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  */
  36
  37 #include "opt_ipsec.h"
  38
  39 #include <sys/param.h>
  40 #include <sys/systm.h>
  41 #include <sys/mbuf.h>
  42 #include <sys/protosw.h>
  43 #include <sys/socket.h>
  44 #include <sys/socketops.h>
  45 #include <sys/time.h>
  46 #include <sys/kernel.h>
  47 #include <sys/sysctl.h>
  48 #include <sys/in_cksum.h>
  49
  50 #include <machine/stdarg.h>
  51
  52 #include <net/if.h>
  53 #include <net/if_types.h>
  54 #include <net/route.h>
  55
  56 #define _IP_VHL
  57 #include <netinet/in.h>
  58 #include <netinet/in_systm.h>
  59 #include <netinet/in_var.h>
  60 #include <netinet/ip.h>
  61 #include <netinet/ip_icmp.h>
  62 #include <netinet/ip_var.h>
  63 #include <netinet/icmp_var.h>
  64
  65 #ifdef IPSEC
  66 #include <netinet6/ipsec.h>
  67 #include <netproto/key/key.h>
  68 #endif
  69
  70 #ifdef FAST_IPSEC
  71 #include <netproto/ipsec/ipsec.h>
  72 #include <netproto/ipsec/key.h>
  73 #define IPSEC
  74 #endif
  75
  76 /*
  77  * ICMP routines: error generation, receive packet processing, and
  78  * routines to turnaround packets back to the originator, and
  79  * host table maintenance routines.
  80  */
  81
  82 struct icmpstat icmpstat;
  83 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
  84         &icmpstat, icmpstat, "ICMP statistics");
  85
  86 static int      icmpmaskrepl = 0;
  87 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
  88         &icmpmaskrepl, 0, "Allow replies to netmask requests");
  89
  90 static int      drop_redirect = 0;
  91 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
  92         &drop_redirect, 0, "Ignore ICMP redirects");
  93
  94 static int      log_redirect = 0;
  95 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
  96         &log_redirect, 0, "Enable output about ICMP redirects");
  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 = 1;
 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 int     icmpprintfs = 0;
 139 #endif
 140
 141 static void     icmp_reflect (struct mbuf *);
 142 static void     icmp_send (struct mbuf *, struct mbuf *, struct route *);
 143
 144 extern  struct protosw inetsw[];
 145
 146 /*
 147  * Generate an error packet of type error
 148  * in response to bad packet ip.
 149  */
 150 void
 151 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
 152 {
 153         struct ip *oip = mtod(n, struct ip *), *nip;
 154         unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
 155         struct icmp *icp;
 156         struct mbuf *m;
 157         unsigned icmplen;
 158
 159 #ifdef ICMPPRINTFS
 160         if (icmpprintfs)
 161                 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code);
 162 #endif
 163         if (type != ICMP_REDIRECT)
 164                 icmpstat.icps_error++;
 165         /*
 166          * Don't send error if the original packet was encrypted.
 167          * Don't send error if not the first fragment of message.
 168          * Don't error if the old packet protocol was ICMP
 169          * error message, only known informational types.
 170          */
 171         if (n->m_flags & M_DECRYPTED)
 172                 goto freeit;
 173         if (oip->ip_off &~ (IP_MF|IP_DF))
 174                 goto freeit;
 175         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
 176           n->m_len >= oiplen + ICMP_MINLEN &&
 177           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
 178                 icmpstat.icps_oldicmp++;
 179                 goto freeit;
 180         }
 181         /* Don't send error in response to a multicast or broadcast packet */
 182         if (n->m_flags & (M_BCAST|M_MCAST))
 183                 goto freeit;
 184         /*
 185          * First, formulate icmp message
 186          */
 187         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
 188         if (m == NULL)
 189                 goto freeit;
 190         icmplen = min(oiplen + 8, oip->ip_len);
 191         if (icmplen < sizeof(struct ip))
 192                 panic("icmp_error: bad length");
 193         m->m_len = icmplen + ICMP_MINLEN;
 194         MH_ALIGN(m, m->m_len);
 195         icp = mtod(m, struct icmp *);
 196         if ((u_int)type > ICMP_MAXTYPE)
 197                 panic("icmp_error");
 198         icmpstat.icps_outhist[type]++;
 199         icp->icmp_type = type;
 200         if (type == ICMP_REDIRECT)
 201                 icp->icmp_gwaddr.s_addr = dest;
 202         else {
 203                 icp->icmp_void = 0;
 204                 /*
 205                  * The following assignments assume an overlay with the
 206                  * zeroed icmp_void field.
 207                  */
 208                 if (type == ICMP_PARAMPROB) {
 209                         icp->icmp_pptr = code;
 210                         code = 0;
 211                 } else if (type == ICMP_UNREACH &&
 212                         code == ICMP_UNREACH_NEEDFRAG && destmtu) {
 213                         icp->icmp_nextmtu = htons(destmtu);
 214                 }
 215         }
 216
 217         icp->icmp_code = code;
 218         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
 219         nip = &icp->icmp_ip;
 220
 221         /*
 222          * Convert fields to network representation.
 223          */
 224         nip->ip_len = htons(nip->ip_len);
 225         nip->ip_off = htons(nip->ip_off);
 226
 227         /*
 228          * Now, copy old ip header (without options)
 229          * in front of icmp message.
 230          */
 231         if (m->m_data - sizeof(struct ip) < m->m_pktdat)
 232                 panic("icmp len");
 233         m->m_data -= sizeof(struct ip);
 234         m->m_len += sizeof(struct ip);
 235         m->m_pkthdr.len = m->m_len;
 236         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
 237         nip = mtod(m, struct ip *);
 238         bcopy(oip, nip, sizeof(struct ip));
 239         nip->ip_len = m->m_len;
 240         nip->ip_vhl = IP_VHL_BORING;
 241         nip->ip_p = IPPROTO_ICMP;
 242         nip->ip_tos = 0;
 243         m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED;
 244         icmp_reflect(m);
 245
 246 freeit:
 247         m_freem(n);
 248 }
 249
 250 /*
 251  * Process a received ICMP message.
 252  */
 253 int
 254 icmp_input(struct mbuf **mp, int *offp, int proto)
 255 {
 256         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 257         struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
 258         struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
 259         struct icmp *icp;
 260         struct in_ifaddr *ia;
 261         struct mbuf *m = *mp;
 262         struct ip *ip = mtod(m, struct ip *);
 263         int icmplen = ip->ip_len;
 264         int i, hlen;
 265         int code;
 266
 267         *mp = NULL;
 268         hlen = *offp;
 269
 270         /*
 271          * Locate icmp structure in mbuf, and check
 272          * that not corrupted and of at least minimum length.
 273          */
 274 #ifdef ICMPPRINTFS
 275         if (icmpprintfs) {
 276                 char buf[sizeof "aaa.bbb.ccc.ddd"];
 277
 278                 strcpy(buf, inet_ntoa(ip->ip_src));
 279                 kprintf("icmp_input from %s to %s, len %d\n",
 280                        buf, inet_ntoa(ip->ip_dst), icmplen);
 281         }
 282 #endif
 283         if (icmplen < ICMP_MINLEN) {
 284                 icmpstat.icps_tooshort++;
 285                 goto freeit;
 286         }
 287         i = hlen + min(icmplen, ICMP_ADVLENMIN);
 288         if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
 289                 icmpstat.icps_tooshort++;
 290                 return(IPPROTO_DONE);
 291         }
 292         ip = mtod(m, struct ip *);
 293         m->m_len -= hlen;
 294         m->m_data += hlen;
 295         icp = mtod(m, struct icmp *);
 296         if (in_cksum(m, icmplen)) {
 297                 icmpstat.icps_checksum++;
 298                 goto freeit;
 299         }
 300         m->m_len += hlen;
 301         m->m_data -= 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                 code = PRC_QUENCH;
 386 deliver:
 387                 /*
 388                  * Problem with datagram; advise higher level routines.
 389                  */
 390                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 391                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 392                         icmpstat.icps_badlen++;
 393                         goto freeit;
 394                 }
 395                 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
 396                 /* Discard ICMP's in response to multicast packets */
 397                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
 398                         goto badcode;
 399 #ifdef ICMPPRINTFS
 400                 if (icmpprintfs)
 401                         kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
 402 #endif
 403                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 404 #if 1
 405                 /*
 406                  * MTU discovery:
 407                  * If we got a needfrag and there is a host route to the
 408                  * original destination, and the MTU is not locked, then
 409                  * set the MTU in the route to the suggested new value
 410                  * (if given) and then notify as usual.  The ULPs will
 411                  * notice that the MTU has changed and adapt accordingly.
 412                  * If no new MTU was suggested, then we guess a new one
 413                  * less than the current value.  If the new MTU is
 414                  * unreasonably small (arbitrarily set at 296), then
 415                  * we reset the MTU to the interface value and enable the
 416                  * lock bit, indicating that we are no longer doing MTU
 417                  * discovery.
 418                  */
 419                 if (code == PRC_MSGSIZE) {
 420                         struct rtentry *rt;
 421                         int mtu;
 422
 423                         rt = rtpurelookup((struct sockaddr *)&icmpsrc);
 424                         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
 425                             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
 426                                 mtu = ntohs(icp->icmp_nextmtu);
 427                                 if (!mtu)
 428                                         mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
 429                                                           1);
 430 #ifdef DEBUG_MTUDISC
 431                                 kprintf("MTU for %s reduced to %d\n",
 432                                         inet_ntoa(icmpsrc.sin_addr), mtu);
 433 #endif
 434                                 if (mtu < 296) {
 435                                         /* rt->rt_rmx.rmx_mtu =
 436                                                 rt->rt_ifp->if_mtu; */
 437                                         rt->rt_rmx.rmx_locks |= RTV_MTU;
 438                                 } else if (rt->rt_rmx.rmx_mtu > mtu) {
 439                                         rt->rt_rmx.rmx_mtu = mtu;
 440                                 }
 441                         }
 442                         if (rt != NULL)
 443                                 --rt->rt_refcnt;
 444                 }
 445 #endif
 446                 /*
 447                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
 448                  * notification to TCP layer.
 449                  */
 450                 so_pru_ctlinput(
 451                         &inetsw[ip_protox[icp->icmp_ip.ip_p]],
 452                         code, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
 453                 break;
 454
 455 badcode:
 456                 icmpstat.icps_badcode++;
 457                 break;
 458
 459         case ICMP_ECHO:
 460                 if (!icmpbmcastecho
 461                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 462                         icmpstat.icps_bmcastecho++;
 463                         break;
 464                 }
 465                 icp->icmp_type = ICMP_ECHOREPLY;
 466 #ifdef ICMP_BANDLIM
 467                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 468                         goto freeit;
 469                 else
 470 #endif
 471                         goto reflect;
 472
 473         case ICMP_TSTAMP:
 474                 if (!icmpbmcastecho
 475                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 476                         icmpstat.icps_bmcasttstamp++;
 477                         break;
 478                 }
 479                 if (icmplen < ICMP_TSLEN) {
 480                         icmpstat.icps_badlen++;
 481                         break;
 482                 }
 483                 icp->icmp_type = ICMP_TSTAMPREPLY;
 484                 icp->icmp_rtime = iptime();
 485                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 486 #ifdef ICMP_BANDLIM
 487                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 488                         goto freeit;
 489                 else
 490 #endif
 491                         goto reflect;
 492
 493         case ICMP_MASKREQ:
 494                 if (icmpmaskrepl == 0)
 495                         break;
 496                 /*
 497                  * We are not able to respond with all ones broadcast
 498                  * unless we receive it over a point-to-point interface.
 499                  */
 500                 if (icmplen < ICMP_MASKLEN)
 501                         break;
 502                 switch (ip->ip_dst.s_addr) {
 503
 504                 case INADDR_BROADCAST:
 505                 case INADDR_ANY:
 506                         icmpdst.sin_addr = ip->ip_src;
 507                         break;
 508
 509                 default:
 510                         icmpdst.sin_addr = ip->ip_dst;
 511                 }
 512                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 513                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 514                 if (ia == NULL)
 515                         break;
 516                 if (ia->ia_ifp == 0)
 517                         break;
 518                 icp->icmp_type = ICMP_MASKREPLY;
 519                 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 520                 if (ip->ip_src.s_addr == 0) {
 521                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 522                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 523                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 524                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 525                 }
 526 reflect:
 527                 ip->ip_len += hlen;     /* since ip_input deducts this */
 528                 icmpstat.icps_reflect++;
 529                 icmpstat.icps_outhist[icp->icmp_type]++;
 530                 icmp_reflect(m);
 531                 return(IPPROTO_DONE);
 532
 533         case ICMP_REDIRECT:
 534                 if (log_redirect) {
 535                         u_long src, dst, gw;
 536
 537                         src = ntohl(ip->ip_src.s_addr);
 538                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
 539                         gw = ntohl(icp->icmp_gwaddr.s_addr);
 540                         kprintf("icmp redirect from %d.%d.%d.%d: "
 541                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
 542                                (int)(src >> 24), (int)((src >> 16) & 0xff),
 543                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
 544                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
 545                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
 546                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
 547                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
 548                 }
 549                 if (drop_redirect)
 550                         break;
 551                 if (code > 3)
 552                         goto badcode;
 553                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 554                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 555                         icmpstat.icps_badlen++;
 556                         break;
 557                 }
 558                 /*
 559                  * Short circuit routing redirects to force
 560                  * immediate change in the kernel's routing
 561                  * tables.  The message is also handed to anyone
 562                  * listening on a raw socket (e.g. the routing
 563                  * daemon for use in updating its tables).
 564                  */
 565                 icmpgw.sin_addr = ip->ip_src;
 566                 icmpdst.sin_addr = icp->icmp_gwaddr;
 567 #ifdef  ICMPPRINTFS
 568                 if (icmpprintfs) {
 569                         char buf[sizeof "aaa.bbb.ccc.ddd"];
 570
 571                         strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
 572                         kprintf("redirect dst %s to %s\n",
 573                                buf, inet_ntoa(icp->icmp_gwaddr));
 574                 }
 575 #endif
 576                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 577                 rtredirect((struct sockaddr *)&icmpsrc,
 578                   (struct sockaddr *)&icmpdst,
 579                   NULL, RTF_GATEWAY | RTF_HOST,
 580                   (struct sockaddr *)&icmpgw);
 581                 kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 582 #ifdef IPSEC
 583                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 584 #endif
 585                 break;
 586
 587         /*
 588          * No kernel processing for the following;
 589          * just fall through to send to raw listener.
 590          */
 591         case ICMP_ECHOREPLY:
 592         case ICMP_ROUTERADVERT:
 593         case ICMP_ROUTERSOLICIT:
 594         case ICMP_TSTAMPREPLY:
 595         case ICMP_IREQREPLY:
 596         case ICMP_MASKREPLY:
 597         default:
 598                 break;
 599         }
 600
 601 raw:
 602         *mp = m;
 603         rip_input(mp, offp, proto);
 604         return(IPPROTO_DONE);
 605
 606 freeit:
 607         m_freem(m);
 608         return(IPPROTO_DONE);
 609 }
 610
 611 /*
 612  * Reflect the ip packet back to the source
 613  */
 614 static void
 615 icmp_reflect(struct mbuf *m)
 616 {
 617         struct ip *ip = mtod(m, struct ip *);
 618         struct in_ifaddr *ia;
 619         struct in_ifaddr_container *iac;
 620         struct ifaddr_container *ifac;
 621         struct ifnet *ifp;
 622         struct in_addr t;
 623         struct mbuf *opts = NULL;
 624         int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
 625         struct route *ro = NULL, rt;
 626
 627         if (!in_canforward(ip->ip_src) &&
 628             ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
 629              (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
 630                 m_freem(m);     /* Bad return address */
 631                 icmpstat.icps_badaddr++;
 632                 goto done;      /* Ip_output() will check for broadcast */
 633         }
 634         t = ip->ip_dst;
 635         ip->ip_dst = ip->ip_src;
 636
 637         ro = &rt;
 638         bzero(ro, sizeof *ro);
 639
 640         /*
 641          * If the incoming packet was addressed directly to us,
 642          * use dst as the src for the reply.  Otherwise (broadcast
 643          * or anonymous), use the address which corresponds
 644          * to the incoming interface.
 645          */
 646         ia = NULL;
 647         LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
 648                 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
 649                         ia = iac->ia;
 650                         goto match;
 651                 }
 652         }
 653         ifp = m->m_pkthdr.rcvif;
 654         if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) {
 655                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 656                         struct ifaddr *ifa = ifac->ifa;
 657
 658                         if (ifa->ifa_addr->sa_family != AF_INET)
 659                                 continue;
 660                         ia = ifatoia(ifa);
 661                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 662                             t.s_addr)
 663                                 goto match;
 664                 }
 665         }
 666         /*
 667          * If the packet was transiting through us, use the address of
 668          * the interface the packet came through in.  If that interface
 669          * doesn't have a suitable IP address, the normal selection
 670          * criteria apply.
 671          */
 672         if (icmp_rfi && ifp != NULL) {
 673                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 674                         struct ifaddr *ifa = ifac->ifa;
 675
 676                         if (ifa->ifa_addr->sa_family != AF_INET)
 677                                 continue;
 678                         ia = ifatoia(ifa);
 679                         goto match;
 680                 }
 681         }
 682         /*
 683          * If the incoming packet was not addressed directly to us, use
 684          * designated interface for icmp replies specified by sysctl
 685          * net.inet.icmp.reply_src (default not set). Otherwise continue
 686          * with normal source selection.
 687          */
 688         if (icmp_reply_src[0] != '\0' && (ifp = ifunit(icmp_reply_src))) {
 689                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 690                         struct ifaddr *ifa = ifac->ifa;
 691
 692                         if (ifa->ifa_addr->sa_family != AF_INET)
 693                                 continue;
 694                         ia = ifatoia(ifa);
 695                         goto match;
 696                 }
 697         }
 698         /*
 699          * If the packet was transiting through us, use the address of
 700          * the interface that is the closest to the packet source.
 701          * When we don't have a route back to the packet source, stop here
 702          * and drop the packet.
 703          */
 704         ia = ip_rtaddr(ip->ip_dst, ro);
 705         if (ia == NULL) {
 706                 m_freem(m);
 707                 icmpstat.icps_noroute++;
 708                 goto done;
 709         }
 710 match:
 711         t = IA_SIN(ia)->sin_addr;
 712         ip->ip_src = t;
 713         ip->ip_ttl = ip_defttl;
 714
 715         if (optlen > 0) {
 716                 u_char *cp;
 717                 int opt, cnt;
 718                 u_int len;
 719
 720                 /*
 721                  * Retrieve any source routing from the incoming packet;
 722                  * add on any record-route or timestamp options.
 723                  */
 724                 cp = (u_char *) (ip + 1);
 725                 if ((opts = ip_srcroute(m)) == NULL &&
 726                     (opts = m_gethdr(MB_DONTWAIT, MT_HEADER))) {
 727                         opts->m_len = sizeof(struct in_addr);
 728                         mtod(opts, struct in_addr *)->s_addr = 0;
 729                 }
 730                 if (opts) {
 731 #ifdef ICMPPRINTFS
 732                         if (icmpprintfs)
 733                                 kprintf("icmp_reflect optlen %d rt %d => ",
 734                                        optlen, opts->m_len);
 735 #endif
 736                         for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
 737                                 opt = cp[IPOPT_OPTVAL];
 738                                 if (opt == IPOPT_EOL)
 739                                         break;
 740                                 if (opt == IPOPT_NOP)
 741                                         len = 1;
 742                                 else {
 743                                         if (cnt < IPOPT_OLEN + sizeof *cp)
 744                                                 break;
 745                                         len = cp[IPOPT_OLEN];
 746                                         if (len < IPOPT_OLEN + sizeof *cp ||
 747                                             len > cnt)
 748                                         break;
 749                                 }
 750                                 /*
 751                                  * Should check for overflow, but it
 752                                  * "can't happen".
 753                                  */
 754                                 if (opt == IPOPT_RR || opt == IPOPT_TS ||
 755                                     opt == IPOPT_SECURITY) {
 756                                         bcopy(cp,
 757                                               mtod(opts, caddr_t) + opts->m_len,
 758                                               len);
 759                                         opts->m_len += len;
 760                                 }
 761                         }
 762                         /* Terminate & pad, if necessary */
 763                         cnt = opts->m_len % 4;
 764                         if (cnt) {
 765                                 for (; cnt < 4; cnt++) {
 766                                         *(mtod(opts, caddr_t) + opts->m_len) =
 767                                             IPOPT_EOL;
 768                                         opts->m_len++;
 769                                 }
 770                         }
 771 #ifdef ICMPPRINTFS
 772                         if (icmpprintfs)
 773                                 kprintf("%d\n", opts->m_len);
 774 #endif
 775                 }
 776                 /*
 777                  * Now strip out original options by copying rest of first
 778                  * mbuf's data back, and adjust the IP length.
 779                  */
 780                 ip->ip_len -= optlen;
 781                 ip->ip_vhl = IP_VHL_BORING;
 782                 m->m_len -= optlen;
 783                 if (m->m_flags & M_PKTHDR)
 784                         m->m_pkthdr.len -= optlen;
 785                 optlen += sizeof(struct ip);
 786                 bcopy((caddr_t)ip + optlen, ip + 1,
 787                       m->m_len - sizeof(struct ip));
 788         }
 789         m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
 790         m->m_flags &= ~(M_BCAST|M_MCAST);
 791         icmp_send(m, opts, ro);
 792 done:
 793         if (opts)
 794                 m_free(opts);
 795         if (ro && ro->ro_rt)
 796                 RTFREE(ro->ro_rt);
 797 }
 798
 799 /*
 800  * Send an icmp packet back to the ip level,
 801  * after supplying a checksum.
 802  */
 803 static void
 804 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
 805 {
 806         struct ip *ip = mtod(m, struct ip *);
 807         struct icmp *icp;
 808         int hlen;
 809
 810         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 811         m->m_data += hlen;
 812         m->m_len -= hlen;
 813         icp = mtod(m, struct icmp *);
 814         icp->icmp_cksum = 0;
 815         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
 816         m->m_data -= hlen;
 817         m->m_len += hlen;
 818         m->m_pkthdr.rcvif = NULL;
 819 #ifdef ICMPPRINTFS
 820         if (icmpprintfs) {
 821                 char buf[sizeof "aaa.bbb.ccc.ddd"];
 822
 823                 strcpy(buf, inet_ntoa(ip->ip_dst));
 824                 kprintf("icmp_send dst %s src %s\n", buf, inet_ntoa(ip->ip_src));
 825         }
 826 #endif
 827         ip_output(m, opts, rt, 0, NULL, NULL);
 828 }
 829
 830 n_time
 831 iptime(void)
 832 {
 833         struct timeval atv;
 834         u_long t;
 835
 836         getmicrotime(&atv);
 837         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 838         return (htonl(t));
 839 }
 840
 841 #if 1
 842 /*
 843  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 844  * given current value MTU.  If DIR is less than zero, a larger plateau
 845  * is returned; otherwise, a smaller value is returned.
 846  */
 847 int
 848 ip_next_mtu(int mtu, int dir)
 849 {
 850         static int mtutab[] = {
 851                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
 852                 68, 0
 853         };
 854         int i;
 855
 856         for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
 857                 if (mtu >= mtutab[i])
 858                         break;
 859         }
 860
 861         if (dir < 0) {
 862                 if (i == 0) {
 863                         return 0;
 864                 } else {
 865                         return mtutab[i - 1];
 866                 }
 867         } else {
 868                 if (mtutab[i] == 0) {
 869                         return 0;
 870                 } else if(mtu > mtutab[i]) {
 871                         return mtutab[i];
 872                 } else {
 873                         return mtutab[i + 1];
 874                 }
 875         }
 876 }
 877 #endif
 878
 879 #ifdef ICMP_BANDLIM
 880 /*
 881  * badport_bandlim() - check for ICMP bandwidth limit
 882  *
 883  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 884  *      hit our bandwidth limit and it is not ok.
 885  *
 886  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 887  *
 888  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 889  *      values.  We may eventually remove this separation (and simplify the
 890  *      code further).
 891  *
 892  *      Note that the printing of the error message is delayed so we can
 893  *      properly print the icmp error rate that the system was trying to do
 894  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 895  *      the 'final' error, but it doesn't make sense to solve the printing
 896  *      delay with more complex code.
 897  */
 898 int
 899 badport_bandlim(int which)
 900 {
 901         static int lticks[BANDLIM_MAX + 1];
 902         static int lpackets[BANDLIM_MAX + 1];
 903         int dticks;
 904         const char *bandlimittype[] = {
 905                 "Limiting icmp unreach response",
 906                 "Limiting icmp ping response",
 907                 "Limiting icmp tstamp response",
 908                 "Limiting closed port RST response",
 909                 "Limiting open port RST response"
 910                 };
 911
 912         /*
 913          * Return ok status if feature disabled or argument out of
 914          * ranage.
 915          */
 916
 917         if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
 918                 return(0);
 919         dticks = ticks - lticks[which];
 920
 921         /*
 922          * reset stats when cumulative dt exceeds one second.
 923          */
 924
 925         if ((unsigned int)dticks > hz) {
 926                 if (lpackets[which] > icmplim && icmplim_output) {
 927                         kprintf("%s from %d to %d packets per second\n",
 928                                 bandlimittype[which],
 929                                 lpackets[which],
 930                                 icmplim
 931                         );
 932                 }
 933                 lticks[which] = ticks;
 934                 lpackets[which] = 0;
 935         }
 936
 937         /*
 938          * bump packet count
 939          */
 940
 941         if (++lpackets[which] > icmplim) {
 942                 return(-1);
 943         }
 944         return(0);
 945 }
 946 #endif