sys/netinet/ip_input.c

   1 /*
   2  * Copyright (c) 1982, 1986, 1988, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. All advertising materials mentioning features or use of this software
  14  *    must display the following acknowledgement:
  15  *      This product includes software developed by the University of
  16  *      California, Berkeley and its contributors.
  17  * 4. Neither the name of the University nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  *
  33  *      @(#)ip_input.c  8.2 (Berkeley) 1/4/94
  34  * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
  35  * $DragonFly: src/sys/netinet/ip_input.c,v 1.3 2003/07/26 21:00:04 rob Exp $
  36  */
  37
  38 #define _IP_VHL
  39
  40 #include "opt_bootp.h"
  41 #include "opt_ipfw.h"
  42 #include "opt_ipdn.h"
  43 #include "opt_ipdivert.h"
  44 #include "opt_ipfilter.h"
  45 #include "opt_ipstealth.h"
  46 #include "opt_ipsec.h"
  47 #include "opt_random_ip_id.h"
  48
  49 #include <sys/param.h>
  50 #include <sys/systm.h>
  51 #include <sys/mbuf.h>
  52 #include <sys/malloc.h>
  53 #include <sys/domain.h>
  54 #include <sys/protosw.h>
  55 #include <sys/socket.h>
  56 #include <sys/time.h>
  57 #include <sys/kernel.h>
  58 #include <sys/syslog.h>
  59 #include <sys/sysctl.h>
  60
  61 #include <net/if.h>
  62 #include <net/if_types.h>
  63 #include <net/if_var.h>
  64 #include <net/if_dl.h>
  65 #include <net/route.h>
  66 #include <net/netisr.h>
  67 #include <net/intrq.h>
  68
  69 #include <netinet/in.h>
  70 #include <netinet/in_systm.h>
  71 #include <netinet/in_var.h>
  72 #include <netinet/ip.h>
  73 #include <netinet/in_pcb.h>
  74 #include <netinet/ip_var.h>
  75 #include <netinet/ip_icmp.h>
  76 #include <machine/in_cksum.h>
  77
  78 #include <netinet/ipprotosw.h>
  79
  80 #include <sys/socketvar.h>
  81
  82 #include <netinet/ip_fw.h>
  83 #include <netinet/ip_dummynet.h>
  84
  85 #ifdef IPSEC
  86 #include <netinet6/ipsec.h>
  87 #include <netkey/key.h>
  88 #endif
  89
  90 #ifdef FAST_IPSEC
  91 #include <netipsec/ipsec.h>
  92 #include <netipsec/key.h>
  93 #endif
  94
  95 int rsvp_on = 0;
  96 static int ip_rsvp_on;
  97 struct socket *ip_rsvpd;
  98
  99 int     ipforwarding = 0;
 100 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
 101     &ipforwarding, 0, "Enable IP forwarding between interfaces");
 102
 103 static int      ipsendredirects = 1; /* XXX */
 104 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
 105     &ipsendredirects, 0, "Enable sending IP redirects");
 106
 107 int     ip_defttl = IPDEFTTL;
 108 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
 109     &ip_defttl, 0, "Maximum TTL on IP packets");
 110
 111 static int      ip_dosourceroute = 0;
 112 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
 113     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
 114
 115 static int      ip_acceptsourceroute = 0;
 116 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
 117     CTLFLAG_RW, &ip_acceptsourceroute, 0,
 118     "Enable accepting source routed IP packets");
 119
 120 static int      ip_keepfaith = 0;
 121 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
 122         &ip_keepfaith,  0,
 123         "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
 124
 125 static int      nipq = 0;       /* total # of reass queues */
 126 static int      maxnipq;
 127 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
 128         &maxnipq, 0,
 129         "Maximum number of IPv4 fragment reassembly queue entries");
 130
 131 static int    maxfragsperpacket;
 132 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
 133         &maxfragsperpacket, 0,
 134         "Maximum number of IPv4 fragments allowed per packet");
 135
 136 static int      ip_sendsourcequench = 0;
 137 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
 138         &ip_sendsourcequench, 0,
 139         "Enable the transmission of source quench packets");
 140
 141 /*
 142  * XXX - Setting ip_checkinterface mostly implements the receive side of
 143  * the Strong ES model described in RFC 1122, but since the routing table
 144  * and transmit implementation do not implement the Strong ES model,
 145  * setting this to 1 results in an odd hybrid.
 146  *
 147  * XXX - ip_checkinterface currently must be disabled if you use ipnat
 148  * to translate the destination address to another local interface.
 149  *
 150  * XXX - ip_checkinterface must be disabled if you add IP aliases
 151  * to the loopback interface instead of the interface where the
 152  * packets for those addresses are received.
 153  */
 154 static int      ip_checkinterface = 0;
 155 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
 156     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
 157
 158 #ifdef DIAGNOSTIC
 159 static int      ipprintfs = 0;
 160 #endif
 161
 162 static int      ipqmaxlen = IFQ_MAXLEN;
 163
 164 extern  struct domain inetdomain;
 165 extern  struct ipprotosw inetsw[];
 166 u_char  ip_protox[IPPROTO_MAX];
 167 struct  in_ifaddrhead in_ifaddrhead;            /* first inet address */
 168 struct  in_ifaddrhashhead *in_ifaddrhashtbl;    /* inet addr hash table */
 169 u_long  in_ifaddrhmask;                         /* mask for hash table */
 170 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
 171     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
 172 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
 173     &ipintrq.ifq_drops, 0, "Number of packets dropped from the IP input queue");
 174
 175 struct ipstat ipstat;
 176 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
 177     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
 178
 179 /* Packet reassembly stuff */
 180 #define IPREASS_NHASH_LOG2      6
 181 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
 182 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
 183 #define IPREASS_HASH(x,y) \
 184         (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
 185
 186 static struct ipq ipq[IPREASS_NHASH];
 187 const  int    ipintrq_present = 1;
 188
 189 #ifdef IPCTL_DEFMTU
 190 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
 191     &ip_mtu, 0, "Default MTU");
 192 #endif
 193
 194 #ifdef IPSTEALTH
 195 static int      ipstealth = 0;
 196 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
 197     &ipstealth, 0, "");
 198 #endif
 199
 200
 201 /* Firewall hooks */
 202 ip_fw_chk_t *ip_fw_chk_ptr;
 203 int fw_enable = 1 ;
 204 int fw_one_pass = 1;
 205
 206 /* Dummynet hooks */
 207 ip_dn_io_t *ip_dn_io_ptr;
 208
 209 int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **)) = NULL;
 210
 211 /*
 212  * XXX this is ugly -- the following two global variables are
 213  * used to store packet state while it travels through the stack.
 214  * Note that the code even makes assumptions on the size and
 215  * alignment of fields inside struct ip_srcrt so e.g. adding some
 216  * fields will break the code. This needs to be fixed.
 217  *
 218  * We need to save the IP options in case a protocol wants to respond
 219  * to an incoming packet over the same route if the packet got here
 220  * using IP source routing.  This allows connection establishment and
 221  * maintenance when the remote end is on a network that is not known
 222  * to us.
 223  */
 224 static int      ip_nhops = 0;
 225 static  struct ip_srcrt {
 226         struct  in_addr dst;                    /* final destination */
 227         char    nop;                            /* one NOP to align */
 228         char    srcopt[IPOPT_OFFSET + 1];       /* OPTVAL, OLEN and OFFSET */
 229         struct  in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
 230 } ip_srcrt;
 231
 232 static void     save_rte(u_char *, struct in_addr);
 233 static int      ip_dooptions(struct mbuf *m, int,
 234                         struct sockaddr_in *next_hop);
 235 static void     ip_forward(struct mbuf *m, int srcrt,
 236                         struct sockaddr_in *next_hop);
 237 static void     ip_freef(struct ipq *);
 238 static struct   mbuf *ip_reass(struct mbuf *, struct ipq *,
 239                 struct ipq *, u_int32_t *, u_int16_t *);
 240 static void     ipintr(void);
 241
 242 /*
 243  * IP initialization: fill in IP protocol switch table.
 244  * All protocols not implemented in kernel go to raw IP protocol handler.
 245  */
 246 void
 247 ip_init()
 248 {
 249         struct ipprotosw *pr;
 250         int i;
 251
 252         TAILQ_INIT(&in_ifaddrhead);
 253         in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
 254         pr = (struct ipprotosw *)pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
 255         if (pr == 0)
 256                 panic("ip_init");
 257         for (i = 0; i < IPPROTO_MAX; i++)
 258                 ip_protox[i] = pr - inetsw;
 259         for (pr = (struct ipprotosw *)inetdomain.dom_protosw;
 260             pr < (struct ipprotosw *)inetdomain.dom_protoswNPROTOSW; pr++)
 261                 if (pr->pr_domain->dom_family == PF_INET &&
 262                     pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
 263                         ip_protox[pr->pr_protocol] = pr - inetsw;
 264
 265         for (i = 0; i < IPREASS_NHASH; i++)
 266             ipq[i].next = ipq[i].prev = &ipq[i];
 267
 268         maxnipq = nmbclusters / 32;
 269         maxfragsperpacket = 16;
 270
 271 #ifndef RANDOM_IP_ID
 272         ip_id = time_second & 0xffff;
 273 #endif
 274         ipintrq.ifq_maxlen = ipqmaxlen;
 275
 276         register_netisr(NETISR_IP, ipintr);
 277 }
 278
 279 /*
 280  * XXX watch out this one. It is perhaps used as a cache for
 281  * the most recently used route ? it is cleared in in_addroute()
 282  * when a new route is successfully created.
 283  */
 284 struct  route ipforward_rt;
 285 static struct   sockaddr_in ipaddr = { sizeof(ipaddr), AF_INET };
 286
 287 /*
 288  * Ip input routine.  Checksum and byte swap header.  If fragmented
 289  * try to reassemble.  Process options.  Pass to next level.
 290  */
 291 void
 292 ip_input(struct mbuf *m)
 293 {
 294         struct ip *ip;
 295         struct ipq *fp;
 296         struct in_ifaddr *ia = NULL;
 297         struct ifaddr *ifa;
 298         int    i, hlen, checkif;
 299         u_short sum;
 300         struct in_addr pkt_dst;
 301         u_int32_t divert_info = 0;              /* packet divert/tee info */
 302         struct ip_fw_args args;
 303 #ifdef FAST_IPSEC
 304         struct m_tag *mtag;
 305         struct tdb_ident *tdbi;
 306         struct secpolicy *sp;
 307         int s, error;
 308 #endif /* FAST_IPSEC */
 309
 310         args.eh = NULL;
 311         args.oif = NULL;
 312         args.rule = NULL;
 313         args.divert_rule = 0;                   /* divert cookie */
 314         args.next_hop = NULL;
 315
 316         /* Grab info from MT_TAG mbufs prepended to the chain.  */
 317         for (; m && m->m_type == MT_TAG; m = m->m_next) {
 318                 switch(m->_m_tag_id) {
 319                 default:
 320                         printf("ip_input: unrecognised MT_TAG tag %d\n",
 321                             m->_m_tag_id);
 322                         break;
 323
 324                 case PACKET_TAG_DUMMYNET:
 325                         args.rule = ((struct dn_pkt *)m)->rule;
 326                         break;
 327
 328                 case PACKET_TAG_DIVERT:
 329                         args.divert_rule = (int)m->m_hdr.mh_data & 0xffff;
 330                         break;
 331
 332                 case PACKET_TAG_IPFORWARD:
 333                         args.next_hop = (struct sockaddr_in *)m->m_hdr.mh_data;
 334                         break;
 335                 }
 336         }
 337
 338         KASSERT(m != NULL && (m->m_flags & M_PKTHDR) != 0,
 339             ("ip_input: no HDR"));
 340
 341         if (args.rule) {        /* dummynet already filtered us */
 342                 ip = mtod(m, struct ip *);
 343                 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 344                 goto iphack ;
 345         }
 346
 347         ipstat.ips_total++;
 348
 349         if (m->m_pkthdr.len < sizeof(struct ip))
 350                 goto tooshort;
 351
 352         if (m->m_len < sizeof (struct ip) &&
 353             (m = m_pullup(m, sizeof (struct ip))) == 0) {
 354                 ipstat.ips_toosmall++;
 355                 return;
 356         }
 357         ip = mtod(m, struct ip *);
 358
 359         if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
 360                 ipstat.ips_badvers++;
 361                 goto bad;
 362         }
 363
 364         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 365         if (hlen < sizeof(struct ip)) { /* minimum header length */
 366                 ipstat.ips_badhlen++;
 367                 goto bad;
 368         }
 369         if (hlen > m->m_len) {
 370                 if ((m = m_pullup(m, hlen)) == 0) {
 371                         ipstat.ips_badhlen++;
 372                         return;
 373                 }
 374                 ip = mtod(m, struct ip *);
 375         }
 376
 377         /* 127/8 must not appear on wire - RFC1122 */
 378         if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
 379             (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
 380                 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
 381                         ipstat.ips_badaddr++;
 382                         goto bad;
 383                 }
 384         }
 385
 386         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
 387                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
 388         } else {
 389                 if (hlen == sizeof(struct ip)) {
 390                         sum = in_cksum_hdr(ip);
 391                 } else {
 392                         sum = in_cksum(m, hlen);
 393                 }
 394         }
 395         if (sum) {
 396                 ipstat.ips_badsum++;
 397                 goto bad;
 398         }
 399
 400         /*
 401          * Convert fields to host representation.
 402          */
 403         ip->ip_len = ntohs(ip->ip_len);
 404         if (ip->ip_len < hlen) {
 405                 ipstat.ips_badlen++;
 406                 goto bad;
 407         }
 408         ip->ip_off = ntohs(ip->ip_off);
 409
 410         /*
 411          * Check that the amount of data in the buffers
 412          * is as at least much as the IP header would have us expect.
 413          * Trim mbufs if longer than we expect.
 414          * Drop packet if shorter than we expect.
 415          */
 416         if (m->m_pkthdr.len < ip->ip_len) {
 417 tooshort:
 418                 ipstat.ips_tooshort++;
 419                 goto bad;
 420         }
 421         if (m->m_pkthdr.len > ip->ip_len) {
 422                 if (m->m_len == m->m_pkthdr.len) {
 423                         m->m_len = ip->ip_len;
 424                         m->m_pkthdr.len = ip->ip_len;
 425                 } else
 426                         m_adj(m, ip->ip_len - m->m_pkthdr.len);
 427         }
 428 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
 429         /*
 430          * Bypass packet filtering for packets from a tunnel (gif).
 431          */
 432         if (ipsec_gethist(m, NULL))
 433                 goto pass;
 434 #endif
 435
 436         /*
 437          * IpHack's section.
 438          * Right now when no processing on packet has done
 439          * and it is still fresh out of network we do our black
 440          * deals with it.
 441          * - Firewall: deny/allow/divert
 442          * - Xlate: translate packet's addr/port (NAT).
 443          * - Pipe: pass pkt through dummynet.
 444          * - Wrap: fake packet's addr/port <unimpl.>
 445          * - Encapsulate: put it in another IP and send out. <unimp.>
 446          */
 447
 448 iphack:
 449         /*
 450          * Check if we want to allow this packet to be processed.
 451          * Consider it to be bad if not.
 452          */
 453         if (fr_checkp) {
 454                 struct  mbuf    *m1 = m;
 455
 456                 if ((*fr_checkp)(ip, hlen, m->m_pkthdr.rcvif, 0, &m1) || !m1)
 457                         return;
 458                 ip = mtod(m = m1, struct ip *);
 459         }
 460         if (fw_enable && IPFW_LOADED) {
 461                 /*
 462                  * If we've been forwarded from the output side, then
 463                  * skip the firewall a second time
 464                  */
 465                 if (args.next_hop)
 466                         goto ours;
 467
 468                 args.m = m;
 469                 i = ip_fw_chk_ptr(&args);
 470                 m = args.m;
 471
 472                 if ( (i & IP_FW_PORT_DENY_FLAG) || m == NULL) { /* drop */
 473                         if (m)
 474                                 m_freem(m);
 475                         return;
 476                 }
 477                 ip = mtod(m, struct ip *); /* just in case m changed */
 478                 if (i == 0 && args.next_hop == NULL)    /* common case */
 479                         goto pass;
 480                 if (DUMMYNET_LOADED && (i & IP_FW_PORT_DYNT_FLAG) != 0) {
 481                         /* Send packet to the appropriate pipe */
 482                         ip_dn_io_ptr(m, i&0xffff, DN_TO_IP_IN, &args);
 483                         return;
 484                 }
 485 #ifdef IPDIVERT
 486                 if (i != 0 && (i & IP_FW_PORT_DYNT_FLAG) == 0) {
 487                         /* Divert or tee packet */
 488                         divert_info = i;
 489                         goto ours;
 490                 }
 491 #endif
 492                 if (i == 0 && args.next_hop != NULL)
 493                         goto pass;
 494                 /*
 495                  * if we get here, the packet must be dropped
 496                  */
 497                 m_freem(m);
 498                 return;
 499         }
 500 pass:
 501
 502         /*
 503          * Process options and, if not destined for us,
 504          * ship it on.  ip_dooptions returns 1 when an
 505          * error was detected (causing an icmp message
 506          * to be sent and the original packet to be freed).
 507          */
 508         ip_nhops = 0;           /* for source routed packets */
 509         if (hlen > sizeof (struct ip) && ip_dooptions(m, 0, args.next_hop))
 510                 return;
 511
 512         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
 513          * matter if it is destined to another node, or whether it is
 514          * a multicast one, RSVP wants it! and prevents it from being forwarded
 515          * anywhere else. Also checks if the rsvp daemon is running before
 516          * grabbing the packet.
 517          */
 518         if (rsvp_on && ip->ip_p==IPPROTO_RSVP)
 519                 goto ours;
 520
 521         /*
 522          * Check our list of addresses, to see if the packet is for us.
 523          * If we don't have any addresses, assume any unicast packet
 524          * we receive might be for us (and let the upper layers deal
 525          * with it).
 526          */
 527         if (TAILQ_EMPTY(&in_ifaddrhead) &&
 528             (m->m_flags & (M_MCAST|M_BCAST)) == 0)
 529                 goto ours;
 530
 531         /*
 532          * Cache the destination address of the packet; this may be
 533          * changed by use of 'ipfw fwd'.
 534          */
 535         pkt_dst = args.next_hop ? args.next_hop->sin_addr : ip->ip_dst;
 536
 537         /*
 538          * Enable a consistency check between the destination address
 539          * and the arrival interface for a unicast packet (the RFC 1122
 540          * strong ES model) if IP forwarding is disabled and the packet
 541          * is not locally generated and the packet is not subject to
 542          * 'ipfw fwd'.
 543          *
 544          * XXX - Checking also should be disabled if the destination
 545          * address is ipnat'ed to a different interface.
 546          *
 547          * XXX - Checking is incompatible with IP aliases added
 548          * to the loopback interface instead of the interface where
 549          * the packets are received.
 550          */
 551         checkif = ip_checkinterface && (ipforwarding == 0) &&
 552             m->m_pkthdr.rcvif != NULL &&
 553             ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
 554             (args.next_hop == NULL);
 555
 556         /*
 557          * Check for exact addresses in the hash bucket.
 558          */
 559         LIST_FOREACH(ia, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
 560                 /*
 561                  * If the address matches, verify that the packet
 562                  * arrived via the correct interface if checking is
 563                  * enabled.
 564                  */
 565                 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
 566                     (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
 567                         goto ours;
 568         }
 569         /*
 570          * Check for broadcast addresses.
 571          *
 572          * Only accept broadcast packets that arrive via the matching
 573          * interface.  Reception of forwarded directed broadcasts would
 574          * be handled via ip_forward() and ether_output() with the loopback
 575          * into the stack for SIMPLEX interfaces handled by ether_output().
 576          */
 577         if (m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
 578                 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
 579                         if (ifa->ifa_addr->sa_family != AF_INET)
 580                                 continue;
 581                         ia = ifatoia(ifa);
 582                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 583                             pkt_dst.s_addr)
 584                                 goto ours;
 585                         if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
 586                                 goto ours;
 587 #ifdef BOOTP_COMPAT
 588                         if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
 589                                 goto ours;
 590 #endif
 591                 }
 592         }
 593         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
 594                 struct in_multi *inm;
 595                 if (ip_mrouter) {
 596                         /*
 597                          * If we are acting as a multicast router, all
 598                          * incoming multicast packets are passed to the
 599                          * kernel-level multicast forwarding function.
 600                          * The packet is returned (relatively) intact; if
 601                          * ip_mforward() returns a non-zero value, the packet
 602                          * must be discarded, else it may be accepted below.
 603                          */
 604                         if (ip_mforward &&
 605                             ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
 606                                 ipstat.ips_cantforward++;
 607                                 m_freem(m);
 608                                 return;
 609                         }
 610
 611                         /*
 612                          * The process-level routing daemon needs to receive
 613                          * all multicast IGMP packets, whether or not this
 614                          * host belongs to their destination groups.
 615                          */
 616                         if (ip->ip_p == IPPROTO_IGMP)
 617                                 goto ours;
 618                         ipstat.ips_forward++;
 619                 }
 620                 /*
 621                  * See if we belong to the destination multicast group on the
 622                  * arrival interface.
 623                  */
 624                 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
 625                 if (inm == NULL) {
 626                         ipstat.ips_notmember++;
 627                         m_freem(m);
 628                         return;
 629                 }
 630                 goto ours;
 631         }
 632         if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
 633                 goto ours;
 634         if (ip->ip_dst.s_addr == INADDR_ANY)
 635                 goto ours;
 636
 637         /*
 638          * FAITH(Firewall Aided Internet Translator)
 639          */
 640         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
 641                 if (ip_keepfaith) {
 642                         if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
 643                                 goto ours;
 644                 }
 645                 m_freem(m);
 646                 return;
 647         }
 648
 649         /*
 650          * Not for us; forward if possible and desirable.
 651          */
 652         if (ipforwarding == 0) {
 653                 ipstat.ips_cantforward++;
 654                 m_freem(m);
 655         } else {
 656 #ifdef IPSEC
 657                 /*
 658                  * Enforce inbound IPsec SPD.
 659                  */
 660                 if (ipsec4_in_reject(m, NULL)) {
 661                         ipsecstat.in_polvio++;
 662                         goto bad;
 663                 }
 664 #endif /* IPSEC */
 665 #ifdef FAST_IPSEC
 666                 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
 667                 s = splnet();
 668                 if (mtag != NULL) {
 669                         tdbi = (struct tdb_ident *)(mtag + 1);
 670                         sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
 671                 } else {
 672                         sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
 673                                                    IP_FORWARDING, &error);
 674                 }
 675                 if (sp == NULL) {       /* NB: can happen if error */
 676                         splx(s);
 677                         /*XXX error stat???*/
 678                         DPRINTF(("ip_input: no SP for forwarding\n"));  /*XXX*/
 679                         goto bad;
 680                 }
 681
 682                 /*
 683                  * Check security policy against packet attributes.
 684                  */
 685                 error = ipsec_in_reject(sp, m);
 686                 KEY_FREESP(&sp);
 687                 splx(s);
 688                 if (error) {
 689                         ipstat.ips_cantforward++;
 690                         goto bad;
 691                 }
 692 #endif /* FAST_IPSEC */
 693                 ip_forward(m, 0, args.next_hop);
 694         }
 695         return;
 696
 697 ours:
 698 #ifdef IPSTEALTH
 699         /*
 700          * IPSTEALTH: Process non-routing options only
 701          * if the packet is destined for us.
 702          */
 703         if (ipstealth && hlen > sizeof (struct ip) &&
 704             ip_dooptions(m, 1, args.next_hop))
 705                 return;
 706 #endif /* IPSTEALTH */
 707
 708         /* Count the packet in the ip address stats */
 709         if (ia != NULL) {
 710                 ia->ia_ifa.if_ipackets++;
 711                 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
 712         }
 713
 714         /*
 715          * If offset or IP_MF are set, must reassemble.
 716          * Otherwise, nothing need be done.
 717          * (We could look in the reassembly queue to see
 718          * if the packet was previously fragmented,
 719          * but it's not worth the time; just let them time out.)
 720          */
 721         if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
 722
 723                 /* If maxnipq is 0, never accept fragments. */
 724                 if (maxnipq == 0) {
 725                         ipstat.ips_fragments++;
 726                         ipstat.ips_fragdropped++;
 727                         goto bad;
 728                 }
 729
 730                 sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
 731                 /*
 732                  * Look for queue of fragments
 733                  * of this datagram.
 734                  */
 735                 for (fp = ipq[sum].next; fp != &ipq[sum]; fp = fp->next)
 736                         if (ip->ip_id == fp->ipq_id &&
 737                             ip->ip_src.s_addr == fp->ipq_src.s_addr &&
 738                             ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
 739                             ip->ip_p == fp->ipq_p)
 740                                 goto found;
 741
 742                 fp = 0;
 743
 744                 /*
 745                  * Enforce upper bound on number of fragmented packets
 746                  * for which we attempt reassembly;
 747                  * If maxnipq is -1, accept all fragments without limitation.
 748                  */
 749                 if ((nipq > maxnipq) && (maxnipq > 0)) {
 750                     /*
 751                      * drop something from the tail of the current queue
 752                      * before proceeding further
 753                      */
 754                     if (ipq[sum].prev == &ipq[sum]) {   /* gak */
 755                         for (i = 0; i < IPREASS_NHASH; i++) {
 756                             if (ipq[i].prev != &ipq[i]) {
 757                                 ipstat.ips_fragtimeout +=
 758                                     ipq[i].prev->ipq_nfrags;
 759                                 ip_freef(ipq[i].prev);
 760                                 break;
 761                             }
 762                         }
 763                     } else {
 764                         ipstat.ips_fragtimeout += ipq[sum].prev->ipq_nfrags;
 765                         ip_freef(ipq[sum].prev);
 766                     }
 767                 }
 768 found:
 769                 /*
 770                  * Adjust ip_len to not reflect header,
 771                  * convert offset of this to bytes.
 772                  */
 773                 ip->ip_len -= hlen;
 774                 if (ip->ip_off & IP_MF) {
 775                         /*
 776                          * Make sure that fragments have a data length
 777                          * that's a non-zero multiple of 8 bytes.
 778                          */
 779                         if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
 780                                 ipstat.ips_toosmall++; /* XXX */
 781                                 goto bad;
 782                         }
 783                         m->m_flags |= M_FRAG;
 784                 } else
 785                         m->m_flags &= ~M_FRAG;
 786                 ip->ip_off <<= 3;
 787
 788                 /*
 789                  * Attempt reassembly; if it succeeds, proceed.
 790                  * ip_reass() will return a different mbuf, and update
 791                  * the divert info in divert_info and args.divert_rule.
 792                  */
 793                 ipstat.ips_fragments++;
 794                 m->m_pkthdr.header = ip;
 795                 m = ip_reass(m,
 796                     fp, &ipq[sum], &divert_info, &args.divert_rule);
 797                 if (m == 0)
 798                         return;
 799                 ipstat.ips_reassembled++;
 800                 ip = mtod(m, struct ip *);
 801                 /* Get the header length of the reassembled packet */
 802                 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 803 #ifdef IPDIVERT
 804                 /* Restore original checksum before diverting packet */
 805                 if (divert_info != 0) {
 806                         ip->ip_len += hlen;
 807                         ip->ip_len = htons(ip->ip_len);
 808                         ip->ip_off = htons(ip->ip_off);
 809                         ip->ip_sum = 0;
 810                         if (hlen == sizeof(struct ip))
 811                                 ip->ip_sum = in_cksum_hdr(ip);
 812                         else
 813                                 ip->ip_sum = in_cksum(m, hlen);
 814                         ip->ip_off = ntohs(ip->ip_off);
 815                         ip->ip_len = ntohs(ip->ip_len);
 816                         ip->ip_len -= hlen;
 817                 }
 818 #endif
 819         } else
 820                 ip->ip_len -= hlen;
 821
 822 #ifdef IPDIVERT
 823         /*
 824          * Divert or tee packet to the divert protocol if required.
 825          */
 826         if (divert_info != 0) {
 827                 struct mbuf *clone = NULL;
 828
 829                 /* Clone packet if we're doing a 'tee' */
 830                 if ((divert_info & IP_FW_PORT_TEE_FLAG) != 0)
 831                         clone = m_dup(m, M_DONTWAIT);
 832
 833                 /* Restore packet header fields to original values */
 834                 ip->ip_len += hlen;
 835                 ip->ip_len = htons(ip->ip_len);
 836                 ip->ip_off = htons(ip->ip_off);
 837
 838                 /* Deliver packet to divert input routine */
 839                 divert_packet(m, 1, divert_info & 0xffff, args.divert_rule);
 840                 ipstat.ips_delivered++;
 841
 842                 /* If 'tee', continue with original packet */
 843                 if (clone == NULL)
 844                         return;
 845                 m = clone;
 846                 ip = mtod(m, struct ip *);
 847                 ip->ip_len += hlen;
 848                 /*
 849                  * Jump backwards to complete processing of the
 850                  * packet. But first clear divert_info to avoid
 851                  * entering this block again.
 852                  * We do not need to clear args.divert_rule
 853                  * or args.next_hop as they will not be used.
 854                  */
 855                 divert_info = 0;
 856                 goto pass;
 857         }
 858 #endif
 859
 860 #ifdef IPSEC
 861         /*
 862          * enforce IPsec policy checking if we are seeing last header.
 863          * note that we do not visit this with protocols with pcb layer
 864          * code - like udp/tcp/raw ip.
 865          */
 866         if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0 &&
 867             ipsec4_in_reject(m, NULL)) {
 868                 ipsecstat.in_polvio++;
 869                 goto bad;
 870         }
 871 #endif
 872 #if FAST_IPSEC
 873         /*
 874          * enforce IPsec policy checking if we are seeing last header.
 875          * note that we do not visit this with protocols with pcb layer
 876          * code - like udp/tcp/raw ip.
 877          */
 878         if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) != 0) {
 879                 /*
 880                  * Check if the packet has already had IPsec processing
 881                  * done.  If so, then just pass it along.  This tag gets
 882                  * set during AH, ESP, etc. input handling, before the
 883                  * packet is returned to the ip input queue for delivery.
 884                  */
 885                 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
 886                 s = splnet();
 887                 if (mtag != NULL) {
 888                         tdbi = (struct tdb_ident *)(mtag + 1);
 889                         sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
 890                 } else {
 891                         sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
 892                                                    IP_FORWARDING, &error);
 893                 }
 894                 if (sp != NULL) {
 895                         /*
 896                          * Check security policy against packet attributes.
 897                          */
 898                         error = ipsec_in_reject(sp, m);
 899                         KEY_FREESP(&sp);
 900                 } else {
 901                         /* XXX error stat??? */
 902                         error = EINVAL;
 903 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
 904                         goto bad;
 905                 }
 906                 splx(s);
 907                 if (error)
 908                         goto bad;
 909         }
 910 #endif /* FAST_IPSEC */
 911
 912         /*
 913          * Switch out to protocol's input routine.
 914          */
 915         ipstat.ips_delivered++;
 916         if (args.next_hop && ip->ip_p == IPPROTO_TCP) {
 917                 /* TCP needs IPFORWARD info if available */
 918                 struct m_hdr tag;
 919
 920                 tag.mh_type = MT_TAG;
 921                 tag.mh_flags = PACKET_TAG_IPFORWARD;
 922                 tag.mh_data = (caddr_t)args.next_hop;
 923                 tag.mh_next = m;
 924
 925                 (*inetsw[ip_protox[ip->ip_p]].pr_input)(
 926                         (struct mbuf *)&tag, hlen, ip->ip_p);
 927         } else
 928                 (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen, ip->ip_p);
 929         return;
 930 bad:
 931         m_freem(m);
 932 }
 933
 934 /*
 935  * IP software interrupt routine - to go away sometime soon
 936  */
 937 static void
 938 ipintr(void)
 939 {
 940         int s;
 941         struct mbuf *m;
 942
 943         while(1) {
 944                 s = splimp();
 945                 IF_DEQUEUE(&ipintrq, m);
 946                 splx(s);
 947                 if (m == 0)
 948                         return;
 949                 ip_input(m);
 950         }
 951 }
 952
 953 /*
 954  * Take incoming datagram fragment and try to reassemble it into
 955  * whole datagram.  If a chain for reassembly of this datagram already
 956  * exists, then it is given as fp; otherwise have to make a chain.
 957  *
 958  * When IPDIVERT enabled, keep additional state with each packet that
 959  * tells us if we need to divert or tee the packet we're building.
 960  * In particular, *divinfo includes the port and TEE flag,
 961  * *divert_rule is the number of the matching rule.
 962  */
 963
 964 static struct mbuf *
 965 ip_reass(struct mbuf *m, struct ipq *fp, struct ipq *where,
 966         u_int32_t *divinfo, u_int16_t *divert_rule)
 967 {
 968         struct ip *ip = mtod(m, struct ip *);
 969         struct mbuf *p = 0, *q, *nq;
 970         struct mbuf *t;
 971         int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 972         int i, next;
 973
 974         /*
 975          * Presence of header sizes in mbufs
 976          * would confuse code below.
 977          */
 978         m->m_data += hlen;
 979         m->m_len -= hlen;
 980
 981         /*
 982          * If first fragment to arrive, create a reassembly queue.
 983          */
 984         if (fp == 0) {
 985                 if ((t = m_get(M_DONTWAIT, MT_FTABLE)) == NULL)
 986                         goto dropfrag;
 987                 fp = mtod(t, struct ipq *);
 988                 insque(fp, where);
 989                 nipq++;
 990                 fp->ipq_nfrags = 1;
 991                 fp->ipq_ttl = IPFRAGTTL;
 992                 fp->ipq_p = ip->ip_p;
 993                 fp->ipq_id = ip->ip_id;
 994                 fp->ipq_src = ip->ip_src;
 995                 fp->ipq_dst = ip->ip_dst;
 996                 fp->ipq_frags = m;
 997                 m->m_nextpkt = NULL;
 998 #ifdef IPDIVERT
 999                 fp->ipq_div_info = 0;
1000                 fp->ipq_div_cookie = 0;
1001 #endif
1002                 goto inserted;
1003         } else {
1004                 fp->ipq_nfrags++;
1005         }
1006
1007 #define GETIP(m)        ((struct ip*)((m)->m_pkthdr.header))
1008
1009         /*
1010          * Find a segment which begins after this one does.
1011          */
1012         for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
1013                 if (GETIP(q)->ip_off > ip->ip_off)
1014                         break;
1015
1016         /*
1017          * If there is a preceding segment, it may provide some of
1018          * our data already.  If so, drop the data from the incoming
1019          * segment.  If it provides all of our data, drop us, otherwise
1020          * stick new segment in the proper place.
1021          *
1022          * If some of the data is dropped from the the preceding
1023          * segment, then it's checksum is invalidated.
1024          */
1025         if (p) {
1026                 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1027                 if (i > 0) {
1028                         if (i >= ip->ip_len)
1029                                 goto dropfrag;
1030                         m_adj(m, i);
1031                         m->m_pkthdr.csum_flags = 0;
1032                         ip->ip_off += i;
1033                         ip->ip_len -= i;
1034                 }
1035                 m->m_nextpkt = p->m_nextpkt;
1036                 p->m_nextpkt = m;
1037         } else {
1038                 m->m_nextpkt = fp->ipq_frags;
1039                 fp->ipq_frags = m;
1040         }
1041
1042         /*
1043          * While we overlap succeeding segments trim them or,
1044          * if they are completely covered, dequeue them.
1045          */
1046         for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1047              q = nq) {
1048                 i = (ip->ip_off + ip->ip_len) -
1049                     GETIP(q)->ip_off;
1050                 if (i < GETIP(q)->ip_len) {
1051                         GETIP(q)->ip_len -= i;
1052                         GETIP(q)->ip_off += i;
1053                         m_adj(q, i);
1054                         q->m_pkthdr.csum_flags = 0;
1055                         break;
1056                 }
1057                 nq = q->m_nextpkt;
1058                 m->m_nextpkt = nq;
1059                 ipstat.ips_fragdropped++;
1060                 fp->ipq_nfrags--;
1061                 m_freem(q);
1062         }
1063
1064 inserted:
1065
1066 #ifdef IPDIVERT
1067         /*
1068          * Transfer firewall instructions to the fragment structure.
1069          * Only trust info in the fragment at offset 0.
1070          */
1071         if (ip->ip_off == 0) {
1072                 fp->ipq_div_info = *divinfo;
1073                 fp->ipq_div_cookie = *divert_rule;
1074         }
1075         *divinfo = 0;
1076         *divert_rule = 0;
1077 #endif
1078
1079         /*
1080          * Check for complete reassembly and perform frag per packet
1081          * limiting.
1082          *
1083          * Frag limiting is performed here so that the nth frag has
1084          * a chance to complete the packet before we drop the packet.
1085          * As a result, n+1 frags are actually allowed per packet, but
1086          * only n will ever be stored. (n = maxfragsperpacket.)
1087          *
1088          */
1089         next = 0;
1090         for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1091                 if (GETIP(q)->ip_off != next) {
1092                         if (fp->ipq_nfrags > maxfragsperpacket) {
1093                                 ipstat.ips_fragdropped += fp->ipq_nfrags;
1094                                 ip_freef(fp);
1095                         }
1096                         return (0);
1097                 }
1098                 next += GETIP(q)->ip_len;
1099         }
1100         /* Make sure the last packet didn't have the IP_MF flag */
1101         if (p->m_flags & M_FRAG) {
1102                 if (fp->ipq_nfrags > maxfragsperpacket) {
1103                         ipstat.ips_fragdropped += fp->ipq_nfrags;
1104                         ip_freef(fp);
1105                 }
1106                 return (0);
1107         }
1108
1109         /*
1110          * Reassembly is complete.  Make sure the packet is a sane size.
1111          */
1112         q = fp->ipq_frags;
1113         ip = GETIP(q);
1114         if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1115                 ipstat.ips_toolong++;
1116                 ipstat.ips_fragdropped += fp->ipq_nfrags;
1117                 ip_freef(fp);
1118                 return (0);
1119         }
1120
1121         /*
1122          * Concatenate fragments.
1123          */
1124         m = q;
1125         t = m->m_next;
1126         m->m_next = 0;
1127         m_cat(m, t);
1128         nq = q->m_nextpkt;
1129         q->m_nextpkt = 0;
1130         for (q = nq; q != NULL; q = nq) {
1131                 nq = q->m_nextpkt;
1132                 q->m_nextpkt = NULL;
1133                 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1134                 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1135                 m_cat(m, q);
1136         }
1137
1138 #ifdef IPDIVERT
1139         /*
1140          * Extract firewall instructions from the fragment structure.
1141          */
1142         *divinfo = fp->ipq_div_info;
1143         *divert_rule = fp->ipq_div_cookie;
1144 #endif
1145
1146         /*
1147          * Create header for new ip packet by
1148          * modifying header of first packet;
1149          * dequeue and discard fragment reassembly header.
1150          * Make header visible.
1151          */
1152         ip->ip_len = next;
1153         ip->ip_src = fp->ipq_src;
1154         ip->ip_dst = fp->ipq_dst;
1155         remque(fp);
1156         nipq--;
1157         (void) m_free(dtom(fp));
1158         m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1159         m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1160         /* some debugging cruft by sklower, below, will go away soon */
1161         if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1162                 int plen = 0;
1163                 for (t = m; t; t = t->m_next)
1164                         plen += t->m_len;
1165                 m->m_pkthdr.len = plen;
1166         }
1167         return (m);
1168
1169 dropfrag:
1170 #ifdef IPDIVERT
1171         *divinfo = 0;
1172         *divert_rule = 0;
1173 #endif
1174         ipstat.ips_fragdropped++;
1175         if (fp != 0)
1176                 fp->ipq_nfrags--;
1177         m_freem(m);
1178         return (0);
1179
1180 #undef GETIP
1181 }
1182
1183 /*
1184  * Free a fragment reassembly header and all
1185  * associated datagrams.
1186  */
1187 static void
1188 ip_freef(fp)
1189         struct ipq *fp;
1190 {
1191         struct mbuf *q;
1192
1193         while (fp->ipq_frags) {
1194                 q = fp->ipq_frags;
1195                 fp->ipq_frags = q->m_nextpkt;
1196                 m_freem(q);
1197         }
1198         remque(fp);
1199         (void) m_free(dtom(fp));
1200         nipq--;
1201 }
1202
1203 /*
1204  * IP timer processing;
1205  * if a timer expires on a reassembly
1206  * queue, discard it.
1207  */
1208 void
1209 ip_slowtimo()
1210 {
1211         struct ipq *fp;
1212         int s = splnet();
1213         int i;
1214
1215         for (i = 0; i < IPREASS_NHASH; i++) {
1216                 fp = ipq[i].next;
1217                 if (fp == 0)
1218                         continue;
1219                 while (fp != &ipq[i]) {
1220                         --fp->ipq_ttl;
1221                         fp = fp->next;
1222                         if (fp->prev->ipq_ttl == 0) {
1223                                 ipstat.ips_fragtimeout += fp->prev->ipq_nfrags;
1224                                 ip_freef(fp->prev);
1225                         }
1226                 }
1227         }
1228         /*
1229          * If we are over the maximum number of fragments
1230          * (due to the limit being lowered), drain off
1231          * enough to get down to the new limit.
1232          */
1233         if (maxnipq >= 0 && nipq > maxnipq) {
1234                 for (i = 0; i < IPREASS_NHASH; i++) {
1235                         while (nipq > maxnipq &&
1236                                 (ipq[i].next != &ipq[i])) {
1237                                 ipstat.ips_fragdropped +=
1238                                     ipq[i].next->ipq_nfrags;
1239                                 ip_freef(ipq[i].next);
1240                         }
1241                 }
1242         }
1243         ipflow_slowtimo();
1244         splx(s);
1245 }
1246
1247 /*
1248  * Drain off all datagram fragments.
1249  */
1250 void
1251 ip_drain()
1252 {
1253         int     i;
1254
1255         for (i = 0; i < IPREASS_NHASH; i++) {
1256                 while (ipq[i].next != &ipq[i]) {
1257                         ipstat.ips_fragdropped += ipq[i].next->ipq_nfrags;
1258                         ip_freef(ipq[i].next);
1259                 }
1260         }
1261         in_rtqdrain();
1262 }
1263
1264 /*
1265  * Do option processing on a datagram,
1266  * possibly discarding it if bad options are encountered,
1267  * or forwarding it if source-routed.
1268  * The pass argument is used when operating in the IPSTEALTH
1269  * mode to tell what options to process:
1270  * [LS]SRR (pass 0) or the others (pass 1).
1271  * The reason for as many as two passes is that when doing IPSTEALTH,
1272  * non-routing options should be processed only if the packet is for us.
1273  * Returns 1 if packet has been forwarded/freed,
1274  * 0 if the packet should be processed further.
1275  */
1276 static int
1277 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1278 {
1279         struct ip *ip = mtod(m, struct ip *);
1280         u_char *cp;
1281         struct in_ifaddr *ia;
1282         int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB, forward = 0;
1283         struct in_addr *sin, dst;
1284         n_time ntime;
1285
1286         dst = ip->ip_dst;
1287         cp = (u_char *)(ip + 1);
1288         cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1289         for (; cnt > 0; cnt -= optlen, cp += optlen) {
1290                 opt = cp[IPOPT_OPTVAL];
1291                 if (opt == IPOPT_EOL)
1292                         break;
1293                 if (opt == IPOPT_NOP)
1294                         optlen = 1;
1295                 else {
1296                         if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1297                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1298                                 goto bad;
1299                         }
1300                         optlen = cp[IPOPT_OLEN];
1301                         if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1302                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1303                                 goto bad;
1304                         }
1305                 }
1306                 switch (opt) {
1307
1308                 default:
1309                         break;
1310
1311                 /*
1312                  * Source routing with record.
1313                  * Find interface with current destination address.
1314                  * If none on this machine then drop if strictly routed,
1315                  * or do nothing if loosely routed.
1316                  * Record interface address and bring up next address
1317                  * component.  If strictly routed make sure next
1318                  * address is on directly accessible net.
1319                  */
1320                 case IPOPT_LSRR:
1321                 case IPOPT_SSRR:
1322 #ifdef IPSTEALTH
1323                         if (ipstealth && pass > 0)
1324                                 break;
1325 #endif
1326                         if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1327                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1328                                 goto bad;
1329                         }
1330                         if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1331                                 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1332                                 goto bad;
1333                         }
1334                         ipaddr.sin_addr = ip->ip_dst;
1335                         ia = (struct in_ifaddr *)
1336                                 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1337                         if (ia == 0) {
1338                                 if (opt == IPOPT_SSRR) {
1339                                         type = ICMP_UNREACH;
1340                                         code = ICMP_UNREACH_SRCFAIL;
1341                                         goto bad;
1342                                 }
1343                                 if (!ip_dosourceroute)
1344                                         goto nosourcerouting;
1345                                 /*
1346                                  * Loose routing, and not at next destination
1347                                  * yet; nothing to do except forward.
1348                                  */
1349                                 break;
1350                         }
1351                         off--;                  /* 0 origin */
1352                         if (off > optlen - (int)sizeof(struct in_addr)) {
1353                                 /*
1354                                  * End of source route.  Should be for us.
1355                                  */
1356                                 if (!ip_acceptsourceroute)
1357                                         goto nosourcerouting;
1358                                 save_rte(cp, ip->ip_src);
1359                                 break;
1360                         }
1361 #ifdef IPSTEALTH
1362                         if (ipstealth)
1363                                 goto dropit;
1364 #endif
1365                         if (!ip_dosourceroute) {
1366                                 if (ipforwarding) {
1367                                         char buf[16]; /* aaa.bbb.ccc.ddd\0 */
1368                                         /*
1369                                          * Acting as a router, so generate ICMP
1370                                          */
1371 nosourcerouting:
1372                                         strcpy(buf, inet_ntoa(ip->ip_dst));
1373                                         log(LOG_WARNING,
1374                                             "attempted source route from %s to %s\n",
1375                                             inet_ntoa(ip->ip_src), buf);
1376                                         type = ICMP_UNREACH;
1377                                         code = ICMP_UNREACH_SRCFAIL;
1378                                         goto bad;
1379                                 } else {
1380                                         /*
1381                                          * Not acting as a router, so silently drop.
1382                                          */
1383 #ifdef IPSTEALTH
1384 dropit:
1385 #endif
1386                                         ipstat.ips_cantforward++;
1387                                         m_freem(m);
1388                                         return (1);
1389                                 }
1390                         }
1391
1392                         /*
1393                          * locate outgoing interface
1394                          */
1395                         (void)memcpy(&ipaddr.sin_addr, cp + off,
1396                             sizeof(ipaddr.sin_addr));
1397
1398                         if (opt == IPOPT_SSRR) {
1399 #define INA     struct in_ifaddr *
1400 #define SA      struct sockaddr *
1401                             if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr)) == 0)
1402                                 ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1403                         } else
1404                                 ia = ip_rtaddr(ipaddr.sin_addr, &ipforward_rt);
1405                         if (ia == 0) {
1406                                 type = ICMP_UNREACH;
1407                                 code = ICMP_UNREACH_SRCFAIL;
1408                                 goto bad;
1409                         }
1410                         ip->ip_dst = ipaddr.sin_addr;
1411                         (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1412                             sizeof(struct in_addr));
1413                         cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1414                         /*
1415                          * Let ip_intr's mcast routing check handle mcast pkts
1416                          */
1417                         forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1418                         break;
1419
1420                 case IPOPT_RR:
1421 #ifdef IPSTEALTH
1422                         if (ipstealth && pass == 0)
1423                                 break;
1424 #endif
1425                         if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1426                                 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1427                                 goto bad;
1428                         }
1429                         if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1430                                 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1431                                 goto bad;
1432                         }
1433                         /*
1434                          * If no space remains, ignore.
1435                          */
1436                         off--;                  /* 0 origin */
1437                         if (off > optlen - (int)sizeof(struct in_addr))
1438                                 break;
1439                         (void)memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1440                             sizeof(ipaddr.sin_addr));
1441                         /*
1442                          * locate outgoing interface; if we're the destination,
1443                          * use the incoming interface (should be same).
1444                          */
1445                         if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == 0 &&
1446                             (ia = ip_rtaddr(ipaddr.sin_addr,
1447                             &ipforward_rt)) == 0) {
1448                                 type = ICMP_UNREACH;
1449                                 code = ICMP_UNREACH_HOST;
1450                                 goto bad;
1451                         }
1452                         (void)memcpy(cp + off, &(IA_SIN(ia)->sin_addr),
1453                             sizeof(struct in_addr));
1454                         cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1455                         break;
1456
1457                 case IPOPT_TS:
1458 #ifdef IPSTEALTH
1459                         if (ipstealth && pass == 0)
1460                                 break;
1461 #endif
1462                         code = cp - (u_char *)ip;
1463                         if (optlen < 4 || optlen > 40) {
1464                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1465                                 goto bad;
1466                         }
1467                         if ((off = cp[IPOPT_OFFSET]) < 5) {
1468                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1469                                 goto bad;
1470                         }
1471                         if (off > optlen - (int)sizeof(int32_t)) {
1472                                 cp[IPOPT_OFFSET + 1] += (1 << 4);
1473                                 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1474                                         code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1475                                         goto bad;
1476                                 }
1477                                 break;
1478                         }
1479                         off--;                          /* 0 origin */
1480                         sin = (struct in_addr *)(cp + off);
1481                         switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1482
1483                         case IPOPT_TS_TSONLY:
1484                                 break;
1485
1486                         case IPOPT_TS_TSANDADDR:
1487                                 if (off + sizeof(n_time) +
1488                                     sizeof(struct in_addr) > optlen) {
1489                                         code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1490                                         goto bad;
1491                                 }
1492                                 ipaddr.sin_addr = dst;
1493                                 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1494                                                             m->m_pkthdr.rcvif);
1495                                 if (ia == 0)
1496                                         continue;
1497                                 (void)memcpy(sin, &IA_SIN(ia)->sin_addr,
1498                                     sizeof(struct in_addr));
1499                                 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1500                                 off += sizeof(struct in_addr);
1501                                 break;
1502
1503                         case IPOPT_TS_PRESPEC:
1504                                 if (off + sizeof(n_time) +
1505                                     sizeof(struct in_addr) > optlen) {
1506                                         code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1507                                         goto bad;
1508                                 }
1509                                 (void)memcpy(&ipaddr.sin_addr, sin,
1510                                     sizeof(struct in_addr));
1511                                 if (ifa_ifwithaddr((SA)&ipaddr) == 0)
1512                                         continue;
1513                                 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1514                                 off += sizeof(struct in_addr);
1515                                 break;
1516
1517                         default:
1518                                 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1519                                 goto bad;
1520                         }
1521                         ntime = iptime();
1522                         (void)memcpy(cp + off, &ntime, sizeof(n_time));
1523                         cp[IPOPT_OFFSET] += sizeof(n_time);
1524                 }
1525         }
1526         if (forward && ipforwarding) {
1527                 ip_forward(m, 1, next_hop);
1528                 return (1);
1529         }
1530         return (0);
1531 bad:
1532         icmp_error(m, type, code, 0, 0);
1533         ipstat.ips_badoptions++;
1534         return (1);
1535 }
1536
1537 /*
1538  * Given address of next destination (final or next hop),
1539  * return internet address info of interface to be used to get there.
1540  */
1541 struct in_ifaddr *
1542 ip_rtaddr(dst, rt)
1543         struct in_addr dst;
1544         struct route *rt;
1545 {
1546         struct sockaddr_in *sin;
1547
1548         sin = (struct sockaddr_in *)&rt->ro_dst;
1549
1550         if (rt->ro_rt == 0 ||
1551             dst.s_addr != sin->sin_addr.s_addr) {
1552                 if (rt->ro_rt) {
1553                         RTFREE(rt->ro_rt);
1554                         rt->ro_rt = 0;
1555                 }
1556                 sin->sin_family = AF_INET;
1557                 sin->sin_len = sizeof(*sin);
1558                 sin->sin_addr = dst;
1559
1560                 rtalloc_ign(rt, RTF_PRCLONING);
1561         }
1562         if (rt->ro_rt == 0)
1563                 return ((struct in_ifaddr *)0);
1564         return (ifatoia(rt->ro_rt->rt_ifa));
1565 }
1566
1567 /*
1568  * Save incoming source route for use in replies,
1569  * to be picked up later by ip_srcroute if the receiver is interested.
1570  */
1571 void
1572 save_rte(option, dst)
1573         u_char *option;
1574         struct in_addr dst;
1575 {
1576         unsigned olen;
1577
1578         olen = option[IPOPT_OLEN];
1579 #ifdef DIAGNOSTIC
1580         if (ipprintfs)
1581                 printf("save_rte: olen %d\n", olen);
1582 #endif
1583         if (olen > sizeof(ip_srcrt) - (1 + sizeof(dst)))
1584                 return;
1585         bcopy(option, ip_srcrt.srcopt, olen);
1586         ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1587         ip_srcrt.dst = dst;
1588 }
1589
1590 /*
1591  * Retrieve incoming source route for use in replies,
1592  * in the same form used by setsockopt.
1593  * The first hop is placed before the options, will be removed later.
1594  */
1595 struct mbuf *
1596 ip_srcroute()
1597 {
1598         struct in_addr *p, *q;
1599         struct mbuf *m;
1600
1601         if (ip_nhops == 0)
1602                 return ((struct mbuf *)0);
1603         m = m_get(M_DONTWAIT, MT_HEADER);
1604         if (m == 0)
1605                 return ((struct mbuf *)0);
1606
1607 #define OPTSIZ  (sizeof(ip_srcrt.nop) + sizeof(ip_srcrt.srcopt))
1608
1609         /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1610         m->m_len = ip_nhops * sizeof(struct in_addr) + sizeof(struct in_addr) +
1611             OPTSIZ;
1612 #ifdef DIAGNOSTIC
1613         if (ipprintfs)
1614                 printf("ip_srcroute: nhops %d mlen %d", ip_nhops, m->m_len);
1615 #endif
1616
1617         /*
1618          * First save first hop for return route
1619          */
1620         p = &ip_srcrt.route[ip_nhops - 1];
1621         *(mtod(m, struct in_addr *)) = *p--;
1622 #ifdef DIAGNOSTIC
1623         if (ipprintfs)
1624                 printf(" hops %lx", (u_long)ntohl(mtod(m, struct in_addr *)->s_addr));
1625 #endif
1626
1627         /*
1628          * Copy option fields and padding (nop) to mbuf.
1629          */
1630         ip_srcrt.nop = IPOPT_NOP;
1631         ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1632         (void)memcpy(mtod(m, caddr_t) + sizeof(struct in_addr),
1633             &ip_srcrt.nop, OPTSIZ);
1634         q = (struct in_addr *)(mtod(m, caddr_t) +
1635             sizeof(struct in_addr) + OPTSIZ);
1636 #undef OPTSIZ
1637         /*
1638          * Record return path as an IP source route,
1639          * reversing the path (pointers are now aligned).
1640          */
1641         while (p >= ip_srcrt.route) {
1642 #ifdef DIAGNOSTIC
1643                 if (ipprintfs)
1644                         printf(" %lx", (u_long)ntohl(q->s_addr));
1645 #endif
1646                 *q++ = *p--;
1647         }
1648         /*
1649          * Last hop goes to final destination.
1650          */
1651         *q = ip_srcrt.dst;
1652 #ifdef DIAGNOSTIC
1653         if (ipprintfs)
1654                 printf(" %lx\n", (u_long)ntohl(q->s_addr));
1655 #endif
1656         return (m);
1657 }
1658
1659 /*
1660  * Strip out IP options, at higher
1661  * level protocol in the kernel.
1662  * Second argument is buffer to which options
1663  * will be moved, and return value is their length.
1664  * XXX should be deleted; last arg currently ignored.
1665  */
1666 void
1667 ip_stripoptions(m, mopt)
1668         struct mbuf *m;
1669         struct mbuf *mopt;
1670 {
1671         int i;
1672         struct ip *ip = mtod(m, struct ip *);
1673         caddr_t opts;
1674         int olen;
1675
1676         olen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof (struct ip);
1677         opts = (caddr_t)(ip + 1);
1678         i = m->m_len - (sizeof (struct ip) + olen);
1679         bcopy(opts + olen, opts, (unsigned)i);
1680         m->m_len -= olen;
1681         if (m->m_flags & M_PKTHDR)
1682                 m->m_pkthdr.len -= olen;
1683         ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1684 }
1685
1686 u_char inetctlerrmap[PRC_NCMDS] = {
1687         0,              0,              0,              0,
1688         0,              EMSGSIZE,       EHOSTDOWN,      EHOSTUNREACH,
1689         EHOSTUNREACH,   EHOSTUNREACH,   ECONNREFUSED,   ECONNREFUSED,
1690         EMSGSIZE,       EHOSTUNREACH,   0,              0,
1691         0,              0,              0,              0,
1692         ENOPROTOOPT,    ECONNREFUSED
1693 };
1694
1695 /*
1696  * Forward a packet.  If some error occurs return the sender
1697  * an icmp packet.  Note we can't always generate a meaningful
1698  * icmp message because icmp doesn't have a large enough repertoire
1699  * of codes and types.
1700  *
1701  * If not forwarding, just drop the packet.  This could be confusing
1702  * if ipforwarding was zero but some routing protocol was advancing
1703  * us as a gateway to somewhere.  However, we must let the routing
1704  * protocol deal with that.
1705  *
1706  * The srcrt parameter indicates whether the packet is being forwarded
1707  * via a source route.
1708  */
1709 static void
1710 ip_forward(struct mbuf *m, int srcrt, struct sockaddr_in *next_hop)
1711 {
1712         struct ip *ip = mtod(m, struct ip *);
1713         struct sockaddr_in *sin;
1714         struct rtentry *rt;
1715         int error, type = 0, code = 0;
1716         struct mbuf *mcopy;
1717         n_long dest;
1718         struct in_addr pkt_dst;
1719         struct ifnet *destifp;
1720 #if defined(IPSEC) || defined(FAST_IPSEC)
1721         struct ifnet dummyifp;
1722 #endif
1723
1724         dest = 0;
1725         /*
1726          * Cache the destination address of the packet; this may be
1727          * changed by use of 'ipfw fwd'.
1728          */
1729         pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
1730
1731 #ifdef DIAGNOSTIC
1732         if (ipprintfs)
1733                 printf("forward: src %lx dst %lx ttl %x\n",
1734                     (u_long)ip->ip_src.s_addr, (u_long)pkt_dst.s_addr,
1735                     ip->ip_ttl);
1736 #endif
1737
1738
1739         if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(pkt_dst) == 0) {
1740                 ipstat.ips_cantforward++;
1741                 m_freem(m);
1742                 return;
1743         }
1744 #ifdef IPSTEALTH
1745         if (!ipstealth) {
1746 #endif
1747                 if (ip->ip_ttl <= IPTTLDEC) {
1748                         icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
1749                             dest, 0);
1750                         return;
1751                 }
1752 #ifdef IPSTEALTH
1753         }
1754 #endif
1755
1756         sin = (struct sockaddr_in *)&ipforward_rt.ro_dst;
1757         if ((rt = ipforward_rt.ro_rt) == 0 ||
1758             pkt_dst.s_addr != sin->sin_addr.s_addr) {
1759                 if (ipforward_rt.ro_rt) {
1760                         RTFREE(ipforward_rt.ro_rt);
1761                         ipforward_rt.ro_rt = 0;
1762                 }
1763                 sin->sin_family = AF_INET;
1764                 sin->sin_len = sizeof(*sin);
1765                 sin->sin_addr = pkt_dst;
1766
1767                 rtalloc_ign(&ipforward_rt, RTF_PRCLONING);
1768                 if (ipforward_rt.ro_rt == 0) {
1769                         icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1770                         return;
1771                 }
1772                 rt = ipforward_rt.ro_rt;
1773         }
1774
1775         /*
1776          * Save the IP header and at most 8 bytes of the payload,
1777          * in case we need to generate an ICMP message to the src.
1778          *
1779          * XXX this can be optimized a lot by saving the data in a local
1780          * buffer on the stack (72 bytes at most), and only allocating the
1781          * mbuf if really necessary. The vast majority of the packets
1782          * are forwarded without having to send an ICMP back (either
1783          * because unnecessary, or because rate limited), so we are
1784          * really we are wasting a lot of work here.
1785          *
1786          * We don't use m_copy() because it might return a reference
1787          * to a shared cluster. Both this function and ip_output()
1788          * assume exclusive access to the IP header in `m', so any
1789          * data in a cluster may change before we reach icmp_error().
1790          */
1791         MGET(mcopy, M_DONTWAIT, m->m_type);
1792         if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
1793                 /*
1794                  * It's probably ok if the pkthdr dup fails (because
1795                  * the deep copy of the tag chain failed), but for now
1796                  * be conservative and just discard the copy since
1797                  * code below may some day want the tags.
1798                  */
1799                 m_free(mcopy);
1800                 mcopy = NULL;
1801         }
1802         if (mcopy != NULL) {
1803                 mcopy->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
1804                     (int)ip->ip_len);
1805                 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
1806         }
1807
1808 #ifdef IPSTEALTH
1809         if (!ipstealth) {
1810 #endif
1811                 ip->ip_ttl -= IPTTLDEC;
1812 #ifdef IPSTEALTH
1813         }
1814 #endif
1815
1816         /*
1817          * If forwarding packet using same interface that it came in on,
1818          * perhaps should send a redirect to sender to shortcut a hop.
1819          * Only send redirect if source is sending directly to us,
1820          * and if packet was not source routed (or has any options).
1821          * Also, don't send redirect if forwarding using a default route
1822          * or a route modified by a redirect.
1823          */
1824         if (rt->rt_ifp == m->m_pkthdr.rcvif &&
1825             (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
1826             satosin(rt_key(rt))->sin_addr.s_addr != 0 &&
1827             ipsendredirects && !srcrt && !next_hop) {
1828 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
1829                 u_long src = ntohl(ip->ip_src.s_addr);
1830
1831                 if (RTA(rt) &&
1832                     (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
1833                     if (rt->rt_flags & RTF_GATEWAY)
1834                         dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
1835                     else
1836                         dest = pkt_dst.s_addr;
1837                     /* Router requirements says to only send host redirects */
1838                     type = ICMP_REDIRECT;
1839                     code = ICMP_REDIRECT_HOST;
1840 #ifdef DIAGNOSTIC
1841                     if (ipprintfs)
1842                         printf("redirect (%d) to %lx\n", code, (u_long)dest);
1843 #endif
1844                 }
1845         }
1846
1847     {
1848         struct m_hdr tag;
1849
1850         if (next_hop) {
1851                 /* Pass IPFORWARD info if available */
1852
1853                 tag.mh_type = MT_TAG;
1854                 tag.mh_flags = PACKET_TAG_IPFORWARD;
1855                 tag.mh_data = (caddr_t)next_hop;
1856                 tag.mh_next = m;
1857                 m = (struct mbuf *)&tag;
1858         }
1859         error = ip_output(m, (struct mbuf *)0, &ipforward_rt,
1860                           IP_FORWARDING, 0, NULL);
1861     }
1862         if (error)
1863                 ipstat.ips_cantforward++;
1864         else {
1865                 ipstat.ips_forward++;
1866                 if (type)
1867                         ipstat.ips_redirectsent++;
1868                 else {
1869                         if (mcopy) {
1870                                 ipflow_create(&ipforward_rt, mcopy);
1871                                 m_freem(mcopy);
1872                         }
1873                         return;
1874                 }
1875         }
1876         if (mcopy == NULL)
1877                 return;
1878         destifp = NULL;
1879
1880         switch (error) {
1881
1882         case 0:                         /* forwarded, but need redirect */
1883                 /* type, code set above */
1884                 break;
1885
1886         case ENETUNREACH:               /* shouldn't happen, checked above */
1887         case EHOSTUNREACH:
1888         case ENETDOWN:
1889         case EHOSTDOWN:
1890         default:
1891                 type = ICMP_UNREACH;
1892                 code = ICMP_UNREACH_HOST;
1893                 break;
1894
1895         case EMSGSIZE:
1896                 type = ICMP_UNREACH;
1897                 code = ICMP_UNREACH_NEEDFRAG;
1898 #ifdef IPSEC
1899                 /*
1900                  * If the packet is routed over IPsec tunnel, tell the
1901                  * originator the tunnel MTU.
1902                  *      tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1903                  * XXX quickhack!!!
1904                  */
1905                 if (ipforward_rt.ro_rt) {
1906                         struct secpolicy *sp = NULL;
1907                         int ipsecerror;
1908                         int ipsechdr;
1909                         struct route *ro;
1910
1911                         sp = ipsec4_getpolicybyaddr(mcopy,
1912                                                     IPSEC_DIR_OUTBOUND,
1913                                                     IP_FORWARDING,
1914                                                     &ipsecerror);
1915
1916                         if (sp == NULL)
1917                                 destifp = ipforward_rt.ro_rt->rt_ifp;
1918                         else {
1919                                 /* count IPsec header size */
1920                                 ipsechdr = ipsec4_hdrsiz(mcopy,
1921                                                          IPSEC_DIR_OUTBOUND,
1922                                                          NULL);
1923
1924                                 /*
1925                                  * find the correct route for outer IPv4
1926                                  * header, compute tunnel MTU.
1927                                  *
1928                                  * XXX BUG ALERT
1929                                  * The "dummyifp" code relies upon the fact
1930                                  * that icmp_error() touches only ifp->if_mtu.
1931                                  */
1932                                 /*XXX*/
1933                                 destifp = NULL;
1934                                 if (sp->req != NULL
1935                                  && sp->req->sav != NULL
1936                                  && sp->req->sav->sah != NULL) {
1937                                         ro = &sp->req->sav->sah->sa_route;
1938                                         if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1939                                                 dummyifp.if_mtu =
1940                                                     ro->ro_rt->rt_ifp->if_mtu;
1941                                                 dummyifp.if_mtu -= ipsechdr;
1942                                                 destifp = &dummyifp;
1943                                         }
1944                                 }
1945
1946                                 key_freesp(sp);
1947                         }
1948                 }
1949 #elif FAST_IPSEC
1950                 /*
1951                  * If the packet is routed over IPsec tunnel, tell the
1952                  * originator the tunnel MTU.
1953                  *      tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
1954                  * XXX quickhack!!!
1955                  */
1956                 if (ipforward_rt.ro_rt) {
1957                         struct secpolicy *sp = NULL;
1958                         int ipsecerror;
1959                         int ipsechdr;
1960                         struct route *ro;
1961
1962                         sp = ipsec_getpolicybyaddr(mcopy,
1963                                                    IPSEC_DIR_OUTBOUND,
1964                                                    IP_FORWARDING,
1965                                                    &ipsecerror);
1966
1967                         if (sp == NULL)
1968                                 destifp = ipforward_rt.ro_rt->rt_ifp;
1969                         else {
1970                                 /* count IPsec header size */
1971                                 ipsechdr = ipsec4_hdrsiz(mcopy,
1972                                                          IPSEC_DIR_OUTBOUND,
1973                                                          NULL);
1974
1975                                 /*
1976                                  * find the correct route for outer IPv4
1977                                  * header, compute tunnel MTU.
1978                                  *
1979                                  * XXX BUG ALERT
1980                                  * The "dummyifp" code relies upon the fact
1981                                  * that icmp_error() touches only ifp->if_mtu.
1982                                  */
1983                                 /*XXX*/
1984                                 destifp = NULL;
1985                                 if (sp->req != NULL
1986                                  && sp->req->sav != NULL
1987                                  && sp->req->sav->sah != NULL) {
1988                                         ro = &sp->req->sav->sah->sa_route;
1989                                         if (ro->ro_rt && ro->ro_rt->rt_ifp) {
1990                                                 dummyifp.if_mtu =
1991                                                     ro->ro_rt->rt_ifp->if_mtu;
1992                                                 dummyifp.if_mtu -= ipsechdr;
1993                                                 destifp = &dummyifp;
1994                                         }
1995                                 }
1996
1997                                 KEY_FREESP(&sp);
1998                         }
1999                 }
2000 #else /* !IPSEC && !FAST_IPSEC */
2001                 if (ipforward_rt.ro_rt)
2002                         destifp = ipforward_rt.ro_rt->rt_ifp;
2003 #endif /*IPSEC*/
2004                 ipstat.ips_cantfrag++;
2005                 break;
2006
2007         case ENOBUFS:
2008                 /*
2009                  * A router should not generate ICMP_SOURCEQUENCH as
2010                  * required in RFC1812 Requirements for IP Version 4 Routers.
2011                  * Source quench could be a big problem under DoS attacks,
2012                  * or if the underlying interface is rate-limited.
2013                  * Those who need source quench packets may re-enable them
2014                  * via the net.inet.ip.sendsourcequench sysctl.
2015                  */
2016                 if (ip_sendsourcequench == 0) {
2017                         m_freem(mcopy);
2018                         return;
2019                 } else {
2020                         type = ICMP_SOURCEQUENCH;
2021                         code = 0;
2022                 }
2023                 break;
2024
2025         case EACCES:                    /* ipfw denied packet */
2026                 m_freem(mcopy);
2027                 return;
2028         }
2029         icmp_error(mcopy, type, code, dest, destifp);
2030 }
2031
2032 void
2033 ip_savecontrol(inp, mp, ip, m)
2034         struct inpcb *inp;
2035         struct mbuf **mp;
2036         struct ip *ip;
2037         struct mbuf *m;
2038 {
2039         if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2040                 struct timeval tv;
2041
2042                 microtime(&tv);
2043                 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2044                         SCM_TIMESTAMP, SOL_SOCKET);
2045                 if (*mp)
2046                         mp = &(*mp)->m_next;
2047         }
2048         if (inp->inp_flags & INP_RECVDSTADDR) {
2049                 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2050                     sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2051                 if (*mp)
2052                         mp = &(*mp)->m_next;
2053         }
2054 #ifdef notyet
2055         /* XXX
2056          * Moving these out of udp_input() made them even more broken
2057          * than they already were.
2058          */
2059         /* options were tossed already */
2060         if (inp->inp_flags & INP_RECVOPTS) {
2061                 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2062                     sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2063                 if (*mp)
2064                         mp = &(*mp)->m_next;
2065         }
2066         /* ip_srcroute doesn't do what we want here, need to fix */
2067         if (inp->inp_flags & INP_RECVRETOPTS) {
2068                 *mp = sbcreatecontrol((caddr_t) ip_srcroute(),
2069                     sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2070                 if (*mp)
2071                         mp = &(*mp)->m_next;
2072         }
2073 #endif
2074         if (inp->inp_flags & INP_RECVIF) {
2075                 struct ifnet *ifp;
2076                 struct sdlbuf {
2077                         struct sockaddr_dl sdl;
2078                         u_char  pad[32];
2079                 } sdlbuf;
2080                 struct sockaddr_dl *sdp;
2081                 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2082
2083                 if (((ifp = m->m_pkthdr.rcvif))
2084                 && ( ifp->if_index && (ifp->if_index <= if_index))) {
2085                         sdp = (struct sockaddr_dl *)(ifnet_addrs
2086                                         [ifp->if_index - 1]->ifa_addr);
2087                         /*
2088                          * Change our mind and don't try copy.
2089                          */
2090                         if ((sdp->sdl_family != AF_LINK)
2091                         || (sdp->sdl_len > sizeof(sdlbuf))) {
2092                                 goto makedummy;
2093                         }
2094                         bcopy(sdp, sdl2, sdp->sdl_len);
2095                 } else {
2096 makedummy:
2097                         sdl2->sdl_len
2098                                 = offsetof(struct sockaddr_dl, sdl_data[0]);
2099                         sdl2->sdl_family = AF_LINK;
2100                         sdl2->sdl_index = 0;
2101                         sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2102                 }
2103                 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2104                         IP_RECVIF, IPPROTO_IP);
2105                 if (*mp)
2106                         mp = &(*mp)->m_next;
2107         }
2108 }
2109
2110 /*
2111  * XXX these routines are called from the upper part of the kernel.
2112  *
2113  * They could also be moved to ip_mroute.c, since all the RSVP
2114  *  handling is done there already.
2115  */
2116 int
2117 ip_rsvp_init(struct socket *so)
2118 {
2119         if (so->so_type != SOCK_RAW ||
2120             so->so_proto->pr_protocol != IPPROTO_RSVP)
2121                 return EOPNOTSUPP;
2122
2123         if (ip_rsvpd != NULL)
2124                 return EADDRINUSE;
2125
2126         ip_rsvpd = so;
2127         /*
2128          * This may seem silly, but we need to be sure we don't over-increment
2129          * the RSVP counter, in case something slips up.
2130          */
2131         if (!ip_rsvp_on) {
2132                 ip_rsvp_on = 1;
2133                 rsvp_on++;
2134         }
2135
2136         return 0;
2137 }
2138
2139 int
2140 ip_rsvp_done(void)
2141 {
2142         ip_rsvpd = NULL;
2143         /*
2144          * This may seem silly, but we need to be sure we don't over-decrement
2145          * the RSVP counter, in case something slips up.
2146          */
2147         if (ip_rsvp_on) {
2148                 ip_rsvp_on = 0;
2149                 rsvp_on--;
2150         }
2151         return 0;
2152 }
2153
2154 void
2155 rsvp_input(struct mbuf *m, int off, int proto)  /* XXX must fixup manually */
2156 {
2157         if (rsvp_input_p) { /* call the real one if loaded */
2158                 rsvp_input_p(m, off, proto);
2159                 return;
2160         }
2161
2162         /* Can still get packets with rsvp_on = 0 if there is a local member
2163          * of the group to which the RSVP packet is addressed.  But in this
2164          * case we want to throw the packet away.
2165          */
2166
2167         if (!rsvp_on) {
2168                 m_freem(m);
2169                 return;
2170         }
2171
2172         if (ip_rsvpd != NULL) {
2173                 rip_input(m, off, proto);
2174                 return;
2175         }
2176         /* Drop the packet */
2177         m_freem(m);
2178 }
2179