ip_output.c revision 105194
1177633Sdfr/* 2177633Sdfr * Copyright (c) 1982, 1986, 1988, 1990, 1993 3177633Sdfr * The Regents of the University of California. All rights reserved. 4177633Sdfr * 5177633Sdfr * Redistribution and use in source and binary forms, with or without 6177633Sdfr * modification, are permitted provided that the following conditions 7177633Sdfr * are met: 8177633Sdfr * 1. Redistributions of source code must retain the above copyright 9177633Sdfr * notice, this list of conditions and the following disclaimer. 10177633Sdfr * 2. Redistributions in binary form must reproduce the above copyright 11177633Sdfr * notice, this list of conditions and the following disclaimer in the 12177633Sdfr * documentation and/or other materials provided with the distribution. 13177633Sdfr * 3. All advertising materials mentioning features or use of this software 14177633Sdfr * must display the following acknowledgement: 15177633Sdfr * This product includes software developed by the University of 16177633Sdfr * California, Berkeley and its contributors. 17177633Sdfr * 4. Neither the name of the University nor the names of its contributors 18177633Sdfr * may be used to endorse or promote products derived from this software 19177633Sdfr * without specific prior written permission. 20177633Sdfr * 21177633Sdfr * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22177633Sdfr * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23177633Sdfr * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24177633Sdfr * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25177633Sdfr * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26177633Sdfr * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27177633Sdfr * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28177633Sdfr * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29177633Sdfr * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30177633Sdfr * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31177633Sdfr * SUCH DAMAGE. 32177633Sdfr * 33177633Sdfr * @(#)ip_output.c 8.3 (Berkeley) 1/21/94 34177633Sdfr * $FreeBSD: head/sys/netinet/ip_output.c 105194 2002-10-16 01:54:46Z sam $ 35177633Sdfr */ 36177633Sdfr 37177633Sdfr#define _IP_VHL 38177633Sdfr 39177633Sdfr#include "opt_ipfw.h" 40177633Sdfr#include "opt_ipdn.h" 41177633Sdfr#include "opt_ipdivert.h" 42177633Sdfr#include "opt_ipfilter.h" 43177633Sdfr#include "opt_ipsec.h" 44177633Sdfr#include "opt_mac.h" 45177633Sdfr#include "opt_pfil_hooks.h" 46177633Sdfr#include "opt_random_ip_id.h" 47177633Sdfr 48180025Sdfr#include <sys/param.h> 49177633Sdfr#include <sys/systm.h> 50177633Sdfr#include <sys/kernel.h> 51177633Sdfr#include <sys/mac.h> 52177633Sdfr#include <sys/malloc.h> 53177633Sdfr#include <sys/mbuf.h> 54177633Sdfr#include <sys/protosw.h> 55177633Sdfr#include <sys/socket.h> 56177633Sdfr#include <sys/socketvar.h> 57177633Sdfr 58177633Sdfr#include <net/if.h> 59177633Sdfr#include <net/route.h> 60196503Szec 61196503Szec#include <netinet/in.h> 62177633Sdfr#include <netinet/in_systm.h> 63177685Sdfr#include <netinet/ip.h> 64177633Sdfr#include <netinet/in_pcb.h> 65177633Sdfr#include <netinet/in_var.h> 66177633Sdfr#include <netinet/ip_var.h> 67177633Sdfr 68177633Sdfr#include <machine/in_cksum.h> 69177633Sdfr 70177633Sdfrstatic MALLOC_DEFINE(M_IPMOPTS, "ip_moptions", "internet multicast options"); 71177633Sdfr 72177633Sdfr#ifdef IPSEC 73177633Sdfr#include <netinet6/ipsec.h> 74177633Sdfr#include <netkey/key.h> 75177633Sdfr#ifdef IPSEC_DEBUG 76180025Sdfr#include <netkey/key_debug.h> 77184588Sdfr#else 78177633Sdfr#define KEYDEBUG(lev,arg) 79177633Sdfr#endif 80177633Sdfr#endif /*IPSEC*/ 81177633Sdfr 82184588Sdfr#include <netinet/ip_fw.h> 83177633Sdfr#include <netinet/ip_dummynet.h> 84193272Sjhb 85177633Sdfr#define print_ip(x, a, y) printf("%s %d.%d.%d.%d%s",\ 86177633Sdfr x, (ntohl(a.s_addr)>>24)&0xFF,\ 87177633Sdfr (ntohl(a.s_addr)>>16)&0xFF,\ 88177633Sdfr (ntohl(a.s_addr)>>8)&0xFF,\ 89177633Sdfr (ntohl(a.s_addr))&0xFF, y); 90177633Sdfr 91184588Sdfru_short ip_id; 92177633Sdfr 93177633Sdfrstatic struct mbuf *ip_insertoptions(struct mbuf *, struct mbuf *, int *); 94177633Sdfrstatic struct ifnet *ip_multicast_if(struct in_addr *, int *); 95177633Sdfrstatic void ip_mloopback 96177633Sdfr (struct ifnet *, struct mbuf *, struct sockaddr_in *, int); 97177633Sdfrstatic int ip_getmoptions 98177633Sdfr (struct sockopt *, struct ip_moptions *); 99180025Sdfrstatic int ip_pcbopts(int, struct mbuf **, struct mbuf *); 100180025Sdfrstatic int ip_setmoptions 101180025Sdfr (struct sockopt *, struct ip_moptions **); 102177633Sdfr 103177633Sdfrint ip_optcopy(struct ip *, struct ip *); 104177633Sdfr 105180025Sdfr 106177633Sdfrextern struct protosw inetsw[]; 107177633Sdfr 108177633Sdfr/* 109184588Sdfr * IP output. The packet in mbuf chain m contains a skeletal IP 110177633Sdfr * header (with len, off, ttl, proto, tos, src, dst). 111177633Sdfr * The mbuf chain containing the packet will be freed. 112177633Sdfr * The mbuf opt, if present, will not be freed. 113177633Sdfr */ 114177633Sdfrint 115177633Sdfrip_output(m0, opt, ro, flags, imo, inp) 116177633Sdfr struct mbuf *m0; 117193272Sjhb struct mbuf *opt; 118177633Sdfr struct route *ro; 119177633Sdfr int flags; 120177633Sdfr struct ip_moptions *imo; 121177633Sdfr struct inpcb *inp; 122177633Sdfr{ 123177633Sdfr struct ip *ip, *mhip; 124177633Sdfr struct ifnet *ifp = NULL; /* keep compiler happy */ 125177633Sdfr struct mbuf *m; 126177633Sdfr int hlen = sizeof (struct ip); 127177633Sdfr int len, off, error = 0; 128193437Srmacklem struct sockaddr_in *dst = NULL; /* keep compiler happy */ 129177633Sdfr struct in_ifaddr *ia = NULL; 130177633Sdfr int isbroadcast, sw_csum; 131193437Srmacklem struct in_addr pkt_dst; 132193437Srmacklem#ifdef IPSEC 133177633Sdfr struct route iproute; 134177633Sdfr struct secpolicy *sp = NULL; 135177633Sdfr struct socket *so = inp ? inp->inp_socket : NULL; 136177633Sdfr#endif 137180025Sdfr struct ip_fw_args args; 138184588Sdfr int src_was_INADDR_ANY = 0; /* as the name says... */ 139184588Sdfr#ifdef PFIL_HOOKS 140177633Sdfr struct packet_filter_hook *pfh; 141177633Sdfr struct mbuf *m1; 142177633Sdfr int rv; 143177633Sdfr#endif /* PFIL_HOOKS */ 144177633Sdfr 145177633Sdfr args.eh = NULL; 146177633Sdfr args.rule = NULL; 147177633Sdfr args.next_hop = NULL; 148177633Sdfr args.divert_rule = 0; /* divert cookie */ 149177633Sdfr 150177633Sdfr /* Grab info from MT_TAG mbufs prepended to the chain. */ 151177633Sdfr for (; m0 && m0->m_type == MT_TAG; m0 = m0->m_next) { 152177633Sdfr switch(m0->_m_tag_id) { 153177633Sdfr default: 154177633Sdfr printf("ip_output: unrecognised MT_TAG tag %d\n", 155177633Sdfr m0->_m_tag_id); 156177633Sdfr break; 157184588Sdfr 158184588Sdfr case PACKET_TAG_DUMMYNET: 159184588Sdfr /* 160177633Sdfr * the packet was already tagged, so part of the 161177633Sdfr * processing was already done, and we need to go down. 162184588Sdfr * Get parameters from the header. 163184588Sdfr */ 164184588Sdfr args.rule = ((struct dn_pkt *)m0)->rule; 165177633Sdfr opt = NULL ; 166177633Sdfr ro = & ( ((struct dn_pkt *)m0)->ro ) ; 167177633Sdfr imo = NULL ; 168177633Sdfr dst = ((struct dn_pkt *)m0)->dn_dst ; 169177633Sdfr ifp = ((struct dn_pkt *)m0)->ifp ; 170177633Sdfr flags = ((struct dn_pkt *)m0)->flags ; 171177633Sdfr break; 172177633Sdfr 173177633Sdfr case PACKET_TAG_DIVERT: 174177633Sdfr args.divert_rule = (intptr_t)m0->m_data & 0xffff; 175177633Sdfr break; 176177633Sdfr 177177633Sdfr case PACKET_TAG_IPFORWARD: 178177633Sdfr args.next_hop = (struct sockaddr_in *)m0->m_data; 179177633Sdfr break; 180177633Sdfr } 181177633Sdfr } 182177633Sdfr m = m0; 183177633Sdfr 184177633Sdfr KASSERT(!m || (m->m_flags & M_PKTHDR) != 0, ("ip_output: no HDR")); 185177633Sdfr 186177633Sdfr if (args.rule != NULL) { /* dummynet already saw us */ 187177633Sdfr ip = mtod(m, struct ip *); 188177633Sdfr hlen = IP_VHL_HL(ip->ip_vhl) << 2 ; 189177633Sdfr if (ro->ro_rt) 190177633Sdfr ia = ifatoia(ro->ro_rt->rt_ifa); 191193272Sjhb goto sendit; 192177633Sdfr } 193177633Sdfr 194177633Sdfr if (opt) { 195177633Sdfr len = 0; 196213756Srmacklem m = ip_insertoptions(m, opt, &len); 197177633Sdfr if (len != 0) 198177633Sdfr hlen = len; 199177633Sdfr } 200177633Sdfr ip = mtod(m, struct ip *); 201177633Sdfr pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst; 202177633Sdfr 203177633Sdfr /* 204177633Sdfr * Fill in IP header. 205177633Sdfr */ 206177633Sdfr if ((flags & (IP_FORWARDING|IP_RAWOUTPUT)) == 0) { 207177633Sdfr ip->ip_vhl = IP_MAKE_VHL(IPVERSION, hlen >> 2); 208177633Sdfr ip->ip_off &= IP_DF; 209177633Sdfr#ifdef RANDOM_IP_ID 210177633Sdfr ip->ip_id = ip_randomid(); 211177633Sdfr#else 212177633Sdfr ip->ip_id = htons(ip_id++); 213177633Sdfr#endif 214177633Sdfr ipstat.ips_localout++; 215177633Sdfr } else { 216177633Sdfr hlen = IP_VHL_HL(ip->ip_vhl) << 2; 217177633Sdfr } 218177633Sdfr 219177633Sdfr dst = (struct sockaddr_in *)&ro->ro_dst; 220177633Sdfr /* 221177633Sdfr * If there is a cached route, 222177633Sdfr * check that it is to the same destination 223177633Sdfr * and is still up. If not, free it and try again. 224177633Sdfr * The address family should also be checked in case of sharing the 225177633Sdfr * cache with IPv6. 226177633Sdfr */ 227177633Sdfr if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 || 228180025Sdfr dst->sin_family != AF_INET || 229180025Sdfr dst->sin_addr.s_addr != pkt_dst.s_addr)) { 230184588Sdfr RTFREE(ro->ro_rt); 231177633Sdfr ro->ro_rt = (struct rtentry *)0; 232177633Sdfr } 233177633Sdfr if (ro->ro_rt == 0) { 234180025Sdfr bzero(dst, sizeof(*dst)); 235177633Sdfr dst->sin_family = AF_INET; 236177633Sdfr dst->sin_len = sizeof(*dst); 237177633Sdfr dst->sin_addr = pkt_dst; 238177633Sdfr } 239177633Sdfr /* 240177633Sdfr * If routing to interface only, 241177633Sdfr * short circuit routing lookup. 242177633Sdfr */ 243177633Sdfr if (flags & IP_ROUTETOIF) { 244177633Sdfr if ((ia = ifatoia(ifa_ifwithdstaddr(sintosa(dst)))) == 0 && 245184588Sdfr (ia = ifatoia(ifa_ifwithnet(sintosa(dst)))) == 0) { 246184588Sdfr ipstat.ips_noroute++; 247184588Sdfr error = ENETUNREACH; 248177633Sdfr goto bad; 249177633Sdfr } 250177633Sdfr ifp = ia->ia_ifp; 251177633Sdfr ip->ip_ttl = 1; 252177633Sdfr isbroadcast = in_broadcast(dst->sin_addr, ifp); 253177633Sdfr } else if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 254177633Sdfr imo != NULL && imo->imo_multicast_ifp != NULL) { 255177633Sdfr /* 256177633Sdfr * Bypass the normal routing lookup for multicast 257177633Sdfr * packets if the interface is specified. 258177633Sdfr */ 259201758Smbr ifp = imo->imo_multicast_ifp; 260177633Sdfr IFP_TO_IA(ifp, ia); 261177633Sdfr isbroadcast = 0; /* fool gcc */ 262177633Sdfr } else { 263177633Sdfr /* 264177633Sdfr * If this is the case, we probably don't want to allocate 265177633Sdfr * a protocol-cloned route since we didn't get one from the 266177633Sdfr * ULP. This lets TCP do its thing, while not burdening 267177633Sdfr * forwarding or ICMP with the overhead of cloning a route. 268213756Srmacklem * Of course, we still want to do any cloning requested by 269213756Srmacklem * the link layer, as this is probably required in all cases 270213756Srmacklem * for correct operation (as it is for ARP). 271213756Srmacklem */ 272213756Srmacklem if (ro->ro_rt == 0) 273213756Srmacklem rtalloc_ign(ro, RTF_PRCLONING); 274177633Sdfr if (ro->ro_rt == 0) { 275193272Sjhb ipstat.ips_noroute++; 276177633Sdfr error = EHOSTUNREACH; 277177633Sdfr goto bad; 278193272Sjhb } 279193272Sjhb ia = ifatoia(ro->ro_rt->rt_ifa); 280177633Sdfr ifp = ro->ro_rt->rt_ifp; 281177633Sdfr ro->ro_rt->rt_use++; 282177633Sdfr if (ro->ro_rt->rt_flags & RTF_GATEWAY) 283177633Sdfr dst = (struct sockaddr_in *)ro->ro_rt->rt_gateway; 284193272Sjhb if (ro->ro_rt->rt_flags & RTF_HOST) 285177633Sdfr isbroadcast = (ro->ro_rt->rt_flags & RTF_BROADCAST); 286177633Sdfr else 287177633Sdfr isbroadcast = in_broadcast(dst->sin_addr, ifp); 288177633Sdfr } 289177633Sdfr if (IN_MULTICAST(ntohl(pkt_dst.s_addr))) { 290177633Sdfr struct in_multi *inm; 291177633Sdfr 292177633Sdfr m->m_flags |= M_MCAST; 293193272Sjhb /* 294177633Sdfr * IP destination address is multicast. Make sure "dst" 295193272Sjhb * still points to the address in "ro". (It may have been 296193272Sjhb * changed to point to a gateway address, above.) 297177633Sdfr */ 298177633Sdfr dst = (struct sockaddr_in *)&ro->ro_dst; 299177633Sdfr /* 300177633Sdfr * See if the caller provided any multicast options 301177633Sdfr */ 302177633Sdfr if (imo != NULL) { 303177633Sdfr ip->ip_ttl = imo->imo_multicast_ttl; 304177633Sdfr if (imo->imo_multicast_vif != -1) 305193437Srmacklem ip->ip_src.s_addr = 306177633Sdfr ip_mcast_src(imo->imo_multicast_vif); 307193272Sjhb } else 308177633Sdfr ip->ip_ttl = IP_DEFAULT_MULTICAST_TTL; 309177633Sdfr /* 310177633Sdfr * Confirm that the outgoing interface supports multicast. 311180025Sdfr */ 312177633Sdfr if ((imo == NULL) || (imo->imo_multicast_vif == -1)) { 313177633Sdfr if ((ifp->if_flags & IFF_MULTICAST) == 0) { 314177633Sdfr ipstat.ips_noroute++; 315177633Sdfr error = ENETUNREACH; 316177633Sdfr goto bad; 317177633Sdfr } 318177633Sdfr } 319177633Sdfr /* 320177633Sdfr * If source address not specified yet, use address 321177633Sdfr * of outgoing interface. 322177633Sdfr */ 323177633Sdfr if (ip->ip_src.s_addr == INADDR_ANY) { 324177633Sdfr /* Interface may have no addresses. */ 325177633Sdfr if (ia != NULL) 326177633Sdfr ip->ip_src = IA_SIN(ia)->sin_addr; 327177633Sdfr } 328177633Sdfr 329180025Sdfr if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 330180025Sdfr /* 331177633Sdfr * XXX 332184588Sdfr * delayed checksums are not currently 333184588Sdfr * compatible with IP multicast routing 334177633Sdfr */ 335177633Sdfr if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 336177633Sdfr in_delayed_cksum(m); 337193272Sjhb m->m_pkthdr.csum_flags &= 338184588Sdfr ~CSUM_DELAY_DATA; 339180025Sdfr } 340184588Sdfr } 341184588Sdfr IN_LOOKUP_MULTI(pkt_dst, ifp, inm); 342177633Sdfr if (inm != NULL && 343177633Sdfr (imo == NULL || imo->imo_multicast_loop)) { 344177633Sdfr /* 345180025Sdfr * If we belong to the destination multicast group 346177633Sdfr * on the outgoing interface, and the caller did not 347180025Sdfr * forbid loopback, loop back a copy. 348180025Sdfr */ 349180025Sdfr ip_mloopback(ifp, m, dst, hlen); 350184588Sdfr } 351177633Sdfr else { 352177633Sdfr /* 353184588Sdfr * If we are acting as a multicast router, perform 354184588Sdfr * multicast forwarding as if the packet had just 355180025Sdfr * arrived on the interface to which we are about 356177633Sdfr * to send. The multicast forwarding function 357177633Sdfr * recursively calls this function, using the 358193272Sjhb * IP_FORWARDING flag to prevent infinite recursion. 359180025Sdfr * 360180025Sdfr * Multicasts that are looped back by ip_mloopback(), 361177633Sdfr * above, will be forwarded by the ip_input() routine, 362177633Sdfr * if necessary. 363184588Sdfr */ 364180025Sdfr if (ip_mrouter && (flags & IP_FORWARDING) == 0) { 365180025Sdfr /* 366180025Sdfr * Check if rsvp daemon is running. If not, don't 367180025Sdfr * set ip_moptions. This ensures that the packet 368180025Sdfr * is multicast and not just sent down one link 369177633Sdfr * as prescribed by rsvpd. 370184588Sdfr */ 371180025Sdfr if (!rsvp_on) 372184588Sdfr imo = NULL; 373184588Sdfr if (ip_mforward(ip, ifp, m, imo) != 0) { 374180025Sdfr m_freem(m); 375184588Sdfr goto done; 376184588Sdfr } 377180025Sdfr } 378180025Sdfr } 379180025Sdfr 380180025Sdfr /* 381180025Sdfr * Multicasts with a time-to-live of zero may be looped- 382177633Sdfr * back, above, but must not be transmitted on a network. 383180025Sdfr * Also, multicasts addressed to the loopback interface 384177633Sdfr * are not sent -- the above call to ip_mloopback() will 385180025Sdfr * loop back a copy if this host actually belongs to the 386177633Sdfr * destination group on the loopback interface. 387180025Sdfr */ 388180025Sdfr if (ip->ip_ttl == 0 || ifp->if_flags & IFF_LOOPBACK) { 389180025Sdfr m_freem(m); 390180025Sdfr goto done; 391177633Sdfr } 392177633Sdfr 393177633Sdfr goto sendit; 394177633Sdfr } 395177633Sdfr#ifndef notdef 396177633Sdfr /* 397177633Sdfr * If the source address is not specified yet, use the address 398184588Sdfr * of the outoing interface. In case, keep note we did that, so 399184588Sdfr * if the the firewall changes the next-hop causing the output 400177633Sdfr * interface to change, we can fix that. 401177633Sdfr */ 402177633Sdfr if (ip->ip_src.s_addr == INADDR_ANY) { 403177633Sdfr /* Interface may have no addresses. */ 404177633Sdfr if (ia != NULL) { 405177633Sdfr ip->ip_src = IA_SIN(ia)->sin_addr; 406177633Sdfr src_was_INADDR_ANY = 1; 407177633Sdfr } 408177633Sdfr } 409177633Sdfr#endif /* notdef */ 410177633Sdfr /* 411180025Sdfr * Verify that we have any chance at all of being able to queue 412180025Sdfr * the packet or packet fragments 413184588Sdfr */ 414184588Sdfr if ((ifp->if_snd.ifq_len + ip->ip_len / ifp->if_mtu + 1) >= 415184588Sdfr ifp->if_snd.ifq_maxlen) { 416184588Sdfr error = ENOBUFS; 417184588Sdfr ipstat.ips_odropped++; 418184588Sdfr goto bad; 419184588Sdfr } 420184588Sdfr 421184588Sdfr /* 422177633Sdfr * Look for broadcast address and 423180025Sdfr * verify user is allowed to send 424177633Sdfr * such a packet. 425177633Sdfr */ 426177633Sdfr if (isbroadcast) { 427177633Sdfr if ((ifp->if_flags & IFF_BROADCAST) == 0) { 428177633Sdfr error = EADDRNOTAVAIL; 429177633Sdfr goto bad; 430177633Sdfr } 431177633Sdfr if ((flags & IP_ALLOWBROADCAST) == 0) { 432177633Sdfr error = EACCES; 433177633Sdfr goto bad; 434248195Sglebius } 435184588Sdfr /* don't allow broadcast messages to be fragmented */ 436184588Sdfr if ((u_short)ip->ip_len > ifp->if_mtu) { 437184588Sdfr error = EMSGSIZE; 438177633Sdfr goto bad; 439177633Sdfr } 440177633Sdfr m->m_flags |= M_BCAST; 441177633Sdfr } else { 442177633Sdfr m->m_flags &= ~M_BCAST; 443177633Sdfr } 444177633Sdfr 445177633Sdfrsendit: 446184588Sdfr#ifdef IPSEC 447177633Sdfr /* get SP for this packet */ 448177633Sdfr if (so == NULL) 449177633Sdfr sp = ipsec4_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, flags, &error); 450184588Sdfr else 451184588Sdfr sp = ipsec4_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error); 452184588Sdfr 453177633Sdfr if (sp == NULL) { 454177633Sdfr ipsecstat.out_inval++; 455177633Sdfr goto bad; 456184588Sdfr } 457177633Sdfr 458180025Sdfr error = 0; 459177633Sdfr 460184588Sdfr /* check policy */ 461184588Sdfr switch (sp->policy) { 462184588Sdfr case IPSEC_POLICY_DISCARD: 463184588Sdfr /* 464184588Sdfr * This packet is just discarded. 465184588Sdfr */ 466184588Sdfr ipsecstat.out_polvio++; 467184588Sdfr goto bad; 468184588Sdfr 469184588Sdfr case IPSEC_POLICY_BYPASS: 470225234Sart case IPSEC_POLICY_NONE: 471225234Sart /* no need to do IPsec. */ 472225234Sart goto skip_ipsec; 473225234Sart 474184588Sdfr case IPSEC_POLICY_IPSEC: 475184588Sdfr if (sp->req == NULL) { 476184588Sdfr /* acquire a policy */ 477184588Sdfr error = key_spdacquire(sp); 478184588Sdfr goto bad; 479180025Sdfr } 480177633Sdfr break; 481177633Sdfr 482177633Sdfr case IPSEC_POLICY_ENTRUST: 483177633Sdfr default: 484177633Sdfr printf("ip_output: Invalid policy found. %d\n", sp->policy); 485177633Sdfr } 486177633Sdfr { 487177633Sdfr struct ipsec_output_state state; 488177633Sdfr bzero(&state, sizeof(state)); 489177633Sdfr state.m = m; 490177633Sdfr if (flags & IP_ROUTETOIF) { 491177633Sdfr state.ro = &iproute; 492177633Sdfr bzero(&iproute, sizeof(iproute)); 493184588Sdfr } else 494184588Sdfr state.ro = ro; 495184588Sdfr state.dst = (struct sockaddr *)dst; 496184588Sdfr 497184588Sdfr ip->ip_sum = 0; 498177633Sdfr 499177633Sdfr /* 500177633Sdfr * XXX 501180025Sdfr * delayed checksums are not currently compatible with IPsec 502184588Sdfr */ 503184588Sdfr if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 504184588Sdfr in_delayed_cksum(m); 505184588Sdfr m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 506184588Sdfr } 507184588Sdfr 508184588Sdfr ip->ip_len = htons(ip->ip_len); 509177633Sdfr ip->ip_off = htons(ip->ip_off); 510177633Sdfr 511177633Sdfr error = ipsec4_output(&state, sp, flags); 512177633Sdfr 513177633Sdfr m = state.m; 514180025Sdfr if (flags & IP_ROUTETOIF) { 515177633Sdfr /* 516180025Sdfr * if we have tunnel mode SA, we may need to ignore 517180025Sdfr * IP_ROUTETOIF. 518184588Sdfr */ 519184588Sdfr if (state.ro != &iproute || state.ro->ro_rt != NULL) { 520184588Sdfr flags &= ~IP_ROUTETOIF; 521184588Sdfr ro = state.ro; 522184588Sdfr } 523184588Sdfr } else 524184588Sdfr ro = state.ro; 525177633Sdfr dst = (struct sockaddr_in *)state.dst; 526177633Sdfr if (error) { 527180025Sdfr /* mbuf is already reclaimed in ipsec4_output. */ 528180025Sdfr m0 = NULL; 529184588Sdfr switch (error) { 530184588Sdfr case EHOSTUNREACH: 531184588Sdfr case ENETUNREACH: 532184588Sdfr case EMSGSIZE: 533184588Sdfr case ENOBUFS: 534177633Sdfr case ENOMEM: 535177633Sdfr break; 536177633Sdfr default: 537177633Sdfr printf("ip4_output (ipsec): error code %d\n", error); 538177633Sdfr /*fall through*/ 539177633Sdfr case ENOENT: 540180025Sdfr /* don't show these error codes to the user */ 541180025Sdfr error = 0; 542184588Sdfr break; 543184588Sdfr } 544184588Sdfr goto bad; 545184588Sdfr } 546184588Sdfr } 547184588Sdfr 548177633Sdfr /* be sure to update variables that are affected by ipsec4_output() */ 549177633Sdfr ip = mtod(m, struct ip *); 550177633Sdfr#ifdef _IP_VHL 551177633Sdfr hlen = IP_VHL_HL(ip->ip_vhl) << 2; 552177633Sdfr#else 553177633Sdfr hlen = ip->ip_hl << 2; 554180025Sdfr#endif 555177633Sdfr if (ro->ro_rt == NULL) { 556180025Sdfr if ((flags & IP_ROUTETOIF) == 0) { 557180025Sdfr printf("ip_output: " 558180025Sdfr "can't update route after IPsec processing\n"); 559180025Sdfr error = EHOSTUNREACH; /*XXX*/ 560180025Sdfr goto bad; 561195245Srmacklem } 562180025Sdfr } else { 563195245Srmacklem ia = ifatoia(ro->ro_rt->rt_ifa); 564195245Srmacklem ifp = ro->ro_rt->rt_ifp; 565180025Sdfr } 566180025Sdfr 567195245Srmacklem /* make it flipped, again. */ 568177633Sdfr ip->ip_len = ntohs(ip->ip_len); 569180025Sdfr ip->ip_off = ntohs(ip->ip_off); 570177633Sdfrskip_ipsec: 571177633Sdfr#endif /*IPSEC*/ 572180025Sdfr 573184588Sdfr /* 574184588Sdfr * IpHack's section. 575184588Sdfr * - Xlate: translate packet's addr/port (NAT). 576184588Sdfr * - Firewall: deny/allow/etc. 577184588Sdfr * - Wrap: fake packet's addr/port <unimpl.> 578177633Sdfr * - Encapsulate: put it in another IP and send out. <unimp.> 579177633Sdfr */ 580177633Sdfr#ifdef PFIL_HOOKS 581177633Sdfr /* 582177633Sdfr * Run through list of hooks for output packets. 583177633Sdfr */ 584177633Sdfr m1 = m; 585180025Sdfr pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IP]].pr_pfh); 586184588Sdfr for (; pfh; pfh = TAILQ_NEXT(pfh, pfil_link)) 587184588Sdfr if (pfh->pfil_func) { 588177633Sdfr rv = pfh->pfil_func(ip, hlen, ifp, 1, &m1); 589177633Sdfr if (rv) { 590184588Sdfr error = EHOSTUNREACH; 591184588Sdfr goto done; 592184588Sdfr } 593184588Sdfr m = m1; 594184588Sdfr if (m == NULL) 595184588Sdfr goto done; 596184588Sdfr ip = mtod(m, struct ip *); 597184588Sdfr } 598184588Sdfr#endif /* PFIL_HOOKS */ 599184588Sdfr 600184588Sdfr /* 601184588Sdfr * Check with the firewall... 602184588Sdfr * but not if we are already being fwd'd from a firewall. 603184588Sdfr */ 604184588Sdfr if (fw_enable && IPFW_LOADED && !args.next_hop) { 605184588Sdfr struct sockaddr_in *old = dst; 606184588Sdfr 607184588Sdfr args.m = m; 608184588Sdfr args.next_hop = dst; 609184588Sdfr args.oif = ifp; 610184588Sdfr off = ip_fw_chk_ptr(&args); 611184588Sdfr m = args.m; 612184588Sdfr dst = args.next_hop; 613184588Sdfr 614184588Sdfr /* 615184588Sdfr * On return we must do the following: 616184588Sdfr * m == NULL -> drop the pkt (old interface, deprecated) 617184588Sdfr * (off & IP_FW_PORT_DENY_FLAG) -> drop the pkt (new interface) 618184588Sdfr * 1<=off<= 0xffff -> DIVERT 619184588Sdfr * (off & IP_FW_PORT_DYNT_FLAG) -> send to a DUMMYNET pipe 620184588Sdfr * (off & IP_FW_PORT_TEE_FLAG) -> TEE the packet 621184588Sdfr * dst != old -> IPFIREWALL_FORWARD 622184588Sdfr * off==0, dst==old -> accept 623184588Sdfr * If some of the above modules are not compiled in, then 624184588Sdfr * we should't have to check the corresponding condition 625184588Sdfr * (because the ipfw control socket should not accept 626184588Sdfr * unsupported rules), but better play safe and drop 627184588Sdfr * packets in case of doubt. 628184588Sdfr */ 629184588Sdfr if ( (off & IP_FW_PORT_DENY_FLAG) || m == NULL) { 630184588Sdfr if (m) 631184588Sdfr m_freem(m); 632177633Sdfr error = EACCES; 633177633Sdfr goto done; 634177633Sdfr } 635177633Sdfr ip = mtod(m, struct ip *); 636177633Sdfr if (off == 0 && dst == old) /* common case */ 637177633Sdfr goto pass; 638177633Sdfr if (DUMMYNET_LOADED && (off & IP_FW_PORT_DYNT_FLAG) != 0) { 639177633Sdfr /* 640177633Sdfr * pass the pkt to dummynet. Need to include 641184588Sdfr * pipe number, m, ifp, ro, dst because these are 642225234Sart * not recomputed in the next pass. 643184588Sdfr * All other parameters have been already used and 644177633Sdfr * so they are not needed anymore. 645184588Sdfr * XXX note: if the ifp or ro entry are deleted 646177633Sdfr * while a pkt is in dummynet, we are in trouble! 647177633Sdfr */ 648177633Sdfr args.ro = ro; 649180025Sdfr args.dst = dst; 650177633Sdfr args.flags = flags; 651177633Sdfr 652180025Sdfr error = ip_dn_io_ptr(m, off & 0xffff, DN_TO_IP_OUT, 653184588Sdfr &args); 654184588Sdfr goto done; 655177633Sdfr } 656177633Sdfr#ifdef IPDIVERT 657177633Sdfr if (off != 0 && (off & IP_FW_PORT_DYNT_FLAG) == 0) { 658177633Sdfr struct mbuf *clone = NULL; 659180025Sdfr 660184588Sdfr /* Clone packet if we're doing a 'tee' */ 661184588Sdfr if ((off & IP_FW_PORT_TEE_FLAG) != 0) 662184588Sdfr clone = m_dup(m, M_DONTWAIT); 663180025Sdfr 664180025Sdfr /* 665180025Sdfr * XXX 666180025Sdfr * delayed checksums are not currently compatible 667180025Sdfr * with divert sockets. 668180025Sdfr */ 669180025Sdfr if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 670180025Sdfr in_delayed_cksum(m); 671180025Sdfr m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 672180025Sdfr } 673184588Sdfr 674184588Sdfr /* Restore packet header fields to original values */ 675184588Sdfr ip->ip_len = htons(ip->ip_len); 676180025Sdfr ip->ip_off = htons(ip->ip_off); 677180025Sdfr 678180025Sdfr /* Deliver packet to divert input routine */ 679177633Sdfr divert_packet(m, 0, off & 0xffff, args.divert_rule); 680180025Sdfr 681180025Sdfr /* If 'tee', continue with original packet */ 682180025Sdfr if (clone != NULL) { 683177633Sdfr m = clone; 684177633Sdfr ip = mtod(m, struct ip *); 685255284Srmacklem goto pass; 686180025Sdfr } 687177633Sdfr goto done; 688177633Sdfr } 689177633Sdfr#endif 690177633Sdfr 691177633Sdfr /* IPFIREWALL_FORWARD */ 692177633Sdfr /* 693177633Sdfr * Check dst to make sure it is directly reachable on the 694177633Sdfr * interface we previously thought it was. 695177633Sdfr * If it isn't (which may be likely in some situations) we have 696180025Sdfr * to re-route it (ie, find a route for the next-hop and the 697180025Sdfr * associated interface) and set them here. This is nested 698180025Sdfr * forwarding which in most cases is undesirable, except where 699180025Sdfr * such control is nigh impossible. So we do it here. 700177633Sdfr * And I'm babbling. 701180025Sdfr */ 702177633Sdfr if (off == 0 && old != dst) { /* FORWARD, dst has changed */ 703177633Sdfr#if 0 704177633Sdfr /* 705184588Sdfr * XXX To improve readability, this block should be 706184588Sdfr * changed into a function call as below: 707177633Sdfr */ 708184588Sdfr error = ip_ipforward(&m, &dst, &ifp); 709177633Sdfr if (error) 710184588Sdfr goto bad; 711184588Sdfr if (m == NULL) /* ip_input consumed the mbuf */ 712184588Sdfr goto done; 713184588Sdfr#else 714184588Sdfr struct in_ifaddr *ia; 715184588Sdfr 716184588Sdfr /* 717184588Sdfr * XXX sro_fwd below is static, and a pointer 718184588Sdfr * to it gets passed to routines downstream. 719184588Sdfr * This could have surprisingly bad results in 720184588Sdfr * practice, because its content is overwritten 721184588Sdfr * by subsequent packets. 722184588Sdfr */ 723184588Sdfr /* There must be a better way to do this next line... */ 724184588Sdfr static struct route sro_fwd; 725184588Sdfr struct route *ro_fwd = &sro_fwd; 726184588Sdfr 727184588Sdfr#if 0 728184588Sdfr print_ip("IPFIREWALL_FORWARD: New dst ip: ", 729184588Sdfr dst->sin_addr, "\n"); 730184588Sdfr#endif 731184588Sdfr 732184588Sdfr /* 733184588Sdfr * We need to figure out if we have been forwarded 734184588Sdfr * to a local socket. If so, then we should somehow 735184588Sdfr * "loop back" to ip_input, and get directed to the 736184588Sdfr * PCB as if we had received this packet. This is 737255284Srmacklem * because it may be dificult to identify the packets 738184588Sdfr * you want to forward until they are being output 739184588Sdfr * and have selected an interface. (e.g. locally 740184588Sdfr * initiated packets) If we used the loopback inteface, 741184588Sdfr * we would not be able to control what happens 742184588Sdfr * as the packet runs through ip_input() as 743184588Sdfr * it is done through a ISR. 744184588Sdfr */ 745177633Sdfr LIST_FOREACH(ia, 746177633Sdfr INADDR_HASH(dst->sin_addr.s_addr), ia_hash) { 747177633Sdfr /* 748177633Sdfr * If the addr to forward to is one 749177633Sdfr * of ours, we pretend to 750177633Sdfr * be the destination for this packet. 751184588Sdfr */ 752177633Sdfr if (IA_SIN(ia)->sin_addr.s_addr == 753177633Sdfr dst->sin_addr.s_addr) 754184588Sdfr break; 755177633Sdfr } 756184588Sdfr if (ia) { /* tell ip_input "dont filter" */ 757184588Sdfr struct m_hdr tag; 758177633Sdfr 759177633Sdfr tag.mh_type = MT_TAG; 760177633Sdfr tag.mh_flags = PACKET_TAG_IPFORWARD; 761177633Sdfr tag.mh_data = (caddr_t)args.next_hop; 762177633Sdfr tag.mh_next = m; 763184588Sdfr 764177633Sdfr if (m->m_pkthdr.rcvif == NULL) 765177633Sdfr m->m_pkthdr.rcvif = ifunit("lo0"); 766184588Sdfr if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 767184588Sdfr m->m_pkthdr.csum_flags |= 768177633Sdfr CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 769177633Sdfr m0->m_pkthdr.csum_data = 0xffff; 770177633Sdfr } 771177633Sdfr m->m_pkthdr.csum_flags |= 772177633Sdfr CSUM_IP_CHECKED | CSUM_IP_VALID; 773180025Sdfr ip->ip_len = htons(ip->ip_len); 774180025Sdfr ip->ip_off = htons(ip->ip_off); 775177633Sdfr ip_input((struct mbuf *)&tag); 776180025Sdfr goto done; 777180025Sdfr } 778180025Sdfr /* Some of the logic for this was 779180025Sdfr * nicked from above. 780177633Sdfr * 781180025Sdfr * This rewrites the cached route in a local PCB. 782184588Sdfr * Is this what we want to do? 783184588Sdfr */ 784184588Sdfr bcopy(dst, &ro_fwd->ro_dst, sizeof(*dst)); 785180025Sdfr 786180025Sdfr ro_fwd->ro_rt = 0; 787184588Sdfr rtalloc_ign(ro_fwd, RTF_PRCLONING); 788177633Sdfr 789177633Sdfr if (ro_fwd->ro_rt == 0) { 790177633Sdfr ipstat.ips_noroute++; 791177633Sdfr error = EHOSTUNREACH; 792177633Sdfr goto bad; 793177633Sdfr } 794177633Sdfr 795177633Sdfr ia = ifatoia(ro_fwd->ro_rt->rt_ifa); 796177633Sdfr ifp = ro_fwd->ro_rt->rt_ifp; 797177633Sdfr ro_fwd->ro_rt->rt_use++; 798177633Sdfr if (ro_fwd->ro_rt->rt_flags & RTF_GATEWAY) 799177633Sdfr dst = (struct sockaddr_in *) 800177633Sdfr ro_fwd->ro_rt->rt_gateway; 801177633Sdfr if (ro_fwd->ro_rt->rt_flags & RTF_HOST) 802177633Sdfr isbroadcast = 803177633Sdfr (ro_fwd->ro_rt->rt_flags & RTF_BROADCAST); 804177633Sdfr else 805177633Sdfr isbroadcast = in_broadcast(dst->sin_addr, ifp); 806177633Sdfr if (ro->ro_rt) 807177633Sdfr RTFREE(ro->ro_rt); 808177633Sdfr ro->ro_rt = ro_fwd->ro_rt; 809177633Sdfr dst = (struct sockaddr_in *)&ro_fwd->ro_dst; 810177633Sdfr 811177633Sdfr#endif /* ... block to be put into a function */ 812177633Sdfr /* 813177633Sdfr * If we added a default src ip earlier, 814177633Sdfr * which would have been gotten from the-then 815177633Sdfr * interface, do it again, from the new one. 816177633Sdfr */ 817177633Sdfr if (src_was_INADDR_ANY) 818177633Sdfr ip->ip_src = IA_SIN(ia)->sin_addr; 819177633Sdfr goto pass ; 820193272Sjhb } 821177633Sdfr 822177633Sdfr /* 823193272Sjhb * if we get here, none of the above matches, and 824177633Sdfr * we have to drop the pkt 825177633Sdfr */ 826177633Sdfr m_freem(m); 827177633Sdfr error = EACCES; /* not sure this is the right error msg */ 828177633Sdfr goto done; 829177633Sdfr } 830177633Sdfr 831177633Sdfrpass: 832177633Sdfr /* 127/8 must not appear on wire - RFC1122. */ 833177633Sdfr if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 834177633Sdfr (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 835177633Sdfr if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 836177633Sdfr ipstat.ips_badaddr++; 837177633Sdfr error = EADDRNOTAVAIL; 838177633Sdfr goto bad; 839177633Sdfr } 840177633Sdfr } 841177633Sdfr 842177633Sdfr m->m_pkthdr.csum_flags |= CSUM_IP; 843177633Sdfr sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_hwassist; 844177633Sdfr if (sw_csum & CSUM_DELAY_DATA) { 845177633Sdfr in_delayed_cksum(m); 846177633Sdfr sw_csum &= ~CSUM_DELAY_DATA; 847177633Sdfr } 848177633Sdfr m->m_pkthdr.csum_flags &= ifp->if_hwassist; 849177633Sdfr 850177633Sdfr /* 851177633Sdfr * If small enough for interface, or the interface will take 852177633Sdfr * care of the fragmentation for us, can just send directly. 853177633Sdfr */ 854177633Sdfr if ((u_short)ip->ip_len <= ifp->if_mtu || 855177633Sdfr ifp->if_hwassist & CSUM_FRAGMENT) { 856177633Sdfr ip->ip_len = htons(ip->ip_len); 857177633Sdfr ip->ip_off = htons(ip->ip_off); 858177633Sdfr ip->ip_sum = 0; 859177633Sdfr if (sw_csum & CSUM_DELAY_IP) { 860177633Sdfr if (ip->ip_vhl == IP_VHL_BORING) { 861177633Sdfr ip->ip_sum = in_cksum_hdr(ip); 862177633Sdfr } else { 863177633Sdfr ip->ip_sum = in_cksum(m, hlen); 864177633Sdfr } 865177633Sdfr } 866177633Sdfr 867177633Sdfr /* Record statistics for this interface address. */ 868177633Sdfr if (!(flags & IP_FORWARDING) && ia) { 869177633Sdfr ia->ia_ifa.if_opackets++; 870177633Sdfr ia->ia_ifa.if_obytes += m->m_pkthdr.len; 871177633Sdfr } 872177633Sdfr 873177633Sdfr#ifdef IPSEC 874177633Sdfr /* clean ipsec history once it goes out of the node */ 875177633Sdfr ipsec_delaux(m); 876177633Sdfr#endif 877177633Sdfr 878177633Sdfr error = (*ifp->if_output)(ifp, m, 879177633Sdfr (struct sockaddr *)dst, ro->ro_rt); 880177633Sdfr goto done; 881177633Sdfr } 882177633Sdfr /* 883177633Sdfr * Too large for interface; fragment if possible. 884177633Sdfr * Must be able to put at least 8 bytes per fragment. 885177633Sdfr */ 886177633Sdfr if (ip->ip_off & IP_DF) { 887177633Sdfr error = EMSGSIZE; 888177633Sdfr /* 889177633Sdfr * This case can happen if the user changed the MTU 890177633Sdfr * of an interface after enabling IP on it. Because 891177633Sdfr * most netifs don't keep track of routes pointing to 892177633Sdfr * them, there is no way for one to update all its 893177633Sdfr * routes when the MTU is changed. 894177633Sdfr */ 895177633Sdfr if ((ro->ro_rt->rt_flags & (RTF_UP | RTF_HOST)) 896177633Sdfr && !(ro->ro_rt->rt_rmx.rmx_locks & RTV_MTU) 897177633Sdfr && (ro->ro_rt->rt_rmx.rmx_mtu > ifp->if_mtu)) { 898177633Sdfr ro->ro_rt->rt_rmx.rmx_mtu = ifp->if_mtu; 899177633Sdfr } 900177633Sdfr ipstat.ips_cantfrag++; 901177633Sdfr goto bad; 902177633Sdfr } 903177633Sdfr len = (ifp->if_mtu - hlen) &~ 7; 904177633Sdfr if (len < 8) { 905177633Sdfr error = EMSGSIZE; 906177633Sdfr goto bad; 907177633Sdfr } 908177633Sdfr 909177633Sdfr /* 910177633Sdfr * if the interface will not calculate checksums on 911177633Sdfr * fragmented packets, then do it here. 912177633Sdfr */ 913177633Sdfr if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA && 914177633Sdfr (ifp->if_hwassist & CSUM_IP_FRAGS) == 0) { 915177633Sdfr in_delayed_cksum(m); 916177633Sdfr m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 917177633Sdfr } 918177633Sdfr 919177633Sdfr if (len > PAGE_SIZE) { 920177633Sdfr /* 921177633Sdfr * Fragement large datagrams such that each segment 922177633Sdfr * contains a multiple of PAGE_SIZE amount of data, 923177633Sdfr * plus headers. This enables a receiver to perform 924184588Sdfr * page-flipping zero-copy optimizations. 925177633Sdfr */ 926177633Sdfr 927177633Sdfr int newlen; 928177633Sdfr struct mbuf *mtmp; 929177633Sdfr 930177633Sdfr for (mtmp = m, off = 0; 931177633Sdfr mtmp && ((off + mtmp->m_len) <= ifp->if_mtu); 932177633Sdfr mtmp = mtmp->m_next) { 933177633Sdfr off += mtmp->m_len; 934177633Sdfr } 935177633Sdfr /* 936177633Sdfr * firstlen (off - hlen) must be aligned on an 937177633Sdfr * 8-byte boundary 938177633Sdfr */ 939177633Sdfr if (off < hlen) 940177633Sdfr goto smart_frag_failure; 941177633Sdfr off = ((off - hlen) & ~7) + hlen; 942177633Sdfr newlen = (~PAGE_MASK) & ifp->if_mtu; 943177633Sdfr if ((newlen + sizeof (struct ip)) > ifp->if_mtu) { 944177633Sdfr /* we failed, go back the default */ 945177633Sdfrsmart_frag_failure: 946177633Sdfr newlen = len; 947177633Sdfr off = hlen + len; 948177633Sdfr } 949177633Sdfr 950184588Sdfr/* printf("ipfrag: len = %d, hlen = %d, mhlen = %d, newlen = %d, off = %d\n", 951177633Sdfr len, hlen, sizeof (struct ip), newlen, off);*/ 952177633Sdfr 953193272Sjhb len = newlen; 954180025Sdfr 955177633Sdfr } else { 956193272Sjhb off = hlen + len; 957180025Sdfr } 958177633Sdfr 959184588Sdfr 960184588Sdfr 961184588Sdfr { 962184588Sdfr int mhlen, firstlen = off - hlen; 963184588Sdfr struct mbuf **mnext = &m->m_nextpkt; 964184588Sdfr int nfrags = 1; 965184588Sdfr 966184588Sdfr /* 967184588Sdfr * Loop through length of segment after first fragment, 968184588Sdfr * make new header and copy data of each part and link onto chain. 969184588Sdfr */ 970184588Sdfr m0 = m; 971184588Sdfr mhlen = sizeof (struct ip); 972180025Sdfr for (; off < (u_short)ip->ip_len; off += len) { 973180025Sdfr MGETHDR(m, M_DONTWAIT, MT_HEADER); 974180025Sdfr if (m == 0) { 975180025Sdfr error = ENOBUFS; 976180025Sdfr ipstat.ips_odropped++; 977180025Sdfr goto sendorfree; 978180025Sdfr } 979180025Sdfr m->m_flags |= (m0->m_flags & M_MCAST) | M_FRAG; 980180025Sdfr m->m_data += max_linkhdr; 981180025Sdfr mhip = mtod(m, struct ip *); 982180025Sdfr *mhip = *ip; 983180025Sdfr if (hlen > sizeof (struct ip)) { 984180025Sdfr mhlen = ip_optcopy(ip, mhip) + sizeof (struct ip); 985180025Sdfr mhip->ip_vhl = IP_MAKE_VHL(IPVERSION, mhlen >> 2); 986180025Sdfr } 987180025Sdfr m->m_len = mhlen; 988184588Sdfr mhip->ip_off = ((off - hlen) >> 3) + ip->ip_off; 989184588Sdfr if (off + len >= (u_short)ip->ip_len) 990184588Sdfr len = (u_short)ip->ip_len - off; 991184588Sdfr else 992184588Sdfr mhip->ip_off |= IP_MF; 993184588Sdfr mhip->ip_len = htons((u_short)(len + mhlen)); 994184588Sdfr m->m_next = m_copy(m0, off, len); 995184588Sdfr if (m->m_next == 0) { 996184588Sdfr (void) m_free(m); 997184588Sdfr error = ENOBUFS; /* ??? */ 998184588Sdfr ipstat.ips_odropped++; 999193272Sjhb goto sendorfree; 1000184588Sdfr } 1001184588Sdfr m->m_pkthdr.len = mhlen + len; 1002184588Sdfr m->m_pkthdr.rcvif = (struct ifnet *)0; 1003193272Sjhb#ifdef MAC 1004184588Sdfr mac_create_fragment(m0, m); 1005184588Sdfr#endif 1006227059Srmacklem m->m_pkthdr.csum_flags = m0->m_pkthdr.csum_flags; 1007184588Sdfr mhip->ip_off = htons(mhip->ip_off); 1008184588Sdfr mhip->ip_sum = 0; 1009177633Sdfr if (sw_csum & CSUM_DELAY_IP) { 1010177633Sdfr if (mhip->ip_vhl == IP_VHL_BORING) { 1011193437Srmacklem mhip->ip_sum = in_cksum_hdr(mhip); 1012193272Sjhb } else { 1013193437Srmacklem mhip->ip_sum = in_cksum(m, mhlen); 1014177633Sdfr } 1015193437Srmacklem } 1016177633Sdfr *mnext = m; 1017177633Sdfr mnext = &m->m_nextpkt; 1018177633Sdfr nfrags++; 1019177633Sdfr } 1020227059Srmacklem ipstat.ips_ofragments += nfrags; 1021177633Sdfr 1022177633Sdfr /* set first/last markers for fragment chain */ 1023177633Sdfr m->m_flags |= M_LASTFRAG; 1024180025Sdfr m0->m_flags |= M_FIRSTFRAG | M_FRAG; 1025177633Sdfr m0->m_pkthdr.csum_data = nfrags; 1026177633Sdfr 1027177633Sdfr /* 1028177633Sdfr * Update first fragment by trimming what's been copied out 1029177633Sdfr * and updating header, then send each fragment (in order). 1030177633Sdfr */ 1031177633Sdfr m = m0; 1032177633Sdfr m_adj(m, hlen + firstlen - (u_short)ip->ip_len); 1033177633Sdfr m->m_pkthdr.len = hlen + firstlen; 1034177633Sdfr ip->ip_len = htons((u_short)m->m_pkthdr.len); 1035177633Sdfr ip->ip_off |= IP_MF; 1036177633Sdfr ip->ip_off = htons(ip->ip_off); 1037177633Sdfr ip->ip_sum = 0; 1038177633Sdfr if (sw_csum & CSUM_DELAY_IP) { 1039177633Sdfr if (ip->ip_vhl == IP_VHL_BORING) { 1040177633Sdfr ip->ip_sum = in_cksum_hdr(ip); 1041177633Sdfr } else { 1042177633Sdfr ip->ip_sum = in_cksum(m, hlen); 1043177633Sdfr } 1044177633Sdfr } 1045177633Sdfrsendorfree: 1046177633Sdfr for (m = m0; m; m = m0) { 1047177633Sdfr m0 = m->m_nextpkt; 1048177633Sdfr m->m_nextpkt = 0; 1049177633Sdfr#ifdef IPSEC 1050193272Sjhb /* clean ipsec history once it goes out of the node */ 1051177633Sdfr ipsec_delaux(m); 1052177633Sdfr#endif 1053177633Sdfr if (error == 0) { 1054177633Sdfr /* Record statistics for this interface address. */ 1055177633Sdfr if (ia != NULL) { 1056177633Sdfr ia->ia_ifa.if_opackets++; 1057177633Sdfr ia->ia_ifa.if_obytes += m->m_pkthdr.len; 1058177633Sdfr } 1059177633Sdfr 1060177633Sdfr error = (*ifp->if_output)(ifp, m, 1061193437Srmacklem (struct sockaddr *)dst, ro->ro_rt); 1062177633Sdfr } else 1063177633Sdfr m_freem(m); 1064177633Sdfr } 1065193272Sjhb 1066177633Sdfr if (error == 0) 1067177633Sdfr ipstat.ips_fragmented++; 1068177633Sdfr } 1069177633Sdfrdone: 1070177633Sdfr#ifdef IPSEC 1071177633Sdfr if (ro == &iproute && ro->ro_rt) { 1072193272Sjhb RTFREE(ro->ro_rt); 1073177633Sdfr ro->ro_rt = NULL; 1074177633Sdfr } 1075177633Sdfr if (sp != NULL) { 1076177633Sdfr KEYDEBUG(KEYDEBUG_IPSEC_STAMP, 1077177633Sdfr printf("DP ip_output call free SP:%p\n", sp)); 1078177633Sdfr key_freesp(sp); 1079177633Sdfr } 1080177633Sdfr#endif /* IPSEC */ 1081177633Sdfr return (error); 1082177633Sdfrbad: 1083177633Sdfr m_freem(m); 1084180025Sdfr goto done; 1085177633Sdfr} 1086177633Sdfr 1087177633Sdfrvoid 1088177633Sdfrin_delayed_cksum(struct mbuf *m) 1089177633Sdfr{ 1090177633Sdfr struct ip *ip; 1091177633Sdfr u_short csum, offset; 1092177633Sdfr 1093177633Sdfr ip = mtod(m, struct ip *); 1094177633Sdfr offset = IP_VHL_HL(ip->ip_vhl) << 2 ; 1095220585Srmacklem csum = in_cksum_skip(m, ip->ip_len, offset); 1096180025Sdfr if (m->m_pkthdr.csum_flags & CSUM_UDP && csum == 0) 1097180025Sdfr csum = 0xffff; 1098180025Sdfr offset += m->m_pkthdr.csum_data; /* checksum offset */ 1099220585Srmacklem 1100180025Sdfr if (offset + sizeof(u_short) > m->m_len) { 1101220585Srmacklem printf("delayed m_pullup, m->len: %d off: %d p: %d\n", 1102180025Sdfr m->m_len, offset, ip->ip_p); 1103217242Srmacklem /* 1104217242Srmacklem * XXX 1105177633Sdfr * this shouldn't happen, but if it does, the 1106177633Sdfr * correct behavior may be to insert the checksum 1107177633Sdfr * in the existing chain instead of rearranging it. 1108177633Sdfr */ 1109177633Sdfr m = m_pullup(m, offset + sizeof(u_short)); 1110177633Sdfr } 1111177633Sdfr *(u_short *)(m->m_data + offset) = csum; 1112177633Sdfr} 1113177633Sdfr 1114180025Sdfr/* 1115177633Sdfr * Insert IP options into preformed packet. 1116180025Sdfr * Adjust IP destination as required for IP source routing, 1117180025Sdfr * as indicated by a non-zero in_addr at the start of the options. 1118180025Sdfr * 1119180025Sdfr * XXX This routine assumes that the packet has no options in place. 1120177633Sdfr */ 1121177633Sdfrstatic struct mbuf * 1122177633Sdfrip_insertoptions(m, opt, phlen) 1123177633Sdfr register struct mbuf *m; 1124177633Sdfr struct mbuf *opt; 1125177633Sdfr int *phlen; 1126177633Sdfr{ 1127177633Sdfr register struct ipoption *p = mtod(opt, struct ipoption *); 1128177633Sdfr struct mbuf *n; 1129177633Sdfr register struct ip *ip = mtod(m, struct ip *); 1130177633Sdfr unsigned optlen; 1131177633Sdfr 1132177633Sdfr optlen = opt->m_len - sizeof(p->ipopt_dst); 1133177633Sdfr if (optlen + (u_short)ip->ip_len > IP_MAXPACKET) { 1134177633Sdfr *phlen = 0; 1135177633Sdfr return (m); /* XXX should fail */ 1136177633Sdfr } 1137177633Sdfr if (p->ipopt_dst.s_addr) 1138193437Srmacklem ip->ip_dst = p->ipopt_dst; 1139193437Srmacklem if (m->m_flags & M_EXT || m->m_data - optlen < m->m_pktdat) { 1140193437Srmacklem MGETHDR(n, M_DONTWAIT, MT_HEADER); 1141193437Srmacklem if (n == 0) { 1142193437Srmacklem *phlen = 0; 1143193272Sjhb return (m); 1144177633Sdfr } 1145177633Sdfr n->m_pkthdr.rcvif = (struct ifnet *)0; 1146193437Srmacklem#ifdef MAC 1147193437Srmacklem mac_create_mbuf_from_mbuf(m, n); 1148193437Srmacklem#endif 1149193437Srmacklem n->m_pkthdr.len = m->m_pkthdr.len + optlen; 1150193437Srmacklem m->m_len -= sizeof(struct ip); 1151193437Srmacklem m->m_data += sizeof(struct ip); 1152193437Srmacklem n->m_next = m; 1153193437Srmacklem m = n; 1154193437Srmacklem m->m_len = optlen + sizeof(struct ip); 1155193437Srmacklem m->m_data += max_linkhdr; 1156193437Srmacklem (void)memcpy(mtod(m, void *), ip, sizeof(struct ip)); 1157193437Srmacklem } else { 1158193437Srmacklem m->m_data -= optlen; 1159 m->m_len += optlen; 1160 m->m_pkthdr.len += optlen; 1161 ovbcopy((caddr_t)ip, mtod(m, caddr_t), sizeof(struct ip)); 1162 } 1163 ip = mtod(m, struct ip *); 1164 bcopy(p->ipopt_list, ip + 1, optlen); 1165 *phlen = sizeof(struct ip) + optlen; 1166 ip->ip_vhl = IP_MAKE_VHL(IPVERSION, *phlen >> 2); 1167 ip->ip_len += optlen; 1168 return (m); 1169} 1170 1171/* 1172 * Copy options from ip to jp, 1173 * omitting those not copied during fragmentation. 1174 */ 1175int 1176ip_optcopy(ip, jp) 1177 struct ip *ip, *jp; 1178{ 1179 register u_char *cp, *dp; 1180 int opt, optlen, cnt; 1181 1182 cp = (u_char *)(ip + 1); 1183 dp = (u_char *)(jp + 1); 1184 cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip); 1185 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1186 opt = cp[0]; 1187 if (opt == IPOPT_EOL) 1188 break; 1189 if (opt == IPOPT_NOP) { 1190 /* Preserve for IP mcast tunnel's LSRR alignment. */ 1191 *dp++ = IPOPT_NOP; 1192 optlen = 1; 1193 continue; 1194 } 1195 1196 KASSERT(cnt >= IPOPT_OLEN + sizeof(*cp), 1197 ("ip_optcopy: malformed ipv4 option")); 1198 optlen = cp[IPOPT_OLEN]; 1199 KASSERT(optlen >= IPOPT_OLEN + sizeof(*cp) && optlen <= cnt, 1200 ("ip_optcopy: malformed ipv4 option")); 1201 1202 /* bogus lengths should have been caught by ip_dooptions */ 1203 if (optlen > cnt) 1204 optlen = cnt; 1205 if (IPOPT_COPIED(opt)) { 1206 bcopy(cp, dp, optlen); 1207 dp += optlen; 1208 } 1209 } 1210 for (optlen = dp - (u_char *)(jp+1); optlen & 0x3; optlen++) 1211 *dp++ = IPOPT_EOL; 1212 return (optlen); 1213} 1214 1215/* 1216 * IP socket option processing. 1217 */ 1218int 1219ip_ctloutput(so, sopt) 1220 struct socket *so; 1221 struct sockopt *sopt; 1222{ 1223 struct inpcb *inp = sotoinpcb(so); 1224 int error, optval; 1225 1226 error = optval = 0; 1227 if (sopt->sopt_level != IPPROTO_IP) { 1228 return (EINVAL); 1229 } 1230 1231 switch (sopt->sopt_dir) { 1232 case SOPT_SET: 1233 switch (sopt->sopt_name) { 1234 case IP_OPTIONS: 1235#ifdef notyet 1236 case IP_RETOPTS: 1237#endif 1238 { 1239 struct mbuf *m; 1240 if (sopt->sopt_valsize > MLEN) { 1241 error = EMSGSIZE; 1242 break; 1243 } 1244 MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER); 1245 if (m == 0) { 1246 error = ENOBUFS; 1247 break; 1248 } 1249 m->m_len = sopt->sopt_valsize; 1250 error = sooptcopyin(sopt, mtod(m, char *), m->m_len, 1251 m->m_len); 1252 1253 return (ip_pcbopts(sopt->sopt_name, &inp->inp_options, 1254 m)); 1255 } 1256 1257 case IP_TOS: 1258 case IP_TTL: 1259 case IP_RECVOPTS: 1260 case IP_RECVRETOPTS: 1261 case IP_RECVDSTADDR: 1262 case IP_RECVIF: 1263 case IP_FAITH: 1264 error = sooptcopyin(sopt, &optval, sizeof optval, 1265 sizeof optval); 1266 if (error) 1267 break; 1268 1269 switch (sopt->sopt_name) { 1270 case IP_TOS: 1271 inp->inp_ip_tos = optval; 1272 break; 1273 1274 case IP_TTL: 1275 inp->inp_ip_ttl = optval; 1276 break; 1277#define OPTSET(bit) \ 1278 if (optval) \ 1279 inp->inp_flags |= bit; \ 1280 else \ 1281 inp->inp_flags &= ~bit; 1282 1283 case IP_RECVOPTS: 1284 OPTSET(INP_RECVOPTS); 1285 break; 1286 1287 case IP_RECVRETOPTS: 1288 OPTSET(INP_RECVRETOPTS); 1289 break; 1290 1291 case IP_RECVDSTADDR: 1292 OPTSET(INP_RECVDSTADDR); 1293 break; 1294 1295 case IP_RECVIF: 1296 OPTSET(INP_RECVIF); 1297 break; 1298 1299 case IP_FAITH: 1300 OPTSET(INP_FAITH); 1301 break; 1302 } 1303 break; 1304#undef OPTSET 1305 1306 case IP_MULTICAST_IF: 1307 case IP_MULTICAST_VIF: 1308 case IP_MULTICAST_TTL: 1309 case IP_MULTICAST_LOOP: 1310 case IP_ADD_MEMBERSHIP: 1311 case IP_DROP_MEMBERSHIP: 1312 error = ip_setmoptions(sopt, &inp->inp_moptions); 1313 break; 1314 1315 case IP_PORTRANGE: 1316 error = sooptcopyin(sopt, &optval, sizeof optval, 1317 sizeof optval); 1318 if (error) 1319 break; 1320 1321 switch (optval) { 1322 case IP_PORTRANGE_DEFAULT: 1323 inp->inp_flags &= ~(INP_LOWPORT); 1324 inp->inp_flags &= ~(INP_HIGHPORT); 1325 break; 1326 1327 case IP_PORTRANGE_HIGH: 1328 inp->inp_flags &= ~(INP_LOWPORT); 1329 inp->inp_flags |= INP_HIGHPORT; 1330 break; 1331 1332 case IP_PORTRANGE_LOW: 1333 inp->inp_flags &= ~(INP_HIGHPORT); 1334 inp->inp_flags |= INP_LOWPORT; 1335 break; 1336 1337 default: 1338 error = EINVAL; 1339 break; 1340 } 1341 break; 1342 1343#ifdef IPSEC 1344 case IP_IPSEC_POLICY: 1345 { 1346 caddr_t req; 1347 size_t len = 0; 1348 int priv; 1349 struct mbuf *m; 1350 int optname; 1351 1352 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */ 1353 break; 1354 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */ 1355 break; 1356 priv = (sopt->sopt_td != NULL && 1357 suser(sopt->sopt_td) != 0) ? 0 : 1; 1358 req = mtod(m, caddr_t); 1359 len = m->m_len; 1360 optname = sopt->sopt_name; 1361 error = ipsec4_set_policy(inp, optname, req, len, priv); 1362 m_freem(m); 1363 break; 1364 } 1365#endif /*IPSEC*/ 1366 1367 default: 1368 error = ENOPROTOOPT; 1369 break; 1370 } 1371 break; 1372 1373 case SOPT_GET: 1374 switch (sopt->sopt_name) { 1375 case IP_OPTIONS: 1376 case IP_RETOPTS: 1377 if (inp->inp_options) 1378 error = sooptcopyout(sopt, 1379 mtod(inp->inp_options, 1380 char *), 1381 inp->inp_options->m_len); 1382 else 1383 sopt->sopt_valsize = 0; 1384 break; 1385 1386 case IP_TOS: 1387 case IP_TTL: 1388 case IP_RECVOPTS: 1389 case IP_RECVRETOPTS: 1390 case IP_RECVDSTADDR: 1391 case IP_RECVIF: 1392 case IP_PORTRANGE: 1393 case IP_FAITH: 1394 switch (sopt->sopt_name) { 1395 1396 case IP_TOS: 1397 optval = inp->inp_ip_tos; 1398 break; 1399 1400 case IP_TTL: 1401 optval = inp->inp_ip_ttl; 1402 break; 1403 1404#define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0) 1405 1406 case IP_RECVOPTS: 1407 optval = OPTBIT(INP_RECVOPTS); 1408 break; 1409 1410 case IP_RECVRETOPTS: 1411 optval = OPTBIT(INP_RECVRETOPTS); 1412 break; 1413 1414 case IP_RECVDSTADDR: 1415 optval = OPTBIT(INP_RECVDSTADDR); 1416 break; 1417 1418 case IP_RECVIF: 1419 optval = OPTBIT(INP_RECVIF); 1420 break; 1421 1422 case IP_PORTRANGE: 1423 if (inp->inp_flags & INP_HIGHPORT) 1424 optval = IP_PORTRANGE_HIGH; 1425 else if (inp->inp_flags & INP_LOWPORT) 1426 optval = IP_PORTRANGE_LOW; 1427 else 1428 optval = 0; 1429 break; 1430 1431 case IP_FAITH: 1432 optval = OPTBIT(INP_FAITH); 1433 break; 1434 } 1435 error = sooptcopyout(sopt, &optval, sizeof optval); 1436 break; 1437 1438 case IP_MULTICAST_IF: 1439 case IP_MULTICAST_VIF: 1440 case IP_MULTICAST_TTL: 1441 case IP_MULTICAST_LOOP: 1442 case IP_ADD_MEMBERSHIP: 1443 case IP_DROP_MEMBERSHIP: 1444 error = ip_getmoptions(sopt, inp->inp_moptions); 1445 break; 1446 1447#ifdef IPSEC 1448 case IP_IPSEC_POLICY: 1449 { 1450 struct mbuf *m = NULL; 1451 caddr_t req = NULL; 1452 size_t len = 0; 1453 1454 if (m != 0) { 1455 req = mtod(m, caddr_t); 1456 len = m->m_len; 1457 } 1458 error = ipsec4_get_policy(sotoinpcb(so), req, len, &m); 1459 if (error == 0) 1460 error = soopt_mcopyout(sopt, m); /* XXX */ 1461 if (error == 0) 1462 m_freem(m); 1463 break; 1464 } 1465#endif /*IPSEC*/ 1466 1467 default: 1468 error = ENOPROTOOPT; 1469 break; 1470 } 1471 break; 1472 } 1473 return (error); 1474} 1475 1476/* 1477 * Set up IP options in pcb for insertion in output packets. 1478 * Store in mbuf with pointer in pcbopt, adding pseudo-option 1479 * with destination address if source routed. 1480 */ 1481static int 1482ip_pcbopts(optname, pcbopt, m) 1483 int optname; 1484 struct mbuf **pcbopt; 1485 register struct mbuf *m; 1486{ 1487 register int cnt, optlen; 1488 register u_char *cp; 1489 u_char opt; 1490 1491 /* turn off any old options */ 1492 if (*pcbopt) 1493 (void)m_free(*pcbopt); 1494 *pcbopt = 0; 1495 if (m == (struct mbuf *)0 || m->m_len == 0) { 1496 /* 1497 * Only turning off any previous options. 1498 */ 1499 if (m) 1500 (void)m_free(m); 1501 return (0); 1502 } 1503 1504 if (m->m_len % sizeof(int32_t)) 1505 goto bad; 1506 /* 1507 * IP first-hop destination address will be stored before 1508 * actual options; move other options back 1509 * and clear it when none present. 1510 */ 1511 if (m->m_data + m->m_len + sizeof(struct in_addr) >= &m->m_dat[MLEN]) 1512 goto bad; 1513 cnt = m->m_len; 1514 m->m_len += sizeof(struct in_addr); 1515 cp = mtod(m, u_char *) + sizeof(struct in_addr); 1516 ovbcopy(mtod(m, caddr_t), (caddr_t)cp, (unsigned)cnt); 1517 bzero(mtod(m, caddr_t), sizeof(struct in_addr)); 1518 1519 for (; cnt > 0; cnt -= optlen, cp += optlen) { 1520 opt = cp[IPOPT_OPTVAL]; 1521 if (opt == IPOPT_EOL) 1522 break; 1523 if (opt == IPOPT_NOP) 1524 optlen = 1; 1525 else { 1526 if (cnt < IPOPT_OLEN + sizeof(*cp)) 1527 goto bad; 1528 optlen = cp[IPOPT_OLEN]; 1529 if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) 1530 goto bad; 1531 } 1532 switch (opt) { 1533 1534 default: 1535 break; 1536 1537 case IPOPT_LSRR: 1538 case IPOPT_SSRR: 1539 /* 1540 * user process specifies route as: 1541 * ->A->B->C->D 1542 * D must be our final destination (but we can't 1543 * check that since we may not have connected yet). 1544 * A is first hop destination, which doesn't appear in 1545 * actual IP option, but is stored before the options. 1546 */ 1547 if (optlen < IPOPT_MINOFF - 1 + sizeof(struct in_addr)) 1548 goto bad; 1549 m->m_len -= sizeof(struct in_addr); 1550 cnt -= sizeof(struct in_addr); 1551 optlen -= sizeof(struct in_addr); 1552 cp[IPOPT_OLEN] = optlen; 1553 /* 1554 * Move first hop before start of options. 1555 */ 1556 bcopy((caddr_t)&cp[IPOPT_OFFSET+1], mtod(m, caddr_t), 1557 sizeof(struct in_addr)); 1558 /* 1559 * Then copy rest of options back 1560 * to close up the deleted entry. 1561 */ 1562 ovbcopy((caddr_t)(&cp[IPOPT_OFFSET+1] + 1563 sizeof(struct in_addr)), 1564 (caddr_t)&cp[IPOPT_OFFSET+1], 1565 (unsigned)cnt + sizeof(struct in_addr)); 1566 break; 1567 } 1568 } 1569 if (m->m_len > MAX_IPOPTLEN + sizeof(struct in_addr)) 1570 goto bad; 1571 *pcbopt = m; 1572 return (0); 1573 1574bad: 1575 (void)m_free(m); 1576 return (EINVAL); 1577} 1578 1579/* 1580 * XXX 1581 * The whole multicast option thing needs to be re-thought. 1582 * Several of these options are equally applicable to non-multicast 1583 * transmission, and one (IP_MULTICAST_TTL) totally duplicates a 1584 * standard option (IP_TTL). 1585 */ 1586 1587/* 1588 * following RFC1724 section 3.3, 0.0.0.0/8 is interpreted as interface index. 1589 */ 1590static struct ifnet * 1591ip_multicast_if(a, ifindexp) 1592 struct in_addr *a; 1593 int *ifindexp; 1594{ 1595 int ifindex; 1596 struct ifnet *ifp; 1597 1598 if (ifindexp) 1599 *ifindexp = 0; 1600 if (ntohl(a->s_addr) >> 24 == 0) { 1601 ifindex = ntohl(a->s_addr) & 0xffffff; 1602 if (ifindex < 0 || if_index < ifindex) 1603 return NULL; 1604 ifp = ifnet_byindex(ifindex); 1605 if (ifindexp) 1606 *ifindexp = ifindex; 1607 } else { 1608 INADDR_TO_IFP(*a, ifp); 1609 } 1610 return ifp; 1611} 1612 1613/* 1614 * Set the IP multicast options in response to user setsockopt(). 1615 */ 1616static int 1617ip_setmoptions(sopt, imop) 1618 struct sockopt *sopt; 1619 struct ip_moptions **imop; 1620{ 1621 int error = 0; 1622 int i; 1623 struct in_addr addr; 1624 struct ip_mreq mreq; 1625 struct ifnet *ifp; 1626 struct ip_moptions *imo = *imop; 1627 struct route ro; 1628 struct sockaddr_in *dst; 1629 int ifindex; 1630 int s; 1631 1632 if (imo == NULL) { 1633 /* 1634 * No multicast option buffer attached to the pcb; 1635 * allocate one and initialize to default values. 1636 */ 1637 imo = (struct ip_moptions*)malloc(sizeof(*imo), M_IPMOPTS, 1638 M_WAITOK); 1639 1640 if (imo == NULL) 1641 return (ENOBUFS); 1642 *imop = imo; 1643 imo->imo_multicast_ifp = NULL; 1644 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1645 imo->imo_multicast_vif = -1; 1646 imo->imo_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1647 imo->imo_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1648 imo->imo_num_memberships = 0; 1649 } 1650 1651 switch (sopt->sopt_name) { 1652 /* store an index number for the vif you wanna use in the send */ 1653 case IP_MULTICAST_VIF: 1654 if (legal_vif_num == 0) { 1655 error = EOPNOTSUPP; 1656 break; 1657 } 1658 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 1659 if (error) 1660 break; 1661 if (!legal_vif_num(i) && (i != -1)) { 1662 error = EINVAL; 1663 break; 1664 } 1665 imo->imo_multicast_vif = i; 1666 break; 1667 1668 case IP_MULTICAST_IF: 1669 /* 1670 * Select the interface for outgoing multicast packets. 1671 */ 1672 error = sooptcopyin(sopt, &addr, sizeof addr, sizeof addr); 1673 if (error) 1674 break; 1675 /* 1676 * INADDR_ANY is used to remove a previous selection. 1677 * When no interface is selected, a default one is 1678 * chosen every time a multicast packet is sent. 1679 */ 1680 if (addr.s_addr == INADDR_ANY) { 1681 imo->imo_multicast_ifp = NULL; 1682 break; 1683 } 1684 /* 1685 * The selected interface is identified by its local 1686 * IP address. Find the interface and confirm that 1687 * it supports multicasting. 1688 */ 1689 s = splimp(); 1690 ifp = ip_multicast_if(&addr, &ifindex); 1691 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1692 splx(s); 1693 error = EADDRNOTAVAIL; 1694 break; 1695 } 1696 imo->imo_multicast_ifp = ifp; 1697 if (ifindex) 1698 imo->imo_multicast_addr = addr; 1699 else 1700 imo->imo_multicast_addr.s_addr = INADDR_ANY; 1701 splx(s); 1702 break; 1703 1704 case IP_MULTICAST_TTL: 1705 /* 1706 * Set the IP time-to-live for outgoing multicast packets. 1707 * The original multicast API required a char argument, 1708 * which is inconsistent with the rest of the socket API. 1709 * We allow either a char or an int. 1710 */ 1711 if (sopt->sopt_valsize == 1) { 1712 u_char ttl; 1713 error = sooptcopyin(sopt, &ttl, 1, 1); 1714 if (error) 1715 break; 1716 imo->imo_multicast_ttl = ttl; 1717 } else { 1718 u_int ttl; 1719 error = sooptcopyin(sopt, &ttl, sizeof ttl, 1720 sizeof ttl); 1721 if (error) 1722 break; 1723 if (ttl > 255) 1724 error = EINVAL; 1725 else 1726 imo->imo_multicast_ttl = ttl; 1727 } 1728 break; 1729 1730 case IP_MULTICAST_LOOP: 1731 /* 1732 * Set the loopback flag for outgoing multicast packets. 1733 * Must be zero or one. The original multicast API required a 1734 * char argument, which is inconsistent with the rest 1735 * of the socket API. We allow either a char or an int. 1736 */ 1737 if (sopt->sopt_valsize == 1) { 1738 u_char loop; 1739 error = sooptcopyin(sopt, &loop, 1, 1); 1740 if (error) 1741 break; 1742 imo->imo_multicast_loop = !!loop; 1743 } else { 1744 u_int loop; 1745 error = sooptcopyin(sopt, &loop, sizeof loop, 1746 sizeof loop); 1747 if (error) 1748 break; 1749 imo->imo_multicast_loop = !!loop; 1750 } 1751 break; 1752 1753 case IP_ADD_MEMBERSHIP: 1754 /* 1755 * Add a multicast group membership. 1756 * Group must be a valid IP multicast address. 1757 */ 1758 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1759 if (error) 1760 break; 1761 1762 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1763 error = EINVAL; 1764 break; 1765 } 1766 s = splimp(); 1767 /* 1768 * If no interface address was provided, use the interface of 1769 * the route to the given multicast address. 1770 */ 1771 if (mreq.imr_interface.s_addr == INADDR_ANY) { 1772 bzero((caddr_t)&ro, sizeof(ro)); 1773 dst = (struct sockaddr_in *)&ro.ro_dst; 1774 dst->sin_len = sizeof(*dst); 1775 dst->sin_family = AF_INET; 1776 dst->sin_addr = mreq.imr_multiaddr; 1777 rtalloc(&ro); 1778 if (ro.ro_rt == NULL) { 1779 error = EADDRNOTAVAIL; 1780 splx(s); 1781 break; 1782 } 1783 ifp = ro.ro_rt->rt_ifp; 1784 rtfree(ro.ro_rt); 1785 } 1786 else { 1787 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1788 } 1789 1790 /* 1791 * See if we found an interface, and confirm that it 1792 * supports multicast. 1793 */ 1794 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) { 1795 error = EADDRNOTAVAIL; 1796 splx(s); 1797 break; 1798 } 1799 /* 1800 * See if the membership already exists or if all the 1801 * membership slots are full. 1802 */ 1803 for (i = 0; i < imo->imo_num_memberships; ++i) { 1804 if (imo->imo_membership[i]->inm_ifp == ifp && 1805 imo->imo_membership[i]->inm_addr.s_addr 1806 == mreq.imr_multiaddr.s_addr) 1807 break; 1808 } 1809 if (i < imo->imo_num_memberships) { 1810 error = EADDRINUSE; 1811 splx(s); 1812 break; 1813 } 1814 if (i == IP_MAX_MEMBERSHIPS) { 1815 error = ETOOMANYREFS; 1816 splx(s); 1817 break; 1818 } 1819 /* 1820 * Everything looks good; add a new record to the multicast 1821 * address list for the given interface. 1822 */ 1823 if ((imo->imo_membership[i] = 1824 in_addmulti(&mreq.imr_multiaddr, ifp)) == NULL) { 1825 error = ENOBUFS; 1826 splx(s); 1827 break; 1828 } 1829 ++imo->imo_num_memberships; 1830 splx(s); 1831 break; 1832 1833 case IP_DROP_MEMBERSHIP: 1834 /* 1835 * Drop a multicast group membership. 1836 * Group must be a valid IP multicast address. 1837 */ 1838 error = sooptcopyin(sopt, &mreq, sizeof mreq, sizeof mreq); 1839 if (error) 1840 break; 1841 1842 if (!IN_MULTICAST(ntohl(mreq.imr_multiaddr.s_addr))) { 1843 error = EINVAL; 1844 break; 1845 } 1846 1847 s = splimp(); 1848 /* 1849 * If an interface address was specified, get a pointer 1850 * to its ifnet structure. 1851 */ 1852 if (mreq.imr_interface.s_addr == INADDR_ANY) 1853 ifp = NULL; 1854 else { 1855 ifp = ip_multicast_if(&mreq.imr_interface, NULL); 1856 if (ifp == NULL) { 1857 error = EADDRNOTAVAIL; 1858 splx(s); 1859 break; 1860 } 1861 } 1862 /* 1863 * Find the membership in the membership array. 1864 */ 1865 for (i = 0; i < imo->imo_num_memberships; ++i) { 1866 if ((ifp == NULL || 1867 imo->imo_membership[i]->inm_ifp == ifp) && 1868 imo->imo_membership[i]->inm_addr.s_addr == 1869 mreq.imr_multiaddr.s_addr) 1870 break; 1871 } 1872 if (i == imo->imo_num_memberships) { 1873 error = EADDRNOTAVAIL; 1874 splx(s); 1875 break; 1876 } 1877 /* 1878 * Give up the multicast address record to which the 1879 * membership points. 1880 */ 1881 in_delmulti(imo->imo_membership[i]); 1882 /* 1883 * Remove the gap in the membership array. 1884 */ 1885 for (++i; i < imo->imo_num_memberships; ++i) 1886 imo->imo_membership[i-1] = imo->imo_membership[i]; 1887 --imo->imo_num_memberships; 1888 splx(s); 1889 break; 1890 1891 default: 1892 error = EOPNOTSUPP; 1893 break; 1894 } 1895 1896 /* 1897 * If all options have default values, no need to keep the mbuf. 1898 */ 1899 if (imo->imo_multicast_ifp == NULL && 1900 imo->imo_multicast_vif == -1 && 1901 imo->imo_multicast_ttl == IP_DEFAULT_MULTICAST_TTL && 1902 imo->imo_multicast_loop == IP_DEFAULT_MULTICAST_LOOP && 1903 imo->imo_num_memberships == 0) { 1904 free(*imop, M_IPMOPTS); 1905 *imop = NULL; 1906 } 1907 1908 return (error); 1909} 1910 1911/* 1912 * Return the IP multicast options in response to user getsockopt(). 1913 */ 1914static int 1915ip_getmoptions(sopt, imo) 1916 struct sockopt *sopt; 1917 register struct ip_moptions *imo; 1918{ 1919 struct in_addr addr; 1920 struct in_ifaddr *ia; 1921 int error, optval; 1922 u_char coptval; 1923 1924 error = 0; 1925 switch (sopt->sopt_name) { 1926 case IP_MULTICAST_VIF: 1927 if (imo != NULL) 1928 optval = imo->imo_multicast_vif; 1929 else 1930 optval = -1; 1931 error = sooptcopyout(sopt, &optval, sizeof optval); 1932 break; 1933 1934 case IP_MULTICAST_IF: 1935 if (imo == NULL || imo->imo_multicast_ifp == NULL) 1936 addr.s_addr = INADDR_ANY; 1937 else if (imo->imo_multicast_addr.s_addr) { 1938 /* return the value user has set */ 1939 addr = imo->imo_multicast_addr; 1940 } else { 1941 IFP_TO_IA(imo->imo_multicast_ifp, ia); 1942 addr.s_addr = (ia == NULL) ? INADDR_ANY 1943 : IA_SIN(ia)->sin_addr.s_addr; 1944 } 1945 error = sooptcopyout(sopt, &addr, sizeof addr); 1946 break; 1947 1948 case IP_MULTICAST_TTL: 1949 if (imo == 0) 1950 optval = coptval = IP_DEFAULT_MULTICAST_TTL; 1951 else 1952 optval = coptval = imo->imo_multicast_ttl; 1953 if (sopt->sopt_valsize == 1) 1954 error = sooptcopyout(sopt, &coptval, 1); 1955 else 1956 error = sooptcopyout(sopt, &optval, sizeof optval); 1957 break; 1958 1959 case IP_MULTICAST_LOOP: 1960 if (imo == 0) 1961 optval = coptval = IP_DEFAULT_MULTICAST_LOOP; 1962 else 1963 optval = coptval = imo->imo_multicast_loop; 1964 if (sopt->sopt_valsize == 1) 1965 error = sooptcopyout(sopt, &coptval, 1); 1966 else 1967 error = sooptcopyout(sopt, &optval, sizeof optval); 1968 break; 1969 1970 default: 1971 error = ENOPROTOOPT; 1972 break; 1973 } 1974 return (error); 1975} 1976 1977/* 1978 * Discard the IP multicast options. 1979 */ 1980void 1981ip_freemoptions(imo) 1982 register struct ip_moptions *imo; 1983{ 1984 register int i; 1985 1986 if (imo != NULL) { 1987 for (i = 0; i < imo->imo_num_memberships; ++i) 1988 in_delmulti(imo->imo_membership[i]); 1989 free(imo, M_IPMOPTS); 1990 } 1991} 1992 1993/* 1994 * Routine called from ip_output() to loop back a copy of an IP multicast 1995 * packet to the input queue of a specified interface. Note that this 1996 * calls the output routine of the loopback "driver", but with an interface 1997 * pointer that might NOT be a loopback interface -- evil, but easier than 1998 * replicating that code here. 1999 */ 2000static void 2001ip_mloopback(ifp, m, dst, hlen) 2002 struct ifnet *ifp; 2003 register struct mbuf *m; 2004 register struct sockaddr_in *dst; 2005 int hlen; 2006{ 2007 register struct ip *ip; 2008 struct mbuf *copym; 2009 2010 copym = m_copy(m, 0, M_COPYALL); 2011 if (copym != NULL && (copym->m_flags & M_EXT || copym->m_len < hlen)) 2012 copym = m_pullup(copym, hlen); 2013 if (copym != NULL) { 2014 /* 2015 * We don't bother to fragment if the IP length is greater 2016 * than the interface's MTU. Can this possibly matter? 2017 */ 2018 ip = mtod(copym, struct ip *); 2019 ip->ip_len = htons(ip->ip_len); 2020 ip->ip_off = htons(ip->ip_off); 2021 ip->ip_sum = 0; 2022 if (ip->ip_vhl == IP_VHL_BORING) { 2023 ip->ip_sum = in_cksum_hdr(ip); 2024 } else { 2025 ip->ip_sum = in_cksum(copym, hlen); 2026 } 2027 /* 2028 * NB: 2029 * It's not clear whether there are any lingering 2030 * reentrancy problems in other areas which might 2031 * be exposed by using ip_input directly (in 2032 * particular, everything which modifies the packet 2033 * in-place). Yet another option is using the 2034 * protosw directly to deliver the looped back 2035 * packet. For the moment, we'll err on the side 2036 * of safety by using if_simloop(). 2037 */ 2038#if 1 /* XXX */ 2039 if (dst->sin_family != AF_INET) { 2040 printf("ip_mloopback: bad address family %d\n", 2041 dst->sin_family); 2042 dst->sin_family = AF_INET; 2043 } 2044#endif 2045 2046#ifdef notdef 2047 copym->m_pkthdr.rcvif = ifp; 2048 ip_input(copym); 2049#else 2050 /* if the checksum hasn't been computed, mark it as valid */ 2051 if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2052 copym->m_pkthdr.csum_flags |= 2053 CSUM_DATA_VALID | CSUM_PSEUDO_HDR; 2054 copym->m_pkthdr.csum_data = 0xffff; 2055 } 2056 if_simloop(ifp, copym, dst->sin_family, 0); 2057#endif 2058 } 2059} 2060