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. All advertising materials mentioning features or use of this software
  14  *    must display the following acknowledgement:
  15  *      This product includes software developed by the University of
  16  *      California, Berkeley and its contributors.
  17  * 4. Neither the name of the University nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  *
  33  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  34  * $FreeBSD: src/sys/netinet/ip_icmp.c,v 1.39.2.19 2003/01/24 05:11:34 sam Exp $
  35  * $DragonFly: src/sys/netinet/ip_icmp.c,v 1.32 2008/10/27 02:56:30 sephe Exp $
  36  */
  37
  38 #include "opt_ipsec.h"
  39
  40 #include <sys/param.h>
  41 #include <sys/systm.h>
  42 #include <sys/mbuf.h>
  43 #include <sys/protosw.h>
  44 #include <sys/socket.h>
  45 #include <sys/socketops.h>
  46 #include <sys/time.h>
  47 #include <sys/kernel.h>
  48 #include <sys/sysctl.h>
  49 #include <sys/in_cksum.h>
  50
  51 #include <machine/stdarg.h>
  52
  53 #include <net/if.h>
  54 #include <net/if_types.h>
  55 #include <net/route.h>
  56
  57 #define _IP_VHL
  58 #include <netinet/in.h>
  59 #include <netinet/in_systm.h>
  60 #include <netinet/in_var.h>
  61 #include <netinet/ip.h>
  62 #include <netinet/ip_icmp.h>
  63 #include <netinet/ip_var.h>
  64 #include <netinet/icmp_var.h>
  65
  66 #ifdef IPSEC
  67 #include <netinet6/ipsec.h>
  68 #include <netproto/key/key.h>
  69 #endif
  70
  71 #ifdef FAST_IPSEC
  72 #include <netproto/ipsec/ipsec.h>
  73 #include <netproto/ipsec/key.h>
  74 #define IPSEC
  75 #endif
  76
  77 /*
  78  * ICMP routines: error generation, receive packet processing, and
  79  * routines to turnaround packets back to the originator, and
  80  * host table maintenance routines.
  81  */
  82
  83 struct icmpstat icmpstat;
  84 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
  85         &icmpstat, icmpstat, "ICMP statistics");
  86
  87 static int      icmpmaskrepl = 0;
  88 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
  89         &icmpmaskrepl, 0, "Allow replies to netmask requests");
  90
  91 static int      drop_redirect = 0;
  92 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
  93         &drop_redirect, 0, "Ignore ICMP redirects");
  94
  95 static int      log_redirect = 0;
  96 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
  97         &log_redirect, 0, "Enable output about ICMP redirects");
  98
  99 #ifdef ICMP_BANDLIM
 100
 101 /*
 102  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
 103  *      variable content is -1 and read-only.
 104  */
 105
 106 static int      icmplim = 200;
 107 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
 108         &icmplim, 0, "ICMP bandwidth limit");
 109 #else
 110
 111 static int      icmplim = -1;
 112 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
 113         &icmplim, 0, "ICMP bandwidth limit");
 114
 115 #endif
 116
 117 static int      icmplim_output = 1;
 118 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
 119         &icmplim_output, 0, "Enable output about ICMP bandwidth limits");
 120
 121 /*
 122  * ICMP broadcast echo sysctl
 123  */
 124
 125 static int      icmpbmcastecho = 0;
 126 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
 127     &icmpbmcastecho, 0, "");
 128
 129 static char     icmp_reply_src[IFNAMSIZ];
 130 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
 131         icmp_reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
 132
 133 static int      icmp_rfi;
 134 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
 135         &icmp_rfi, 0, "ICMP reply from incoming interface for "
 136         "non-local packets");
 137
 138 #ifdef ICMPPRINTFS
 139 int     icmpprintfs = 0;
 140 #endif
 141
 142 static void     icmp_reflect (struct mbuf *);
 143 static void     icmp_send (struct mbuf *, struct mbuf *, struct route *);
 144
 145 extern  struct protosw inetsw[];
 146
 147 /*
 148  * Generate an error packet of type error
 149  * in response to bad packet ip.
 150  */
 151 void
 152 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
 153 {
 154         struct ip *oip = mtod(n, struct ip *), *nip;
 155         unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
 156         struct icmp *icp;
 157         struct mbuf *m;
 158         unsigned icmplen;
 159
 160 #ifdef ICMPPRINTFS
 161         if (icmpprintfs)
 162                 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code);
 163 #endif
 164         if (type != ICMP_REDIRECT)
 165                 icmpstat.icps_error++;
 166         /*
 167          * Don't send error if the original packet was encrypted.
 168          * Don't send error if not the first fragment of message.
 169          * Don't error if the old packet protocol was ICMP
 170          * error message, only known informational types.
 171          */
 172         if (n->m_flags & M_DECRYPTED)
 173                 goto freeit;
 174         if (oip->ip_off &~ (IP_MF|IP_DF))
 175                 goto freeit;
 176         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
 177           n->m_len >= oiplen + ICMP_MINLEN &&
 178           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
 179                 icmpstat.icps_oldicmp++;
 180                 goto freeit;
 181         }
 182         /* Don't send error in response to a multicast or broadcast packet */
 183         if (n->m_flags & (M_BCAST|M_MCAST))
 184                 goto freeit;
 185         /*
 186          * First, formulate icmp message
 187          */
 188         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
 189         if (m == NULL)
 190                 goto freeit;
 191         icmplen = min(oiplen + 8, oip->ip_len);
 192         if (icmplen < sizeof(struct ip))
 193                 panic("icmp_error: bad length");
 194         m->m_len = icmplen + ICMP_MINLEN;
 195         MH_ALIGN(m, m->m_len);
 196         icp = mtod(m, struct icmp *);
 197         if ((u_int)type > ICMP_MAXTYPE)
 198                 panic("icmp_error");
 199         icmpstat.icps_outhist[type]++;
 200         icp->icmp_type = type;
 201         if (type == ICMP_REDIRECT)
 202                 icp->icmp_gwaddr.s_addr = dest;
 203         else {
 204                 icp->icmp_void = 0;
 205                 /*
 206                  * The following assignments assume an overlay with the
 207                  * zeroed icmp_void field.
 208                  */
 209                 if (type == ICMP_PARAMPROB) {
 210                         icp->icmp_pptr = code;
 211                         code = 0;
 212                 } else if (type == ICMP_UNREACH &&
 213                         code == ICMP_UNREACH_NEEDFRAG && destmtu) {
 214                         icp->icmp_nextmtu = htons(destmtu);
 215                 }
 216         }
 217
 218         icp->icmp_code = code;
 219         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
 220         nip = &icp->icmp_ip;
 221
 222         /*
 223          * Convert fields to network representation.
 224          */
 225         nip->ip_len = htons(nip->ip_len);
 226         nip->ip_off = htons(nip->ip_off);
 227
 228         /*
 229          * Now, copy old ip header (without options)
 230          * in front of icmp message.
 231          */
 232         if (m->m_data - sizeof(struct ip) < m->m_pktdat)
 233                 panic("icmp len");
 234         m->m_data -= sizeof(struct ip);
 235         m->m_len += sizeof(struct ip);
 236         m->m_pkthdr.len = m->m_len;
 237         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
 238         nip = mtod(m, struct ip *);
 239         bcopy(oip, nip, sizeof(struct ip));
 240         nip->ip_len = m->m_len;
 241         nip->ip_vhl = IP_VHL_BORING;
 242         nip->ip_p = IPPROTO_ICMP;
 243         nip->ip_tos = 0;
 244         m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED;
 245         icmp_reflect(m);
 246
 247 freeit:
 248         m_freem(n);
 249 }
 250
 251 /*
 252  * Process a received ICMP message.
 253  */
 254 int
 255 icmp_input(struct mbuf **mp, int *offp, int proto)
 256 {
 257         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 258         struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
 259         struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
 260         struct icmp *icp;
 261         struct in_ifaddr *ia;
 262         struct mbuf *m = *mp;
 263         struct ip *ip = mtod(m, struct ip *);
 264         int icmplen = ip->ip_len;
 265         int i, hlen;
 266         int code;
 267
 268         *mp = NULL;
 269         hlen = *offp;
 270
 271         /*
 272          * Locate icmp structure in mbuf, and check
 273          * that not corrupted and of at least minimum length.
 274          */
 275 #ifdef ICMPPRINTFS
 276         if (icmpprintfs) {
 277                 char buf[sizeof "aaa.bbb.ccc.ddd"];
 278
 279                 strcpy(buf, inet_ntoa(ip->ip_src));
 280                 kprintf("icmp_input from %s to %s, len %d\n",
 281                        buf, inet_ntoa(ip->ip_dst), icmplen);
 282         }
 283 #endif
 284         if (icmplen < ICMP_MINLEN) {
 285                 icmpstat.icps_tooshort++;
 286                 goto freeit;
 287         }
 288         i = hlen + min(icmplen, ICMP_ADVLENMIN);
 289         if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
 290                 icmpstat.icps_tooshort++;
 291                 return(IPPROTO_DONE);
 292         }
 293         ip = mtod(m, struct ip *);
 294         m->m_len -= hlen;
 295         m->m_data += hlen;
 296         icp = mtod(m, struct icmp *);
 297         if (in_cksum(m, icmplen)) {
 298                 icmpstat.icps_checksum++;
 299                 goto freeit;
 300         }
 301         m->m_len += hlen;
 302         m->m_data -= hlen;
 303
 304         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
 305                 /*
 306                  * Deliver very specific ICMP type only.
 307                  */
 308                 switch (icp->icmp_type) {
 309                 case ICMP_UNREACH:
 310                 case ICMP_TIMXCEED:
 311                         break;
 312                 default:
 313                         goto freeit;
 314                 }
 315         }
 316
 317 #ifdef ICMPPRINTFS
 318         if (icmpprintfs)
 319                 kprintf("icmp_input, type %d code %d\n", icp->icmp_type,
 320                     icp->icmp_code);
 321 #endif
 322
 323         /*
 324          * Message type specific processing.
 325          */
 326         if (icp->icmp_type > ICMP_MAXTYPE)
 327                 goto raw;
 328         icmpstat.icps_inhist[icp->icmp_type]++;
 329         code = icp->icmp_code;
 330         switch (icp->icmp_type) {
 331
 332         case ICMP_UNREACH:
 333                 switch (code) {
 334                         case ICMP_UNREACH_NET:
 335                         case ICMP_UNREACH_HOST:
 336                         case ICMP_UNREACH_SRCFAIL:
 337                         case ICMP_UNREACH_NET_UNKNOWN:
 338                         case ICMP_UNREACH_HOST_UNKNOWN:
 339                         case ICMP_UNREACH_ISOLATED:
 340                         case ICMP_UNREACH_TOSNET:
 341                         case ICMP_UNREACH_TOSHOST:
 342                         case ICMP_UNREACH_HOST_PRECEDENCE:
 343                         case ICMP_UNREACH_PRECEDENCE_CUTOFF:
 344                                 code = PRC_UNREACH_NET;
 345                                 break;
 346
 347                         case ICMP_UNREACH_NEEDFRAG:
 348                                 code = PRC_MSGSIZE;
 349                                 break;
 350
 351                         /*
 352                          * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
 353                          * Treat subcodes 2,3 as immediate RST
 354                          */
 355                         case ICMP_UNREACH_PROTOCOL:
 356                         case ICMP_UNREACH_PORT:
 357                                 code = PRC_UNREACH_PORT;
 358                                 break;
 359
 360                         case ICMP_UNREACH_NET_PROHIB:
 361                         case ICMP_UNREACH_HOST_PROHIB:
 362                         case ICMP_UNREACH_FILTER_PROHIB:
 363                                 code = PRC_UNREACH_ADMIN_PROHIB;
 364                                 break;
 365
 366                         default:
 367                                 goto badcode;
 368                 }
 369                 goto deliver;
 370
 371         case ICMP_TIMXCEED:
 372                 if (code > 1)
 373                         goto badcode;
 374                 code += PRC_TIMXCEED_INTRANS;
 375                 goto deliver;
 376
 377         case ICMP_PARAMPROB:
 378                 if (code > 1)
 379                         goto badcode;
 380                 code = PRC_PARAMPROB;
 381                 goto deliver;
 382
 383         case ICMP_SOURCEQUENCH:
 384                 if (code)
 385                         goto badcode;
 386                 code = PRC_QUENCH;
 387 deliver:
 388                 /*
 389                  * Problem with datagram; advise higher level routines.
 390                  */
 391                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 392                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 393                         icmpstat.icps_badlen++;
 394                         goto freeit;
 395                 }
 396                 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
 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                                 mtu = ntohs(icp->icmp_nextmtu);
 428                                 if (!mtu)
 429                                         mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
 430                                                           1);
 431 #ifdef DEBUG_MTUDISC
 432                                 kprintf("MTU for %s reduced to %d\n",
 433                                         inet_ntoa(icmpsrc.sin_addr), mtu);
 434 #endif
 435                                 if (mtu < 296) {
 436                                         /* rt->rt_rmx.rmx_mtu =
 437                                                 rt->rt_ifp->if_mtu; */
 438                                         rt->rt_rmx.rmx_locks |= RTV_MTU;
 439                                 } else if (rt->rt_rmx.rmx_mtu > mtu) {
 440                                         rt->rt_rmx.rmx_mtu = mtu;
 441                                 }
 442                         }
 443                         if (rt != NULL)
 444                                 --rt->rt_refcnt;
 445                 }
 446 #endif
 447                 /*
 448                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
 449                  * notification to TCP layer.
 450                  */
 451                 so_pru_ctlinput(
 452                         &inetsw[ip_protox[icp->icmp_ip.ip_p]],
 453                         code, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
 454                 break;
 455
 456 badcode:
 457                 icmpstat.icps_badcode++;
 458                 break;
 459
 460         case ICMP_ECHO:
 461                 if (!icmpbmcastecho
 462                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 463                         icmpstat.icps_bmcastecho++;
 464                         break;
 465                 }
 466                 icp->icmp_type = ICMP_ECHOREPLY;
 467 #ifdef ICMP_BANDLIM
 468                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 469                         goto freeit;
 470                 else
 471 #endif
 472                         goto reflect;
 473
 474         case ICMP_TSTAMP:
 475                 if (!icmpbmcastecho
 476                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 477                         icmpstat.icps_bmcasttstamp++;
 478                         break;
 479                 }
 480                 if (icmplen < ICMP_TSLEN) {
 481                         icmpstat.icps_badlen++;
 482                         break;
 483                 }
 484                 icp->icmp_type = ICMP_TSTAMPREPLY;
 485                 icp->icmp_rtime = iptime();
 486                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 487 #ifdef ICMP_BANDLIM
 488                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 489                         goto freeit;
 490                 else
 491 #endif
 492                         goto reflect;
 493
 494         case ICMP_MASKREQ:
 495                 if (icmpmaskrepl == 0)
 496                         break;
 497                 /*
 498                  * We are not able to respond with all ones broadcast
 499                  * unless we receive it over a point-to-point interface.
 500                  */
 501                 if (icmplen < ICMP_MASKLEN)
 502                         break;
 503                 switch (ip->ip_dst.s_addr) {
 504
 505                 case INADDR_BROADCAST:
 506                 case INADDR_ANY:
 507                         icmpdst.sin_addr = ip->ip_src;
 508                         break;
 509
 510                 default:
 511                         icmpdst.sin_addr = ip->ip_dst;
 512                 }
 513                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 514                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 515                 if (ia == 0)
 516                         break;
 517                 if (ia->ia_ifp == 0)
 518                         break;
 519                 icp->icmp_type = ICMP_MASKREPLY;
 520                 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 521                 if (ip->ip_src.s_addr == 0) {
 522                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 523                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 524                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 525                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 526                 }
 527 reflect:
 528                 ip->ip_len += hlen;     /* since ip_input deducts this */
 529                 icmpstat.icps_reflect++;
 530                 icmpstat.icps_outhist[icp->icmp_type]++;
 531                 icmp_reflect(m);
 532                 return(IPPROTO_DONE);
 533
 534         case ICMP_REDIRECT:
 535                 if (log_redirect) {
 536                         u_long src, dst, gw;
 537
 538                         src = ntohl(ip->ip_src.s_addr);
 539                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
 540                         gw = ntohl(icp->icmp_gwaddr.s_addr);
 541                         kprintf("icmp redirect from %d.%d.%d.%d: "
 542                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
 543                                (int)(src >> 24), (int)((src >> 16) & 0xff),
 544                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
 545                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
 546                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
 547                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
 548                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
 549                 }
 550                 if (drop_redirect)
 551                         break;
 552                 if (code > 3)
 553                         goto badcode;
 554                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 555                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 556                         icmpstat.icps_badlen++;
 557                         break;
 558                 }
 559                 /*
 560                  * Short circuit routing redirects to force
 561                  * immediate change in the kernel's routing
 562                  * tables.  The message is also handed to anyone
 563                  * listening on a raw socket (e.g. the routing
 564                  * daemon for use in updating its tables).
 565                  */
 566                 icmpgw.sin_addr = ip->ip_src;
 567                 icmpdst.sin_addr = icp->icmp_gwaddr;
 568 #ifdef  ICMPPRINTFS
 569                 if (icmpprintfs) {
 570                         char buf[sizeof "aaa.bbb.ccc.ddd"];
 571
 572                         strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
 573                         kprintf("redirect dst %s to %s\n",
 574                                buf, inet_ntoa(icp->icmp_gwaddr));
 575                 }
 576 #endif
 577                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 578                 rtredirect((struct sockaddr *)&icmpsrc,
 579                   (struct sockaddr *)&icmpdst,
 580                   NULL, RTF_GATEWAY | RTF_HOST,
 581                   (struct sockaddr *)&icmpgw);
 582                 kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 583 #ifdef IPSEC
 584                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 585 #endif
 586                 break;
 587
 588         /*
 589          * No kernel processing for the following;
 590          * just fall through to send to raw listener.
 591          */
 592         case ICMP_ECHOREPLY:
 593         case ICMP_ROUTERADVERT:
 594         case ICMP_ROUTERSOLICIT:
 595         case ICMP_TSTAMPREPLY:
 596         case ICMP_IREQREPLY:
 597         case ICMP_MASKREPLY:
 598         default:
 599                 break;
 600         }
 601
 602 raw:
 603         *mp = m;
 604         rip_input(mp, offp, proto);
 605         return(IPPROTO_DONE);
 606
 607 freeit:
 608         m_freem(m);
 609         return(IPPROTO_DONE);
 610 }
 611
 612 /*
 613  * Reflect the ip packet back to the source
 614  */
 615 static void
 616 icmp_reflect(struct mbuf *m)
 617 {
 618         struct ip *ip = mtod(m, struct ip *);
 619         struct in_ifaddr *ia;
 620         struct in_ifaddr_container *iac;
 621         struct ifaddr_container *ifac;
 622         struct ifnet *ifp;
 623         struct in_addr t;
 624         struct mbuf *opts = 0;
 625         int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
 626         struct route *ro = NULL, rt;
 627
 628         if (!in_canforward(ip->ip_src) &&
 629             ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
 630              (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
 631                 m_freem(m);     /* Bad return address */
 632                 icmpstat.icps_badaddr++;
 633                 goto done;      /* Ip_output() will check for broadcast */
 634         }
 635         t = ip->ip_dst;
 636         ip->ip_dst = ip->ip_src;
 637
 638         ro = &rt;
 639         bzero(ro, sizeof *ro);
 640
 641         /*
 642          * If the incoming packet was addressed directly to us,
 643          * use dst as the src for the reply.  Otherwise (broadcast
 644          * or anonymous), use the address which corresponds
 645          * to the incoming interface.
 646          */
 647         ia = NULL;
 648         LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
 649                 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
 650                         ia = iac->ia;
 651                         goto match;
 652                 }
 653         }
 654         ifp = m->m_pkthdr.rcvif;
 655         if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) {
 656                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 657                         struct ifaddr *ifa = ifac->ifa;
 658
 659                         if (ifa->ifa_addr->sa_family != AF_INET)
 660                                 continue;
 661                         ia = ifatoia(ifa);
 662                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 663                             t.s_addr)
 664                                 goto match;
 665                 }
 666         }
 667         /*
 668          * If the packet was transiting through us, use the address of
 669          * the interface the packet came through in.  If that interface
 670          * doesn't have a suitable IP address, the normal selection
 671          * criteria apply.
 672          */
 673         if (icmp_rfi && ifp != NULL) {
 674                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 675                         struct ifaddr *ifa = ifac->ifa;
 676
 677                         if (ifa->ifa_addr->sa_family != AF_INET)
 678                                 continue;
 679                         ia = ifatoia(ifa);
 680                         goto match;
 681                 }
 682         }
 683         /*
 684          * If the incoming packet was not addressed directly to us, use
 685          * designated interface for icmp replies specified by sysctl
 686          * net.inet.icmp.reply_src (default not set). Otherwise continue
 687          * with normal source selection.
 688          */
 689         if (icmp_reply_src[0] != '\0' && (ifp = ifunit(icmp_reply_src))) {
 690                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 691                         struct ifaddr *ifa = ifac->ifa;
 692
 693                         if (ifa->ifa_addr->sa_family != AF_INET)
 694                                 continue;
 695                         ia = ifatoia(ifa);
 696                         goto match;
 697                 }
 698         }
 699         /*
 700          * If the packet was transiting through us, use the address of
 701          * the interface that is the closest to the packet source.
 702          * When we don't have a route back to the packet source, stop here
 703          * and drop the packet.
 704          */
 705         ia = ip_rtaddr(ip->ip_dst, ro);
 706         if (ia == NULL) {
 707                 m_freem(m);
 708                 icmpstat.icps_noroute++;
 709                 goto done;
 710         }
 711 match:
 712         t = IA_SIN(ia)->sin_addr;
 713         ip->ip_src = t;
 714         ip->ip_ttl = ip_defttl;
 715
 716         if (optlen > 0) {
 717                 u_char *cp;
 718                 int opt, cnt;
 719                 u_int len;
 720
 721                 /*
 722                  * Retrieve any source routing from the incoming packet;
 723                  * add on any record-route or timestamp options.
 724                  */
 725                 cp = (u_char *) (ip + 1);
 726                 if ((opts = ip_srcroute(m)) == 0 &&
 727                     (opts = m_gethdr(MB_DONTWAIT, MT_HEADER))) {
 728                         opts->m_len = sizeof(struct in_addr);
 729                         mtod(opts, struct in_addr *)->s_addr = 0;
 730                 }
 731                 if (opts) {
 732 #ifdef ICMPPRINTFS
 733                         if (icmpprintfs)
 734                                 kprintf("icmp_reflect optlen %d rt %d => ",
 735                                        optlen, opts->m_len);
 736 #endif
 737                         for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
 738                                 opt = cp[IPOPT_OPTVAL];
 739                                 if (opt == IPOPT_EOL)
 740                                         break;
 741                                 if (opt == IPOPT_NOP)
 742                                         len = 1;
 743                                 else {
 744                                         if (cnt < IPOPT_OLEN + sizeof *cp)
 745                                                 break;
 746                                         len = cp[IPOPT_OLEN];
 747                                         if (len < IPOPT_OLEN + sizeof *cp ||
 748                                             len > cnt)
 749                                         break;
 750                                 }
 751                                 /*
 752                                  * Should check for overflow, but it
 753                                  * "can't happen".
 754                                  */
 755                                 if (opt == IPOPT_RR || opt == IPOPT_TS ||
 756                                     opt == IPOPT_SECURITY) {
 757                                         bcopy(cp,
 758                                               mtod(opts, caddr_t) + opts->m_len,
 759                                               len);
 760                                         opts->m_len += len;
 761                                 }
 762                         }
 763                         /* Terminate & pad, if necessary */
 764                         cnt = opts->m_len % 4;
 765                         if (cnt) {
 766                                 for (; cnt < 4; cnt++) {
 767                                         *(mtod(opts, caddr_t) + opts->m_len) =
 768                                             IPOPT_EOL;
 769                                         opts->m_len++;
 770                                 }
 771                         }
 772 #ifdef ICMPPRINTFS
 773                         if (icmpprintfs)
 774                                 kprintf("%d\n", opts->m_len);
 775 #endif
 776                 }
 777                 /*
 778                  * Now strip out original options by copying rest of first
 779                  * mbuf's data back, and adjust the IP length.
 780                  */
 781                 ip->ip_len -= optlen;
 782                 ip->ip_vhl = IP_VHL_BORING;
 783                 m->m_len -= optlen;
 784                 if (m->m_flags & M_PKTHDR)
 785                         m->m_pkthdr.len -= optlen;
 786                 optlen += sizeof(struct ip);
 787                 bcopy((caddr_t)ip + optlen, ip + 1,
 788                       m->m_len - sizeof(struct ip));
 789         }
 790         m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
 791         m->m_flags &= ~(M_BCAST|M_MCAST);
 792         icmp_send(m, opts, ro);
 793 done:
 794         if (opts)
 795                 m_free(opts);
 796         if (ro && ro->ro_rt)
 797                 RTFREE(ro->ro_rt);
 798 }
 799
 800 /*
 801  * Send an icmp packet back to the ip level,
 802  * after supplying a checksum.
 803  */
 804 static void
 805 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
 806 {
 807         struct ip *ip = mtod(m, struct ip *);
 808         struct icmp *icp;
 809         int hlen;
 810
 811         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 812         m->m_data += hlen;
 813         m->m_len -= hlen;
 814         icp = mtod(m, struct icmp *);
 815         icp->icmp_cksum = 0;
 816         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
 817         m->m_data -= hlen;
 818         m->m_len += hlen;
 819         m->m_pkthdr.rcvif = NULL;
 820 #ifdef ICMPPRINTFS
 821         if (icmpprintfs) {
 822                 char buf[sizeof "aaa.bbb.ccc.ddd"];
 823
 824                 strcpy(buf, inet_ntoa(ip->ip_dst));
 825                 kprintf("icmp_send dst %s src %s\n", buf, inet_ntoa(ip->ip_src));
 826         }
 827 #endif
 828         ip_output(m, opts, rt, 0, NULL, NULL);
 829 }
 830
 831 n_time
 832 iptime(void)
 833 {
 834         struct timeval atv;
 835         u_long t;
 836
 837         getmicrotime(&atv);
 838         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 839         return (htonl(t));
 840 }
 841
 842 #if 1
 843 /*
 844  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 845  * given current value MTU.  If DIR is less than zero, a larger plateau
 846  * is returned; otherwise, a smaller value is returned.
 847  */
 848 int
 849 ip_next_mtu(int mtu, int dir)
 850 {
 851         static int mtutab[] = {
 852                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
 853                 68, 0
 854         };
 855         int i;
 856
 857         for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
 858                 if (mtu >= mtutab[i])
 859                         break;
 860         }
 861
 862         if (dir < 0) {
 863                 if (i == 0) {
 864                         return 0;
 865                 } else {
 866                         return mtutab[i - 1];
 867                 }
 868         } else {
 869                 if (mtutab[i] == 0) {
 870                         return 0;
 871                 } else if(mtu > mtutab[i]) {
 872                         return mtutab[i];
 873                 } else {
 874                         return mtutab[i + 1];
 875                 }
 876         }
 877 }
 878 #endif
 879
 880 #ifdef ICMP_BANDLIM
 881 /*
 882  * badport_bandlim() - check for ICMP bandwidth limit
 883  *
 884  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 885  *      hit our bandwidth limit and it is not ok.
 886  *
 887  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 888  *
 889  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 890  *      values.  We may eventually remove this separation (and simplify the
 891  *      code further).
 892  *
 893  *      Note that the printing of the error message is delayed so we can
 894  *      properly print the icmp error rate that the system was trying to do
 895  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 896  *      the 'final' error, but it doesn't make sense to solve the printing
 897  *      delay with more complex code.
 898  */
 899 int
 900 badport_bandlim(int which)
 901 {
 902         static int lticks[BANDLIM_MAX + 1];
 903         static int lpackets[BANDLIM_MAX + 1];
 904         int dticks;
 905         const char *bandlimittype[] = {
 906                 "Limiting icmp unreach response",
 907                 "Limiting icmp ping response",
 908                 "Limiting icmp tstamp response",
 909                 "Limiting closed port RST response",
 910                 "Limiting open port RST response"
 911                 };
 912
 913         /*
 914          * Return ok status if feature disabled or argument out of
 915          * ranage.
 916          */
 917
 918         if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
 919                 return(0);
 920         dticks = ticks - lticks[which];
 921
 922         /*
 923          * reset stats when cumulative dt exceeds one second.
 924          */
 925
 926         if ((unsigned int)dticks > hz) {
 927                 if (lpackets[which] > icmplim && icmplim_output) {
 928                         kprintf("%s from %d to %d packets per second\n",
 929                                 bandlimittype[which],
 930                                 lpackets[which],
 931                                 icmplim
 932                         );
 933                 }
 934                 lticks[which] = ticks;
 935                 lpackets[which] = 0;
 936         }
 937
 938         /*
 939          * bump packet count
 940          */
 941
 942         if (++lpackets[which] > icmplim) {
 943                 return(-1);
 944         }
 945         return(0);
 946 }
 947 #endif