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 = (struct icmp *)((caddr_t)ip + hlen);
 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                 /* Discard ICMP's in response to multicast packets */
 398                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
 399                         goto badcode;
 400 #ifdef ICMPPRINTFS
 401                 if (icmpprintfs)
 402                         kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
 403 #endif
 404                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 405 #if 1
 406                 /*
 407                  * MTU discovery:
 408                  * If we got a needfrag and there is a host route to the
 409                  * original destination, and the MTU is not locked, then
 410                  * set the MTU in the route to the suggested new value
 411                  * (if given) and then notify as usual.  The ULPs will
 412                  * notice that the MTU has changed and adapt accordingly.
 413                  * If no new MTU was suggested, then we guess a new one
 414                  * less than the current value.  If the new MTU is
 415                  * unreasonably small (arbitrarily set at 296), then
 416                  * we reset the MTU to the interface value and enable the
 417                  * lock bit, indicating that we are no longer doing MTU
 418                  * discovery.
 419                  */
 420                 if (code == PRC_MSGSIZE) {
 421                         struct rtentry *rt;
 422                         int mtu;
 423
 424                         rt = rtpurelookup((struct sockaddr *)&icmpsrc);
 425                         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
 426                             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
 427 #ifdef DEBUG_MTUDISC
 428                                 char src_buf[INET_ADDRSTRLEN];
 429 #endif
 430                                 mtu = ntohs(icp->icmp_nextmtu);
 431                                 if (!mtu)
 432                                         mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
 433                                                           1);
 434 #ifdef DEBUG_MTUDISC
 435                                 kprintf("MTU for %s reduced to %d\n",
 436                                     inet_ntop(AF_INET, &icmpsrc.sin_addr,
 437                                         src_buf, INET_ADDRSTRLEN), mtu);
 438 #endif
 439                                 if (mtu < 296) {
 440                                         /* rt->rt_rmx.rmx_mtu =
 441                                                 rt->rt_ifp->if_mtu; */
 442                                         rt->rt_rmx.rmx_locks |= RTV_MTU;
 443                                 } else if (rt->rt_rmx.rmx_mtu > mtu) {
 444                                         rt->rt_rmx.rmx_mtu = mtu;
 445                                 }
 446                         }
 447                         if (rt != NULL)
 448                                 --rt->rt_refcnt;
 449                 }
 450 #endif
 451                 /*
 452                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
 453                  * notification to TCP layer.
 454                  */
 455                 so_pr_ctlinput(
 456                         &inetsw[ip_protox[icp->icmp_ip.ip_p]],
 457                         code, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
 458                 break;
 459
 460 badcode:
 461                 icmpstat.icps_badcode++;
 462                 break;
 463
 464         case ICMP_ECHO:
 465                 if (!icmpbmcastecho
 466                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 467                         icmpstat.icps_bmcastecho++;
 468                         break;
 469                 }
 470                 icp->icmp_type = ICMP_ECHOREPLY;
 471 #ifdef ICMP_BANDLIM
 472                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 473                         goto freeit;
 474                 else
 475 #endif
 476                         goto reflect;
 477
 478         case ICMP_TSTAMP:
 479                 if (!icmpbmcastecho
 480                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 481                         icmpstat.icps_bmcasttstamp++;
 482                         break;
 483                 }
 484                 if (icmplen < ICMP_TSLEN) {
 485                         icmpstat.icps_badlen++;
 486                         break;
 487                 }
 488                 icp->icmp_type = ICMP_TSTAMPREPLY;
 489                 icp->icmp_rtime = iptime();
 490                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 491 #ifdef ICMP_BANDLIM
 492                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 493                         goto freeit;
 494                 else
 495 #endif
 496                         goto reflect;
 497
 498         case ICMP_MASKREQ:
 499                 if (icmpmaskrepl == 0)
 500                         break;
 501                 /*
 502                  * We are not able to respond with all ones broadcast
 503                  * unless we receive it over a point-to-point interface.
 504                  */
 505                 if (icmplen < ICMP_MASKLEN)
 506                         break;
 507                 switch (ip->ip_dst.s_addr) {
 508
 509                 case INADDR_BROADCAST:
 510                 case INADDR_ANY:
 511                         icmpdst.sin_addr = ip->ip_src;
 512                         break;
 513
 514                 default:
 515                         icmpdst.sin_addr = ip->ip_dst;
 516                 }
 517                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 518                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 519                 if (ia == NULL)
 520                         break;
 521                 if (ia->ia_ifp == 0)
 522                         break;
 523                 icp->icmp_type = ICMP_MASKREPLY;
 524                 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 525                 if (ip->ip_src.s_addr == 0) {
 526                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 527                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 528                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 529                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 530                 }
 531 reflect:
 532                 ip->ip_len += hlen;     /* since ip_input deducts this */
 533                 icmpstat.icps_reflect++;
 534                 icmpstat.icps_outhist[icp->icmp_type]++;
 535                 icmp_reflect(m);
 536                 return(IPPROTO_DONE);
 537
 538         case ICMP_REDIRECT:
 539                 if (log_redirect) {
 540                         char src_buf[INET_ADDRSTRLEN];
 541                         char dst_buf[INET_ADDRSTRLEN];
 542                         char gwy_buf[INET_ADDRSTRLEN];
 543
 544                         kprintf("icmp redirect from %s: %s => %s\n",
 545                             inet_ntop(AF_INET, &ip->ip_src,
 546                                 src_buf, INET_ADDRSTRLEN),
 547                             inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
 548                                 dst_buf, INET_ADDRSTRLEN),
 549                             inet_ntop(AF_INET, &icp->icmp_gwaddr,
 550                                 gwy_buf, INET_ADDRSTRLEN));
 551                 }
 552                 if (drop_redirect)
 553                         break;
 554                 if (code > 3)
 555                         goto badcode;
 556                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 557                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 558                         icmpstat.icps_badlen++;
 559                         break;
 560                 }
 561                 /*
 562                  * Short circuit routing redirects to force
 563                  * immediate change in the kernel's routing
 564                  * tables.  The message is also handed to anyone
 565                  * listening on a raw socket (e.g. the routing
 566                  * daemon for use in updating its tables).
 567                  */
 568                 icmpgw.sin_addr = ip->ip_src;
 569                 icmpdst.sin_addr = icp->icmp_gwaddr;
 570 #ifdef  ICMPPRINTFS
 571                 if (icmpprintfs) {
 572                         char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN];
 573
 574                         kprintf("redirect dst %s to %s\n",
 575                             inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
 576                                 dst_buf, INET_ADDRSTRLEN),
 577                             inet_ntop(AF_INET, &icp->icmp_gwaddr,
 578                                 gw_buf, INET_ADDRSTRLEN));
 579                 }
 580 #endif
 581                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 582                 rtredirect((struct sockaddr *)&icmpsrc,
 583                   (struct sockaddr *)&icmpdst,
 584                   NULL, RTF_GATEWAY | RTF_HOST,
 585                   (struct sockaddr *)&icmpgw);
 586                 kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 587 #ifdef IPSEC
 588                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 589 #endif
 590                 break;
 591
 592         /*
 593          * No kernel processing for the following;
 594          * just fall through to send to raw listener.
 595          */
 596         case ICMP_ECHOREPLY:
 597         case ICMP_ROUTERADVERT:
 598         case ICMP_ROUTERSOLICIT:
 599         case ICMP_TSTAMPREPLY:
 600         case ICMP_IREQREPLY:
 601         case ICMP_MASKREPLY:
 602         default:
 603                 break;
 604         }
 605
 606 raw:
 607         *mp = m;
 608         rip_input(mp, offp, proto);
 609         return(IPPROTO_DONE);
 610
 611 freeit:
 612         m_freem(m);
 613         return(IPPROTO_DONE);
 614 }
 615
 616 /*
 617  * Reflect the ip packet back to the source
 618  */
 619 static void
 620 icmp_reflect(struct mbuf *m)
 621 {
 622         struct ip *ip = mtod(m, struct ip *);
 623         struct in_ifaddr *ia;
 624         struct in_ifaddr_container *iac;
 625         struct ifaddr_container *ifac;
 626         struct ifnet *ifp;
 627         struct in_addr t;
 628         struct mbuf *opts = NULL;
 629         int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
 630         struct route *ro = NULL, rt;
 631
 632         if (!in_canforward(ip->ip_src) &&
 633             ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
 634              (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
 635                 m_freem(m);     /* Bad return address */
 636                 icmpstat.icps_badaddr++;
 637                 goto done;      /* Ip_output() will check for broadcast */
 638         }
 639         t = ip->ip_dst;
 640         ip->ip_dst = ip->ip_src;
 641
 642         ro = &rt;
 643         bzero(ro, sizeof *ro);
 644
 645         /*
 646          * If the incoming packet was addressed directly to us,
 647          * use dst as the src for the reply.  Otherwise (broadcast
 648          * or anonymous), use the address which corresponds
 649          * to the incoming interface.
 650          */
 651         ia = NULL;
 652         LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
 653                 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
 654                         ia = iac->ia;
 655                         goto match;
 656                 }
 657         }
 658         ifp = m->m_pkthdr.rcvif;
 659         if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) {
 660                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 661                         struct ifaddr *ifa = ifac->ifa;
 662
 663                         if (ifa->ifa_addr->sa_family != AF_INET)
 664                                 continue;
 665                         ia = ifatoia(ifa);
 666                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 667                             t.s_addr)
 668                                 goto match;
 669                 }
 670         }
 671         /*
 672          * If the packet was transiting through us, use the address of
 673          * the interface the packet came through in.  If that interface
 674          * doesn't have a suitable IP address, the normal selection
 675          * criteria apply.
 676          */
 677         if (icmp_rfi && ifp != NULL) {
 678                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 679                         struct ifaddr *ifa = ifac->ifa;
 680
 681                         if (ifa->ifa_addr->sa_family != AF_INET)
 682                                 continue;
 683                         ia = ifatoia(ifa);
 684                         goto match;
 685                 }
 686         }
 687         /*
 688          * If the incoming packet was not addressed directly to us, use
 689          * designated interface for icmp replies specified by sysctl
 690          * net.inet.icmp.reply_src (default not set). Otherwise continue
 691          * with normal source selection.
 692          */
 693         if (icmp_reply_src[0] != '\0' && (ifp = ifunit(icmp_reply_src))) {
 694                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 695                         struct ifaddr *ifa = ifac->ifa;
 696
 697                         if (ifa->ifa_addr->sa_family != AF_INET)
 698                                 continue;
 699                         ia = ifatoia(ifa);
 700                         goto match;
 701                 }
 702         }
 703         /*
 704          * If the packet was transiting through us, use the address of
 705          * the interface that is the closest to the packet source.
 706          * When we don't have a route back to the packet source, stop here
 707          * and drop the packet.
 708          */
 709         ia = ip_rtaddr(ip->ip_dst, ro);
 710         if (ia == NULL) {
 711                 m_freem(m);
 712                 icmpstat.icps_noroute++;
 713                 goto done;
 714         }
 715 match:
 716         t = IA_SIN(ia)->sin_addr;
 717         ip->ip_src = t;
 718         ip->ip_ttl = ip_defttl;
 719
 720         if (optlen > 0) {
 721                 u_char *cp;
 722                 int opt, cnt;
 723                 u_int len;
 724
 725                 /*
 726                  * Retrieve any source routing from the incoming packet;
 727                  * add on any record-route or timestamp options.
 728                  */
 729                 cp = (u_char *) (ip + 1);
 730                 if ((opts = ip_srcroute(m)) == NULL &&
 731                     (opts = m_gethdr(MB_DONTWAIT, MT_HEADER))) {
 732                         opts->m_len = sizeof(struct in_addr);
 733                         mtod(opts, struct in_addr *)->s_addr = 0;
 734                 }
 735                 if (opts) {
 736 #ifdef ICMPPRINTFS
 737                         if (icmpprintfs)
 738                                 kprintf("icmp_reflect optlen %d rt %d => ",
 739                                        optlen, opts->m_len);
 740 #endif
 741                         for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
 742                                 opt = cp[IPOPT_OPTVAL];
 743                                 if (opt == IPOPT_EOL)
 744                                         break;
 745                                 if (opt == IPOPT_NOP)
 746                                         len = 1;
 747                                 else {
 748                                         if (cnt < IPOPT_OLEN + sizeof *cp)
 749                                                 break;
 750                                         len = cp[IPOPT_OLEN];
 751                                         if (len < IPOPT_OLEN + sizeof *cp ||
 752                                             len > cnt)
 753                                         break;
 754                                 }
 755                                 /*
 756                                  * Should check for overflow, but it
 757                                  * "can't happen".
 758                                  */
 759                                 if (opt == IPOPT_RR || opt == IPOPT_TS ||
 760                                     opt == IPOPT_SECURITY) {
 761                                         bcopy(cp,
 762                                               mtod(opts, caddr_t) + opts->m_len,
 763                                               len);
 764                                         opts->m_len += len;
 765                                 }
 766                         }
 767                         /* Terminate & pad, if necessary */
 768                         cnt = opts->m_len % 4;
 769                         if (cnt) {
 770                                 for (; cnt < 4; cnt++) {
 771                                         *(mtod(opts, caddr_t) + opts->m_len) =
 772                                             IPOPT_EOL;
 773                                         opts->m_len++;
 774                                 }
 775                         }
 776 #ifdef ICMPPRINTFS
 777                         if (icmpprintfs)
 778                                 kprintf("%d\n", opts->m_len);
 779 #endif
 780                 }
 781                 /*
 782                  * Now strip out original options by copying rest of first
 783                  * mbuf's data back, and adjust the IP length.
 784                  */
 785                 ip->ip_len -= optlen;
 786                 ip->ip_vhl = IP_VHL_BORING;
 787                 m->m_len -= optlen;
 788                 if (m->m_flags & M_PKTHDR)
 789                         m->m_pkthdr.len -= optlen;
 790                 optlen += sizeof(struct ip);
 791                 bcopy((caddr_t)ip + optlen, ip + 1,
 792                       m->m_len - sizeof(struct ip));
 793         }
 794         m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
 795         m->m_flags &= ~(M_BCAST|M_MCAST);
 796         icmp_send(m, opts, ro);
 797 done:
 798         if (opts)
 799                 m_free(opts);
 800         if (ro && ro->ro_rt)
 801                 RTFREE(ro->ro_rt);
 802 }
 803
 804 /*
 805  * Send an icmp packet back to the ip level,
 806  * after supplying a checksum.
 807  */
 808 static void
 809 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
 810 {
 811         struct ip *ip = mtod(m, struct ip *);
 812         struct icmp *icp;
 813         int hlen;
 814
 815         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 816         m->m_data += hlen;
 817         m->m_len -= hlen;
 818         icp = mtod(m, struct icmp *);
 819         icp->icmp_cksum = 0;
 820         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
 821         m->m_data -= hlen;
 822         m->m_len += hlen;
 823         m->m_pkthdr.rcvif = NULL;
 824 #ifdef ICMPPRINTFS
 825         if (icmpprintfs) {
 826                 char dst_buf[INET_ADDRSTRLEN], src_buf[INET_ADDRSTRLEN];
 827
 828                 kprintf("icmp_send dst %s src %s\n",
 829                     inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN),
 830                     inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN));
 831         }
 832 #endif
 833         ip_output(m, opts, rt, 0, NULL, NULL);
 834 }
 835
 836 n_time
 837 iptime(void)
 838 {
 839         struct timeval atv;
 840         u_long t;
 841
 842         getmicrotime(&atv);
 843         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 844         return (htonl(t));
 845 }
 846
 847 #if 1
 848 /*
 849  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 850  * given current value MTU.  If DIR is less than zero, a larger plateau
 851  * is returned; otherwise, a smaller value is returned.
 852  */
 853 int
 854 ip_next_mtu(int mtu, int dir)
 855 {
 856         static int mtutab[] = {
 857                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
 858                 68, 0
 859         };
 860         int i;
 861
 862         for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
 863                 if (mtu >= mtutab[i])
 864                         break;
 865         }
 866
 867         if (dir < 0) {
 868                 if (i == 0) {
 869                         return 0;
 870                 } else {
 871                         return mtutab[i - 1];
 872                 }
 873         } else {
 874                 if (mtutab[i] == 0) {
 875                         return 0;
 876                 } else if(mtu > mtutab[i]) {
 877                         return mtutab[i];
 878                 } else {
 879                         return mtutab[i + 1];
 880                 }
 881         }
 882 }
 883 #endif
 884
 885 #ifdef ICMP_BANDLIM
 886 /*
 887  * badport_bandlim() - check for ICMP bandwidth limit
 888  *
 889  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 890  *      hit our bandwidth limit and it is not ok.
 891  *
 892  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 893  *
 894  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 895  *      values.  We may eventually remove this separation (and simplify the
 896  *      code further).
 897  *
 898  *      Note that the printing of the error message is delayed so we can
 899  *      properly print the icmp error rate that the system was trying to do
 900  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 901  *      the 'final' error, but it doesn't make sense to solve the printing
 902  *      delay with more complex code.
 903  */
 904 int
 905 badport_bandlim(int which)
 906 {
 907         static int lticks[BANDLIM_MAX + 1];
 908         static int lpackets[BANDLIM_MAX + 1];
 909         int dticks;
 910         const char *bandlimittype[] = {
 911                 "Limiting icmp unreach response",
 912                 "Limiting icmp ping response",
 913                 "Limiting icmp tstamp response",
 914                 "Limiting closed port RST response",
 915                 "Limiting open port RST response"
 916                 };
 917
 918         /*
 919          * Return ok status if feature disabled or argument out of
 920          * ranage.
 921          */
 922
 923         if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
 924                 return(0);
 925         dticks = ticks - lticks[which];
 926
 927         /*
 928          * reset stats when cumulative dt exceeds one second.
 929          */
 930
 931         if ((unsigned int)dticks > hz) {
 932                 if (lpackets[which] > icmplim && icmplim_output) {
 933                         kprintf("%s from %d to %d packets per second\n",
 934                                 bandlimittype[which],
 935                                 lpackets[which],
 936                                 icmplim
 937                         );
 938                 }
 939                 lticks[which] = ticks;
 940                 lpackets[which] = 0;
 941         }
 942
 943         /*
 944          * bump packet count
 945          */
 946
 947         if (++lpackets[which] > icmplim) {
 948                 return(-1);
 949         }
 950         return(0);
 951 }
 952 #endif