sys/netinet/ip_icmp.c

   1 /*
   2  * Copyright (c) 1982, 1986, 1988, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. Neither the name of the University nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  30  * $FreeBSD: src/sys/netinet/ip_icmp.c,v 1.39.2.19 2003/01/24 05:11:34 sam Exp $
  31  */
  32
  33 #include "opt_ipsec.h"
  34
  35 #include <sys/param.h>
  36 #include <sys/systm.h>
  37 #include <sys/mbuf.h>
  38 #include <sys/protosw.h>
  39 #include <sys/socket.h>
  40 #include <sys/socketops.h>
  41 #include <sys/time.h>
  42 #include <sys/kernel.h>
  43 #include <sys/sysctl.h>
  44 #include <sys/in_cksum.h>
  45
  46 #include <machine/stdarg.h>
  47
  48 #include <net/if.h>
  49 #include <net/if_types.h>
  50 #include <net/route.h>
  51
  52 #define _IP_VHL
  53 #include <netinet/in.h>
  54 #include <netinet/in_systm.h>
  55 #include <netinet/in_var.h>
  56 #include <netinet/ip.h>
  57 #include <netinet/ip_icmp.h>
  58 #include <netinet/ip_var.h>
  59 #include <netinet/icmp_var.h>
  60
  61 #ifdef IPSEC
  62 #include <netinet6/ipsec.h>
  63 #include <netproto/key/key.h>
  64 #endif
  65
  66 #ifdef FAST_IPSEC
  67 #include <netproto/ipsec/ipsec.h>
  68 #include <netproto/ipsec/key.h>
  69 #define IPSEC
  70 #endif
  71
  72 /*
  73  * ICMP routines: error generation, receive packet processing, and
  74  * routines to turnaround packets back to the originator, and
  75  * host table maintenance routines.
  76  */
  77
  78 struct icmpstat icmpstat;
  79 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
  80         &icmpstat, icmpstat, "ICMP statistics");
  81
  82 static int      icmpmaskrepl = 0;
  83 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
  84         &icmpmaskrepl, 0, "Allow replies to netmask requests");
  85
  86 static int      drop_redirect = 0;
  87 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
  88         &drop_redirect, 0, "Ignore ICMP redirects");
  89
  90 static int      log_redirect = 0;
  91 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
  92         &log_redirect, 0, "Enable output about ICMP redirects");
  93
  94 static int      discard_sourcequench = 1;
  95 SYSCTL_INT(_net_inet_icmp, OID_AUTO, discard_sourcequench, CTLFLAG_RW,
  96         &discard_sourcequench, 0, "Discard ICMP Source Quench");
  97
  98 #ifdef ICMP_BANDLIM
  99
 100 /*
 101  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
 102  *      variable content is -1 and read-only.
 103  */
 104
 105 static int      icmplim = 200;
 106 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
 107         &icmplim, 0, "ICMP bandwidth limit");
 108 #else
 109
 110 static int      icmplim = -1;
 111 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
 112         &icmplim, 0, "ICMP bandwidth limit");
 113
 114 #endif
 115
 116 static int      icmplim_output = 0;
 117 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
 118         &icmplim_output, 0, "Enable output about ICMP bandwidth limits");
 119
 120 /*
 121  * ICMP broadcast echo sysctl
 122  */
 123
 124 static int      icmpbmcastecho = 0;
 125 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
 126     &icmpbmcastecho, 0, "");
 127
 128 static char     icmp_reply_src[IFNAMSIZ];
 129 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
 130         icmp_reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
 131
 132 static int      icmp_rfi;
 133 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
 134         &icmp_rfi, 0, "ICMP reply from incoming interface for "
 135         "non-local packets");
 136
 137 #ifdef ICMPPRINTFS
 138 static int      icmpprintfs = 0;
 139 SYSCTL_INT(_net_inet_icmp, OID_AUTO, debug_prints, CTLFLAG_RW,
 140         &icmpprintfs, 0, "extra ICMP debug prints");
 141 #endif
 142
 143 static void     icmp_reflect (struct mbuf *);
 144 static void     icmp_send (struct mbuf *, struct mbuf *, struct route *);
 145
 146 extern  struct protosw inetsw[];
 147
 148 /*
 149  * Generate an error packet of type error
 150  * in response to bad packet ip.
 151  */
 152 void
 153 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
 154 {
 155         struct ip *oip = mtod(n, struct ip *), *nip;
 156         unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
 157         struct icmp *icp;
 158         struct mbuf *m;
 159         unsigned icmplen;
 160
 161 #ifdef ICMPPRINTFS
 162         if (icmpprintfs)
 163                 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code);
 164 #endif
 165         if (type != ICMP_REDIRECT)
 166                 icmpstat.icps_error++;
 167         /*
 168          * Don't send error if the original packet was encrypted.
 169          * Don't send error if not the first fragment of message.
 170          * Don't error if the old packet protocol was ICMP
 171          * error message, only known informational types.
 172          */
 173         if (n->m_flags & M_DECRYPTED)
 174                 goto freeit;
 175         if (oip->ip_off &~ (IP_MF|IP_DF))
 176                 goto freeit;
 177         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
 178           n->m_len >= oiplen + ICMP_MINLEN &&
 179           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
 180                 icmpstat.icps_oldicmp++;
 181                 goto freeit;
 182         }
 183         /* Don't send error in response to a multicast or broadcast packet */
 184         if (n->m_flags & (M_BCAST|M_MCAST))
 185                 goto freeit;
 186         /*
 187          * First, formulate icmp message
 188          */
 189         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
 190         if (m == NULL)
 191                 goto freeit;
 192         icmplen = min(oiplen + 8, oip->ip_len);
 193         if (icmplen < sizeof(struct ip))
 194                 panic("icmp_error: bad length");
 195         m->m_len = icmplen + ICMP_MINLEN;
 196         MH_ALIGN(m, m->m_len);
 197         icp = mtod(m, struct icmp *);
 198         if ((u_int)type > ICMP_MAXTYPE)
 199                 panic("icmp_error");
 200         icmpstat.icps_outhist[type]++;
 201         icp->icmp_type = type;
 202         if (type == ICMP_REDIRECT)
 203                 icp->icmp_gwaddr.s_addr = dest;
 204         else {
 205                 icp->icmp_void = 0;
 206                 /*
 207                  * The following assignments assume an overlay with the
 208                  * zeroed icmp_void field.
 209                  */
 210                 if (type == ICMP_PARAMPROB) {
 211                         icp->icmp_pptr = code;
 212                         code = 0;
 213                 } else if (type == ICMP_UNREACH &&
 214                         code == ICMP_UNREACH_NEEDFRAG && destmtu) {
 215                         icp->icmp_nextmtu = htons(destmtu);
 216                 }
 217         }
 218
 219         icp->icmp_code = code;
 220         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
 221         nip = &icp->icmp_ip;
 222
 223         /*
 224          * Convert fields to network representation.
 225          */
 226         nip->ip_len = htons(nip->ip_len);
 227         nip->ip_off = htons(nip->ip_off);
 228
 229         /*
 230          * Now, copy old ip header (without options)
 231          * in front of icmp message.
 232          */
 233         if (m->m_data - sizeof(struct ip) < m->m_pktdat)
 234                 panic("icmp len");
 235         m->m_data -= sizeof(struct ip);
 236         m->m_len += sizeof(struct ip);
 237         m->m_pkthdr.len = m->m_len;
 238         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
 239         nip = mtod(m, struct ip *);
 240         bcopy(oip, nip, sizeof(struct ip));
 241         nip->ip_len = m->m_len;
 242         nip->ip_vhl = IP_VHL_BORING;
 243         nip->ip_p = IPPROTO_ICMP;
 244         nip->ip_tos = 0;
 245         m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED;
 246         icmp_reflect(m);
 247
 248 freeit:
 249         m_freem(n);
 250 }
 251
 252 /*
 253  * Process a received ICMP message.
 254  */
 255 int
 256 icmp_input(struct mbuf **mp, int *offp, int proto)
 257 {
 258         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 259         struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
 260         struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
 261         struct icmp *icp;
 262         struct in_ifaddr *ia;
 263         struct mbuf *m = *mp;
 264         struct ip *ip = mtod(m, struct ip *);
 265         int icmplen = ip->ip_len;
 266         int i, hlen;
 267         int code;
 268
 269         *mp = NULL;
 270         hlen = *offp;
 271
 272         /*
 273          * Locate icmp structure in mbuf, and check
 274          * that not corrupted and of at least minimum length.
 275          */
 276 #ifdef ICMPPRINTFS
 277         if (icmpprintfs) {
 278                 char src_buf[INET_ADDRSTRLEN], dst_buf[INET_ADDRSTRLEN];
 279
 280                 kprintf("icmp_input from %s to %s, len %d\n",
 281                     inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN),
 282                     inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN),
 283                     icmplen);
 284         }
 285 #endif
 286         if (icmplen < ICMP_MINLEN) {
 287                 icmpstat.icps_tooshort++;
 288                 goto freeit;
 289         }
 290         i = hlen + min(icmplen, ICMP_ADVLENMIN);
 291         if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
 292                 icmpstat.icps_tooshort++;
 293                 return(IPPROTO_DONE);
 294         }
 295         ip = mtod(m, struct ip *);
 296
 297         if (in_cksum_skip(m, hlen + icmplen, hlen)) {
 298                 icmpstat.icps_checksum++;
 299                 goto freeit;
 300         }
 301         icp = mtod(m, struct icmp *);
 302
 303         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
 304                 /*
 305                  * Deliver very specific ICMP type only.
 306                  */
 307                 switch (icp->icmp_type) {
 308                 case ICMP_UNREACH:
 309                 case ICMP_TIMXCEED:
 310                         break;
 311                 default:
 312                         goto freeit;
 313                 }
 314         }
 315
 316 #ifdef ICMPPRINTFS
 317         if (icmpprintfs)
 318                 kprintf("icmp_input, type %d code %d\n", icp->icmp_type,
 319                     icp->icmp_code);
 320 #endif
 321
 322         /*
 323          * Message type specific processing.
 324          */
 325         if (icp->icmp_type > ICMP_MAXTYPE)
 326                 goto raw;
 327         icmpstat.icps_inhist[icp->icmp_type]++;
 328         code = icp->icmp_code;
 329         switch (icp->icmp_type) {
 330
 331         case ICMP_UNREACH:
 332                 switch (code) {
 333                         case ICMP_UNREACH_NET:
 334                         case ICMP_UNREACH_HOST:
 335                         case ICMP_UNREACH_SRCFAIL:
 336                         case ICMP_UNREACH_NET_UNKNOWN:
 337                         case ICMP_UNREACH_HOST_UNKNOWN:
 338                         case ICMP_UNREACH_ISOLATED:
 339                         case ICMP_UNREACH_TOSNET:
 340                         case ICMP_UNREACH_TOSHOST:
 341                         case ICMP_UNREACH_HOST_PRECEDENCE:
 342                         case ICMP_UNREACH_PRECEDENCE_CUTOFF:
 343                                 code = PRC_UNREACH_NET;
 344                                 break;
 345
 346                         case ICMP_UNREACH_NEEDFRAG:
 347                                 code = PRC_MSGSIZE;
 348                                 break;
 349
 350                         /*
 351                          * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
 352                          * Treat subcodes 2,3 as immediate RST
 353                          */
 354                         case ICMP_UNREACH_PROTOCOL:
 355                         case ICMP_UNREACH_PORT:
 356                                 code = PRC_UNREACH_PORT;
 357                                 break;
 358
 359                         case ICMP_UNREACH_NET_PROHIB:
 360                         case ICMP_UNREACH_HOST_PROHIB:
 361                         case ICMP_UNREACH_FILTER_PROHIB:
 362                                 code = PRC_UNREACH_ADMIN_PROHIB;
 363                                 break;
 364
 365                         default:
 366                                 goto badcode;
 367                 }
 368                 goto deliver;
 369
 370         case ICMP_TIMXCEED:
 371                 if (code > 1)
 372                         goto badcode;
 373                 code += PRC_TIMXCEED_INTRANS;
 374                 goto deliver;
 375
 376         case ICMP_PARAMPROB:
 377                 if (code > 1)
 378                         goto badcode;
 379                 code = PRC_PARAMPROB;
 380                 goto deliver;
 381
 382         case ICMP_SOURCEQUENCH:
 383                 if (code)
 384                         goto badcode;
 385                 if (discard_sourcequench)
 386                         break;
 387                 code = PRC_QUENCH;
 388 deliver:
 389                 /*
 390                  * Problem with datagram; advise higher level routines.
 391                  */
 392                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 393                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 394                         icmpstat.icps_badlen++;
 395                         goto freeit;
 396                 }
 397                 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
 398                 /* Discard ICMP's in response to multicast packets */
 399                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
 400                         goto badcode;
 401 #ifdef ICMPPRINTFS
 402                 if (icmpprintfs)
 403                         kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
 404 #endif
 405                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 406 #if 1
 407                 /*
 408                  * MTU discovery:
 409                  * If we got a needfrag and there is a host route to the
 410                  * original destination, and the MTU is not locked, then
 411                  * set the MTU in the route to the suggested new value
 412                  * (if given) and then notify as usual.  The ULPs will
 413                  * notice that the MTU has changed and adapt accordingly.
 414                  * If no new MTU was suggested, then we guess a new one
 415                  * less than the current value.  If the new MTU is
 416                  * unreasonably small (arbitrarily set at 296), then
 417                  * we reset the MTU to the interface value and enable the
 418                  * lock bit, indicating that we are no longer doing MTU
 419                  * discovery.
 420                  */
 421                 if (code == PRC_MSGSIZE) {
 422                         struct rtentry *rt;
 423                         int mtu;
 424
 425                         rt = rtpurelookup((struct sockaddr *)&icmpsrc);
 426                         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
 427                             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
 428 #ifdef DEBUG_MTUDISC
 429                                 char src_buf[INET_ADDRSTRLEN];
 430 #endif
 431                                 mtu = ntohs(icp->icmp_nextmtu);
 432                                 if (!mtu)
 433                                         mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
 434                                                           1);
 435 #ifdef DEBUG_MTUDISC
 436                                 kprintf("MTU for %s reduced to %d\n",
 437                                     inet_ntop(AF_INET, &icmpsrc.sin_addr,
 438                                         src_buf, INET_ADDRSTRLEN), mtu);
 439 #endif
 440                                 if (mtu < 296) {
 441                                         /* rt->rt_rmx.rmx_mtu =
 442                                                 rt->rt_ifp->if_mtu; */
 443                                         rt->rt_rmx.rmx_locks |= RTV_MTU;
 444                                 } else if (rt->rt_rmx.rmx_mtu > mtu) {
 445                                         rt->rt_rmx.rmx_mtu = mtu;
 446                                 }
 447                         }
 448                         if (rt != NULL)
 449                                 --rt->rt_refcnt;
 450                 }
 451 #endif
 452                 /*
 453                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
 454                  * notification to TCP layer.
 455                  */
 456                 so_pr_ctlinput(
 457                         &inetsw[ip_protox[icp->icmp_ip.ip_p]],
 458                         code, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
 459                 break;
 460
 461 badcode:
 462                 icmpstat.icps_badcode++;
 463                 break;
 464
 465         case ICMP_ECHO:
 466                 if (!icmpbmcastecho
 467                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 468                         icmpstat.icps_bmcastecho++;
 469                         break;
 470                 }
 471                 icp->icmp_type = ICMP_ECHOREPLY;
 472 #ifdef ICMP_BANDLIM
 473                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 474                         goto freeit;
 475                 else
 476 #endif
 477                         goto reflect;
 478
 479         case ICMP_TSTAMP:
 480                 if (!icmpbmcastecho
 481                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 482                         icmpstat.icps_bmcasttstamp++;
 483                         break;
 484                 }
 485                 if (icmplen < ICMP_TSLEN) {
 486                         icmpstat.icps_badlen++;
 487                         break;
 488                 }
 489                 icp->icmp_type = ICMP_TSTAMPREPLY;
 490                 icp->icmp_rtime = iptime();
 491                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 492 #ifdef ICMP_BANDLIM
 493                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 494                         goto freeit;
 495                 else
 496 #endif
 497                         goto reflect;
 498
 499         case ICMP_MASKREQ:
 500                 if (icmpmaskrepl == 0)
 501                         break;
 502                 /*
 503                  * We are not able to respond with all ones broadcast
 504                  * unless we receive it over a point-to-point interface.
 505                  */
 506                 if (icmplen < ICMP_MASKLEN)
 507                         break;
 508                 switch (ip->ip_dst.s_addr) {
 509
 510                 case INADDR_BROADCAST:
 511                 case INADDR_ANY:
 512                         icmpdst.sin_addr = ip->ip_src;
 513                         break;
 514
 515                 default:
 516                         icmpdst.sin_addr = ip->ip_dst;
 517                 }
 518                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 519                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 520                 if (ia == NULL)
 521                         break;
 522                 if (ia->ia_ifp == 0)
 523                         break;
 524                 icp->icmp_type = ICMP_MASKREPLY;
 525                 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 526                 if (ip->ip_src.s_addr == 0) {
 527                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 528                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 529                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 530                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 531                 }
 532 reflect:
 533                 ip->ip_len += hlen;     /* since ip_input deducts this */
 534                 icmpstat.icps_reflect++;
 535                 icmpstat.icps_outhist[icp->icmp_type]++;
 536                 icmp_reflect(m);
 537                 return(IPPROTO_DONE);
 538
 539         case ICMP_REDIRECT:
 540                 if (log_redirect) {
 541                         u_long src, dst, gw;
 542
 543                         src = ntohl(ip->ip_src.s_addr);
 544                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
 545                         gw = ntohl(icp->icmp_gwaddr.s_addr);
 546                         kprintf("icmp redirect from %d.%d.%d.%d: "
 547                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
 548                                (int)(src >> 24), (int)((src >> 16) & 0xff),
 549                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
 550                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
 551                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
 552                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
 553                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
 554                 }
 555                 if (drop_redirect)
 556                         break;
 557                 if (code > 3)
 558                         goto badcode;
 559                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 560                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 561                         icmpstat.icps_badlen++;
 562                         break;
 563                 }
 564                 /*
 565                  * Short circuit routing redirects to force
 566                  * immediate change in the kernel's routing
 567                  * tables.  The message is also handed to anyone
 568                  * listening on a raw socket (e.g. the routing
 569                  * daemon for use in updating its tables).
 570                  */
 571                 icmpgw.sin_addr = ip->ip_src;
 572                 icmpdst.sin_addr = icp->icmp_gwaddr;
 573 #ifdef  ICMPPRINTFS
 574                 if (icmpprintfs) {
 575                         char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN];
 576
 577                         kprintf("redirect dst %s to %s\n",
 578                             inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
 579                                 dst_buf, INET_ADDRSTRLEN),
 580                             inet_ntop(AF_INET, &icp->icmp_gwaddr,
 581                                 gw_buf, INET_ADDRSTRLEN));
 582                 }
 583 #endif
 584                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 585                 rtredirect((struct sockaddr *)&icmpsrc,
 586                   (struct sockaddr *)&icmpdst,
 587                   NULL, RTF_GATEWAY | RTF_HOST,
 588                   (struct sockaddr *)&icmpgw);
 589                 kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 590 #ifdef IPSEC
 591                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 592 #endif
 593                 break;
 594
 595         /*
 596          * No kernel processing for the following;
 597          * just fall through to send to raw listener.
 598          */
 599         case ICMP_ECHOREPLY:
 600         case ICMP_ROUTERADVERT:
 601         case ICMP_ROUTERSOLICIT:
 602         case ICMP_TSTAMPREPLY:
 603         case ICMP_IREQREPLY:
 604         case ICMP_MASKREPLY:
 605         default:
 606                 break;
 607         }
 608
 609 raw:
 610         *mp = m;
 611         rip_input(mp, offp, proto);
 612         return(IPPROTO_DONE);
 613
 614 freeit:
 615         m_freem(m);
 616         return(IPPROTO_DONE);
 617 }
 618
 619 /*
 620  * Reflect the ip packet back to the source
 621  */
 622 static void
 623 icmp_reflect(struct mbuf *m)
 624 {
 625         struct ip *ip = mtod(m, struct ip *);
 626         struct in_ifaddr *ia;
 627         struct in_ifaddr_container *iac;
 628         struct ifaddr_container *ifac;
 629         struct ifnet *ifp;
 630         struct in_addr t;
 631         struct mbuf *opts = NULL;
 632         int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
 633         struct route *ro = NULL, rt;
 634
 635         if (!in_canforward(ip->ip_src) &&
 636             ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
 637              (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
 638                 m_freem(m);     /* Bad return address */
 639                 icmpstat.icps_badaddr++;
 640                 goto done;      /* Ip_output() will check for broadcast */
 641         }
 642         t = ip->ip_dst;
 643         ip->ip_dst = ip->ip_src;
 644
 645         ro = &rt;
 646         bzero(ro, sizeof *ro);
 647
 648         /*
 649          * If the incoming packet was addressed directly to us,
 650          * use dst as the src for the reply.  Otherwise (broadcast
 651          * or anonymous), use the address which corresponds
 652          * to the incoming interface.
 653          */
 654         ia = NULL;
 655         LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
 656                 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
 657                         ia = iac->ia;
 658                         goto match;
 659                 }
 660         }
 661         ifp = m->m_pkthdr.rcvif;
 662         if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) {
 663                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 664                         struct ifaddr *ifa = ifac->ifa;
 665
 666                         if (ifa->ifa_addr->sa_family != AF_INET)
 667                                 continue;
 668                         ia = ifatoia(ifa);
 669                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 670                             t.s_addr)
 671                                 goto match;
 672                 }
 673         }
 674         /*
 675          * If the packet was transiting through us, use the address of
 676          * the interface the packet came through in.  If that interface
 677          * doesn't have a suitable IP address, the normal selection
 678          * criteria apply.
 679          */
 680         if (icmp_rfi && ifp != NULL) {
 681                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 682                         struct ifaddr *ifa = ifac->ifa;
 683
 684                         if (ifa->ifa_addr->sa_family != AF_INET)
 685                                 continue;
 686                         ia = ifatoia(ifa);
 687                         goto match;
 688                 }
 689         }
 690         /*
 691          * If the incoming packet was not addressed directly to us, use
 692          * designated interface for icmp replies specified by sysctl
 693          * net.inet.icmp.reply_src (default not set). Otherwise continue
 694          * with normal source selection.
 695          */
 696         if (icmp_reply_src[0] != '\0' && (ifp = ifunit(icmp_reply_src))) {
 697                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 698                         struct ifaddr *ifa = ifac->ifa;
 699
 700                         if (ifa->ifa_addr->sa_family != AF_INET)
 701                                 continue;
 702                         ia = ifatoia(ifa);
 703                         goto match;
 704                 }
 705         }
 706         /*
 707          * If the packet was transiting through us, use the address of
 708          * the interface that is the closest to the packet source.
 709          * When we don't have a route back to the packet source, stop here
 710          * and drop the packet.
 711          */
 712         ia = ip_rtaddr(ip->ip_dst, ro);
 713         if (ia == NULL) {
 714                 m_freem(m);
 715                 icmpstat.icps_noroute++;
 716                 goto done;
 717         }
 718 match:
 719         t = IA_SIN(ia)->sin_addr;
 720         ip->ip_src = t;
 721         ip->ip_ttl = ip_defttl;
 722
 723         if (optlen > 0) {
 724                 u_char *cp;
 725                 int opt, cnt;
 726                 u_int len;
 727
 728                 /*
 729                  * Retrieve any source routing from the incoming packet;
 730                  * add on any record-route or timestamp options.
 731                  */
 732                 cp = (u_char *) (ip + 1);
 733                 if ((opts = ip_srcroute(m)) == NULL &&
 734                     (opts = m_gethdr(MB_DONTWAIT, MT_HEADER))) {
 735                         opts->m_len = sizeof(struct in_addr);
 736                         mtod(opts, struct in_addr *)->s_addr = 0;
 737                 }
 738                 if (opts) {
 739 #ifdef ICMPPRINTFS
 740                         if (icmpprintfs)
 741                                 kprintf("icmp_reflect optlen %d rt %d => ",
 742                                        optlen, opts->m_len);
 743 #endif
 744                         for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
 745                                 opt = cp[IPOPT_OPTVAL];
 746                                 if (opt == IPOPT_EOL)
 747                                         break;
 748                                 if (opt == IPOPT_NOP)
 749                                         len = 1;
 750                                 else {
 751                                         if (cnt < IPOPT_OLEN + sizeof *cp)
 752                                                 break;
 753                                         len = cp[IPOPT_OLEN];
 754                                         if (len < IPOPT_OLEN + sizeof *cp ||
 755                                             len > cnt)
 756                                         break;
 757                                 }
 758                                 /*
 759                                  * Should check for overflow, but it
 760                                  * "can't happen".
 761                                  */
 762                                 if (opt == IPOPT_RR || opt == IPOPT_TS ||
 763                                     opt == IPOPT_SECURITY) {
 764                                         bcopy(cp,
 765                                               mtod(opts, caddr_t) + opts->m_len,
 766                                               len);
 767                                         opts->m_len += len;
 768                                 }
 769                         }
 770                         /* Terminate & pad, if necessary */
 771                         cnt = opts->m_len % 4;
 772                         if (cnt) {
 773                                 for (; cnt < 4; cnt++) {
 774                                         *(mtod(opts, caddr_t) + opts->m_len) =
 775                                             IPOPT_EOL;
 776                                         opts->m_len++;
 777                                 }
 778                         }
 779 #ifdef ICMPPRINTFS
 780                         if (icmpprintfs)
 781                                 kprintf("%d\n", opts->m_len);
 782 #endif
 783                 }
 784                 /*
 785                  * Now strip out original options by copying rest of first
 786                  * mbuf's data back, and adjust the IP length.
 787                  */
 788                 ip->ip_len -= optlen;
 789                 ip->ip_vhl = IP_VHL_BORING;
 790                 m->m_len -= optlen;
 791                 if (m->m_flags & M_PKTHDR)
 792                         m->m_pkthdr.len -= optlen;
 793                 optlen += sizeof(struct ip);
 794                 bcopy((caddr_t)ip + optlen, ip + 1,
 795                       m->m_len - sizeof(struct ip));
 796         }
 797         m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
 798         m->m_flags &= ~(M_BCAST|M_MCAST);
 799         icmp_send(m, opts, ro);
 800 done:
 801         if (opts)
 802                 m_free(opts);
 803         if (ro && ro->ro_rt)
 804                 RTFREE(ro->ro_rt);
 805 }
 806
 807 /*
 808  * Send an icmp packet back to the ip level,
 809  * after supplying a checksum.
 810  */
 811 static void
 812 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
 813 {
 814         struct ip *ip = mtod(m, struct ip *);
 815         struct icmp *icp;
 816         int hlen;
 817
 818         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 819         m->m_data += hlen;
 820         m->m_len -= hlen;
 821         icp = mtod(m, struct icmp *);
 822         icp->icmp_cksum = 0;
 823         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
 824         m->m_data -= hlen;
 825         m->m_len += hlen;
 826         m->m_pkthdr.rcvif = NULL;
 827 #ifdef ICMPPRINTFS
 828         if (icmpprintfs) {
 829                 char dst_buf[INET_ADDRSTRLEN], src_buf[INET_ADDRSTRLEN];
 830
 831                 kprintf("icmp_send dst %s src %s\n",
 832                     inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN),
 833                     inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN));
 834         }
 835 #endif
 836         ip_output(m, opts, rt, 0, NULL, NULL);
 837 }
 838
 839 n_time
 840 iptime(void)
 841 {
 842         struct timeval atv;
 843         u_long t;
 844
 845         getmicrotime(&atv);
 846         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 847         return (htonl(t));
 848 }
 849
 850 #if 1
 851 /*
 852  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 853  * given current value MTU.  If DIR is less than zero, a larger plateau
 854  * is returned; otherwise, a smaller value is returned.
 855  */
 856 int
 857 ip_next_mtu(int mtu, int dir)
 858 {
 859         static int mtutab[] = {
 860                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
 861                 68, 0
 862         };
 863         int i;
 864
 865         for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
 866                 if (mtu >= mtutab[i])
 867                         break;
 868         }
 869
 870         if (dir < 0) {
 871                 if (i == 0) {
 872                         return 0;
 873                 } else {
 874                         return mtutab[i - 1];
 875                 }
 876         } else {
 877                 if (mtutab[i] == 0) {
 878                         return 0;
 879                 } else if(mtu > mtutab[i]) {
 880                         return mtutab[i];
 881                 } else {
 882                         return mtutab[i + 1];
 883                 }
 884         }
 885 }
 886 #endif
 887
 888 #ifdef ICMP_BANDLIM
 889 /*
 890  * badport_bandlim() - check for ICMP bandwidth limit
 891  *
 892  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 893  *      hit our bandwidth limit and it is not ok.
 894  *
 895  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 896  *
 897  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 898  *      values.  We may eventually remove this separation (and simplify the
 899  *      code further).
 900  *
 901  *      Note that the printing of the error message is delayed so we can
 902  *      properly print the icmp error rate that the system was trying to do
 903  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 904  *      the 'final' error, but it doesn't make sense to solve the printing
 905  *      delay with more complex code.
 906  */
 907 int
 908 badport_bandlim(int which)
 909 {
 910         static int lticks[BANDLIM_MAX + 1];
 911         static int lpackets[BANDLIM_MAX + 1];
 912         int dticks;
 913         const char *bandlimittype[] = {
 914                 "Limiting icmp unreach response",
 915                 "Limiting icmp ping response",
 916                 "Limiting icmp tstamp response",
 917                 "Limiting closed port RST response",
 918                 "Limiting open port RST response"
 919                 };
 920
 921         /*
 922          * Return ok status if feature disabled or argument out of
 923          * ranage.
 924          */
 925
 926         if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
 927                 return(0);
 928         dticks = ticks - lticks[which];
 929
 930         /*
 931          * reset stats when cumulative dt exceeds one second.
 932          */
 933
 934         if ((unsigned int)dticks > hz) {
 935                 if (lpackets[which] > icmplim && icmplim_output) {
 936                         kprintf("%s from %d to %d packets per second\n",
 937                                 bandlimittype[which],
 938                                 lpackets[which],
 939                                 icmplim
 940                         );
 941                 }
 942                 lticks[which] = ticks;
 943                 lpackets[which] = 0;
 944         }
 945
 946         /*
 947          * bump packet count
 948          */
 949
 950         if (++lpackets[which] > icmplim) {
 951                 return(-1);
 952         }
 953         return(0);
 954 }
 955 #endif