sys/netinet/ip_input.c

   1 /*
   2  * Copyright (c) 2003, 2004 Jeffrey M. Hsu.  All rights reserved.
   3  * Copyright (c) 2003, 2004 The DragonFly Project.  All rights reserved.
   4  *
   5  * This code is derived from software contributed to The DragonFly Project
   6  * by Jeffrey M. Hsu.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the name of The DragonFly Project nor the names of its
  17  *    contributors may be used to endorse or promote products derived
  18  *    from this software without specific, prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
  24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
  26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
  28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  */
  33
  34 /*
  35  * Copyright (c) 1982, 1986, 1988, 1993
  36  *      The Regents of the University of California.  All rights reserved.
  37  *
  38  * Redistribution and use in source and binary forms, with or without
  39  * modification, are permitted provided that the following conditions
  40  * are met:
  41  * 1. Redistributions of source code must retain the above copyright
  42  *    notice, this list of conditions and the following disclaimer.
  43  * 2. Redistributions in binary form must reproduce the above copyright
  44  *    notice, this list of conditions and the following disclaimer in the
  45  *    documentation and/or other materials provided with the distribution.
  46  * 3. Neither the name of the University nor the names of its contributors
  47  *    may be used to endorse or promote products derived from this software
  48  *    without specific prior written permission.
  49  *
  50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  60  * SUCH DAMAGE.
  61  *
  62  *      @(#)ip_input.c  8.2 (Berkeley) 1/4/94
  63  * $FreeBSD: src/sys/netinet/ip_input.c,v 1.130.2.52 2003/03/07 07:01:28 silby Exp $
  64  */
  65
  66 #define _IP_VHL
  67
  68 #include "opt_bootp.h"
  69 #include "opt_ipdn.h"
  70 #include "opt_ipdivert.h"
  71 #include "opt_ipstealth.h"
  72 #include "opt_ipsec.h"
  73 #include "opt_rss.h"
  74
  75 #include <sys/param.h>
  76 #include <sys/systm.h>
  77 #include <sys/mbuf.h>
  78 #include <sys/malloc.h>
  79 #include <sys/mpipe.h>
  80 #include <sys/domain.h>
  81 #include <sys/protosw.h>
  82 #include <sys/socket.h>
  83 #include <sys/time.h>
  84 #include <sys/globaldata.h>
  85 #include <sys/thread.h>
  86 #include <sys/kernel.h>
  87 #include <sys/syslog.h>
  88 #include <sys/sysctl.h>
  89 #include <sys/in_cksum.h>
  90 #include <sys/lock.h>
  91
  92 #include <sys/mplock2.h>
  93
  94 #include <machine/stdarg.h>
  95
  96 #include <net/if.h>
  97 #include <net/if_types.h>
  98 #include <net/if_var.h>
  99 #include <net/if_dl.h>
 100 #include <net/pfil.h>
 101 #include <net/route.h>
 102 #include <net/netisr2.h>
 103
 104 #include <netinet/in.h>
 105 #include <netinet/in_systm.h>
 106 #include <netinet/in_var.h>
 107 #include <netinet/ip.h>
 108 #include <netinet/in_pcb.h>
 109 #include <netinet/ip_var.h>
 110 #include <netinet/ip_icmp.h>
 111 #include <netinet/ip_divert.h>
 112 #include <netinet/ip_flow.h>
 113
 114 #include <sys/thread2.h>
 115 #include <sys/msgport2.h>
 116 #include <net/netmsg2.h>
 117
 118 #include <sys/socketvar.h>
 119
 120 #include <net/ipfw/ip_fw.h>
 121 #include <net/dummynet/ip_dummynet.h>
 122
 123 #ifdef IPSEC
 124 #include <netinet6/ipsec.h>
 125 #include <netproto/key/key.h>
 126 #endif
 127
 128 #ifdef FAST_IPSEC
 129 #include <netproto/ipsec/ipsec.h>
 130 #include <netproto/ipsec/key.h>
 131 #endif
 132
 133 int rsvp_on = 0;
 134 static int ip_rsvp_on;
 135 struct socket *ip_rsvpd;
 136
 137 int ipforwarding = 0;
 138 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
 139     &ipforwarding, 0, "Enable IP forwarding between interfaces");
 140
 141 static int ipsendredirects = 1; /* XXX */
 142 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
 143     &ipsendredirects, 0, "Enable sending IP redirects");
 144
 145 int ip_defttl = IPDEFTTL;
 146 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
 147     &ip_defttl, 0, "Maximum TTL on IP packets");
 148
 149 static int ip_dosourceroute = 0;
 150 SYSCTL_INT(_net_inet_ip, IPCTL_SOURCEROUTE, sourceroute, CTLFLAG_RW,
 151     &ip_dosourceroute, 0, "Enable forwarding source routed IP packets");
 152
 153 static int ip_acceptsourceroute = 0;
 154 SYSCTL_INT(_net_inet_ip, IPCTL_ACCEPTSOURCEROUTE, accept_sourceroute,
 155     CTLFLAG_RW, &ip_acceptsourceroute, 0,
 156     "Enable accepting source routed IP packets");
 157
 158 static int ip_keepfaith = 0;
 159 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
 160     &ip_keepfaith, 0,
 161     "Enable packet capture for FAITH IPv4->IPv6 translator daemon");
 162
 163 static int maxnipq;
 164 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragpackets, CTLFLAG_RW,
 165     &maxnipq, 0,
 166     "Maximum number of IPv4 fragment reassembly queue entries");
 167
 168 static int maxfragsperpacket;
 169 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
 170     &maxfragsperpacket, 0,
 171     "Maximum number of IPv4 fragments allowed per packet");
 172
 173 static int ip_sendsourcequench = 0;
 174 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
 175     &ip_sendsourcequench, 0,
 176     "Enable the transmission of source quench packets");
 177
 178 int ip_do_randomid = 1;
 179 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
 180     &ip_do_randomid, 0,
 181     "Assign random ip_id values");
 182 /*
 183  * XXX - Setting ip_checkinterface mostly implements the receive side of
 184  * the Strong ES model described in RFC 1122, but since the routing table
 185  * and transmit implementation do not implement the Strong ES model,
 186  * setting this to 1 results in an odd hybrid.
 187  *
 188  * XXX - ip_checkinterface currently must be disabled if you use ipnat
 189  * to translate the destination address to another local interface.
 190  *
 191  * XXX - ip_checkinterface must be disabled if you add IP aliases
 192  * to the loopback interface instead of the interface where the
 193  * packets for those addresses are received.
 194  */
 195 static int ip_checkinterface = 0;
 196 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
 197     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
 198
 199 static u_long ip_hash_count = 0;
 200 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, hash_count, CTLFLAG_RD,
 201     &ip_hash_count, 0, "Number of packets hashed by IP");
 202
 203 #ifdef RSS_DEBUG
 204 static u_long ip_rehash_count = 0;
 205 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, rehash_count, CTLFLAG_RD,
 206     &ip_rehash_count, 0, "Number of packets rehashed by IP");
 207
 208 static u_long ip_dispatch_fast = 0;
 209 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_fast_count, CTLFLAG_RD,
 210     &ip_dispatch_fast, 0, "Number of packets handled on current CPU");
 211
 212 static u_long ip_dispatch_slow = 0;
 213 SYSCTL_ULONG(_net_inet_ip, OID_AUTO, dispatch_slow_count, CTLFLAG_RD,
 214     &ip_dispatch_slow, 0, "Number of packets messaged to another CPU");
 215 #endif
 216
 217 #ifdef DIAGNOSTIC
 218 static int ipprintfs = 0;
 219 #endif
 220
 221 extern  struct domain inetdomain;
 222 extern  struct protosw inetsw[];
 223 u_char  ip_protox[IPPROTO_MAX];
 224 struct  in_ifaddrhead in_ifaddrheads[MAXCPU];   /* first inet address */
 225 struct  in_ifaddrhashhead *in_ifaddrhashtbls[MAXCPU];
 226                                                 /* inet addr hash table */
 227 u_long  in_ifaddrhmask;                         /* mask for hash table */
 228
 229 static struct mbuf *ipforward_mtemp[MAXCPU];
 230
 231 struct ip_stats ipstats_percpu[MAXCPU] __cachealign;
 232
 233 static int
 234 sysctl_ipstats(SYSCTL_HANDLER_ARGS)
 235 {
 236         int cpu, error = 0;
 237
 238         for (cpu = 0; cpu < ncpus; ++cpu) {
 239                 if ((error = SYSCTL_OUT(req, &ipstats_percpu[cpu],
 240                                         sizeof(struct ip_stats))))
 241                         break;
 242                 if ((error = SYSCTL_IN(req, &ipstats_percpu[cpu],
 243                                        sizeof(struct ip_stats))))
 244                         break;
 245         }
 246
 247         return (error);
 248 }
 249 SYSCTL_PROC(_net_inet_ip, IPCTL_STATS, stats, (CTLTYPE_OPAQUE | CTLFLAG_RW),
 250     0, 0, sysctl_ipstats, "S,ip_stats", "IP statistics");
 251
 252 /* Packet reassembly stuff */
 253 #define IPREASS_NHASH_LOG2      6
 254 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
 255 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
 256 #define IPREASS_HASH(x,y)                                               \
 257     (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
 258
 259 TAILQ_HEAD(ipqhead, ipq);
 260 struct ipfrag_queue {
 261         int                     nipq;
 262         int                     timeo_inprog;
 263         struct netmsg_base      timeo_netmsg;
 264         struct netmsg_base      drain_netmsg;
 265         struct ipqhead          ipq[IPREASS_NHASH];
 266 } __cachealign;
 267
 268 static struct ipfrag_queue      ipfrag_queue_pcpu[MAXCPU];
 269
 270 #ifdef IPCTL_DEFMTU
 271 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
 272     &ip_mtu, 0, "Default MTU");
 273 #endif
 274
 275 #ifdef IPSTEALTH
 276 static int ipstealth = 0;
 277 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW, &ipstealth, 0, "");
 278 #else
 279 static const int ipstealth = 0;
 280 #endif
 281
 282 struct mbuf *(*ip_divert_p)(struct mbuf *, int, int);
 283
 284 struct pfil_head inet_pfil_hook;
 285
 286 /*
 287  * struct ip_srcrt_opt is used to store packet state while it travels
 288  * through the stack.
 289  *
 290  * XXX Note that the code even makes assumptions on the size and
 291  * alignment of fields inside struct ip_srcrt so e.g. adding some
 292  * fields will break the code.  This needs to be fixed.
 293  *
 294  * We need to save the IP options in case a protocol wants to respond
 295  * to an incoming packet over the same route if the packet got here
 296  * using IP source routing.  This allows connection establishment and
 297  * maintenance when the remote end is on a network that is not known
 298  * to us.
 299  */
 300 struct ip_srcrt {
 301         struct  in_addr dst;                    /* final destination */
 302         char    nop;                            /* one NOP to align */
 303         char    srcopt[IPOPT_OFFSET + 1];       /* OPTVAL, OLEN and OFFSET */
 304         struct  in_addr route[MAX_IPOPTLEN/sizeof(struct in_addr)];
 305 };
 306
 307 struct ip_srcrt_opt {
 308         int             ip_nhops;
 309         struct ip_srcrt ip_srcrt;
 310 };
 311
 312 #define IPFRAG_MPIPE_MAX        4096
 313 #define MAXIPFRAG_MIN           ((IPFRAG_MPIPE_MAX * 2) / 256)
 314
 315 static MALLOC_DEFINE(M_IPQ, "ipq", "IP Fragment Management");
 316 static struct malloc_pipe ipq_mpipe;
 317
 318 static void             save_rte(struct mbuf *, u_char *, struct in_addr);
 319 static int              ip_dooptions(struct mbuf *m, int, struct sockaddr_in *);
 320 static void             ip_freef(struct ipfrag_queue *, struct ipqhead *,
 321                             struct ipq *);
 322 static void             ip_input_handler(netmsg_t);
 323
 324 static void             ipfrag_timeo_dispatch(netmsg_t);
 325 static void             ipfrag_drain_dispatch(netmsg_t);
 326
 327 /*
 328  * IP initialization: fill in IP protocol switch table.
 329  * All protocols not implemented in kernel go to raw IP protocol handler.
 330  */
 331 void
 332 ip_init(void)
 333 {
 334         struct protosw *pr;
 335         int cpu, i;
 336
 337         /*
 338          * Make sure we can handle a reasonable number of fragments but
 339          * cap it at IPFRAG_MPIPE_MAX.
 340          */
 341         mpipe_init(&ipq_mpipe, M_IPQ, sizeof(struct ipq),
 342             IFQ_MAXLEN, IPFRAG_MPIPE_MAX, 0, NULL, NULL, NULL);
 343         for (cpu = 0; cpu < ncpus; ++cpu) {
 344                 TAILQ_INIT(&in_ifaddrheads[cpu]);
 345                 in_ifaddrhashtbls[cpu] =
 346                     hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
 347         }
 348         pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
 349         if (pr == NULL)
 350                 panic("ip_init");
 351         for (i = 0; i < IPPROTO_MAX; i++)
 352                 ip_protox[i] = pr - inetsw;
 353         for (pr = inetdomain.dom_protosw;
 354              pr < inetdomain.dom_protoswNPROTOSW; pr++) {
 355                 if (pr->pr_domain->dom_family == PF_INET && pr->pr_protocol) {
 356                         if (pr->pr_protocol != IPPROTO_RAW)
 357                                 ip_protox[pr->pr_protocol] = pr - inetsw;
 358                 }
 359         }
 360
 361         inet_pfil_hook.ph_type = PFIL_TYPE_AF;
 362         inet_pfil_hook.ph_af = AF_INET;
 363         if ((i = pfil_head_register(&inet_pfil_hook)) != 0) {
 364                 kprintf("%s: WARNING: unable to register pfil hook, "
 365                         "error %d\n", __func__, i);
 366         }
 367
 368         maxnipq = (nmbclusters / 32) / ncpus;
 369         if (maxnipq < MAXIPFRAG_MIN)
 370                 maxnipq = MAXIPFRAG_MIN;
 371         maxfragsperpacket = 16;
 372
 373         ip_id = time_second & 0xffff;   /* time_second survives reboots */
 374
 375         for (cpu = 0; cpu < ncpus; ++cpu) {
 376                 /*
 377                  * Initialize IP statistics counters for each CPU.
 378                  */
 379                 bzero(&ipstats_percpu[cpu], sizeof(struct ip_stats));
 380
 381                 /*
 382                  * Preallocate mbuf template for forwarding
 383                  */
 384                 MGETHDR(ipforward_mtemp[cpu], M_WAITOK, MT_DATA);
 385
 386                 /*
 387                  * Initialize per-cpu ip fragments queues
 388                  */
 389                 for (i = 0; i < IPREASS_NHASH; i++) {
 390                         struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[cpu];
 391
 392                         TAILQ_INIT(&fragq->ipq[i]);
 393                         netmsg_init(&fragq->timeo_netmsg, NULL,
 394                             &netisr_adone_rport, MSGF_PRIORITY,
 395                             ipfrag_timeo_dispatch);
 396                         netmsg_init(&fragq->drain_netmsg, NULL,
 397                             &netisr_adone_rport, MSGF_PRIORITY,
 398                             ipfrag_drain_dispatch);
 399                 }
 400         }
 401
 402         netisr_register(NETISR_IP, ip_input_handler, ip_hashfn);
 403         netisr_register_hashcheck(NETISR_IP, ip_hashcheck);
 404 }
 405
 406 /* Do transport protocol processing. */
 407 static void
 408 transport_processing_oncpu(struct mbuf *m, int hlen, struct ip *ip)
 409 {
 410         const struct protosw *pr = &inetsw[ip_protox[ip->ip_p]];
 411
 412         /*
 413          * Switch out to protocol's input routine.
 414          */
 415         PR_GET_MPLOCK(pr);
 416         pr->pr_input(&m, &hlen, ip->ip_p);
 417         PR_REL_MPLOCK(pr);
 418 }
 419
 420 static void
 421 transport_processing_handler(netmsg_t msg)
 422 {
 423         struct netmsg_packet *pmsg = &msg->packet;
 424         struct ip *ip;
 425         int hlen;
 426
 427         ip = mtod(pmsg->nm_packet, struct ip *);
 428         hlen = pmsg->base.lmsg.u.ms_result;
 429
 430         transport_processing_oncpu(pmsg->nm_packet, hlen, ip);
 431         /* msg was embedded in the mbuf, do not reply! */
 432 }
 433
 434 static void
 435 ip_input_handler(netmsg_t msg)
 436 {
 437         ip_input(msg->packet.nm_packet);
 438         /* msg was embedded in the mbuf, do not reply! */
 439 }
 440
 441 /*
 442  * IP input routine.  Checksum and byte swap header.  If fragmented
 443  * try to reassemble.  Process options.  Pass to next level.
 444  */
 445 void
 446 ip_input(struct mbuf *m)
 447 {
 448         struct ip *ip;
 449         struct in_ifaddr *ia = NULL;
 450         struct in_ifaddr_container *iac;
 451         int hlen, checkif;
 452         u_short sum;
 453         struct in_addr pkt_dst;
 454         boolean_t using_srcrt = FALSE;          /* forward (by PFIL_HOOKS) */
 455         struct in_addr odst;                    /* original dst address(NAT) */
 456         struct m_tag *mtag;
 457         struct sockaddr_in *next_hop = NULL;
 458         lwkt_port_t port;
 459 #ifdef FAST_IPSEC
 460         struct tdb_ident *tdbi;
 461         struct secpolicy *sp;
 462         int error;
 463 #endif
 464
 465         M_ASSERTPKTHDR(m);
 466
 467         /*
 468          * This routine is called from numerous places which may not have
 469          * characterized the packet.
 470          */
 471         ip = mtod(m, struct ip *);
 472         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
 473             (ntohs(ip->ip_off) & (IP_MF | IP_OFFMASK))) {
 474                 /*
 475                  * Force hash recalculation for fragments and multicast
 476                  * packets; hardware may not do it correctly.
 477                  * XXX add flag to indicate the hash is from hardware
 478                  */
 479                 m->m_flags &= ~M_HASH;
 480         }
 481         if ((m->m_flags & M_HASH) == 0) {
 482                 ip_hashfn(&m, 0);
 483                 if (m == NULL)
 484                         return;
 485                 KKASSERT(m->m_flags & M_HASH);
 486
 487                 if (&curthread->td_msgport !=
 488                     netisr_hashport(m->m_pkthdr.hash)) {
 489                         netisr_queue(NETISR_IP, m);
 490                         /* Requeued to other netisr msgport; done */
 491                         return;
 492                 }
 493
 494                 /* mbuf could have been changed */
 495                 ip = mtod(m, struct ip *);
 496         }
 497
 498         /*
 499          * Pull out certain tags
 500          */
 501         if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
 502                 /* Next hop */
 503                 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
 504                 KKASSERT(mtag != NULL);
 505                 next_hop = m_tag_data(mtag);
 506         }
 507
 508         if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
 509                 /* dummynet already filtered us */
 510                 ip = mtod(m, struct ip *);
 511                 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 512                 goto iphack;
 513         }
 514
 515         ipstat.ips_total++;
 516
 517         /* length checks already done in ip_hashfn() */
 518         KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
 519
 520         if (IP_VHL_V(ip->ip_vhl) != IPVERSION) {
 521                 ipstat.ips_badvers++;
 522                 goto bad;
 523         }
 524
 525         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 526         /* length checks already done in ip_hashfn() */
 527         KASSERT(hlen >= sizeof(struct ip), ("IP header len too small"));
 528         KASSERT(m->m_len >= hlen, ("complete IP header not in one mbuf"));
 529
 530         /* 127/8 must not appear on wire - RFC1122 */
 531         if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
 532             (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
 533                 if (!(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) {
 534                         ipstat.ips_badaddr++;
 535                         goto bad;
 536                 }
 537         }
 538
 539         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
 540                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
 541         } else {
 542                 if (hlen == sizeof(struct ip))
 543                         sum = in_cksum_hdr(ip);
 544                 else
 545                         sum = in_cksum(m, hlen);
 546         }
 547         if (sum != 0) {
 548                 ipstat.ips_badsum++;
 549                 goto bad;
 550         }
 551
 552 #ifdef ALTQ
 553         if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) {
 554                 /* packet is dropped by traffic conditioner */
 555                 return;
 556         }
 557 #endif
 558         /*
 559          * Convert fields to host representation.
 560          */
 561         ip->ip_len = ntohs(ip->ip_len);
 562         ip->ip_off = ntohs(ip->ip_off);
 563
 564         /* length checks already done in ip_hashfn() */
 565         KASSERT(ip->ip_len >= hlen, ("total length less then header length"));
 566         KASSERT(m->m_pkthdr.len >= ip->ip_len, ("mbuf too short"));
 567
 568         /*
 569          * Trim mbufs if longer than the IP header would have us expect.
 570          */
 571         if (m->m_pkthdr.len > ip->ip_len) {
 572                 if (m->m_len == m->m_pkthdr.len) {
 573                         m->m_len = ip->ip_len;
 574                         m->m_pkthdr.len = ip->ip_len;
 575                 } else {
 576                         m_adj(m, ip->ip_len - m->m_pkthdr.len);
 577                 }
 578         }
 579 #if defined(IPSEC) && !defined(IPSEC_FILTERGIF)
 580         /*
 581          * Bypass packet filtering for packets from a tunnel (gif).
 582          */
 583         if (ipsec_gethist(m, NULL))
 584                 goto pass;
 585 #endif
 586
 587         /*
 588          * IpHack's section.
 589          * Right now when no processing on packet has done
 590          * and it is still fresh out of network we do our black
 591          * deals with it.
 592          * - Firewall: deny/allow/divert
 593          * - Xlate: translate packet's addr/port (NAT).
 594          * - Pipe: pass pkt through dummynet.
 595          * - Wrap: fake packet's addr/port <unimpl.>
 596          * - Encapsulate: put it in another IP and send out. <unimp.>
 597          */
 598
 599 iphack:
 600         /*
 601          * If we've been forwarded from the output side, then
 602          * skip the firewall a second time
 603          */
 604         if (next_hop != NULL)
 605                 goto ours;
 606
 607         /* No pfil hooks */
 608         if (!pfil_has_hooks(&inet_pfil_hook)) {
 609                 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
 610                         /*
 611                          * Strip dummynet tags from stranded packets
 612                          */
 613                         mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
 614                         KKASSERT(mtag != NULL);
 615                         m_tag_delete(m, mtag);
 616                         m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
 617                 }
 618                 goto pass;
 619         }
 620
 621         /*
 622          * Run through list of hooks for input packets.
 623          *
 624          * NOTE!  If the packet is rewritten pf/ipfw/whoever must
 625          *        clear M_HASH.
 626          */
 627         odst = ip->ip_dst;
 628         if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN))
 629                 return;
 630         if (m == NULL)  /* consumed by filter */
 631                 return;
 632         ip = mtod(m, struct ip *);
 633         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 634         using_srcrt = (odst.s_addr != ip->ip_dst.s_addr);
 635
 636         if (m->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
 637                 mtag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL);
 638                 KKASSERT(mtag != NULL);
 639                 next_hop = m_tag_data(mtag);
 640         }
 641         if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
 642                 ip_dn_queue(m);
 643                 return;
 644         }
 645         if (m->m_pkthdr.fw_flags & FW_MBUF_REDISPATCH) {
 646                 m->m_pkthdr.fw_flags &= ~FW_MBUF_REDISPATCH;
 647         }
 648 pass:
 649         /*
 650          * Process options and, if not destined for us,
 651          * ship it on.  ip_dooptions returns 1 when an
 652          * error was detected (causing an icmp message
 653          * to be sent and the original packet to be freed).
 654          */
 655         if (hlen > sizeof(struct ip) && ip_dooptions(m, 0, next_hop))
 656                 return;
 657
 658         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
 659          * matter if it is destined to another node, or whether it is
 660          * a multicast one, RSVP wants it! and prevents it from being forwarded
 661          * anywhere else. Also checks if the rsvp daemon is running before
 662          * grabbing the packet.
 663          */
 664         if (rsvp_on && ip->ip_p == IPPROTO_RSVP)
 665                 goto ours;
 666
 667         /*
 668          * Check our list of addresses, to see if the packet is for us.
 669          * If we don't have any addresses, assume any unicast packet
 670          * we receive might be for us (and let the upper layers deal
 671          * with it).
 672          */
 673         if (TAILQ_EMPTY(&in_ifaddrheads[mycpuid]) &&
 674             !(m->m_flags & (M_MCAST | M_BCAST)))
 675                 goto ours;
 676
 677         /*
 678          * Cache the destination address of the packet; this may be
 679          * changed by use of 'ipfw fwd'.
 680          */
 681         pkt_dst = next_hop ? next_hop->sin_addr : ip->ip_dst;
 682
 683         /*
 684          * Enable a consistency check between the destination address
 685          * and the arrival interface for a unicast packet (the RFC 1122
 686          * strong ES model) if IP forwarding is disabled and the packet
 687          * is not locally generated and the packet is not subject to
 688          * 'ipfw fwd'.
 689          *
 690          * XXX - Checking also should be disabled if the destination
 691          * address is ipnat'ed to a different interface.
 692          *
 693          * XXX - Checking is incompatible with IP aliases added
 694          * to the loopback interface instead of the interface where
 695          * the packets are received.
 696          */
 697         checkif = ip_checkinterface &&
 698                   !ipforwarding &&
 699                   m->m_pkthdr.rcvif != NULL &&
 700                   !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) &&
 701                   next_hop == NULL;
 702
 703         /*
 704          * Check for exact addresses in the hash bucket.
 705          */
 706         LIST_FOREACH(iac, INADDR_HASH(pkt_dst.s_addr), ia_hash) {
 707                 ia = iac->ia;
 708
 709                 /*
 710                  * If the address matches, verify that the packet
 711                  * arrived via the correct interface if checking is
 712                  * enabled.
 713                  */
 714                 if (IA_SIN(ia)->sin_addr.s_addr == pkt_dst.s_addr &&
 715                     (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
 716                         goto ours;
 717         }
 718         ia = NULL;
 719
 720         /*
 721          * Check for broadcast addresses.
 722          *
 723          * Only accept broadcast packets that arrive via the matching
 724          * interface.  Reception of forwarded directed broadcasts would
 725          * be handled via ip_forward() and ether_output() with the loopback
 726          * into the stack for SIMPLEX interfaces handled by ether_output().
 727          */
 728         if (m->m_pkthdr.rcvif != NULL &&
 729             m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
 730                 struct ifaddr_container *ifac;
 731
 732                 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
 733                               ifa_link) {
 734                         struct ifaddr *ifa = ifac->ifa;
 735
 736                         if (ifa->ifa_addr == NULL) /* shutdown/startup race */
 737                                 continue;
 738                         if (ifa->ifa_addr->sa_family != AF_INET)
 739                                 continue;
 740                         ia = ifatoia(ifa);
 741                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 742                                                                 pkt_dst.s_addr)
 743                                 goto ours;
 744                         if (ia->ia_netbroadcast.s_addr == pkt_dst.s_addr)
 745                                 goto ours;
 746 #ifdef BOOTP_COMPAT
 747                         if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
 748                                 goto ours;
 749 #endif
 750                 }
 751         }
 752         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
 753                 struct in_multi *inm;
 754
 755                 if (ip_mrouter != NULL) {
 756                         /* XXX Multicast routing is not MPSAFE yet */
 757                         get_mplock();
 758
 759                         /*
 760                          * If we are acting as a multicast router, all
 761                          * incoming multicast packets are passed to the
 762                          * kernel-level multicast forwarding function.
 763                          * The packet is returned (relatively) intact; if
 764                          * ip_mforward() returns a non-zero value, the packet
 765                          * must be discarded, else it may be accepted below.
 766                          */
 767                         if (ip_mforward != NULL &&
 768                             ip_mforward(ip, m->m_pkthdr.rcvif, m, NULL) != 0) {
 769                                 rel_mplock();
 770                                 ipstat.ips_cantforward++;
 771                                 m_freem(m);
 772                                 return;
 773                         }
 774
 775                         rel_mplock();
 776
 777                         /*
 778                          * The process-level routing daemon needs to receive
 779                          * all multicast IGMP packets, whether or not this
 780                          * host belongs to their destination groups.
 781                          */
 782                         if (ip->ip_p == IPPROTO_IGMP)
 783                                 goto ours;
 784                         ipstat.ips_forward++;
 785                 }
 786                 /*
 787                  * See if we belong to the destination multicast group on the
 788                  * arrival interface.
 789                  */
 790                 inm = IN_LOOKUP_MULTI(&ip->ip_dst, m->m_pkthdr.rcvif);
 791                 if (inm == NULL) {
 792                         ipstat.ips_notmember++;
 793                         m_freem(m);
 794                         return;
 795                 }
 796                 goto ours;
 797         }
 798         if (ip->ip_dst.s_addr == INADDR_BROADCAST)
 799                 goto ours;
 800         if (ip->ip_dst.s_addr == INADDR_ANY)
 801                 goto ours;
 802
 803         /*
 804          * FAITH(Firewall Aided Internet Translator)
 805          */
 806         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
 807                 if (ip_keepfaith) {
 808                         if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP)
 809                                 goto ours;
 810                 }
 811                 m_freem(m);
 812                 return;
 813         }
 814
 815         /*
 816          * Not for us; forward if possible and desirable.
 817          */
 818         if (!ipforwarding) {
 819                 ipstat.ips_cantforward++;
 820                 m_freem(m);
 821         } else {
 822 #ifdef IPSEC
 823                 /*
 824                  * Enforce inbound IPsec SPD.
 825                  */
 826                 if (ipsec4_in_reject(m, NULL)) {
 827                         ipsecstat.in_polvio++;
 828                         goto bad;
 829                 }
 830 #endif
 831 #ifdef FAST_IPSEC
 832                 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
 833                 crit_enter();
 834                 if (mtag != NULL) {
 835                         tdbi = (struct tdb_ident *)m_tag_data(mtag);
 836                         sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
 837                 } else {
 838                         sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
 839                                                    IP_FORWARDING, &error);
 840                 }
 841                 if (sp == NULL) {       /* NB: can happen if error */
 842                         crit_exit();
 843                         /*XXX error stat???*/
 844                         DPRINTF(("ip_input: no SP for forwarding\n"));  /*XXX*/
 845                         goto bad;
 846                 }
 847
 848                 /*
 849                  * Check security policy against packet attributes.
 850                  */
 851                 error = ipsec_in_reject(sp, m);
 852                 KEY_FREESP(&sp);
 853                 crit_exit();
 854                 if (error) {
 855                         ipstat.ips_cantforward++;
 856                         goto bad;
 857                 }
 858 #endif
 859                 ip_forward(m, using_srcrt, next_hop);
 860         }
 861         return;
 862
 863 ours:
 864
 865         /*
 866          * IPSTEALTH: Process non-routing options only
 867          * if the packet is destined for us.
 868          */
 869         if (ipstealth &&
 870             hlen > sizeof(struct ip) &&
 871             ip_dooptions(m, 1, next_hop))
 872                 return;
 873
 874         /* Count the packet in the ip address stats */
 875         if (ia != NULL) {
 876                 IFA_STAT_INC(&ia->ia_ifa, ipackets, 1);
 877                 IFA_STAT_INC(&ia->ia_ifa, ibytes, m->m_pkthdr.len);
 878         }
 879
 880         /*
 881          * If offset or IP_MF are set, must reassemble.
 882          * Otherwise, nothing need be done.
 883          * (We could look in the reassembly queue to see
 884          * if the packet was previously fragmented,
 885          * but it's not worth the time; just let them time out.)
 886          */
 887         if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
 888                 /*
 889                  * Attempt reassembly; if it succeeds, proceed.  ip_reass()
 890                  * will return a different mbuf.
 891                  *
 892                  * NOTE: ip_reass() returns m with M_HASH cleared to force
 893                  *       us to recharacterize the packet.
 894                  */
 895                 m = ip_reass(m);
 896                 if (m == NULL)
 897                         return;
 898                 ip = mtod(m, struct ip *);
 899
 900                 /* Get the header length of the reassembled packet */
 901                 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 902         } else {
 903                 ip->ip_len -= hlen;
 904         }
 905
 906 #ifdef IPSEC
 907         /*
 908          * enforce IPsec policy checking if we are seeing last header.
 909          * note that we do not visit this with protocols with pcb layer
 910          * code - like udp/tcp/raw ip.
 911          */
 912         if ((inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) &&
 913             ipsec4_in_reject(m, NULL)) {
 914                 ipsecstat.in_polvio++;
 915                 goto bad;
 916         }
 917 #endif
 918 #ifdef FAST_IPSEC
 919         /*
 920          * enforce IPsec policy checking if we are seeing last header.
 921          * note that we do not visit this with protocols with pcb layer
 922          * code - like udp/tcp/raw ip.
 923          */
 924         if (inetsw[ip_protox[ip->ip_p]].pr_flags & PR_LASTHDR) {
 925                 /*
 926                  * Check if the packet has already had IPsec processing
 927                  * done.  If so, then just pass it along.  This tag gets
 928                  * set during AH, ESP, etc. input handling, before the
 929                  * packet is returned to the ip input queue for delivery.
 930                  */
 931                 mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
 932                 crit_enter();
 933                 if (mtag != NULL) {
 934                         tdbi = (struct tdb_ident *)m_tag_data(mtag);
 935                         sp = ipsec_getpolicy(tdbi, IPSEC_DIR_INBOUND);
 936                 } else {
 937                         sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
 938                                                    IP_FORWARDING, &error);
 939                 }
 940                 if (sp != NULL) {
 941                         /*
 942                          * Check security policy against packet attributes.
 943                          */
 944                         error = ipsec_in_reject(sp, m);
 945                         KEY_FREESP(&sp);
 946                 } else {
 947                         /* XXX error stat??? */
 948                         error = EINVAL;
 949 DPRINTF(("ip_input: no SP, packet discarded\n"));/*XXX*/
 950                         crit_exit();
 951                         goto bad;
 952                 }
 953                 crit_exit();
 954                 if (error)
 955                         goto bad;
 956         }
 957 #endif /* FAST_IPSEC */
 958
 959         /*
 960          * We must forward the packet to the correct protocol thread if
 961          * we are not already in it.
 962          *
 963          * NOTE: ip_len is now in host form.  ip_len is not adjusted
 964          *       further for protocol processing, instead we pass hlen
 965          *       to the protosw and let it deal with it.
 966          */
 967         ipstat.ips_delivered++;
 968
 969         if ((m->m_flags & M_HASH) == 0) {
 970 #ifdef RSS_DEBUG
 971                 atomic_add_long(&ip_rehash_count, 1);
 972 #endif
 973                 ip->ip_len = htons(ip->ip_len + hlen);
 974                 ip->ip_off = htons(ip->ip_off);
 975
 976                 ip_hashfn(&m, 0);
 977                 if (m == NULL)
 978                         return;
 979
 980                 ip = mtod(m, struct ip *);
 981                 ip->ip_len = ntohs(ip->ip_len) - hlen;
 982                 ip->ip_off = ntohs(ip->ip_off);
 983                 KKASSERT(m->m_flags & M_HASH);
 984         }
 985         port = netisr_hashport(m->m_pkthdr.hash);
 986
 987         if (port != &curthread->td_msgport) {
 988                 struct netmsg_packet *pmsg;
 989
 990 #ifdef RSS_DEBUG
 991                 atomic_add_long(&ip_dispatch_slow, 1);
 992 #endif
 993
 994                 pmsg = &m->m_hdr.mh_netmsg;
 995                 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
 996                             0, transport_processing_handler);
 997                 pmsg->nm_packet = m;
 998                 pmsg->base.lmsg.u.ms_result = hlen;
 999                 lwkt_sendmsg(port, &pmsg->base.lmsg);
1000         } else {
1001 #ifdef RSS_DEBUG
1002                 atomic_add_long(&ip_dispatch_fast, 1);
1003 #endif
1004                 transport_processing_oncpu(m, hlen, ip);
1005         }
1006         return;
1007
1008 bad:
1009         m_freem(m);
1010 }
1011
1012 /*
1013  * Take incoming datagram fragment and try to reassemble it into
1014  * whole datagram.  If a chain for reassembly of this datagram already
1015  * exists, then it is given as fp; otherwise have to make a chain.
1016  */
1017 struct mbuf *
1018 ip_reass(struct mbuf *m)
1019 {
1020         struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid];
1021         struct ip *ip = mtod(m, struct ip *);
1022         struct mbuf *p = NULL, *q, *nq;
1023         struct mbuf *n;
1024         struct ipq *fp = NULL;
1025         struct ipqhead *head;
1026         int hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1027         int i, next;
1028         u_short sum;
1029
1030         /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
1031         if (maxnipq == 0 || maxfragsperpacket == 0) {
1032                 ipstat.ips_fragments++;
1033                 ipstat.ips_fragdropped++;
1034                 m_freem(m);
1035                 return NULL;
1036         }
1037
1038         sum = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
1039         /*
1040          * Look for queue of fragments of this datagram.
1041          */
1042         head = &fragq->ipq[sum];
1043         TAILQ_FOREACH(fp, head, ipq_list) {
1044                 if (ip->ip_id == fp->ipq_id &&
1045                     ip->ip_src.s_addr == fp->ipq_src.s_addr &&
1046                     ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
1047                     ip->ip_p == fp->ipq_p)
1048                         goto found;
1049         }
1050
1051         fp = NULL;
1052
1053         /*
1054          * Enforce upper bound on number of fragmented packets
1055          * for which we attempt reassembly;
1056          * If maxnipq is -1, accept all fragments without limitation.
1057          */
1058         if (fragq->nipq > maxnipq && maxnipq > 0) {
1059                 /*
1060                  * drop something from the tail of the current queue
1061                  * before proceeding further
1062                  */
1063                 struct ipq *q = TAILQ_LAST(head, ipqhead);
1064                 if (q == NULL) {
1065                         /*
1066                          * The current queue is empty,
1067                          * so drop from one of the others.
1068                          */
1069                         for (i = 0; i < IPREASS_NHASH; i++) {
1070                                 struct ipq *r = TAILQ_LAST(&fragq->ipq[i],
1071                                     ipqhead);
1072                                 if (r) {
1073                                         ipstat.ips_fragtimeout += r->ipq_nfrags;
1074                                         ip_freef(fragq, &fragq->ipq[i], r);
1075                                         break;
1076                                 }
1077                         }
1078                 } else {
1079                         ipstat.ips_fragtimeout += q->ipq_nfrags;
1080                         ip_freef(fragq, head, q);
1081                 }
1082         }
1083 found:
1084         /*
1085          * Adjust ip_len to not reflect header,
1086          * convert offset of this to bytes.
1087          */
1088         ip->ip_len -= hlen;
1089         if (ip->ip_off & IP_MF) {
1090                 /*
1091                  * Make sure that fragments have a data length
1092                  * that's a non-zero multiple of 8 bytes.
1093                  */
1094                 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
1095                         ipstat.ips_toosmall++; /* XXX */
1096                         m_freem(m);
1097                         goto done;
1098                 }
1099                 m->m_flags |= M_FRAG;
1100         } else {
1101                 m->m_flags &= ~M_FRAG;
1102         }
1103         ip->ip_off <<= 3;
1104
1105         ipstat.ips_fragments++;
1106         m->m_pkthdr.header = ip;
1107
1108         /*
1109          * If the hardware has not done csum over this fragment
1110          * then csum_data is not valid at all.
1111          */
1112         if ((m->m_pkthdr.csum_flags & (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID))
1113             == (CSUM_FRAG_NOT_CHECKED | CSUM_DATA_VALID)) {
1114                 m->m_pkthdr.csum_data = 0;
1115                 m->m_pkthdr.csum_flags &= ~(CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
1116         }
1117
1118         /*
1119          * Presence of header sizes in mbufs
1120          * would confuse code below.
1121          */
1122         m->m_data += hlen;
1123         m->m_len -= hlen;
1124
1125         /*
1126          * If first fragment to arrive, create a reassembly queue.
1127          */
1128         if (fp == NULL) {
1129                 if ((fp = mpipe_alloc_nowait(&ipq_mpipe)) == NULL)
1130                         goto dropfrag;
1131                 TAILQ_INSERT_HEAD(head, fp, ipq_list);
1132                 fragq->nipq++;
1133                 fp->ipq_nfrags = 1;
1134                 fp->ipq_ttl = IPFRAGTTL;
1135                 fp->ipq_p = ip->ip_p;
1136                 fp->ipq_id = ip->ip_id;
1137                 fp->ipq_src = ip->ip_src;
1138                 fp->ipq_dst = ip->ip_dst;
1139                 fp->ipq_frags = m;
1140                 m->m_nextpkt = NULL;
1141                 goto inserted;
1142         }
1143         fp->ipq_nfrags++;
1144
1145 #define GETIP(m)        ((struct ip*)((m)->m_pkthdr.header))
1146
1147         /*
1148          * Find a segment which begins after this one does.
1149          */
1150         for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1151                 if (GETIP(q)->ip_off > ip->ip_off)
1152                         break;
1153         }
1154
1155         /*
1156          * If there is a preceding segment, it may provide some of
1157          * our data already.  If so, drop the data from the incoming
1158          * segment.  If it provides all of our data, drop us, otherwise
1159          * stick new segment in the proper place.
1160          *
1161          * If some of the data is dropped from the the preceding
1162          * segment, then it's checksum is invalidated.
1163          */
1164         if (p) {
1165                 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
1166                 if (i > 0) {
1167                         if (i >= ip->ip_len)
1168                                 goto dropfrag;
1169                         m_adj(m, i);
1170                         m->m_pkthdr.csum_flags = 0;
1171                         ip->ip_off += i;
1172                         ip->ip_len -= i;
1173                 }
1174                 m->m_nextpkt = p->m_nextpkt;
1175                 p->m_nextpkt = m;
1176         } else {
1177                 m->m_nextpkt = fp->ipq_frags;
1178                 fp->ipq_frags = m;
1179         }
1180
1181         /*
1182          * While we overlap succeeding segments trim them or,
1183          * if they are completely covered, dequeue them.
1184          */
1185         for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
1186              q = nq) {
1187                 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
1188                 if (i < GETIP(q)->ip_len) {
1189                         GETIP(q)->ip_len -= i;
1190                         GETIP(q)->ip_off += i;
1191                         m_adj(q, i);
1192                         q->m_pkthdr.csum_flags = 0;
1193                         break;
1194                 }
1195                 nq = q->m_nextpkt;
1196                 m->m_nextpkt = nq;
1197                 ipstat.ips_fragdropped++;
1198                 fp->ipq_nfrags--;
1199                 q->m_nextpkt = NULL;
1200                 m_freem(q);
1201         }
1202
1203 inserted:
1204         /*
1205          * Check for complete reassembly and perform frag per packet
1206          * limiting.
1207          *
1208          * Frag limiting is performed here so that the nth frag has
1209          * a chance to complete the packet before we drop the packet.
1210          * As a result, n+1 frags are actually allowed per packet, but
1211          * only n will ever be stored. (n = maxfragsperpacket.)
1212          *
1213          */
1214         next = 0;
1215         for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
1216                 if (GETIP(q)->ip_off != next) {
1217                         if (fp->ipq_nfrags > maxfragsperpacket) {
1218                                 ipstat.ips_fragdropped += fp->ipq_nfrags;
1219                                 ip_freef(fragq, head, fp);
1220                         }
1221                         goto done;
1222                 }
1223                 next += GETIP(q)->ip_len;
1224         }
1225         /* Make sure the last packet didn't have the IP_MF flag */
1226         if (p->m_flags & M_FRAG) {
1227                 if (fp->ipq_nfrags > maxfragsperpacket) {
1228                         ipstat.ips_fragdropped += fp->ipq_nfrags;
1229                         ip_freef(fragq, head, fp);
1230                 }
1231                 goto done;
1232         }
1233
1234         /*
1235          * Reassembly is complete.  Make sure the packet is a sane size.
1236          */
1237         q = fp->ipq_frags;
1238         ip = GETIP(q);
1239         if (next + (IP_VHL_HL(ip->ip_vhl) << 2) > IP_MAXPACKET) {
1240                 ipstat.ips_toolong++;
1241                 ipstat.ips_fragdropped += fp->ipq_nfrags;
1242                 ip_freef(fragq, head, fp);
1243                 goto done;
1244         }
1245
1246         /*
1247          * Concatenate fragments.
1248          */
1249         m = q;
1250         n = m->m_next;
1251         m->m_next = NULL;
1252         m_cat(m, n);
1253         nq = q->m_nextpkt;
1254         q->m_nextpkt = NULL;
1255         for (q = nq; q != NULL; q = nq) {
1256                 nq = q->m_nextpkt;
1257                 q->m_nextpkt = NULL;
1258                 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
1259                 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
1260                 m_cat(m, q);
1261         }
1262
1263         /*
1264          * Clean up the 1's complement checksum.  Carry over 16 bits must
1265          * be added back.  This assumes no more then 65535 packet fragments
1266          * were reassembled.  A second carry can also occur (but not a third).
1267          */
1268         m->m_pkthdr.csum_data = (m->m_pkthdr.csum_data & 0xffff) +
1269                                 (m->m_pkthdr.csum_data >> 16);
1270         if (m->m_pkthdr.csum_data > 0xFFFF)
1271                 m->m_pkthdr.csum_data -= 0xFFFF;
1272
1273         /*
1274          * Create header for new ip packet by
1275          * modifying header of first packet;
1276          * dequeue and discard fragment reassembly header.
1277          * Make header visible.
1278          */
1279         ip->ip_len = next;
1280         ip->ip_src = fp->ipq_src;
1281         ip->ip_dst = fp->ipq_dst;
1282         TAILQ_REMOVE(head, fp, ipq_list);
1283         fragq->nipq--;
1284         mpipe_free(&ipq_mpipe, fp);
1285         m->m_len += (IP_VHL_HL(ip->ip_vhl) << 2);
1286         m->m_data -= (IP_VHL_HL(ip->ip_vhl) << 2);
1287         /* some debugging cruft by sklower, below, will go away soon */
1288         if (m->m_flags & M_PKTHDR) { /* XXX this should be done elsewhere */
1289                 int plen = 0;
1290
1291                 for (n = m; n; n = n->m_next)
1292                         plen += n->m_len;
1293                 m->m_pkthdr.len = plen;
1294         }
1295
1296         /*
1297          * Reassembly complete, return the next protocol.
1298          *
1299          * Be sure to clear M_HASH to force the packet
1300          * to be re-characterized.
1301          *
1302          * Clear M_FRAG, we are no longer a fragment.
1303          */
1304         m->m_flags &= ~(M_HASH | M_FRAG);
1305
1306         ipstat.ips_reassembled++;
1307         return (m);
1308
1309 dropfrag:
1310         ipstat.ips_fragdropped++;
1311         if (fp != NULL)
1312                 fp->ipq_nfrags--;
1313         m_freem(m);
1314 done:
1315         return (NULL);
1316
1317 #undef GETIP
1318 }
1319
1320 /*
1321  * Free a fragment reassembly header and all
1322  * associated datagrams.
1323  */
1324 static void
1325 ip_freef(struct ipfrag_queue *fragq, struct ipqhead *fhp, struct ipq *fp)
1326 {
1327         struct mbuf *q;
1328
1329         /*
1330          * Remove first to protect against blocking
1331          */
1332         TAILQ_REMOVE(fhp, fp, ipq_list);
1333
1334         /*
1335          * Clean out at our leisure
1336          */
1337         while (fp->ipq_frags) {
1338                 q = fp->ipq_frags;
1339                 fp->ipq_frags = q->m_nextpkt;
1340                 q->m_nextpkt = NULL;
1341                 m_freem(q);
1342         }
1343         mpipe_free(&ipq_mpipe, fp);
1344         fragq->nipq--;
1345 }
1346
1347 /*
1348  * If a timer expires on a reassembly queue, discard it.
1349  */
1350 static void
1351 ipfrag_timeo_dispatch(netmsg_t nmsg)
1352 {
1353         struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid];
1354         struct ipq *fp, *fp_temp;
1355         struct ipqhead *head;
1356         int i;
1357
1358         crit_enter();
1359         lwkt_replymsg(&nmsg->lmsg, 0);  /* reply ASAP */
1360         crit_exit();
1361
1362         for (i = 0; i < IPREASS_NHASH; i++) {
1363                 head = &fragq->ipq[i];
1364                 TAILQ_FOREACH_MUTABLE(fp, head, ipq_list, fp_temp) {
1365                         if (--fp->ipq_ttl == 0) {
1366                                 ipstat.ips_fragtimeout += fp->ipq_nfrags;
1367                                 ip_freef(fragq, head, fp);
1368                         }
1369                 }
1370         }
1371         /*
1372          * If we are over the maximum number of fragments
1373          * (due to the limit being lowered), drain off
1374          * enough to get down to the new limit.
1375          */
1376         if (maxnipq >= 0 && fragq->nipq > maxnipq) {
1377                 for (i = 0; i < IPREASS_NHASH; i++) {
1378                         head = &fragq->ipq[i];
1379                         while (fragq->nipq > maxnipq && !TAILQ_EMPTY(head)) {
1380                                 ipstat.ips_fragdropped +=
1381                                     TAILQ_FIRST(head)->ipq_nfrags;
1382                                 ip_freef(fragq, head, TAILQ_FIRST(head));
1383                         }
1384                 }
1385         }
1386 }
1387
1388 static void
1389 ipfrag_timeo_ipi(void *arg __unused)
1390 {
1391         int cpu = mycpuid;
1392         struct lwkt_msg *msg = &ipfrag_queue_pcpu[cpu].timeo_netmsg.lmsg;
1393
1394         ipfrag_queue_pcpu[cpu].timeo_inprog = 0;
1395         crit_enter();
1396         if (msg->ms_flags & MSGF_DONE)
1397                 lwkt_sendmsg_oncpu(netisr_cpuport(cpu), msg);
1398         crit_exit();
1399 }
1400
1401 static void
1402 ipfrag_slowtimo(void)
1403 {
1404         cpumask_t mask;
1405         int i;
1406
1407         CPUMASK_ASSZERO(mask);
1408         for (i = 0; i < ncpus; ++i) {
1409                 if (ipfrag_queue_pcpu[i].nipq &&
1410                     ipfrag_queue_pcpu[i].timeo_inprog == 0) {
1411                         ipfrag_queue_pcpu[i].timeo_inprog = 1;
1412                         CPUMASK_ORBIT(mask, i);
1413                 }
1414         }
1415         CPUMASK_ANDMASK(mask, smp_active_mask);
1416         if (CPUMASK_TESTNZERO(mask))
1417                 lwkt_send_ipiq_mask(mask, ipfrag_timeo_ipi, NULL);
1418 }
1419
1420 /*
1421  * IP timer processing
1422  */
1423 void
1424 ip_slowtimo(void)
1425 {
1426         ipfrag_slowtimo();
1427         ipflow_slowtimo();
1428 }
1429
1430 /*
1431  * Drain off all datagram fragments.
1432  */
1433 static void
1434 ipfrag_drain_dispatch(netmsg_t nmsg)
1435 {
1436         struct ipfrag_queue *fragq = &ipfrag_queue_pcpu[mycpuid];
1437         struct ipqhead *head;
1438         int i;
1439
1440         crit_enter();
1441         lwkt_replymsg(&nmsg->lmsg, 0);  /* reply ASAP */
1442         crit_exit();
1443
1444         for (i = 0; i < IPREASS_NHASH; i++) {
1445                 head = &fragq->ipq[i];
1446                 while (!TAILQ_EMPTY(head)) {
1447                         ipstat.ips_fragdropped += TAILQ_FIRST(head)->ipq_nfrags;
1448                         ip_freef(fragq, head, TAILQ_FIRST(head));
1449                 }
1450         }
1451 }
1452
1453 static void
1454 ipfrag_drain_ipi(void *arg __unused)
1455 {
1456         int cpu = mycpuid;
1457         struct lwkt_msg *msg = &ipfrag_queue_pcpu[cpu].drain_netmsg.lmsg;
1458
1459         crit_enter();
1460         if (msg->ms_flags & MSGF_DONE)
1461                 lwkt_sendmsg_oncpu(netisr_cpuport(cpu), msg);
1462         crit_exit();
1463 }
1464
1465 static void
1466 ipfrag_drain(void)
1467 {
1468         cpumask_t mask;
1469
1470         CPUMASK_ASSBMASK(mask, ncpus);
1471         CPUMASK_ANDMASK(mask, smp_active_mask);
1472         if (CPUMASK_TESTNZERO(mask))
1473                 lwkt_send_ipiq_mask(mask, ipfrag_drain_ipi, NULL);
1474 }
1475
1476 void
1477 ip_drain(void)
1478 {
1479         ipfrag_drain();
1480         in_rtqdrain();
1481 }
1482
1483 /*
1484  * Do option processing on a datagram,
1485  * possibly discarding it if bad options are encountered,
1486  * or forwarding it if source-routed.
1487  * The pass argument is used when operating in the IPSTEALTH
1488  * mode to tell what options to process:
1489  * [LS]SRR (pass 0) or the others (pass 1).
1490  * The reason for as many as two passes is that when doing IPSTEALTH,
1491  * non-routing options should be processed only if the packet is for us.
1492  * Returns 1 if packet has been forwarded/freed,
1493  * 0 if the packet should be processed further.
1494  */
1495 static int
1496 ip_dooptions(struct mbuf *m, int pass, struct sockaddr_in *next_hop)
1497 {
1498         struct sockaddr_in ipaddr = { sizeof ipaddr, AF_INET };
1499         struct ip *ip = mtod(m, struct ip *);
1500         u_char *cp;
1501         struct in_ifaddr *ia;
1502         int opt, optlen, cnt, off, code, type = ICMP_PARAMPROB;
1503         boolean_t forward = FALSE;
1504         struct in_addr *sin, dst;
1505         n_time ntime;
1506
1507         dst = ip->ip_dst;
1508         cp = (u_char *)(ip + 1);
1509         cnt = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1510         for (; cnt > 0; cnt -= optlen, cp += optlen) {
1511                 opt = cp[IPOPT_OPTVAL];
1512                 if (opt == IPOPT_EOL)
1513                         break;
1514                 if (opt == IPOPT_NOP)
1515                         optlen = 1;
1516                 else {
1517                         if (cnt < IPOPT_OLEN + sizeof(*cp)) {
1518                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1519                                 goto bad;
1520                         }
1521                         optlen = cp[IPOPT_OLEN];
1522                         if (optlen < IPOPT_OLEN + sizeof(*cp) || optlen > cnt) {
1523                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1524                                 goto bad;
1525                         }
1526                 }
1527                 switch (opt) {
1528
1529                 default:
1530                         break;
1531
1532                 /*
1533                  * Source routing with record.
1534                  * Find interface with current destination address.
1535                  * If none on this machine then drop if strictly routed,
1536                  * or do nothing if loosely routed.
1537                  * Record interface address and bring up next address
1538                  * component.  If strictly routed make sure next
1539                  * address is on directly accessible net.
1540                  */
1541                 case IPOPT_LSRR:
1542                 case IPOPT_SSRR:
1543                         if (ipstealth && pass > 0)
1544                                 break;
1545                         if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1546                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1547                                 goto bad;
1548                         }
1549                         if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1550                                 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1551                                 goto bad;
1552                         }
1553                         ipaddr.sin_addr = ip->ip_dst;
1554                         ia = (struct in_ifaddr *)
1555                                 ifa_ifwithaddr((struct sockaddr *)&ipaddr);
1556                         if (ia == NULL) {
1557                                 if (opt == IPOPT_SSRR) {
1558                                         type = ICMP_UNREACH;
1559                                         code = ICMP_UNREACH_SRCFAIL;
1560                                         goto bad;
1561                                 }
1562                                 if (!ip_dosourceroute)
1563                                         goto nosourcerouting;
1564                                 /*
1565                                  * Loose routing, and not at next destination
1566                                  * yet; nothing to do except forward.
1567                                  */
1568                                 break;
1569                         }
1570                         off--;                  /* 0 origin */
1571                         if (off > optlen - (int)sizeof(struct in_addr)) {
1572                                 /*
1573                                  * End of source route.  Should be for us.
1574                                  */
1575                                 if (!ip_acceptsourceroute)
1576                                         goto nosourcerouting;
1577                                 save_rte(m, cp, ip->ip_src);
1578                                 break;
1579                         }
1580                         if (ipstealth)
1581                                 goto dropit;
1582                         if (!ip_dosourceroute) {
1583                                 if (ipforwarding) {
1584                                         char sbuf[INET_ADDRSTRLEN];
1585                                         char dbuf[INET_ADDRSTRLEN];
1586
1587                                         /*
1588                                          * Acting as a router, so generate ICMP
1589                                          */
1590 nosourcerouting:
1591                                         log(LOG_WARNING,
1592                                             "attempted source route from %s to %s\n",
1593                                             kinet_ntoa(ip->ip_src, sbuf),
1594                                             kinet_ntoa(ip->ip_dst, dbuf));
1595                                         type = ICMP_UNREACH;
1596                                         code = ICMP_UNREACH_SRCFAIL;
1597                                         goto bad;
1598                                 } else {
1599                                         /*
1600                                          * Not acting as a router,
1601                                          * so silently drop.
1602                                          */
1603 dropit:
1604                                         ipstat.ips_cantforward++;
1605                                         m_freem(m);
1606                                         return (1);
1607                                 }
1608                         }
1609
1610                         /*
1611                          * locate outgoing interface
1612                          */
1613                         memcpy(&ipaddr.sin_addr, cp + off,
1614                             sizeof ipaddr.sin_addr);
1615
1616                         if (opt == IPOPT_SSRR) {
1617 #define INA     struct in_ifaddr *
1618 #define SA      struct sockaddr *
1619                                 if ((ia = (INA)ifa_ifwithdstaddr((SA)&ipaddr))
1620                                                                         == NULL)
1621                                         ia = (INA)ifa_ifwithnet((SA)&ipaddr);
1622                         } else {
1623                                 ia = ip_rtaddr(ipaddr.sin_addr, NULL);
1624                         }
1625                         if (ia == NULL) {
1626                                 type = ICMP_UNREACH;
1627                                 code = ICMP_UNREACH_SRCFAIL;
1628                                 goto bad;
1629                         }
1630                         ip->ip_dst = ipaddr.sin_addr;
1631                         memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1632                             sizeof(struct in_addr));
1633                         cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1634                         /*
1635                          * Let ip_intr's mcast routing check handle mcast pkts
1636                          */
1637                         forward = !IN_MULTICAST(ntohl(ip->ip_dst.s_addr));
1638                         break;
1639
1640                 case IPOPT_RR:
1641                         if (ipstealth && pass == 0)
1642                                 break;
1643                         if (optlen < IPOPT_OFFSET + sizeof(*cp)) {
1644                                 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1645                                 goto bad;
1646                         }
1647                         if ((off = cp[IPOPT_OFFSET]) < IPOPT_MINOFF) {
1648                                 code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1649                                 goto bad;
1650                         }
1651                         /*
1652                          * If no space remains, ignore.
1653                          */
1654                         off--;                  /* 0 origin */
1655                         if (off > optlen - (int)sizeof(struct in_addr))
1656                                 break;
1657                         memcpy(&ipaddr.sin_addr, &ip->ip_dst,
1658                             sizeof ipaddr.sin_addr);
1659                         /*
1660                          * locate outgoing interface; if we're the destination,
1661                          * use the incoming interface (should be same).
1662                          */
1663                         if ((ia = (INA)ifa_ifwithaddr((SA)&ipaddr)) == NULL &&
1664                             (ia = ip_rtaddr(ipaddr.sin_addr, NULL)) == NULL) {
1665                                 type = ICMP_UNREACH;
1666                                 code = ICMP_UNREACH_HOST;
1667                                 goto bad;
1668                         }
1669                         memcpy(cp + off, &IA_SIN(ia)->sin_addr,
1670                             sizeof(struct in_addr));
1671                         cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1672                         break;
1673
1674                 case IPOPT_TS:
1675                         if (ipstealth && pass == 0)
1676                                 break;
1677                         code = cp - (u_char *)ip;
1678                         if (optlen < 4 || optlen > 40) {
1679                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1680                                 goto bad;
1681                         }
1682                         if ((off = cp[IPOPT_OFFSET]) < 5) {
1683                                 code = &cp[IPOPT_OLEN] - (u_char *)ip;
1684                                 goto bad;
1685                         }
1686                         if (off > optlen - (int)sizeof(int32_t)) {
1687                                 cp[IPOPT_OFFSET + 1] += (1 << 4);
1688                                 if ((cp[IPOPT_OFFSET + 1] & 0xf0) == 0) {
1689                                         code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1690                                         goto bad;
1691                                 }
1692                                 break;
1693                         }
1694                         off--;                          /* 0 origin */
1695                         sin = (struct in_addr *)(cp + off);
1696                         switch (cp[IPOPT_OFFSET + 1] & 0x0f) {
1697
1698                         case IPOPT_TS_TSONLY:
1699                                 break;
1700
1701                         case IPOPT_TS_TSANDADDR:
1702                                 if (off + sizeof(n_time) +
1703                                     sizeof(struct in_addr) > optlen) {
1704                                         code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1705                                         goto bad;
1706                                 }
1707                                 ipaddr.sin_addr = dst;
1708                                 ia = (INA)ifaof_ifpforaddr((SA)&ipaddr,
1709                                                             m->m_pkthdr.rcvif);
1710                                 if (ia == NULL)
1711                                         continue;
1712                                 memcpy(sin, &IA_SIN(ia)->sin_addr,
1713                                     sizeof(struct in_addr));
1714                                 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1715                                 off += sizeof(struct in_addr);
1716                                 break;
1717
1718                         case IPOPT_TS_PRESPEC:
1719                                 if (off + sizeof(n_time) +
1720                                     sizeof(struct in_addr) > optlen) {
1721                                         code = &cp[IPOPT_OFFSET] - (u_char *)ip;
1722                                         goto bad;
1723                                 }
1724                                 memcpy(&ipaddr.sin_addr, sin,
1725                                     sizeof(struct in_addr));
1726                                 if (ifa_ifwithaddr((SA)&ipaddr) == NULL)
1727                                         continue;
1728                                 cp[IPOPT_OFFSET] += sizeof(struct in_addr);
1729                                 off += sizeof(struct in_addr);
1730                                 break;
1731
1732                         default:
1733                                 code = &cp[IPOPT_OFFSET + 1] - (u_char *)ip;
1734                                 goto bad;
1735                         }
1736                         ntime = iptime();
1737                         memcpy(cp + off, &ntime, sizeof(n_time));
1738                         cp[IPOPT_OFFSET] += sizeof(n_time);
1739                 }
1740         }
1741         if (forward && ipforwarding) {
1742                 ip_forward(m, TRUE, next_hop);
1743                 return (1);
1744         }
1745         return (0);
1746 bad:
1747         icmp_error(m, type, code, 0, 0);
1748         ipstat.ips_badoptions++;
1749         return (1);
1750 }
1751
1752 /*
1753  * Given address of next destination (final or next hop),
1754  * return internet address info of interface to be used to get there.
1755  */
1756 struct in_ifaddr *
1757 ip_rtaddr(struct in_addr dst, struct route *ro0)
1758 {
1759         struct route sro, *ro;
1760         struct sockaddr_in *sin;
1761         struct in_ifaddr *ia;
1762
1763         if (ro0 != NULL) {
1764                 ro = ro0;
1765         } else {
1766                 bzero(&sro, sizeof(sro));
1767                 ro = &sro;
1768         }
1769
1770         sin = (struct sockaddr_in *)&ro->ro_dst;
1771
1772         if (ro->ro_rt == NULL || dst.s_addr != sin->sin_addr.s_addr) {
1773                 if (ro->ro_rt != NULL) {
1774                         RTFREE(ro->ro_rt);
1775                         ro->ro_rt = NULL;
1776                 }
1777                 sin->sin_family = AF_INET;
1778                 sin->sin_len = sizeof *sin;
1779                 sin->sin_addr = dst;
1780                 rtalloc_ign(ro, RTF_PRCLONING);
1781         }
1782
1783         if (ro->ro_rt == NULL)
1784                 return (NULL);
1785
1786         ia = ifatoia(ro->ro_rt->rt_ifa);
1787
1788         if (ro == &sro)
1789                 RTFREE(ro->ro_rt);
1790         return ia;
1791 }
1792
1793 /*
1794  * Save incoming source route for use in replies,
1795  * to be picked up later by ip_srcroute if the receiver is interested.
1796  */
1797 static void
1798 save_rte(struct mbuf *m, u_char *option, struct in_addr dst)
1799 {
1800         struct m_tag *mtag;
1801         struct ip_srcrt_opt *opt;
1802         unsigned olen;
1803
1804         mtag = m_tag_get(PACKET_TAG_IPSRCRT, sizeof(*opt), M_NOWAIT);
1805         if (mtag == NULL)
1806                 return;
1807         opt = m_tag_data(mtag);
1808
1809         olen = option[IPOPT_OLEN];
1810 #ifdef DIAGNOSTIC
1811         if (ipprintfs)
1812                 kprintf("save_rte: olen %d\n", olen);
1813 #endif
1814         if (olen > sizeof(opt->ip_srcrt) - (1 + sizeof(dst))) {
1815                 m_tag_free(mtag);
1816                 return;
1817         }
1818         bcopy(option, opt->ip_srcrt.srcopt, olen);
1819         opt->ip_nhops = (olen - IPOPT_OFFSET - 1) / sizeof(struct in_addr);
1820         opt->ip_srcrt.dst = dst;
1821         m_tag_prepend(m, mtag);
1822 }
1823
1824 /*
1825  * Retrieve incoming source route for use in replies,
1826  * in the same form used by setsockopt.
1827  * The first hop is placed before the options, will be removed later.
1828  */
1829 struct mbuf *
1830 ip_srcroute(struct mbuf *m0)
1831 {
1832         struct in_addr *p, *q;
1833         struct mbuf *m;
1834         struct m_tag *mtag;
1835         struct ip_srcrt_opt *opt;
1836
1837         if (m0 == NULL)
1838                 return NULL;
1839
1840         mtag = m_tag_find(m0, PACKET_TAG_IPSRCRT, NULL);
1841         if (mtag == NULL)
1842                 return NULL;
1843         opt = m_tag_data(mtag);
1844
1845         if (opt->ip_nhops == 0)
1846                 return (NULL);
1847         m = m_get(M_NOWAIT, MT_HEADER);
1848         if (m == NULL)
1849                 return (NULL);
1850
1851 #define OPTSIZ  (sizeof(opt->ip_srcrt.nop) + sizeof(opt->ip_srcrt.srcopt))
1852
1853         /* length is (nhops+1)*sizeof(addr) + sizeof(nop + srcrt header) */
1854         m->m_len = opt->ip_nhops * sizeof(struct in_addr) +
1855                    sizeof(struct in_addr) + OPTSIZ;
1856 #ifdef DIAGNOSTIC
1857         if (ipprintfs) {
1858                 kprintf("ip_srcroute: nhops %d mlen %d",
1859                         opt->ip_nhops, m->m_len);
1860         }
1861 #endif
1862
1863         /*
1864          * First save first hop for return route
1865          */
1866         p = &opt->ip_srcrt.route[opt->ip_nhops - 1];
1867         *(mtod(m, struct in_addr *)) = *p--;
1868 #ifdef DIAGNOSTIC
1869         if (ipprintfs)
1870                 kprintf(" hops %x", ntohl(mtod(m, struct in_addr *)->s_addr));
1871 #endif
1872
1873         /*
1874          * Copy option fields and padding (nop) to mbuf.
1875          */
1876         opt->ip_srcrt.nop = IPOPT_NOP;
1877         opt->ip_srcrt.srcopt[IPOPT_OFFSET] = IPOPT_MINOFF;
1878         memcpy(mtod(m, caddr_t) + sizeof(struct in_addr), &opt->ip_srcrt.nop,
1879             OPTSIZ);
1880         q = (struct in_addr *)(mtod(m, caddr_t) +
1881             sizeof(struct in_addr) + OPTSIZ);
1882 #undef OPTSIZ
1883         /*
1884          * Record return path as an IP source route,
1885          * reversing the path (pointers are now aligned).
1886          */
1887         while (p >= opt->ip_srcrt.route) {
1888 #ifdef DIAGNOSTIC
1889                 if (ipprintfs)
1890                         kprintf(" %x", ntohl(q->s_addr));
1891 #endif
1892                 *q++ = *p--;
1893         }
1894         /*
1895          * Last hop goes to final destination.
1896          */
1897         *q = opt->ip_srcrt.dst;
1898         m_tag_delete(m0, mtag);
1899 #ifdef DIAGNOSTIC
1900         if (ipprintfs)
1901                 kprintf(" %x\n", ntohl(q->s_addr));
1902 #endif
1903         return (m);
1904 }
1905
1906 /*
1907  * Strip out IP options.
1908  */
1909 void
1910 ip_stripoptions(struct mbuf *m)
1911 {
1912         int datalen;
1913         struct ip *ip = mtod(m, struct ip *);
1914         caddr_t opts;
1915         int optlen;
1916
1917         optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
1918         opts = (caddr_t)(ip + 1);
1919         datalen = m->m_len - (sizeof(struct ip) + optlen);
1920         bcopy(opts + optlen, opts, datalen);
1921         m->m_len -= optlen;
1922         if (m->m_flags & M_PKTHDR)
1923                 m->m_pkthdr.len -= optlen;
1924         ip->ip_vhl = IP_MAKE_VHL(IPVERSION, sizeof(struct ip) >> 2);
1925 }
1926
1927 u_char inetctlerrmap[PRC_NCMDS] = {
1928         0,              0,              0,              0,
1929         0,              EMSGSIZE,       EHOSTDOWN,      EHOSTUNREACH,
1930         EHOSTUNREACH,   EHOSTUNREACH,   ECONNREFUSED,   ECONNREFUSED,
1931         EMSGSIZE,       EHOSTUNREACH,   0,              0,
1932         0,              0,              0,              0,
1933         ENOPROTOOPT,    ECONNREFUSED
1934 };
1935
1936 /*
1937  * Forward a packet.  If some error occurs return the sender
1938  * an icmp packet.  Note we can't always generate a meaningful
1939  * icmp message because icmp doesn't have a large enough repertoire
1940  * of codes and types.
1941  *
1942  * If not forwarding, just drop the packet.  This could be confusing
1943  * if ipforwarding was zero but some routing protocol was advancing
1944  * us as a gateway to somewhere.  However, we must let the routing
1945  * protocol deal with that.
1946  *
1947  * The using_srcrt parameter indicates whether the packet is being forwarded
1948  * via a source route.
1949  */
1950 void
1951 ip_forward(struct mbuf *m, boolean_t using_srcrt, struct sockaddr_in *next_hop)
1952 {
1953         struct ip *ip = mtod(m, struct ip *);
1954         struct rtentry *rt;
1955         struct route fwd_ro;
1956         int error, type = 0, code = 0, destmtu = 0;
1957         struct mbuf *mcopy, *mtemp = NULL;
1958         n_long dest;
1959         struct in_addr pkt_dst;
1960
1961         dest = INADDR_ANY;
1962         /*
1963          * Cache the destination address of the packet; this may be
1964          * changed by use of 'ipfw fwd'.
1965          */
1966         pkt_dst = (next_hop != NULL) ? next_hop->sin_addr : ip->ip_dst;
1967
1968 #ifdef DIAGNOSTIC
1969         if (ipprintfs)
1970                 kprintf("forward: src %x dst %x ttl %x\n",
1971                        ip->ip_src.s_addr, pkt_dst.s_addr, ip->ip_ttl);
1972 #endif
1973
1974         if (m->m_flags & (M_BCAST | M_MCAST) || !in_canforward(pkt_dst)) {
1975                 ipstat.ips_cantforward++;
1976                 m_freem(m);
1977                 return;
1978         }
1979         if (!ipstealth && ip->ip_ttl <= IPTTLDEC) {
1980                 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, dest, 0);
1981                 return;
1982         }
1983
1984         bzero(&fwd_ro, sizeof(fwd_ro));
1985         ip_rtaddr(pkt_dst, &fwd_ro);
1986         if (fwd_ro.ro_rt == NULL) {
1987                 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, dest, 0);
1988                 return;
1989         }
1990         rt = fwd_ro.ro_rt;
1991
1992         if (curthread->td_type == TD_TYPE_NETISR) {
1993                 /*
1994                  * Save the IP header and at most 8 bytes of the payload,
1995                  * in case we need to generate an ICMP message to the src.
1996                  */
1997                 mtemp = ipforward_mtemp[mycpuid];
1998                 KASSERT((mtemp->m_flags & M_EXT) == 0 &&
1999                     mtemp->m_data == mtemp->m_pktdat &&
2000                     m_tag_first(mtemp) == NULL,
2001                     ("ip_forward invalid mtemp1"));
2002
2003                 if (!m_dup_pkthdr(mtemp, m, M_NOWAIT)) {
2004                         /*
2005                          * It's probably ok if the pkthdr dup fails (because
2006                          * the deep copy of the tag chain failed), but for now
2007                          * be conservative and just discard the copy since
2008                          * code below may some day want the tags.
2009                          */
2010                         mtemp = NULL;
2011                 } else {
2012                         mtemp->m_type = m->m_type;
2013                         mtemp->m_len = imin((IP_VHL_HL(ip->ip_vhl) << 2) + 8,
2014                             (int)ip->ip_len);
2015                         mtemp->m_pkthdr.len = mtemp->m_len;
2016                         m_copydata(m, 0, mtemp->m_len, mtod(mtemp, caddr_t));
2017                 }
2018         }
2019
2020         if (!ipstealth)
2021                 ip->ip_ttl -= IPTTLDEC;
2022
2023         /*
2024          * If forwarding packet using same interface that it came in on,
2025          * perhaps should send a redirect to sender to shortcut a hop.
2026          * Only send redirect if source is sending directly to us,
2027          * and if packet was not source routed (or has any options).
2028          * Also, don't send redirect if forwarding using a default route
2029          * or a route modified by a redirect.
2030          */
2031         if (rt->rt_ifp == m->m_pkthdr.rcvif &&
2032             !(rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) &&
2033             satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY &&
2034             ipsendredirects && !using_srcrt && next_hop == NULL) {
2035                 u_long src = ntohl(ip->ip_src.s_addr);
2036                 struct in_ifaddr *rt_ifa = (struct in_ifaddr *)rt->rt_ifa;
2037
2038                 if (rt_ifa != NULL &&
2039                     (src & rt_ifa->ia_subnetmask) == rt_ifa->ia_subnet) {
2040                         if (rt->rt_flags & RTF_GATEWAY)
2041                                 dest = satosin(rt->rt_gateway)->sin_addr.s_addr;
2042                         else
2043                                 dest = pkt_dst.s_addr;
2044                         /*
2045                          * Router requirements says to only send
2046                          * host redirects.
2047                          */
2048                         type = ICMP_REDIRECT;
2049                         code = ICMP_REDIRECT_HOST;
2050 #ifdef DIAGNOSTIC
2051                         if (ipprintfs)
2052                                 kprintf("redirect (%d) to %x\n", code, dest);
2053 #endif
2054                 }
2055         }
2056
2057         error = ip_output(m, NULL, &fwd_ro, IP_FORWARDING, NULL, NULL);
2058         if (error == 0) {
2059                 ipstat.ips_forward++;
2060                 if (type == 0) {
2061                         if (mtemp)
2062                                 ipflow_create(&fwd_ro, mtemp);
2063                         goto done;
2064                 }
2065                 ipstat.ips_redirectsent++;
2066         } else {
2067                 ipstat.ips_cantforward++;
2068         }
2069
2070         if (mtemp == NULL)
2071                 goto done;
2072
2073         /*
2074          * Errors that do not require generating ICMP message
2075          */
2076         switch (error) {
2077         case ENOBUFS:
2078                 /*
2079                  * A router should not generate ICMP_SOURCEQUENCH as
2080                  * required in RFC1812 Requirements for IP Version 4 Routers.
2081                  * Source quench could be a big problem under DoS attacks,
2082                  * or if the underlying interface is rate-limited.
2083                  * Those who need source quench packets may re-enable them
2084                  * via the net.inet.ip.sendsourcequench sysctl.
2085                  */
2086                 if (!ip_sendsourcequench)
2087                         goto done;
2088                 break;
2089
2090         case EACCES:                    /* ipfw denied packet */
2091                 goto done;
2092         }
2093
2094         KASSERT((mtemp->m_flags & M_EXT) == 0 &&
2095             mtemp->m_data == mtemp->m_pktdat,
2096             ("ip_forward invalid mtemp2"));
2097         mcopy = m_copym(mtemp, 0, mtemp->m_len, M_NOWAIT);
2098         if (mcopy == NULL)
2099                 goto done;
2100
2101         /*
2102          * Send ICMP message.
2103          */
2104         switch (error) {
2105         case 0:                         /* forwarded, but need redirect */
2106                 /* type, code set above */
2107                 break;
2108
2109         case ENETUNREACH:               /* shouldn't happen, checked above */
2110         case EHOSTUNREACH:
2111         case ENETDOWN:
2112         case EHOSTDOWN:
2113         default:
2114                 type = ICMP_UNREACH;
2115                 code = ICMP_UNREACH_HOST;
2116                 break;
2117
2118         case EMSGSIZE:
2119                 type = ICMP_UNREACH;
2120                 code = ICMP_UNREACH_NEEDFRAG;
2121 #ifdef IPSEC
2122                 /*
2123                  * If the packet is routed over IPsec tunnel, tell the
2124                  * originator the tunnel MTU.
2125                  *      tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2126                  * XXX quickhack!!!
2127                  */
2128                 if (fwd_ro.ro_rt != NULL) {
2129                         struct secpolicy *sp = NULL;
2130                         int ipsecerror;
2131                         int ipsechdr;
2132                         struct route *ro;
2133
2134                         sp = ipsec4_getpolicybyaddr(mcopy,
2135                                                     IPSEC_DIR_OUTBOUND,
2136                                                     IP_FORWARDING,
2137                                                     &ipsecerror);
2138
2139                         if (sp == NULL)
2140                                 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2141                         else {
2142                                 /* count IPsec header size */
2143                                 ipsechdr = ipsec4_hdrsiz(mcopy,
2144                                                          IPSEC_DIR_OUTBOUND,
2145                                                          NULL);
2146
2147                                 /*
2148                                  * find the correct route for outer IPv4
2149                                  * header, compute tunnel MTU.
2150                                  *
2151                                  */
2152                                 if (sp->req != NULL && sp->req->sav != NULL &&
2153                                     sp->req->sav->sah != NULL) {
2154                                         ro = &sp->req->sav->sah->sa_route;
2155                                         if (ro->ro_rt != NULL &&
2156                                             ro->ro_rt->rt_ifp != NULL) {
2157                                                 destmtu =
2158                                                     ro->ro_rt->rt_ifp->if_mtu;
2159                                                 destmtu -= ipsechdr;
2160                                         }
2161                                 }
2162
2163                                 key_freesp(sp);
2164                         }
2165                 }
2166 #elif defined(FAST_IPSEC)
2167                 /*
2168                  * If the packet is routed over IPsec tunnel, tell the
2169                  * originator the tunnel MTU.
2170                  *      tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
2171                  * XXX quickhack!!!
2172                  */
2173                 if (fwd_ro.ro_rt != NULL) {
2174                         struct secpolicy *sp = NULL;
2175                         int ipsecerror;
2176                         int ipsechdr;
2177                         struct route *ro;
2178
2179                         sp = ipsec_getpolicybyaddr(mcopy,
2180                                                    IPSEC_DIR_OUTBOUND,
2181                                                    IP_FORWARDING,
2182                                                    &ipsecerror);
2183
2184                         if (sp == NULL)
2185                                 destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2186                         else {
2187                                 /* count IPsec header size */
2188                                 ipsechdr = ipsec4_hdrsiz(mcopy,
2189                                                          IPSEC_DIR_OUTBOUND,
2190                                                          NULL);
2191
2192                                 /*
2193                                  * find the correct route for outer IPv4
2194                                  * header, compute tunnel MTU.
2195                                  */
2196
2197                                 if (sp->req != NULL &&
2198                                     sp->req->sav != NULL &&
2199                                     sp->req->sav->sah != NULL) {
2200                                         ro = &sp->req->sav->sah->sa_route;
2201                                         if (ro->ro_rt != NULL &&
2202                                             ro->ro_rt->rt_ifp != NULL) {
2203                                                 destmtu =
2204                                                     ro->ro_rt->rt_ifp->if_mtu;
2205                                                 destmtu -= ipsechdr;
2206                                         }
2207                                 }
2208
2209                                 KEY_FREESP(&sp);
2210                         }
2211                 }
2212 #else /* !IPSEC && !FAST_IPSEC */
2213                 if (fwd_ro.ro_rt != NULL)
2214                         destmtu = fwd_ro.ro_rt->rt_ifp->if_mtu;
2215 #endif /*IPSEC*/
2216                 ipstat.ips_cantfrag++;
2217                 break;
2218
2219         case ENOBUFS:
2220                 type = ICMP_SOURCEQUENCH;
2221                 code = 0;
2222                 break;
2223
2224         case EACCES:                    /* ipfw denied packet */
2225                 panic("ip_forward EACCES should not reach");
2226         }
2227         icmp_error(mcopy, type, code, dest, destmtu);
2228 done:
2229         if (mtemp != NULL)
2230                 m_tag_delete_chain(mtemp);
2231         if (fwd_ro.ro_rt != NULL)
2232                 RTFREE(fwd_ro.ro_rt);
2233 }
2234
2235 void
2236 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
2237                struct mbuf *m)
2238 {
2239         if (inp->inp_socket->so_options & SO_TIMESTAMP) {
2240                 struct timeval tv;
2241
2242                 microtime(&tv);
2243                 *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
2244                     SCM_TIMESTAMP, SOL_SOCKET);
2245                 if (*mp)
2246                         mp = &(*mp)->m_next;
2247         }
2248         if (inp->inp_flags & INP_RECVDSTADDR) {
2249                 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
2250                     sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
2251                 if (*mp)
2252                         mp = &(*mp)->m_next;
2253         }
2254         if (inp->inp_flags & INP_RECVTTL) {
2255                 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
2256                     sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
2257                 if (*mp)
2258                         mp = &(*mp)->m_next;
2259         }
2260 #ifdef notyet
2261         /* XXX
2262          * Moving these out of udp_input() made them even more broken
2263          * than they already were.
2264          */
2265         /* options were tossed already */
2266         if (inp->inp_flags & INP_RECVOPTS) {
2267                 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
2268                     sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
2269                 if (*mp)
2270                         mp = &(*mp)->m_next;
2271         }
2272         /* ip_srcroute doesn't do what we want here, need to fix */
2273         if (inp->inp_flags & INP_RECVRETOPTS) {
2274                 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
2275                     sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
2276                 if (*mp)
2277                         mp = &(*mp)->m_next;
2278         }
2279 #endif
2280         if (inp->inp_flags & INP_RECVIF) {
2281                 struct ifnet *ifp;
2282                 struct sdlbuf {
2283                         struct sockaddr_dl sdl;
2284                         u_char  pad[32];
2285                 } sdlbuf;
2286                 struct sockaddr_dl *sdp;
2287                 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
2288
2289                 if (((ifp = m->m_pkthdr.rcvif)) &&
2290                     ((ifp->if_index != 0) && (ifp->if_index <= if_index))) {
2291                         sdp = IF_LLSOCKADDR(ifp);
2292                         /*
2293                          * Change our mind and don't try copy.
2294                          */
2295                         if ((sdp->sdl_family != AF_LINK) ||
2296                             (sdp->sdl_len > sizeof(sdlbuf))) {
2297                                 goto makedummy;
2298                         }
2299                         bcopy(sdp, sdl2, sdp->sdl_len);
2300                 } else {
2301 makedummy:
2302                         sdl2->sdl_len =
2303                             offsetof(struct sockaddr_dl, sdl_data[0]);
2304                         sdl2->sdl_family = AF_LINK;
2305                         sdl2->sdl_index = 0;
2306                         sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
2307                 }
2308                 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
2309                         IP_RECVIF, IPPROTO_IP);
2310                 if (*mp)
2311                         mp = &(*mp)->m_next;
2312         }
2313 }
2314
2315 /*
2316  * XXX these routines are called from the upper part of the kernel.
2317  *
2318  * They could also be moved to ip_mroute.c, since all the RSVP
2319  *  handling is done there already.
2320  */
2321 int
2322 ip_rsvp_init(struct socket *so)
2323 {
2324         if (so->so_type != SOCK_RAW ||
2325             so->so_proto->pr_protocol != IPPROTO_RSVP)
2326                 return EOPNOTSUPP;
2327
2328         if (ip_rsvpd != NULL)
2329                 return EADDRINUSE;
2330
2331         ip_rsvpd = so;
2332         /*
2333          * This may seem silly, but we need to be sure we don't over-increment
2334          * the RSVP counter, in case something slips up.
2335          */
2336         if (!ip_rsvp_on) {
2337                 ip_rsvp_on = 1;
2338                 rsvp_on++;
2339         }
2340
2341         return 0;
2342 }
2343
2344 int
2345 ip_rsvp_done(void)
2346 {
2347         ip_rsvpd = NULL;
2348         /*
2349          * This may seem silly, but we need to be sure we don't over-decrement
2350          * the RSVP counter, in case something slips up.
2351          */
2352         if (ip_rsvp_on) {
2353                 ip_rsvp_on = 0;
2354                 rsvp_on--;
2355         }
2356         return 0;
2357 }
2358
2359 int
2360 rsvp_input(struct mbuf **mp, int *offp, int proto)
2361 {
2362         struct mbuf *m = *mp;
2363
2364         *mp = NULL;
2365
2366         if (rsvp_input_p) { /* call the real one if loaded */
2367                 *mp = m;
2368                 rsvp_input_p(mp, offp, proto);
2369                 return(IPPROTO_DONE);
2370         }
2371
2372         /* Can still get packets with rsvp_on = 0 if there is a local member
2373          * of the group to which the RSVP packet is addressed.  But in this
2374          * case we want to throw the packet away.
2375          */
2376
2377         if (!rsvp_on) {
2378                 m_freem(m);
2379                 return(IPPROTO_DONE);
2380         }
2381
2382         if (ip_rsvpd != NULL) {
2383                 *mp = m;
2384                 rip_input(mp, offp, proto);
2385                 return(IPPROTO_DONE);
2386         }
2387         /* Drop the packet */
2388         m_freem(m);
2389         return(IPPROTO_DONE);
2390 }