sadrian args.divert_rule = (intptr_t)m0->m_data & 0xffff; 193296077Sadrian break; 194296077Sadrian 195296077Sadrian case PACKET_TAG_IPFORWARD: 196296077Sadrian args.next_hop = (struct sockaddr_in *)m0->m_data; 197296077Sadrian break; 198296077Sadrian } 199296077Sadrian } 200296077Sadrian m = m0; 201296077Sadrian 202296077Sadrian KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR")); 203296077Sadrian#ifndef FAST_IPSEC 204296077Sadrian KASSERT(ro != NULL, ("ip_output: no route, proto %d", 205296077Sadrian mtod(m, struct ip *)->ip_p)); 206296077Sadrian#endif 207296077Sadrian 208296077Sadrian if (args.rule != NULL) { /* dummynet already saw us */ 209296077Sadrian ip = mtod(m, struct ip *); 210296077Sadrian hlen = ip->ip_hl << 2 ; 211296077Sadrian if (ro->ro_rt) 212296077Sadrian ia = ifatoia(ro->ro_rt->rt_ifa); 213296077Sadrian goto sendit; 214296077Sadrian } 215296077Sadrian 216296077Sadrian if (opt) { 217296077Sadrian len = 0; 218296077Sadrian m = ip_insertoptions(m, opt, &len); 219296077Sadrian if (len != 0) 220296077Sadrian hlen = len; 221296077Sadrian } 222296077Sadrian ip = mtod(m, struct ip *); 223296077Sadrian pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 224296077Sadrian 225296077Sadrian /* 226296077Sadrian * Fill in IP header. 227296077Sadrian */ 228296077Sadrian if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 229296077Sadrian ip->ip_v = IPVERSION; 230296077Sadrian ip->ip_hl = hlen >> 2; 231296077Sadrian ip->ip_off &= IP_DF; 232296077Sadrian#ifdef RANDOM_IP_ID 233296077Sadrian ip->ip_id = ip_randomid(); 234296077Sadrian#else 235296077Sadrian ip->ip_id = htons(ip_id++); 236296077Sadrian#endif 237296077Sadrian ipstat.ips_localout++; 238296077Sadrian } else { 239296077Sadrian hlen = ip->ip_hl << 2; 240296077Sadrian } 241296077Sadrian 242296077Sadrian#ifdef FAST_IPSEC 243296077Sadrian if (ro == NULL) { 244296077Sadrian ro = &iproute; 245296077Sadrian bzero(ro, sizeof (*ro)); 246296077Sadrian } 247296077Sadrian#endif /* FAST_IPSEC */ 248296077Sadrian dst = (struct sockaddr_in *)&ro->ro_dst; 249296077Sadrian /* 250296077Sadrian * If there is a cached route, 251296077Sadrian * check that it is to the same destination 252296077Sadrian * and is still up. If not, free it and try again. 253296077Sadrian * The address family should also be checked in case of sharing the 254296077Sadrian * cache with IPv6. 255296077Sadrian */ 256296077Sadrian if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 257296077Sadrian dst->sin_family != AF_INET || 258296077Sadrian dst->sin_addr.s_addr != pkt_dst.s_addr)) { 259296077Sadrian RTFREE(ro->ro_rt); 260296077Sadrian ro->ro_rt = (struct rtentry *)0; 261296077Sadrian } 262296077Sadrian if (ro->ro_rt == 0) { 263296077Sadrian bzero(dst, sizeof(*dst)); 264296077Sadrian dst->sin_family = AF_INET; 265296077Sadrian dst->sin_len = sizeof(*dst); 266296077Sadrian dst->sin_addr = pkt_dst; 267296077Sadrian } 268296077Sadrian /* 269296077Sadrian * If routing to interface only, 270296077Sadrian * short circuit routing lookup. 271296077Sadrian */ 272296077Sadrian if (flags & IP_ROUTETOIF) { 273296077Sadrian if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 274296077Sadrian (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 275296077Sadrian ipstat.ips_noroute++; 276296077Sadrian error = ENETUNREACH; 277296077Sadrian goto bad; 278296077Sadrian } 279296077Sadrian ifp = ia->ia_ifp; 280296077Sadrian ip->ip_ttl = 1; 281296077Sadrian isbroadcast = in_broadcast(dst->sin_addr, ifp); 282296077Sadrian } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 283296077Sadrian imo != NULL && imo->imo_multicast_ifp != NULL) { 284296077Sadrian /* 285296077Sadrian * Bypass the normal routing lookup for multicast 286296077Sadrian * packets if the interface is specified. 287296077Sadrian */ 288296077Sadrian ifp = imo->imo_multicast_ifp; 289296077Sadrian IFP_TO_IA(ifp, ia); 290296077Sadrian isbroadcast = 0; /* fool gcc */ 291296077Sadrian } else { 292296077Sadrian /* 293296077Sadrian * If this is the case, we probably don't want to allocate 294296077Sadrian * a protocol-cloned route since we didn't get one from the 295296077Sadrian * ULP. This lets TCP do its thing, while not burdening 296296077Sadrian * forwarding or ICMP with the overhead of cloning a route. 297296077Sadrian * Of course, we still want to do any cloning requested by 298296077Sadrian * the link layer, as this is probably required in all cases 299296077Sadrian * for correct operation (as it is for ARP). 300296077Sadrian */ 301296077Sadrian if (ro->ro_rt == 0) 302296077Sadrian rtalloc_ign(ro, RTF_PRCLONING); 303296077Sadrian if (ro->ro_rt == 0) { 304296077Sadrian ipstat.ips_noroute++; 305296077Sadrian error = EHOSTUNREACH; 306296077Sadrian goto bad; 307296077Sadrian } 308296077Sadrian ia = ifatoia(ro->ro_rt->rt_ifa); 309296077Sadrian ifp = ro->ro_rt->rt_ifp; 310296077Sadrian ro->ro_rt->rt_use++; 311296077Sadrian if (ro->ro_rt->rt_flags & RTF_GATEWAY) 312296077Sadrian dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 313296077Sadrian if (ro->ro_rt->rt_flags & RTF_HOST) 314296077Sadrian isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 315296077Sadrian else 316296077Sadrian isbroadcast = in_broadcast(dst->sin_addr, ifp); 317296077Sadrian } 318296077Sadrian if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 319296077Sadrian struct in_multi *inm; 320296077Sadrian 321296077Sadrian m->m_flags |= M_MCAST; 322296077Sadrian /* 323296077Sadrian * IP destination address is multicast. Make sure "dst" 324296077Sadrian * still points to the address in "ro". (It may have been 325296077Sadrian * changed to point to a gateway address, above.) 326296077Sadrian */ 327296077Sadrian dst = (struct sockaddr_in *)&ro->ro_dst; 328296077Sadrian /* 329296077Sadrian * See if the caller provided any multicast options 330296077Sadrian */ 331296077Sadrian if (imo != NULL) { 332296077Sadrian ip->ip_ttl = imo->imo_multicast_ttl; 333296077Sadrian if (imo->imo_multicast_vif != -1) 334296077Sadrian ip->ip_src.s_addr = 335296077Sadrian ip_mcast_src ? 336296077Sadrian ip_mcast_src(imo->imo_multicast_vif) : 337296077Sadrian INADDR_ANY; 338296077Sadrian } else 339296077Sadrian ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 340296077Sadrian /* 341296077Sadrian * Confirm that the outgoing interface supports multicast. 342296077Sadrian */ 343296077Sadrian if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 344296077Sadrian if ((ifp->if_flags & IFF_MULTICAST) == 0) { 345296077Sadrian ipstat.ips_noroute++; 346296077Sadrian error = ENETUNREACH; 347296077Sadrian goto bad; 348296077Sadrian } 349296077Sadrian } 350296077Sadrian /* 351296077Sadrian * If source address not specified yet, use address 352296077Sadrian * of outgoing interface. 353296077Sadrian */ 354296077Sadrian if (ip->ip_src.s_addr == INADDR_ANY) { 355296077Sadrian /* Interface may have no addresses. */ 356296077Sadrian if (ia != NULL) 357296077Sadrian ip->ip_src = IA_SIN(ia)->sin_addr; 358296077Sadrian } 359296077Sadrian 360296077Sadrian if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 361296077Sadrian /* 362296077Sadrian * XXX 363296077Sadrian * delayed checksums are not currently 364296077Sadrian * compatible with IP multicast routing 365296077Sadrian */ 366296077Sadrian if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 367296077Sadrian in_delayed_cksum(m); 368296077Sadrian m->m_pkthdr.csum_flags &= 369296077Sadrian ~CSUM_DELAY_DATA; 370296077Sadrian } 371296077Sadrian } 372296077Sadrian IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 373296077Sadrian if (inm != NULL && 374296077Sadrian (imo == NULL || imo->imo_multicast_loop)) { 375296077Sadrian /* 376296077Sadrian * If we belong to the destination multicast group 377296077Sadrian * on the outgoing interface, and the caller did not 378296077Sadrian * forbid loopback, loop back a copy. 379296077Sadrian */ 380296077Sadrian ip_mloopback(ifp, m, dst, hlen); 381296077Sadrian } 382296077Sadrian else { 383296077Sadrian /* 384296077Sadrian * If we are acting as a multicast router, perform 385296077Sadrian * multicast forwarding as if the packet had just 386296077Sadrian * arrived on the interface to which we are about 387296077Sadrian * to send. The multicast forwarding function 388296077Sadrian * recursively calls this function, using the 389296077Sadrian * IP_FORWARDING flag to prevent infinite recursion. 390296077Sadrian * 391296077Sadrian * Multicasts that are looped back by ip_mloopback(), 392296077Sadrian * above, will be forwarded by the ip_input() routine, 393296077Sadrian * if necessary. 394296077Sadrian */ 395296077Sadrian if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 396296077Sadrian /* 397296077Sadrian * If rsvp daemon is not running, do not 398296077Sadrian * set ip_moptions. This ensures that the packet 399296077Sadrian * is multicast and not just sent down one link 400296077Sadrian * as prescribed by rsvpd. 401296077Sadrian */ 402296077Sadrian if (!rsvp_on) 403296077Sadrian imo = NULL; 404296077Sadrian if (ip_mforward && 405296077Sadrian ip_mforward(ip, ifp, m, imo) != 0) { 406296077Sadrian m_freem(m); 407296077Sadrian goto done; 408296077Sadrian } 409296077Sadrian } 410296077Sadrian } 411296077Sadrian 412296077Sadrian /* 413296077Sadrian * Multicasts with a time-to-live of zero may be looped- 414296077Sadrian * back, above, but must not be transmitted on a network. 415296077Sadrian * Also, multicasts addressed to the loopback interface 416296077Sadrian * are not sent -- the above call to ip_mloopback() will 417296077Sadrian * loop back a copy if this host actually belongs to the 418296077Sadrian * destination group on the loopback interface. 419296077Sadrian */ 420296077Sadrian if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 421296077Sadrian m_freem(m); 422296077Sadrian goto done; 423296077Sadrian } 424296077Sadrian 425296077Sadrian goto sendit; 426296077Sadrian } 427296077Sadrian#ifndef notdef 428296077Sadrian /* 429296077Sadrian * If the source address is not specified yet, use the address 430296077Sadrian * of the outoing interface. In case, keep note we did that, so 431296077Sadrian * if the the firewall changes the next-hop causing the output 432296077Sadrian * interface to change, we can fix that. 433296077Sadrian */ 434296077Sadrian if (ip->ip_src.s_addr == INADDR_ANY) { 435296077Sadrian /* Interface may have no addresses. */ 436296077Sadrian if (ia != NULL) { 437296077Sadrian ip->ip_src = IA_SIN(ia)->sin_addr; 438296077Sadrian src_was_INADDR_ANY = 1; 439296077Sadrian } 440296077Sadrian } 441296077Sadrian#endif /* notdef */ 442296077Sadrian /* 443296077Sadrian * Verify that we have any chance at all of being able to queue 444296077Sadrian * the packet or packet fragments 445296077Sadrian */ 446296077Sadrian if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 447296077Sadrian ifp->if_snd.ifq_maxlen) { 448296077Sadrian error = ENOBUFS; 449296077Sadrian ipstat.ips_odropped++; 450296077Sadrian goto bad; 451296077Sadrian } 452296077Sadrian 453296077Sadrian /* 454296077Sadrian * Look for broadcast address and 455296077Sadrian * verify user is allowed to send 456296077Sadrian * such a packet. 457296077Sadrian */ 458296077Sadrian if (isbroadcast) { 459296077Sadrian if ((ifp->if_flags & IFF_BROADCAST) == 0) { 460296077Sadrian error = EADDRNOTAVAIL; 461296077Sadrian goto bad; 462296077Sadrian } 463296077Sadrian if ((flags & IP_ALLOWBROADCAST) == 0) { 464296077Sadrian error = EACCES; 465296077Sadrian goto bad; 466296077Sadrian } 467296077Sadrian /* don't allow broadcast messages to be fragmented */ 468296077Sadrian if ((u_short)ip->ip_len > ifp->if_mtu) { 469296077Sadrian error = EMSGSIZE; 470296077Sadrian goto bad; 471296077Sadrian } 472296077Sadrian m->m_flags |= M_BCAST; 473296077Sadrian } else { 474296077Sadrian m->m_flags &= ~M_BCAST; 475296077Sadrian } 476296077Sadrian 477296077Sadriansendit: 478296077Sadrian#ifdef IPSEC 479296077Sadrian /* get SP for this packet */ 480296077Sadrian if (inp == NULL) 481296077Sadrian sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 482296077Sadrian else 483296077Sadrian sp = ipsec4_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error); 484296077Sadrian 485296077Sadrian if (sp == NULL) { 486296077Sadrian ipsecstat.out_inval++; 487296077Sadrian goto bad; 488296077Sadrian } 489296077Sadrian 490296077Sadrian error = 0; 491296077Sadrian 492296077Sadrian /* check policy */ 493296077Sadrian switch (sp->policy) { 494296077Sadrian case IPSEC_POLICY_DISCARD: 495296077Sadrian /* 496296077Sadrian * This packet is just discarded. 497296077Sadrian */ 498296077Sadrian ipsecstat.out_polvio++; 499296077Sadrian goto bad; 500296077Sadrian 501296077Sadrian case IPSEC_POLICY_BYPASS: 502296077Sadrian case IPSEC_POLICY_NONE: 503296077Sadrian /* no need to do IPsec. */ 504296077Sadrian goto skip_ipsec; 505296077Sadrian 506296077Sadrian case IPSEC_POLICY_IPSEC: 507296077Sadrian if (sp->req == NULL) { 508296077Sadrian /* acquire a policy */ 509296077Sadrian error = key_spdacquire(sp); 510296077Sadrian goto bad; 511296077Sadrian } 512296077Sadrian break; 513296077Sadrian 514296077Sadrian case IPSEC_POLICY_ENTRUST: 515296077Sadrian default: 516296077Sadrian printf("ip_output: Invalid policy found. %d\n", sp->policy); 517296077Sadrian } 518296077Sadrian { 519296077Sadrian struct ipsec_output_state state; 520296077Sadrian bzero(&state, sizeof(state)); 521296077Sadrian state.m = m; 522296077Sadrian if (flags & IP_ROUTETOIF) { 523296077Sadrian state.ro = &iproute; 524296077Sadrian bzero(&iproute, sizeof(iproute)); 525296077Sadrian } else 526296077Sadrian state.ro = ro; 527296077Sadrian state.dst = (struct sockaddr *)dst; 528296077Sadrian 529296077Sadrian ip->ip_sum = 0; 530296077Sadrian 531296077Sadrian /* 532296077Sadrian * XXX 533296077Sadrian * delayed checksums are not currently compatible with IPsec 534296077Sadrian */ 535296077Sadrian if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 536296077Sadrian in_delayed_cksum(m); 537296077Sadrian m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 538296077Sadrian } 539296077Sadrian 540296077Sadrian ip->ip_len = htons(ip->ip_len); 541296077Sadrian ip->ip_off = htons(ip->ip_off); 542296077Sadrian 543296077Sadrian error = ipsec4_output(&state, sp, flags); 544296077Sadrian 545296077Sadrian m = state.m; 546296077Sadrian if (flags & IP_ROUTETOIF) { 547296077Sadrian /* 548296077Sadrian * if we have tunnel mode SA, we may need to ignore 549296077Sadrian * IP_ROUTETOIF. 550296077Sadrian */ 551296077Sadrian if (state.ro != &iproute || state.ro->ro_rt != NULL) { 552296077Sadrian flags &= ~IP_ROUTETOIF; 553296077Sadrian ro = state.ro; 554296077Sadrian } 555296077Sadrian } else 556296077Sadrian ro = state.ro; 557296077Sadrian dst = (struct sockaddr_in *)state.dst; 558296077Sadrian if (error) { 559296077Sadrian /* mbuf is already reclaimed in ipsec4_output. */ 560296077Sadrian m0 = NULL; 561296077Sadrian switch (error) { 562296077Sadrian case EHOSTUNREACH: 563296077Sadrian case ENETUNREACH: 564296077Sadrian case EMSGSIZE: 565296077Sadrian case ENOBUFS: 566296077Sadrian case ENOMEM: 567296077Sadrian break; 568296077Sadrian default: 569296077Sadrian printf("ip4_output (ipsec): error code %d\n", error); 570297793Spfg /*fall through*/ 571296077Sadrian case ENOENT: 572296077Sadrian /* don't show these error codes to the user */ 573296077Sadrian error = 0; 574296077Sadrian break; 575296077Sadrian } 576296077Sadrian goto bad; 577296077Sadrian } 578296077Sadrian } 579296077Sadrian 580296077Sadrian /* be sure to update variables that are affected by ipsec4_output() */ 581296077Sadrian ip = mtod(m, struct ip *); 582296077Sadrian hlen = ip->ip_hl << 2; 583296077Sadrian if (ro->ro_rt == NULL) { 584296077Sadrian if ((flags & IP_ROUTETOIF) == 0) { 585296077Sadrian printf("ip_output: " 586296077Sadrian "can't update route after IPsec processing\n"); 587296077Sadrian error = EHOSTUNREACH; /*XXX*/ 588296077Sadrian goto bad; 589296077Sadrian } 590296077Sadrian } else { 591296077Sadrian ia = ifatoia(ro->ro_rt->rt_ifa); 592296077Sadrian ifp = ro->ro_rt->rt_ifp; 593296077Sadrian } 594296077Sadrian 595296077Sadrian /* make it flipped, again. */ 596296077Sadrian ip->ip_len = ntohs(ip->ip_len); 597296077Sadrian ip->ip_off = ntohs(ip->ip_off); 598296077Sadrianskip_ipsec: 599296077Sadrian#endif /*IPSEC*/ 600296077Sadrian#ifdef FAST_IPSEC 601296077Sadrian /* 602296077Sadrian * Check the security policy (SP) for the packet and, if 603296077Sadrian * required, do IPsec-related processing. There are two 604296077Sadrian * cases here; the first time a packet is sent through 605296077Sadrian * it will be untagged and handled by ipsec4_checkpolicy. 606296077Sadrian * If the packet is resubmitted to ip_output (e.g. after 607296077Sadrian * AH, ESP, etc. processing), there will be a tag to bypass 608296077Sadrian * the lookup and related policy checking. 609296077Sadrian */ 610296077Sadrian mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL); 611296077Sadrian s = splnet(); 612296077Sadrian if (mtag != NULL) { 613296077Sadrian tdbi = (struct tdb_ident *)(mtag + 1); 614296077Sadrian sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND); 615296077Sadrian if (sp == NULL) 616296077Sadrian error = -EINVAL; /* force silent drop */ 617296077Sadrian m_tag_delete(m, mtag); 618296077Sadrian } else { 619296077Sadrian sp = ipsec4_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, 620296077Sadrian &error, inp); 621296077Sadrian } 622296077Sadrian /* 623296077Sadrian * There are four return cases: 624296077Sadrian * sp != NULL apply IPsec policy 625296077Sadrian * sp == NULL, error == 0 no IPsec handling needed 626296077Sadrian * sp == NULL, error == -EINVAL discard packet w/o error 627296077Sadrian * sp == NULL, error != 0 discard packet, report error 628297793Spfg */ 629296077Sadrian if (sp != NULL) { 630296077Sadrian /* Loop detection, check if ipsec processing already done */ 631296077Sadrian KASSERT(sp->req != NULL, ("ip_output: no ipsec request")); 632296077Sadrian for (mtag = m_tag_first(m); mtag != NULL; 633296077Sadrian mtag = m_tag_next(m, mtag)) { 634296077Sadrian if (mtag->m_tag_cookie != MTAG_ABI_COMPAT) 635296077Sadrian continue; 636296077Sadrian if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE && 637296077Sadrian mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED) 638296077Sadrian continue; 639296077Sadrian /* 640296077Sadrian * Check if policy has an SA associated with it. 641296077Sadrian * This can happen when an SP has yet to acquire 642296077Sadrian * an SA; e.g. on first reference. If it occurs, 643296077Sadrian * then we let ipsec4_process_packet do its thing. 644296077Sadrian */ 645296077Sadrian if (sp->req->sav == NULL) 646296077Sadrian break; 647296077Sadrian tdbi = (struct tdb_ident *)(mtag + 1); 648296077Sadrian if (tdbi->spi == sp->req->sav->spi && 649296077Sadrian tdbi->proto == sp->req->sav->sah->saidx.proto && 650296077Sadrian bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst, 651296077Sadrian sizeof (union sockaddr_union)) == 0) { 652296077Sadrian /* 653296077Sadrian * No IPsec processing is needed, free 654296077Sadrian * reference to SP. 655296077Sadrian * 656296077Sadrian * NB: null pointer to avoid free at 657296077Sadrian * done: below. 658296077Sadrian */ 659296077Sadrian KEY_FREESP(&sp), sp = NULL; 660296077Sadrian splx(s); 661296077Sadrian goto spd_done; 662296077Sadrian } 663296077Sadrian } 664296077Sadrian 665296077Sadrian /* 666296077Sadrian * Do delayed checksums now because we send before 667296077Sadrian * this is done in the normal processing path. 668296077Sadrian */ 669296077Sadrian if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 670296077Sadrian in_delayed_cksum(m); 671296077Sadrian m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 672296077Sadrian } 673296077Sadrian 674296077Sadrian ip->ip_len = htons(ip->ip_len); 675296077Sadrian ip->ip_off = htons(ip->ip_off); 676296077Sadrian 677296077Sadrian /* NB: callee frees mbuf */ 678296077Sadrian error = ipsec4_process_packet(m, sp->req, flags, 0); 679296077Sadrian /* 680296077Sadrian * Preserve KAME behaviour: ENOENT can be returned 681296077Sadrian * when an SA acquire is in progress. Don't propagate 682296077Sadrian * this to user-level; it confuses applications. 683296077Sadrian * 684296077Sadrian * XXX this will go away when the SADB is redone. 685296077Sadrian */ 686296077Sadrian if (error == ENOENT) 687296077Sadrian error = 0; 688296077Sadrian splx(s); 689296077Sadrian goto done; 690296077Sadrian } else { 691296077Sadrian splx(s); 692296077Sadrian 693296077Sadrian if (error != 0) { 694296077Sadrian /* 695296077Sadrian * Hack: -EINVAL is used to signal that a packet 696296077Sadrian * should be silently discarded. This is typically 697296077Sadrian * because we asked key management for an SA and 698296077Sadrian * it was delayed (e.g. kicked up to IKE). 699296077Sadrian */ 700296077Sadrian if (error == -EINVAL) 701296077Sadrian error = 0; 702296077Sadrian goto bad; 703296077Sadrian } else { 704296077Sadrian /* No IPsec processing for this packet. */ 705296077Sadrian } 706296077Sadrian#ifdef notyet 707296077Sadrian /* 708296077Sadrian * If deferred crypto processing is needed, check that 709296077Sadrian * the interface supports it. 710296077Sadrian */ 711296077Sadrian mtag = m_tag_find(m, PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED, NULL); 712296077Sadrian if (mtag != NULL && (ifp->if_capenable & IFCAP_IPSEC) == 0) { 713296077Sadrian /* notify IPsec to do its own crypto */ 714296077Sadrian ipsp_skipcrypto_unmark((struct tdb_ident *)(mtag + 1)); 715296077Sadrian error = EHOSTUNREACH; 716296077Sadrian goto bad; 717296077Sadrian } 718296077Sadrian#endif 719296077Sadrian } 720296077Sadrianspd_done: 721296077Sadrian#endif /* FAST_IPSEC */ 722296077Sadrian 723296077Sadrian /* 724296077Sadrian * IpHack's section. 725296077Sadrian * - Xlate: translate packet's addr/port (NAT). 726296077Sadrian * - Firewall: deny/allow/etc. 727296077Sadrian * - Wrap: fake packet's addr/port <unimpl.> 728296077Sadrian * - Encapsulate: put it in another IP and send out. <unimp.> 729296077Sadrian */ 730296077Sadrian#ifdef PFIL_HOOKS 731296077Sadrian /* 732296077Sadrian * Run through list of hooks for output packets. 733296077Sadrian */ 734296077Sadrian m1 = m; 735296077Sadrian pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 736296077Sadrian for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 737296077Sadrian if (pfh->pfil_func) { 738296077Sadrian rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1); 739296077Sadrian if (rv) { 740296077Sadrian error = EHOSTUNREACH; 741296077Sadrian goto done; 742296077Sadrian } 743296077Sadrian m = m1; 744296077Sadrian if (m == NULL) 745296077Sadrian goto done; 746296077Sadrian ip = mtod(m, struct ip *); 747296077Sadrian } 748296077Sadrian#endif /* PFIL_HOOKS */ 749296077Sadrian 750296077Sadrian /* 751296077Sadrian * Check with the firewall... 752296077Sadrian * but not if we are already being fwd'd from a firewall. 753296077Sadrian */ 754296077Sadrian if (fw_enable && IPFW_LOADED && !args.next_hop) { 755296077Sadrian struct sockaddr_in *old = dst; 756296077Sadrian 757296077Sadrian args.m = m; 758296077Sadrian args.next_hop = dst; 759296077Sadrian args.oif = ifp; 760296077Sadrian off = ip_fw_chk_ptr(&args); 761296077Sadrian m = args.m; 762296077Sadrian dst = args.next_hop; 763296077Sadrian 764296077Sadrian /* 765296077Sadrian * On return we must do the following: 766296077Sadrian * m == NULL -> drop the pkt (old interface, deprecated) 767296077Sadrian * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 768296077Sadrian * 1<=off<= 0xffff -> DIVERT 769296077Sadrian * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 770296077Sadrian * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 771296077Sadrian * dst != old -> IPFIREWALL_FORWARD 772296077Sadrian * off==0, dst==old -> accept 773296077Sadrian * If some of the above modules are not compiled in, then 774296077Sadrian * we should't have to check the corresponding condition 775296077Sadrian * (because the ipfw control socket should not accept 776296077Sadrian * unsupported rules), but better play safe and drop 777296077Sadrian * packets in case of doubt. 778296077Sadrian */ 779296077Sadrian if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 780296077Sadrian if (m) 781296077Sadrian m_freem(m); 782296077Sadrian error = EACCES; 783296077Sadrian goto done; 784296077Sadrian } 785296077Sadrian ip = mtod(m, struct ip *); 786296077Sadrian if (off == 0 && dst == old) /* common case */ 787296077Sadrian goto pass; 788296077Sadrian if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 789296077Sadrian /* 790296077Sadrian * pass the pkt to dummynet. Need to include 791296077Sadrian * pipe number, m, ifp, ro, dst because these are 792296077Sadrian * not recomputed in the next pass. 793296077Sadrian * All other parameters have been already used and 794296077Sadrian * so they are not needed anymore. 795296077Sadrian * XXX note: if the ifp or ro entry are deleted 796296077Sadrian * while a pkt is in dummynet, we are in trouble! 797296077Sadrian */ 798296077Sadrian args.ro = ro; 799296077Sadrian args.dst = dst; 800296077Sadrian args.flags = flags; 801296077Sadrian 802296077Sadrian error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 803296077Sadrian &args); 804296077Sadrian goto done; 805296077Sadrian } 806296077Sadrian#ifdef IPDIVERT 807296077Sadrian if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 808296077Sadrian struct mbuf *clone = NULL; 809296077Sadrian 810296077Sadrian /* Clone packet if we're doing a 'tee' */ 811296077Sadrian if ((off & IP_FW_PORT_TEE_FLAG) != 0) 812296077Sadrian clone = m_dup(m, M_DONTWAIT); 813296077Sadrian 814296077Sadrian /* 815296077Sadrian * XXX 816296077Sadrian * delayed checksums are not currently compatible 817296077Sadrian * with divert sockets. 818296077Sadrian */ 819296077Sadrian if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 820296077Sadrian in_delayed_cksum(m); 821296077Sadrian m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 822296077Sadrian } 823296077Sadrian 824296077Sadrian /* Restore packet header fields to original values */ 825296077Sadrian ip->ip_len = htons(ip->ip_len); 826296077Sadrian ip->ip_off = htons(ip->ip_off); 827296077Sadrian 828296077Sadrian /* Deliver packet to divert input routine */ 829296077Sadrian divert_packet(m, 0, off & 0xffff, args.divert_rule); 830296077Sadrian 831296077Sadrian /* If 'tee', continue with original packet */ 832296077Sadrian if (clone != NULL) { 833296077Sadrian m = clone; 834296077Sadrian ip = mtod(m, struct ip *); 835296077Sadrian goto pass; 836296077Sadrian } 837296077Sadrian goto done; 838296077Sadrian } 839296077Sadrian#endif 840296077Sadrian 841296077Sadrian /* IPFIREWALL_FORWARD */ 842296077Sadrian /* 843296077Sadrian * Check dst to make sure it is directly reachable on the 844296077Sadrian * interface we previously thought it was. 845296077Sadrian * If it isn't (which may be likely in some situations) we have 846296077Sadrian * to re-route it (ie, find a route for the next-hop and the 847296077Sadrian * associated interface) and set them here. This is nested 848296077Sadrian * forwarding which in most cases is undesirable, except where 849296077Sadrian * such control is nigh impossible. So we do it here. 850296077Sadrian * And I'm babbling. 851296077Sadrian */ 852296077Sadrian if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 853296077Sadrian#if 0 854296077Sadrian /* 855296077Sadrian * XXX To improve readability, this block should be 856296077Sadrian * changed into a function call as below: 857296077Sadrian */ 858296077Sadrian error = ip_ipforward(&m, &dst, &ifp); 859296077Sadrian if (error) 860296077Sadrian goto bad; 861296077Sadrian if (m == NULL) /* ip_input consumed the mbuf */ 862296077Sadrian goto done; 863296077Sadrian#else 864296077Sadrian struct in_ifaddr *ia; 865296077Sadrian 866296077Sadrian /* 867296077Sadrian * XXX sro_fwd below is static, and a pointer 868296077Sadrian * to it gets passed to routines downstream. 869296077Sadrian * This could have surprisingly bad results in 870296077Sadrian * practice, because its content is overwritten 871296077Sadrian * by subsequent packets. 872296077Sadrian */ 873296077Sadrian /* There must be a better way to do this next line... */ 874296077Sadrian static struct route sro_fwd; 875296077Sadrian struct route *ro_fwd = &sro_fwd; 876296077Sadrian 877296077Sadrian#if 0 878296077Sadrian print_ip("IPFIREWALL_FORWARD: New dst ip: ", 879296077Sadrian dst->sin_addr, "\n"); 880296077Sadrian#endif 881299314Sadrian 882299314Sadrian /* 883299314Sadrian * We need to figure out if we have been forwarded 884299314Sadrian * to a local socket. If so, then we should somehow 885299314Sadrian * "loop back" to ip_input, and get directed to the 886299314Sadrian * PCB as if we had received this packet. This is 887299314Sadrian * because it may be dificult to identify the packets 888299314Sadrian * you want to forward until they are being output 889299314Sadrian * and have selected an interface. (e.g. locally 890299314Sadrian * initiated packets) If we used the loopback inteface, 891299314Sadrian * we would not be able to control what happens 892299314Sadrian * as the packet runs through ip_input() as 893299314Sadrian * it is done through an ISR. 894299314Sadrian */ 895299314Sadrian LIST_FOREACH(ia, 896299314Sadrian INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 897299314Sadrian /* 898299314Sadrian * If the addr to forward to is one 899299314Sadrian * of ours, we pretend to 900299314Sadrian * be the destination for this packet. 901296077Sadrian */ 902296077Sadrian if (IA_SIN(ia)->sin_addr.s_addr == 903296077Sadrian dst->sin_addr.s_addr) 904296077Sadrian break; 905296077Sadrian } 906296077Sadrian if (ia) { /* tell ip_input "dont filter" */ 907296077Sadrian struct m_hdr tag; 908296077Sadrian 909296077Sadrian tag.mh_type = MT_TAG; 910296077Sadrian tag.mh_flags = PACKET_TAG_IPFORWARD; 911296077Sadrian tag.mh_data = (caddr_t)args.next_hop; 912296077Sadrian tag.mh_next = m; 913296077Sadrian 914296077Sadrian if (m->m_pkthdr.rcvif == NULL) 915296077Sadrian m->m_pkthdr.rcvif = ifunit("lo0"); 916296077Sadrian if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 917 m->m_pkthdr.csum_flags |= 918 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 919 m0->m_pkthdr.csum_data = 0xffff; 920 } 921 m->m_pkthdr.csum_flags |= 922 CSUM_IP_CHECKED | CSUM_IP_VALID; 923 ip->ip_len = htons(ip->ip_len); 924 ip->ip_off = htons(ip->ip_off); 925 ip_input((struct mbuf *)&tag); 926 goto done; 927 } 928 /* Some of the logic for this was 929 * nicked from above. 930 * 931 * This rewrites the cached route in a local PCB. 932 * Is this what we want to do? 933 */ 934 bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 935 936 ro_fwd->ro_rt = 0; 937 rtalloc_ign(ro_fwd, RTF_PRCLONING); 938 939 if (ro_fwd->ro_rt == 0) { 940 ipstat.ips_noroute++; 941 error = EHOSTUNREACH; 942 goto bad; 943 } 944 945 ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 946 ifp = ro_fwd->ro_rt->rt_ifp; 947 ro_fwd->ro_rt->rt_use++; 948 if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 949 dst = (struct sockaddr_in *) 950 ro_fwd->ro_rt->rt_gateway; 951 if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 952 isbroadcast = 953 (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 954 else 955 isbroadcast = in_broadcast(dst->sin_addr, ifp); 956 if (ro->ro_rt) 957 RTFREE(ro->ro_rt); 958 ro->ro_rt = ro_fwd->ro_rt; 959 dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 960 961#endif /* ... block to be put into a function */ 962 /* 963 * If we added a default src ip earlier, 964 * which would have been gotten from the-then 965 * interface, do it again, from the new one. 966 */ 967 if (src_was_INADDR_ANY) 968 ip->ip_src = IA_SIN(ia)->sin_addr; 969 goto pass ; 970 } 971 972 /* 973 * if we get here, none of the above matches, and 974 * we have to drop the pkt 975 */ 976 m_freem(m); 977 error = EACCES; /* not sure this is the right error msg */ 978 goto done; 979 } 980 981pass: 982 /* 127/8 must not appear on wire - RFC1122. */ 983 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 984 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 985 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 986 ipstat.ips_badaddr++; 987 error = EADDRNOTAVAIL; 988 goto bad; 989 } 990 } 991 992 m->m_pkthdr.csum_flags |= CSUM_IP; 993 sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 994 if (sw_csum & CSUM_DELAY_DATA) { 995 in_delayed_cksum(m); 996 sw_csum &= ~CSUM_DELAY_DATA; 997 } 998 m->m_pkthdr.csum_flags &= ifp->if_hwassist; 999 1000 /* 1001 * If small enough for interface, or the interface will take 1002 * care of the fragmentation for us, can just send directly. 1003 */ 1004 if ((u_short)ip->ip_len <= ifp->if_mtu || 1005 ifp->if_hwassist & CSUM_FRAGMENT) { 1006 ip->ip_len = htons(ip->ip_len); 1007 ip->ip_off = htons(ip->ip_off); 1008 ip->ip_sum = 0; 1009 if (sw_csum & CSUM_DELAY_IP) 1010 ip->ip_sum = in_cksum(m, hlen); 1011 1012 /* Record statistics for this interface address. */ 1013 if (!(flags & IP_FORWARDING) && ia) { 1014 ia->ia_ifa.if_opackets++; 1015 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1016 } 1017 1018#ifdef IPSEC 1019 /* clean ipsec history once it goes out of the node */ 1020 ipsec_delaux(m); 1021#endif 1022 1023#ifdef MBUF_FRAG_TEST 1024 if (mbuf_frag_size) { 1025 struct mbuf *m1, *m2; 1026 int length; 1027 1028 length = m->m_len; 1029 1030 while (1) { 1031 length -= mbuf_frag_size; 1032 if (length < 1) 1033 break; 1034 m1 = m_split(m, length, M_DONTWAIT); 1035 m2 = m; 1036 while (1) { 1037 if (m2->m_next == NULL) 1038 break; 1039 m2 = m2->m_next; 1040 } 1041 m2->m_next = m1; 1042 } 1043 } 1044#endif 1045 error = (*ifp->if_output)(ifp, m, 1046 (struct sockaddr *)dst, ro->ro_rt); 1047 goto done; 1048 } 1049 /* 1050 * Too large for interface; fragment if possible. 1051 * Must be able to put at least 8 bytes per fragment. 1052 */ 1053 if (ip->ip_off & IP_DF) { 1054 error = EMSGSIZE; 1055 /* 1056 * This case can happen if the user changed the MTU 1057 * of an interface after enabling IP on it. Because 1058 * most netifs don't keep track of routes pointing to 1059 * them, there is no way for one to update all its 1060 * routes when the MTU is changed. 1061 */ 1062 if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 1063 && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 1064 && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 1065 ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 1066 } 1067 ipstat.ips_cantfrag++; 1068 goto bad; 1069 } 1070 len = (ifp->if_mtu - hlen) &~ 7; 1071 if (len < 8) { 1072 error = EMSGSIZE; 1073 goto bad; 1074 } 1075 1076 /* 1077 * if the interface will not calculate checksums on 1078 * fragmented packets, then do it here. 1079 */ 1080 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 1081 (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 1082 in_delayed_cksum(m); 1083 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1084 } 1085 1086 if (len > PAGE_SIZE) { 1087 /* 1088 * Fragement large datagrams such that each segment 1089 * contains a multiple of PAGE_SIZE amount of data, 1090 * plus headers. This enables a receiver to perform 1091 * page-flipping zero-copy optimizations. 1092 */ 1093 1094 int newlen; 1095 struct mbuf *mtmp; 1096 1097 for (mtmp = m, off = 0; 1098 mtmp && ((off + mtmp->m_len) <= ifp->if_mtu); 1099 mtmp = mtmp->m_next) { 1100 off += mtmp->m_len; 1101 } 1102 /* 1103 * firstlen (off - hlen) must be aligned on an 1104 * 8-byte boundary 1105 */ 1106 if (off < hlen) 1107 goto smart_frag_failure; 1108 off = ((off - hlen) & ~7) + hlen; 1109 newlen = (~PAGE_MASK) & ifp->if_mtu; 1110 if ((newlen + sizeof (struct ip)) > ifp->if_mtu) { 1111 /* we failed, go back the default */ 1112smart_frag_failure: 1113 newlen = len; 1114 off = hlen + len; 1115 } 1116 1117/* printf("ipfrag: len = %d, hlen = %d, mhlen = %d, newlen = %d, off = %d\n", 1118 len, hlen, sizeof (struct ip), newlen, off);*/ 1119 1120 len = newlen; 1121 1122 } else { 1123 off = hlen + len; 1124 } 1125 1126 1127 1128 { 1129 int mhlen, firstlen = off - hlen; 1130 struct mbuf **mnext = &m->m_nextpkt; 1131 int nfrags = 1; 1132 1133 /* 1134 * Loop through length of segment after first fragment, 1135 * make new header and copy data of each part and link onto chain. 1136 */ 1137 m0 = m; 1138 mhlen = sizeof (struct ip); 1139 for (; off < (u_short)ip->ip_len; off += len) { 1140 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1141 if (m == 0) { 1142 error = ENOBUFS; 1143 ipstat.ips_odropped++; 1144 goto sendorfree; 1145 } 1146 m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 1147 m->m_data += max_linkhdr; 1148 mhip = mtod(m, struct ip *); 1149 *mhip = *ip; 1150 if (hlen > sizeof (struct ip)) { 1151 mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 1152 mhip->ip_v = IPVERSION; 1153 mhip->ip_hl = mhlen >> 2; 1154 } 1155 m->m_len = mhlen; 1156 mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 1157 if (off + len >= (u_short)ip->ip_len) 1158 len = (u_short)ip->ip_len - off; 1159 else 1160 mhip->ip_off |= IP_MF; 1161 mhip->ip_len = htons((u_short)(len + mhlen)); 1162 m->m_next = m_copy(m0, off, len); 1163 if (m->m_next == 0) { 1164 (void) m_free(m); 1165 error = ENOBUFS; /* ??? */ 1166 ipstat.ips_odropped++; 1167 goto sendorfree; 1168 } 1169 m->m_pkthdr.len = mhlen + len; 1170 m->m_pkthdr.rcvif = (struct ifnet *)0; 1171#ifdef MAC 1172 mac_create_fragment(m0, m); 1173#endif 1174 m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1175 mhip->ip_off = htons(mhip->ip_off); 1176 mhip->ip_sum = 0; 1177 if (sw_csum & CSUM_DELAY_IP) 1178 mhip->ip_sum = in_cksum(m, mhlen); 1179 *mnext = m; 1180 mnext = &m->m_nextpkt; 1181 nfrags++; 1182 } 1183 ipstat.ips_ofragments += nfrags; 1184 1185 /* set first/last markers for fragment chain */ 1186 m->m_flags |= M_LASTFRAG; 1187 m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1188 m0->m_pkthdr.csum_data = nfrags; 1189 1190 /* 1191 * Update first fragment by trimming what's been copied out 1192 * and updating header, then send each fragment (in order). 1193 */ 1194 m = m0; 1195 m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 1196 m->m_pkthdr.len = hlen + firstlen; 1197 ip->ip_len = htons((u_short)m->m_pkthdr.len); 1198 ip->ip_off |= IP_MF; 1199 ip->ip_off = htons(ip->ip_off); 1200 ip->ip_sum = 0; 1201 if (sw_csum & CSUM_DELAY_IP) 1202 ip->ip_sum = in_cksum(m, hlen); 1203sendorfree: 1204 for (m = m0; m; m = m0) { 1205 m0 = m->m_nextpkt; 1206 m->m_nextpkt = 0; 1207#ifdef IPSEC 1208 /* clean ipsec history once it goes out of the node */ 1209 ipsec_delaux(m); 1210#endif 1211 if (error == 0) { 1212 /* Record statistics for this interface address. */ 1213 if (ia != NULL) { 1214 ia->ia_ifa.if_opackets++; 1215 ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1216 } 1217 1218 error = (*ifp->if_output)(ifp, m, 1219 (struct sockaddr *)dst, ro->ro_rt); 1220 } else 1221 m_freem(m); 1222 } 1223 1224 if (error == 0) 1225 ipstat.ips_fragmented++; 1226 } 1227done: 1228#ifdef IPSEC 1229 if (ro == &iproute && ro->ro_rt) { 1230 RTFREE(ro->ro_rt); 1231 ro->ro_rt = NULL; 1232 } 1233 if (sp != NULL) { 1234 KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1235 printf("DP ip_output call free SP:%p\n", sp)); 1236 key_freesp(sp); 1237 } 1238#endif /* IPSEC */ 1239#ifdef FAST_IPSEC 1240 if (ro == &iproute && ro->ro_rt) { 1241 RTFREE(ro->ro_rt); 1242 ro->ro_rt = NULL; 1243 } 1244 if (sp != NULL) 1245 KEY_FREESP(&sp); 1246#endif /* FAST_IPSEC */ 1247 return (error); 1248bad: 1249 m_freem(m); 1250 goto done; 1251} 1252 1253void 1254in_delayed_cksum(struct mbuf *m) 1255{ 1256 struct ip *ip; 1257 u_short csum, offset; 1258 1259 ip = mtod(m, struct ip *); 1260 offset = ip->ip_hl << 2 ; 1261 csum = in_cksum_skip(m, ip->ip_len, offset); 1262 if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1263 csum = 0xffff; 1264 offset += m->m_pkthdr.csum_data; /* checksum offset */ 1265 1266 if (offset + sizeof(u_short) > m->m_len) { 1267 printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1268 m->m_len, offset, ip->ip_p); 1269 /* 1270 * XXX 1271 * this shouldn't happen, but if it does, the 1272 * correct behavior may be to insert the checksum 1273 * in the existing chain instead of rearranging it. 1274 */ 1275 m = m_pullup(m, offset + sizeof(u_short)); 1276 } 1277 *(u_short *)(m->m_data + offset) = csum; 1278} 1279 1280/* 1281 * Insert IP options into preformed packet. 1282 * Adjust IP destination as required for IP source routing, 1283 * as indicated by a non-zero in_addr at the start of the options. 1284 * 1285 * XXX This routine assumes that the packet has no options in place. 1286 */ 1287static struct mbuf * 1288ip_insertoptions(m, opt, phlen) 1289 register struct mbuf *m; 1290 struct mbuf *opt; 1291 int *phlen; 1292{ 1293 register struct ipoption *p = mtod(opt, struct ipoption *); 1294 struct mbuf *n; 1295 register struct ip *ip = mtod(m, struct ip *); 1296 unsigned optlen; 1297 1298 optlen = opt->m_len - sizeof(p->ipopt_dst); 1299 if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) { 1300 *phlen = 0; 1301 return (m); /* XXX should fail */ 1302 } 1303 if (p->ipopt_dst.s_addr) 1304 ip->ip_dst = p->ipopt_dst; 1305 if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1306 MGETHDR(n, M_DONTWAIT, MT_HEADER); 1307 if (n == 0) { 1308 *phlen = 0; 1309 return (m); 1310 } 1311 n->m_pkthdr.rcvif = (struct ifnet *)0; 1312#ifdef MAC 1313 mac_create_mbuf_from_mbuf(m, n); 1314#endif 1315 n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1316 m->m_len -= sizeof(struct ip); 1317 m->m_data += sizeof(struct ip); 1318 n->m_next = m; 1319 m = n; 1320 m->m_len = optlen + sizeof(struct ip); 1321 m->m_data += max_linkhdr; 1322 (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1323 } else { 1324 m->m_data -= optlen; 1325 m->m_len += optlen; 1326 m->m_pkthdr.len += optlen; 1327 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1328 } 1329 ip = mtod(m, struct ip *); 1330 bcopy(p->ipopt_list, ip + 1, optlen); 1331 *phlen = sizeof(struct ip) + optlen; 1332 ip->ip_v = IPVERSION; 1333 ip->ip_hl = *phlen >> 2; 1334 ip->ip_len += optlen; 1335 return (m); 1336} 1337 1338/* 1339 * Copy options from ip to jp, 1340 * omitting those not copied during fragmentation. 1341 */ 1342int 1343ip_optcopy(ip, jp) 1344 struct ip *ip, *jp; 1345{ 1346 register u_char *cp, *dp; 1347 int opt, optlen, cnt; 1348 1349 cp = (u_char *)(ip + 1); 1350 dp = (u_char *)(jp + 1); 1351 cnt = (ip->ip_hl << 2) - sizeof (struct ip); 1352 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1353 opt = cp[0]; 1354 if (opt == IPOPT_EOL) 1355 break; 1356 if (opt == IPOPT_NOP) { 1357 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1358 *dp++ = IPOPT_NOP; 1359 optlen = 1; 1360 continue; 1361 } 1362 1363 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1364 ("ip_optcopy: malformed ipv4 option")); 1365 optlen = cp[IPOPT_OLEN]; 1366 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1367 ("ip_optcopy: malformed ipv4 option")); 1368 1369 /* bogus lengths should have been caught by ip_dooptions */ 1370 if (optlen > cnt) 1371 optlen = cnt; 1372 if (IPOPT_COPIED(opt)) { 1373 bcopy(cp, dp, optlen); 1374 dp += optlen; 1375 } 1376 } 1377 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1378 *dp++ = IPOPT_EOL; 1379 return (optlen); 1380} 1381 1382/* 1383 * IP socket option processing. 1384 */ 1385int 1386ip_ctloutput(so, sopt) 1387 struct socket *so; 1388 struct sockopt *sopt; 1389{ 1390 struct inpcb *inp = sotoinpcb(so); 1391 int error, optval; 1392 1393 error = optval = 0; 1394 if (sopt->sopt_level != IPPROTO_IP) { 1395 return (EINVAL); 1396 } 1397 1398 switch (sopt->sopt_dir) { 1399 case SOPT_SET: 1400 switch (sopt->sopt_name) { 1401 case IP_OPTIONS: 1402#ifdef notyet 1403 case IP_RETOPTS: 1404#endif 1405 { 1406 struct mbuf *m; 1407 if (sopt->sopt_valsize > MLEN) { 1408 error = EMSGSIZE; 1409 break; 1410 } 1411 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1412 if (m == 0) { 1413 error = ENOBUFS; 1414 break; 1415 } 1416 m->m_len = sopt->sopt_valsize; 1417 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1418 m->m_len); 1419 1420 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1421 m)); 1422 } 1423 1424 case IP_TOS: 1425 case IP_TTL: 1426 case IP_RECVOPTS: 1427 case IP_RECVRETOPTS: 1428 case IP_RECVDSTADDR: 1429 case IP_RECVIF: 1430 case IP_FAITH: 1431 error = sooptcopyin(sopt, &optval, sizeof optval, 1432 sizeof optval); 1433 if (error) 1434 break; 1435 1436 switch (sopt->sopt_name) { 1437 case IP_TOS: 1438 inp->inp_ip_tos = optval; 1439 break; 1440 1441 case IP_TTL: 1442 inp->inp_ip_ttl = optval; 1443 break; 1444#define OPTSET(bit) \ 1445 if (optval) \ 1446 inp->inp_flags |= bit; \ 1447 else \ 1448 inp->inp_flags &= ~bit; 1449 1450 case IP_RECVOPTS: 1451 OPTSET(INP_RECVOPTS); 1452 break; 1453 1454 case IP_RECVRETOPTS: 1455 OPTSET(INP_RECVRETOPTS); 1456 break; 1457 1458 case IP_RECVDSTADDR: 1459 OPTSET(INP_RECVDSTADDR); 1460 break; 1461 1462 case IP_RECVIF: 1463 OPTSET(INP_RECVIF); 1464 break; 1465 1466 case IP_FAITH: 1467 OPTSET(INP_FAITH); 1468 break; 1469 } 1470 break; 1471#undef OPTSET 1472 1473 case IP_MULTICAST_IF: 1474 case IP_MULTICAST_VIF: 1475 case IP_MULTICAST_TTL: 1476 case IP_MULTICAST_LOOP: 1477 case IP_ADD_MEMBERSHIP: 1478 case IP_DROP_MEMBERSHIP: 1479 error = ip_setmoptions(sopt, &inp->inp_moptions); 1480 break; 1481 1482 case IP_PORTRANGE: 1483 error = sooptcopyin(sopt, &optval, sizeof optval, 1484 sizeof optval); 1485 if (error) 1486 break; 1487 1488 switch (optval) { 1489 case IP_PORTRANGE_DEFAULT: 1490 inp->inp_flags &= ~(INP_LOWPORT); 1491 inp->inp_flags &= ~(INP_HIGHPORT); 1492 break; 1493 1494 case IP_PORTRANGE_HIGH: 1495 inp->inp_flags &= ~(INP_LOWPORT); 1496 inp->inp_flags |= INP_HIGHPORT; 1497 break; 1498 1499 case IP_PORTRANGE_LOW: 1500 inp->inp_flags &= ~(INP_HIGHPORT); 1501 inp->inp_flags |= INP_LOWPORT; 1502 break; 1503 1504 default: 1505 error = EINVAL; 1506 break; 1507 } 1508 break; 1509 1510#if defined(IPSEC) || defined(FAST_IPSEC) 1511 case IP_IPSEC_POLICY: 1512 { 1513 caddr_t req; 1514 size_t len = 0; 1515 int priv; 1516 struct mbuf *m; 1517 int optname; 1518 1519 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1520 break; 1521 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1522 break; 1523 priv = (sopt->sopt_td != NULL && 1524 suser(sopt->sopt_td) != 0) ? 0 : 1; 1525 req = mtod(m, caddr_t); 1526 len = m->m_len; 1527 optname = sopt->sopt_name; 1528 error = ipsec4_set_policy(inp, optname, req, len, priv); 1529 m_freem(m); 1530 break; 1531 } 1532#endif /*IPSEC*/ 1533 1534 default: 1535 error = ENOPROTOOPT; 1536 break; 1537 } 1538 break; 1539 1540 case SOPT_GET: 1541 switch (sopt->sopt_name) { 1542 case IP_OPTIONS: 1543 case IP_RETOPTS: 1544 if (inp->inp_options) 1545 error = sooptcopyout(sopt, 1546 mtod(inp->inp_options, 1547 char *), 1548 inp->inp_options->m_len); 1549 else 1550 sopt->sopt_valsize = 0; 1551 break; 1552 1553 case IP_TOS: 1554 case IP_TTL: 1555 case IP_RECVOPTS: 1556 case IP_RECVRETOPTS: 1557 case IP_RECVDSTADDR: 1558 case IP_RECVIF: 1559 case IP_PORTRANGE: 1560 case IP_FAITH: 1561 switch (sopt->sopt_name) { 1562 1563 case IP_TOS: 1564 optval = inp->inp_ip_tos; 1565 break; 1566 1567 case IP_TTL: 1568 optval = inp->inp_ip_ttl; 1569 break; 1570 1571#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1572 1573 case IP_RECVOPTS: 1574 optval = OPTBIT(INP_RECVOPTS); 1575 break; 1576 1577 case IP_RECVRETOPTS: 1578 optval = OPTBIT(INP_RECVRETOPTS); 1579 break; 1580 1581 case IP_RECVDSTADDR: 1582 optval = OPTBIT(INP_RECVDSTADDR); 1583 break; 1584 1585 case IP_RECVIF: 1586 optval = OPTBIT(INP_RECVIF); 1587 break; 1588 1589 case IP_PORTRANGE: 1590 if (inp->inp_flags & INP_HIGHPORT) 1591 optval = IP_PORTRANGE_HIGH; 1592 else if (inp->inp_flags & INP_LOWPORT) 1593 optval = IP_PORTRANGE_LOW; 1594 else 1595 optval = 0; 1596 break; 1597 1598 case IP_FAITH: 1599 optval = OPTBIT(INP_FAITH); 1600 break; 1601 } 1602 error = sooptcopyout(sopt, &optval, sizeof optval); 1603 break; 1604 1605 case IP_MULTICAST_IF: 1606 case IP_MULTICAST_VIF: 1607 case IP_MULTICAST_TTL: 1608 case IP_MULTICAST_LOOP: 1609 case IP_ADD_MEMBERSHIP: 1610 case IP_DROP_MEMBERSHIP: 1611 error = ip_getmoptions(sopt, inp->inp_moptions); 1612 break; 1613 1614#if defined(IPSEC) || defined(FAST_IPSEC) 1615 case IP_IPSEC_POLICY: 1616 { 1617 struct mbuf *m = NULL; 1618 caddr_t req = NULL; 1619 size_t len = 0; 1620 1621 if (m != 0) { 1622 req = mtod(m, caddr_t); 1623 len = m->m_len; 1624 } 1625 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1626 if (error == 0) 1627 error = soopt_mcopyout(sopt, m); /* XXX */ 1628 if (error == 0) 1629 m_freem(m); 1630 break; 1631 } 1632#endif /*IPSEC*/ 1633 1634 default: 1635 error = ENOPROTOOPT; 1636 break; 1637 } 1638 break; 1639 } 1640 return (error); 1641} 1642 1643/* 1644 * Set up IP options in pcb for insertion in output packets. 1645 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1646 * with destination address if source routed. 1647 */ 1648static int 1649ip_pcbopts(optname, pcbopt, m) 1650 int optname; 1651 struct mbuf **pcbopt; 1652 register struct mbuf *m; 1653{ 1654 register int cnt, optlen; 1655 register u_char *cp; 1656 u_char opt; 1657 1658 /* turn off any old options */ 1659 if (*pcbopt) 1660 (void)m_free(*pcbopt); 1661 *pcbopt = 0; 1662 if (m == (struct mbuf *)0 || m->m_len == 0) { 1663 /* 1664 * Only turning off any previous options. 1665 */ 1666 if (m) 1667 (void)m_free(m); 1668 return (0); 1669 } 1670 1671 if (m->m_len % sizeof(int32_t)) 1672 goto bad; 1673 /* 1674 * IP first-hop destination address will be stored before 1675 * actual options; move other options back 1676 * and clear it when none present. 1677 */ 1678 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1679 goto bad; 1680 cnt = m->m_len; 1681 m->m_len += sizeof(struct in_addr); 1682 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1683 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1684 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1685 1686 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1687 opt = cp[IPOPT_OPTVAL]; 1688 if (opt == IPOPT_EOL) 1689 break; 1690 if (opt == IPOPT_NOP) 1691 optlen = 1; 1692 else { 1693 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1694 goto bad; 1695 optlen = cp[IPOPT_OLEN]; 1696 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1697 goto bad; 1698 } 1699 switch (opt) { 1700 1701 default: 1702 break; 1703 1704 case IPOPT_LSRR: 1705 case IPOPT_SSRR: 1706 /* 1707 * user process specifies route as: 1708 * ->A->B->C->D 1709 * D must be our final destination (but we can't 1710 * check that since we may not have connected yet). 1711 * A is first hop destination, which doesn't appear in 1712 * actual IP option, but is stored before the options. 1713 */ 1714 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1715 goto bad; 1716 m->m_len -= sizeof(struct in_addr); 1717 cnt -= sizeof(struct in_addr); 1718 optlen -= sizeof(struct in_addr); 1719 cp[IPOPT_OLEN] = optlen; 1720 /* 1721 * Move first hop before start of options. 1722 */ 1723 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1724 sizeof(struct in_addr)); 1725 /* 1726 * Then copy rest of options back 1727 * to close up the deleted entry. 1728 */ 1729 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1730 sizeof(struct in_addr)), 1731 (caddr_t)&cp[IPOPT_OFFSET+1], 1732 (unsigned)cnt + sizeof(struct in_addr)); 1733 break; 1734 } 1735 } 1736 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1737 goto bad; 1738 *pcbopt = m; 1739 return (0); 1740 1741bad: 1742 (void)m_free(m); 1743 return (EINVAL); 1744} 1745 1746/* 1747 * XXX 1748 * The whole multicast option thing needs to be re-thought. 1749 * Several of these options are equally applicable to non-multicast 1750 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1751 * standard option (IP_TTL). 1752 */ 1753 1754/* 1755 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1756 */ 1757static struct ifnet * 1758ip_multicast_if(a, ifindexp) 1759 struct in_addr *a; 1760 int *ifindexp; 1761{ 1762 int ifindex; 1763 struct ifnet *ifp; 1764 1765 if (ifindexp) 1766 *ifindexp = 0; 1767 if (ntohl(a->s_addr) >> 24 == 0) { 1768 ifindex = ntohl(a->s_addr) & 0xffffff; 1769 if (ifindex < 0 || if_index < ifindex) 1770 return NULL; 1771 ifp = ifnet_byindex(ifindex); 1772 if (ifindexp) 1773 *ifindexp = ifindex; 1774 } else { 1775 INADDR_TO_IFP(*a, ifp); 1776 } 1777 return ifp; 1778} 1779 1780/* 1781 * Set the IP multicast options in response to user setsockopt(). 1782 */ 1783static int 1784ip_setmoptions(sopt, imop) 1785 struct sockopt *sopt; 1786 struct ip_moptions **imop; 1787{ 1788 int error = 0; 1789 int i; 1790 struct in_addr addr; 1791 struct ip_mreq mreq; 1792 struct ifnet *ifp; 1793 struct ip_moptions *imo = *imop; 1794 struct route ro; 1795 struct sockaddr_in *dst; 1796 int ifindex; 1797 int s; 1798 1799 if (imo == NULL) { 1800 /* 1801 * No multicast option buffer attached to the pcb; 1802 * allocate one and initialize to default values. 1803 */ 1804 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1805 M_WAITOK); 1806 1807 if (imo == NULL) 1808 return (ENOBUFS); 1809 *imop = imo; 1810 imo->imo_multicast_ifp = NULL; 1811 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1812 imo->imo_multicast_vif = -1; 1813 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1814 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1815 imo->imo_num_memberships = 0; 1816 } 1817 1818 switch (sopt->sopt_name) { 1819 /* store an index number for the vif you wanna use in the send */ 1820 case IP_MULTICAST_VIF: 1821 if (legal_vif_num == 0) { 1822 error = EOPNOTSUPP; 1823 break; 1824 } 1825 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1826 if (error) 1827 break; 1828 if (!legal_vif_num(i) && (i != -1)) { 1829 error = EINVAL; 1830 break; 1831 } 1832 imo->imo_multicast_vif = i; 1833 break; 1834 1835 case IP_MULTICAST_IF: 1836 /* 1837 * Select the interface for outgoing multicast packets. 1838 */ 1839 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1840 if (error) 1841 break; 1842 /* 1843 * INADDR_ANY is used to remove a previous selection. 1844 * When no interface is selected, a default one is 1845 * chosen every time a multicast packet is sent. 1846 */ 1847 if (addr.s_addr == INADDR_ANY) { 1848 imo->imo_multicast_ifp = NULL; 1849 break; 1850 } 1851 /* 1852 * The selected interface is identified by its local 1853 * IP address. Find the interface and confirm that 1854 * it supports multicasting. 1855 */ 1856 s = splimp(); 1857 ifp = ip_multicast_if(&addr, &ifindex); 1858 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1859 splx(s); 1860 error = EADDRNOTAVAIL; 1861 break; 1862 } 1863 imo->imo_multicast_ifp = ifp; 1864 if (ifindex) 1865 imo->imo_multicast_addr = addr; 1866 else 1867 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1868 splx(s); 1869 break; 1870 1871 case IP_MULTICAST_TTL: 1872 /* 1873 * Set the IP time-to-live for outgoing multicast packets. 1874 * The original multicast API required a char argument, 1875 * which is inconsistent with the rest of the socket API. 1876 * We allow either a char or an int. 1877 */ 1878 if (sopt->sopt_valsize == 1) { 1879 u_char ttl; 1880 error = sooptcopyin(sopt, &ttl, 1, 1); 1881 if (error) 1882 break; 1883 imo->imo_multicast_ttl = ttl; 1884 } else { 1885 u_int ttl; 1886 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1887 sizeof ttl); 1888 if (error) 1889 break; 1890 if (ttl > 255) 1891 error = EINVAL; 1892 else 1893 imo->imo_multicast_ttl = ttl; 1894 } 1895 break; 1896 1897 case IP_MULTICAST_LOOP: 1898 /* 1899 * Set the loopback flag for outgoing multicast packets. 1900 * Must be zero or one. The original multicast API required a 1901 * char argument, which is inconsistent with the rest 1902 * of the socket API. We allow either a char or an int. 1903 */ 1904 if (sopt->sopt_valsize == 1) { 1905 u_char loop; 1906 error = sooptcopyin(sopt, &loop, 1, 1); 1907 if (error) 1908 break; 1909 imo->imo_multicast_loop = !!loop; 1910 } else { 1911 u_int loop; 1912 error = sooptcopyin(sopt, &loop, sizeof loop, 1913 sizeof loop); 1914 if (error) 1915 break; 1916 imo->imo_multicast_loop = !!loop; 1917 } 1918 break; 1919 1920 case IP_ADD_MEMBERSHIP: 1921 /* 1922 * Add a multicast group membership. 1923 * Group must be a valid IP multicast address. 1924 */ 1925 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1926 if (error) 1927 break; 1928 1929 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1930 error = EINVAL; 1931 break; 1932 } 1933 s = splimp(); 1934 /* 1935 * If no interface address was provided, use the interface of 1936 * the route to the given multicast address. 1937 */ 1938 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1939 bzero((caddr_t)&ro, sizeof(ro)); 1940 dst = (struct sockaddr_in *)&ro.ro_dst; 1941 dst->sin_len = sizeof(*dst); 1942 dst->sin_family = AF_INET; 1943 dst->sin_addr = mreq.imr_multiaddr; 1944 rtalloc(&ro); 1945 if (ro.ro_rt == NULL) { 1946 error = EADDRNOTAVAIL; 1947 splx(s); 1948 break; 1949 } 1950 ifp = ro.ro_rt->rt_ifp; 1951 rtfree(ro.ro_rt); 1952 } 1953 else { 1954 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1955 } 1956 1957 /* 1958 * See if we found an interface, and confirm that it 1959 * supports multicast. 1960 */ 1961 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1962 error = EADDRNOTAVAIL; 1963 splx(s); 1964 break; 1965 } 1966 /* 1967 * See if the membership already exists or if all the 1968 * membership slots are full. 1969 */ 1970 for (i = 0; i < imo->imo_num_memberships; ++i) { 1971 if (imo->imo_membership[i]->inm_ifp == ifp && 1972 imo->imo_membership[i]->inm_addr.s_addr 1973 == mreq.imr_multiaddr.s_addr) 1974 break; 1975 } 1976 if (i < imo->imo_num_memberships) { 1977 error = EADDRINUSE; 1978 splx(s); 1979 break; 1980 } 1981 if (i == IP_MAX_MEMBERSHIPS) { 1982 error = ETOOMANYREFS; 1983 splx(s); 1984 break; 1985 } 1986 /* 1987 * Everything looks good; add a new record to the multicast 1988 * address list for the given interface. 1989 */ 1990 if ((imo->imo_membership[i] = 1991 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1992 error = ENOBUFS; 1993 splx(s); 1994 break; 1995 } 1996 ++imo->imo_num_memberships; 1997 splx(s); 1998 break; 1999 2000 case IP_DROP_MEMBERSHIP: 2001 /* 2002 * Drop a multicast group membership. 2003 * Group must be a valid IP multicast address. 2004 */ 2005 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 2006 if (error) 2007 break; 2008 2009 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 2010 error = EINVAL; 2011 break; 2012 } 2013 2014 s = splimp(); 2015 /* 2016 * If an interface address was specified, get a pointer 2017 * to its ifnet structure. 2018 */ 2019 if (mreq.imr_interface.s_addr == INADDR_ANY) 2020 ifp = NULL; 2021 else { 2022 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 2023 if (ifp == NULL) { 2024 error = EADDRNOTAVAIL; 2025 splx(s); 2026 break; 2027 } 2028 } 2029 /* 2030 * Find the membership in the membership array. 2031 */ 2032 for (i = 0; i < imo->imo_num_memberships; ++i) { 2033 if ((ifp == NULL || 2034 imo->imo_membership[i]->inm_ifp == ifp) && 2035 imo->imo_membership[i]->inm_addr.s_addr == 2036 mreq.imr_multiaddr.s_addr) 2037 break; 2038 } 2039 if (i == imo->imo_num_memberships) { 2040 error = EADDRNOTAVAIL; 2041 splx(s); 2042 break; 2043 } 2044 /* 2045 * Give up the multicast address record to which the 2046 * membership points. 2047 */ 2048 in_delmulti(imo->imo_membership[i]); 2049 /* 2050 * Remove the gap in the membership array. 2051 */ 2052 for (++i; i < imo->imo_num_memberships; ++i) 2053 imo->imo_membership[i-1] = imo->imo_membership[i]; 2054 --imo->imo_num_memberships; 2055 splx(s); 2056 break; 2057 2058 default: 2059 error = EOPNOTSUPP; 2060 break; 2061 } 2062 2063 /* 2064 * If all options have default values, no need to keep the mbuf. 2065 */ 2066 if (imo->imo_multicast_ifp == NULL && 2067 imo->imo_multicast_vif == -1 && 2068 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 2069 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 2070 imo->imo_num_memberships == 0) { 2071 free(*imop, M_IPMOPTS); 2072 *imop = NULL; 2073 } 2074 2075 return (error); 2076} 2077 2078/* 2079 * Return the IP multicast options in response to user getsockopt(). 2080 */ 2081static int 2082ip_getmoptions(sopt, imo) 2083 struct sockopt *sopt; 2084 register struct ip_moptions *imo; 2085{ 2086 struct in_addr addr; 2087 struct in_ifaddr *ia; 2088 int error, optval; 2089 u_char coptval; 2090 2091 error = 0; 2092 switch (sopt->sopt_name) { 2093 case IP_MULTICAST_VIF: 2094 if (imo != NULL) 2095 optval = imo->imo_multicast_vif; 2096 else 2097 optval = -1; 2098 error = sooptcopyout(sopt, &optval, sizeof optval); 2099 break; 2100 2101 case IP_MULTICAST_IF: 2102 if (imo == NULL || imo->imo_multicast_ifp == NULL) 2103 addr.s_addr = INADDR_ANY; 2104 else if (imo->imo_multicast_addr.s_addr) { 2105 /* return the value user has set */ 2106 addr = imo->imo_multicast_addr; 2107 } else { 2108 IFP_TO_IA(imo->imo_multicast_ifp, ia); 2109 addr.s_addr = (ia == NULL) ? INADDR_ANY 2110 : IA_SIN(ia)->sin_addr.s_addr; 2111 } 2112 error = sooptcopyout(sopt, &addr, sizeof addr); 2113 break; 2114 2115 case IP_MULTICAST_TTL: 2116 if (imo == 0) 2117 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 2118 else 2119 optval = coptval = imo->imo_multicast_ttl; 2120 if (sopt->sopt_valsize == 1) 2121 error = sooptcopyout(sopt, &coptval, 1); 2122 else 2123 error = sooptcopyout(sopt, &optval, sizeof optval); 2124 break; 2125 2126 case IP_MULTICAST_LOOP: 2127 if (imo == 0) 2128 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 2129 else 2130 optval = coptval = imo->imo_multicast_loop; 2131 if (sopt->sopt_valsize == 1) 2132 error = sooptcopyout(sopt, &coptval, 1); 2133 else 2134 error = sooptcopyout(sopt, &optval, sizeof optval); 2135 break; 2136 2137 default: 2138 error = ENOPROTOOPT; 2139 break; 2140 } 2141 return (error); 2142} 2143 2144/* 2145 * Discard the IP multicast options. 2146 */ 2147void 2148ip_freemoptions(imo) 2149 register struct ip_moptions *imo; 2150{ 2151 register int i; 2152 2153 if (imo != NULL) { 2154 for (i = 0; i < imo->imo_num_memberships; ++i) 2155 in_delmulti(imo->imo_membership[i]); 2156 free(imo, M_IPMOPTS); 2157 } 2158} 2159 2160/* 2161 * Routine called from ip_output() to loop back a copy of an IP multicast 2162 * packet to the input queue of a specified interface. Note that this 2163 * calls the output routine of the loopback "driver", but with an interface 2164 * pointer that might NOT be a loopback interface -- evil, but easier than 2165 * replicating that code here. 2166 */ 2167static void 2168ip_mloopback(ifp, m, dst, hlen) 2169 struct ifnet *ifp; 2170 register struct mbuf *m; 2171 register struct sockaddr_in *dst; 2172 int hlen; 2173{ 2174 register struct ip *ip; 2175 struct mbuf *copym; 2176 2177 copym = m_copy(m, 0, M_COPYALL); 2178 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2179 copym = m_pullup(copym, hlen); 2180 if (copym != NULL) { 2181 /* 2182 * We don't bother to fragment if the IP length is greater 2183 * than the interface's MTU. Can this possibly matter? 2184 */ 2185 ip = mtod(copym, struct ip *); 2186 ip->ip_len = htons(ip->ip_len); 2187 ip->ip_off = htons(ip->ip_off); 2188 ip->ip_sum = 0; 2189 ip->ip_sum = in_cksum(copym, hlen); 2190 /* 2191 * NB: 2192 * It's not clear whether there are any lingering 2193 * reentrancy problems in other areas which might 2194 * be exposed by using ip_input directly (in 2195 * particular, everything which modifies the packet 2196 * in-place). Yet another option is using the 2197 * protosw directly to deliver the looped back 2198 * packet. For the moment, we'll err on the side 2199 * of safety by using if_simloop(). 2200 */ 2201#if 1 /* XXX */ 2202 if (dst->sin_family != AF_INET) { 2203 printf("ip_mloopback: bad address family %d\n", 2204 dst->sin_family); 2205 dst->sin_family = AF_INET; 2206 } 2207#endif 2208 2209#ifdef notdef 2210 copym->m_pkthdr.rcvif = ifp; 2211 ip_input(copym); 2212#else 2213 /* if the checksum hasn't been computed, mark it as valid */ 2214 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2215 copym->m_pkthdr.csum_flags |= 2216 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2217 copym->m_pkthdr.csum_data = 0xffff; 2218 } 2219 if_simloop(ifp, copym, dst->sin_family, 0); 2220#endif 2221 } 2222} 2223