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  * 4. 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  */
  31
  32 #include <sys/cdefs.h>
  33 __FBSDID("$FreeBSD$");
  34
  35 #include "opt_ipsec.h"
  36 #include "opt_mac.h"
  37
  38 #include <sys/param.h>
  39 #include <sys/systm.h>
  40 #include <sys/mbuf.h>
  41 #include <sys/protosw.h>
  42 #include <sys/socket.h>
  43 #include <sys/time.h>
  44 #include <sys/kernel.h>
  45 #include <sys/sysctl.h>
  46 #include <sys/vimage.h>
  47
  48 #include <net/if.h>
  49 #include <net/if_types.h>
  50 #include <net/route.h>
  51
  52 #include <netinet/in.h>
  53 #include <netinet/in_pcb.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/ip_options.h>
  60 #include <netinet/tcp.h>
  61 #include <netinet/tcp_var.h>
  62 #include <netinet/tcpip.h>
  63 #include <netinet/icmp_var.h>
  64 #include <netinet/vinet.h>
  65
  66 #ifdef IPSEC
  67 #include <netipsec/ipsec.h>
  68 #include <netipsec/key.h>
  69 #endif
  70
  71 #include <machine/in_cksum.h>
  72
  73 #include <security/mac/mac_framework.h>
  74
  75 /*
  76  * ICMP routines: error generation, receive packet processing, and
  77  * routines to turnaround packets back to the originator, and
  78  * host table maintenance routines.
  79  */
  80
  81 #ifdef VIMAGE_GLOBALS
  82 struct icmpstat icmpstat;
  83 static int      icmpmaskrepl;
  84 static u_int    icmpmaskfake;
  85 static int      drop_redirect;
  86 static int      log_redirect;
  87 static int      icmplim;
  88 static int      icmplim_output;
  89 static char     reply_src[IFNAMSIZ];
  90 static int      icmp_rfi;
  91 static int      icmp_quotelen;
  92 static int      icmpbmcastecho;
  93 #endif
  94
  95 SYSCTL_V_STRUCT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_STATS, stats,
  96         CTLFLAG_RW, icmpstat, icmpstat, "");
  97
  98 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_MASKREPL, maskrepl,
  99         CTLFLAG_RW, icmpmaskrepl, 0,
 100         "Reply to ICMP Address Mask Request packets.");
 101
 102 SYSCTL_V_UINT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
 103         icmpmaskfake, 0, "Fake reply to ICMP Address Mask Request packets.");
 104
 105 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, drop_redirect,
 106         CTLFLAG_RW, drop_redirect, 0, "Ignore ICMP redirects");
 107
 108 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, log_redirect,
 109         CTLFLAG_RW, log_redirect, 0, "Log ICMP redirects to the console");
 110
 111 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, ICMPCTL_ICMPLIM, icmplim,
 112         CTLFLAG_RW, icmplim, 0, "Maximum number of ICMP responses per second");
 113
 114 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, icmplim_output,
 115         CTLFLAG_RW, icmplim_output, 0,
 116         "Enable rate limiting of ICMP responses");
 117
 118 SYSCTL_V_STRING(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, reply_src,
 119         CTLFLAG_RW, reply_src, IFNAMSIZ,
 120         "icmp reply source for non-local packets.");
 121
 122 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, reply_from_interface,
 123         CTLFLAG_RW, icmp_rfi, 0, "ICMP reply from incoming interface for "
 124         "non-local packets");
 125
 126 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
 127         icmp_quotelen, 0, "Number of bytes from original packet to "
 128         "quote in ICMP reply");
 129
 130 /*
 131  * ICMP broadcast echo sysctl
 132  */
 133
 134 SYSCTL_V_INT(V_NET, vnet_inet, _net_inet_icmp, OID_AUTO, bmcastecho,
 135         CTLFLAG_RW, icmpbmcastecho, 0, "");
 136
 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 *);
 144
 145 extern  struct protosw inetsw[];
 146
 147 void
 148 icmp_init(void)
 149 {
 150         INIT_VNET_INET(curvnet);
 151
 152         V_icmpmaskrepl = 0;
 153         V_icmpmaskfake = 0;
 154         V_drop_redirect = 0;
 155         V_log_redirect = 0;
 156         V_icmplim = 200;
 157         V_icmplim_output = 1;
 158         V_icmp_rfi = 0;
 159         V_icmp_quotelen = 8;
 160         V_icmpbmcastecho = 0;
 161 }
 162
 163 /*
 164  * Generate an error packet of type error
 165  * in response to bad packet ip.
 166  */
 167 void
 168 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
 169 {
 170         INIT_VNET_INET(curvnet);
 171         register struct ip *oip = mtod(n, struct ip *), *nip;
 172         register unsigned oiphlen = oip->ip_hl << 2;
 173         register struct icmp *icp;
 174         register struct mbuf *m;
 175         unsigned icmplen, icmpelen, nlen;
 176
 177         KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
 178 #ifdef ICMPPRINTFS
 179         if (icmpprintfs)
 180                 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
 181 #endif
 182         if (type != ICMP_REDIRECT)
 183                 V_icmpstat.icps_error++;
 184         /*
 185          * Don't send error:
 186          *  if the original packet was encrypted.
 187          *  if not the first fragment of message.
 188          *  in response to a multicast or broadcast packet.
 189          *  if the old packet protocol was an ICMP error message.
 190          */
 191         if (n->m_flags & M_DECRYPTED)
 192                 goto freeit;
 193         if (oip->ip_off & ~(IP_MF|IP_DF))
 194                 goto freeit;
 195         if (n->m_flags & (M_BCAST|M_MCAST))
 196                 goto freeit;
 197         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
 198           n->m_len >= oiphlen + ICMP_MINLEN &&
 199           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
 200                 V_icmpstat.icps_oldicmp++;
 201                 goto freeit;
 202         }
 203         /* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
 204         if (oiphlen + 8 > n->m_len)
 205                 goto freeit;
 206         /*
 207          * Calculate length to quote from original packet and
 208          * prevent the ICMP mbuf from overflowing.
 209          * Unfortunatly this is non-trivial since ip_forward()
 210          * sends us truncated packets.
 211          */
 212         nlen = m_length(n, NULL);
 213         if (oip->ip_p == IPPROTO_TCP) {
 214                 struct tcphdr *th;
 215                 int tcphlen;
 216
 217                 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
 218                     n->m_next == NULL)
 219                         goto stdreply;
 220                 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
 221                     ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
 222                         goto freeit;
 223                 th = (struct tcphdr *)((caddr_t)oip + oiphlen);
 224                 tcphlen = th->th_off << 2;
 225                 if (tcphlen < sizeof(struct tcphdr))
 226                         goto freeit;
 227                 if (oip->ip_len < oiphlen + tcphlen)
 228                         goto freeit;
 229                 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
 230                         goto stdreply;
 231                 if (n->m_len < oiphlen + tcphlen &&
 232                     ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
 233                         goto freeit;
 234                 icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
 235         } else
 236 stdreply:       icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
 237
 238         icmplen = min(oiphlen + icmpelen, nlen);
 239         if (icmplen < sizeof(struct ip))
 240                 goto freeit;
 241
 242         if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
 243                 m = m_gethdr(M_DONTWAIT, MT_DATA);
 244         else
 245                 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
 246         if (m == NULL)
 247                 goto freeit;
 248 #ifdef MAC
 249         mac_netinet_icmp_reply(n, m);
 250 #endif
 251         icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
 252         m_align(m, ICMP_MINLEN + icmplen);
 253         m->m_len = ICMP_MINLEN + icmplen;
 254
 255         /* XXX MRT  make the outgoing packet use the same FIB
 256          * that was associated with the incoming packet
 257          */
 258         M_SETFIB(m, M_GETFIB(n));
 259         icp = mtod(m, struct icmp *);
 260         V_icmpstat.icps_outhist[type]++;
 261         icp->icmp_type = type;
 262         if (type == ICMP_REDIRECT)
 263                 icp->icmp_gwaddr.s_addr = dest;
 264         else {
 265                 icp->icmp_void = 0;
 266                 /*
 267                  * The following assignments assume an overlay with the
 268                  * just zeroed icmp_void field.
 269                  */
 270                 if (type == ICMP_PARAMPROB) {
 271                         icp->icmp_pptr = code;
 272                         code = 0;
 273                 } else if (type == ICMP_UNREACH &&
 274                         code == ICMP_UNREACH_NEEDFRAG && mtu) {
 275                         icp->icmp_nextmtu = htons(mtu);
 276                 }
 277         }
 278         icp->icmp_code = code;
 279
 280         /*
 281          * Copy the quotation into ICMP message and
 282          * convert quoted IP header back to network representation.
 283          */
 284         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
 285         nip = &icp->icmp_ip;
 286         nip->ip_len = htons(nip->ip_len);
 287         nip->ip_off = htons(nip->ip_off);
 288
 289         /*
 290          * Set up ICMP message mbuf and copy old IP header (without options
 291          * in front of ICMP message.
 292          * If the original mbuf was meant to bypass the firewall, the error
 293          * reply should bypass as well.
 294          */
 295         m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
 296         m->m_data -= sizeof(struct ip);
 297         m->m_len += sizeof(struct ip);
 298         m->m_pkthdr.len = m->m_len;
 299         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
 300         nip = mtod(m, struct ip *);
 301         bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
 302         nip->ip_len = m->m_len;
 303         nip->ip_v = IPVERSION;
 304         nip->ip_hl = 5;
 305         nip->ip_p = IPPROTO_ICMP;
 306         nip->ip_tos = 0;
 307         icmp_reflect(m);
 308
 309 freeit:
 310         m_freem(n);
 311 }
 312
 313 /*
 314  * Process a received ICMP message.
 315  */
 316 void
 317 icmp_input(struct mbuf *m, int off)
 318 {
 319         INIT_VNET_INET(curvnet);
 320         struct icmp *icp;
 321         struct in_ifaddr *ia;
 322         struct ip *ip = mtod(m, struct ip *);
 323         struct sockaddr_in icmpsrc, icmpdst, icmpgw;
 324         int hlen = off;
 325         int icmplen = ip->ip_len;
 326         int i, code;
 327         void (*ctlfunc)(int, struct sockaddr *, void *);
 328         int fibnum;
 329
 330         /*
 331          * Locate icmp structure in mbuf, and check
 332          * that not corrupted and of at least minimum length.
 333          */
 334 #ifdef ICMPPRINTFS
 335         if (icmpprintfs) {
 336                 char buf[4 * sizeof "123"];
 337                 strcpy(buf, inet_ntoa(ip->ip_src));
 338                 printf("icmp_input from %s to %s, len %d\n",
 339                        buf, inet_ntoa(ip->ip_dst), icmplen);
 340         }
 341 #endif
 342         if (icmplen < ICMP_MINLEN) {
 343                 V_icmpstat.icps_tooshort++;
 344                 goto freeit;
 345         }
 346         i = hlen + min(icmplen, ICMP_ADVLENMIN);
 347         if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
 348                 V_icmpstat.icps_tooshort++;
 349                 return;
 350         }
 351         ip = mtod(m, struct ip *);
 352         m->m_len -= hlen;
 353         m->m_data += hlen;
 354         icp = mtod(m, struct icmp *);
 355         if (in_cksum(m, icmplen)) {
 356                 V_icmpstat.icps_checksum++;
 357                 goto freeit;
 358         }
 359         m->m_len += hlen;
 360         m->m_data -= hlen;
 361
 362         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
 363                 /*
 364                  * Deliver very specific ICMP type only.
 365                  */
 366                 switch (icp->icmp_type) {
 367                 case ICMP_UNREACH:
 368                 case ICMP_TIMXCEED:
 369                         break;
 370                 default:
 371                         goto freeit;
 372                 }
 373         }
 374
 375 #ifdef ICMPPRINTFS
 376         if (icmpprintfs)
 377                 printf("icmp_input, type %d code %d\n", icp->icmp_type,
 378                     icp->icmp_code);
 379 #endif
 380
 381         /*
 382          * Message type specific processing.
 383          */
 384         if (icp->icmp_type > ICMP_MAXTYPE)
 385                 goto raw;
 386
 387         /* Initialize */
 388         bzero(&icmpsrc, sizeof(icmpsrc));
 389         icmpsrc.sin_len = sizeof(struct sockaddr_in);
 390         icmpsrc.sin_family = AF_INET;
 391         bzero(&icmpdst, sizeof(icmpdst));
 392         icmpdst.sin_len = sizeof(struct sockaddr_in);
 393         icmpdst.sin_family = AF_INET;
 394         bzero(&icmpgw, sizeof(icmpgw));
 395         icmpgw.sin_len = sizeof(struct sockaddr_in);
 396         icmpgw.sin_family = AF_INET;
 397
 398         V_icmpstat.icps_inhist[icp->icmp_type]++;
 399         code = icp->icmp_code;
 400         switch (icp->icmp_type) {
 401
 402         case ICMP_UNREACH:
 403                 switch (code) {
 404                         case ICMP_UNREACH_NET:
 405                         case ICMP_UNREACH_HOST:
 406                         case ICMP_UNREACH_SRCFAIL:
 407                         case ICMP_UNREACH_NET_UNKNOWN:
 408                         case ICMP_UNREACH_HOST_UNKNOWN:
 409                         case ICMP_UNREACH_ISOLATED:
 410                         case ICMP_UNREACH_TOSNET:
 411                         case ICMP_UNREACH_TOSHOST:
 412                         case ICMP_UNREACH_HOST_PRECEDENCE:
 413                         case ICMP_UNREACH_PRECEDENCE_CUTOFF:
 414                                 code = PRC_UNREACH_NET;
 415                                 break;
 416
 417                         case ICMP_UNREACH_NEEDFRAG:
 418                                 code = PRC_MSGSIZE;
 419                                 break;
 420
 421                         /*
 422                          * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
 423                          * Treat subcodes 2,3 as immediate RST
 424                          */
 425                         case ICMP_UNREACH_PROTOCOL:
 426                         case ICMP_UNREACH_PORT:
 427                                 code = PRC_UNREACH_PORT;
 428                                 break;
 429
 430                         case ICMP_UNREACH_NET_PROHIB:
 431                         case ICMP_UNREACH_HOST_PROHIB:
 432                         case ICMP_UNREACH_FILTER_PROHIB:
 433                                 code = PRC_UNREACH_ADMIN_PROHIB;
 434                                 break;
 435
 436                         default:
 437                                 goto badcode;
 438                 }
 439                 goto deliver;
 440
 441         case ICMP_TIMXCEED:
 442                 if (code > 1)
 443                         goto badcode;
 444                 code += PRC_TIMXCEED_INTRANS;
 445                 goto deliver;
 446
 447         case ICMP_PARAMPROB:
 448                 if (code > 1)
 449                         goto badcode;
 450                 code = PRC_PARAMPROB;
 451                 goto deliver;
 452
 453         case ICMP_SOURCEQUENCH:
 454                 if (code)
 455                         goto badcode;
 456                 code = PRC_QUENCH;
 457         deliver:
 458                 /*
 459                  * Problem with datagram; advise higher level routines.
 460                  */
 461                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 462                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
 463                         V_icmpstat.icps_badlen++;
 464                         goto freeit;
 465                 }
 466                 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
 467                 /* Discard ICMP's in response to multicast packets */
 468                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
 469                         goto badcode;
 470 #ifdef ICMPPRINTFS
 471                 if (icmpprintfs)
 472                         printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
 473 #endif
 474                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 475                 /*
 476                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
 477                  * notification to TCP layer.
 478                  */
 479                 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
 480                 if (ctlfunc)
 481                         (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
 482                                    (void *)&icp->icmp_ip);
 483                 break;
 484
 485         badcode:
 486                 V_icmpstat.icps_badcode++;
 487                 break;
 488
 489         case ICMP_ECHO:
 490                 if (!V_icmpbmcastecho
 491                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 492                         V_icmpstat.icps_bmcastecho++;
 493                         break;
 494                 }
 495                 icp->icmp_type = ICMP_ECHOREPLY;
 496                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 497                         goto freeit;
 498                 else
 499                         goto reflect;
 500
 501         case ICMP_TSTAMP:
 502                 if (!V_icmpbmcastecho
 503                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 504                         V_icmpstat.icps_bmcasttstamp++;
 505                         break;
 506                 }
 507                 if (icmplen < ICMP_TSLEN) {
 508                         V_icmpstat.icps_badlen++;
 509                         break;
 510                 }
 511                 icp->icmp_type = ICMP_TSTAMPREPLY;
 512                 icp->icmp_rtime = iptime();
 513                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 514                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 515                         goto freeit;
 516                 else
 517                         goto reflect;
 518
 519         case ICMP_MASKREQ:
 520                 if (V_icmpmaskrepl == 0)
 521                         break;
 522                 /*
 523                  * We are not able to respond with all ones broadcast
 524                  * unless we receive it over a point-to-point interface.
 525                  */
 526                 if (icmplen < ICMP_MASKLEN)
 527                         break;
 528                 switch (ip->ip_dst.s_addr) {
 529
 530                 case INADDR_BROADCAST:
 531                 case INADDR_ANY:
 532                         icmpdst.sin_addr = ip->ip_src;
 533                         break;
 534
 535                 default:
 536                         icmpdst.sin_addr = ip->ip_dst;
 537                 }
 538                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 539                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 540                 if (ia == 0)
 541                         break;
 542                 if (ia->ia_ifp == 0)
 543                         break;
 544                 icp->icmp_type = ICMP_MASKREPLY;
 545                 if (V_icmpmaskfake == 0)
 546                         icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 547                 else
 548                         icp->icmp_mask = V_icmpmaskfake;
 549                 if (ip->ip_src.s_addr == 0) {
 550                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 551                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 552                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 553                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 554                 }
 555 reflect:
 556                 ip->ip_len += hlen;     /* since ip_input deducts this */
 557                 V_icmpstat.icps_reflect++;
 558                 V_icmpstat.icps_outhist[icp->icmp_type]++;
 559                 icmp_reflect(m);
 560                 return;
 561
 562         case ICMP_REDIRECT:
 563                 if (V_log_redirect) {
 564                         u_long src, dst, gw;
 565
 566                         src = ntohl(ip->ip_src.s_addr);
 567                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
 568                         gw = ntohl(icp->icmp_gwaddr.s_addr);
 569                         printf("icmp redirect from %d.%d.%d.%d: "
 570                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
 571                                (int)(src >> 24), (int)((src >> 16) & 0xff),
 572                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
 573                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
 574                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
 575                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
 576                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
 577                 }
 578                 /*
 579                  * RFC1812 says we must ignore ICMP redirects if we
 580                  * are acting as router.
 581                  */
 582                 if (V_drop_redirect || V_ipforwarding)
 583                         break;
 584                 if (code > 3)
 585                         goto badcode;
 586                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 587                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
 588                         V_icmpstat.icps_badlen++;
 589                         break;
 590                 }
 591                 /*
 592                  * Short circuit routing redirects to force
 593                  * immediate change in the kernel's routing
 594                  * tables.  The message is also handed to anyone
 595                  * listening on a raw socket (e.g. the routing
 596                  * daemon for use in updating its tables).
 597                  */
 598                 icmpgw.sin_addr = ip->ip_src;
 599                 icmpdst.sin_addr = icp->icmp_gwaddr;
 600 #ifdef  ICMPPRINTFS
 601                 if (icmpprintfs) {
 602                         char buf[4 * sizeof "123"];
 603                         strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
 604
 605                         printf("redirect dst %s to %s\n",
 606                                buf, inet_ntoa(icp->icmp_gwaddr));
 607                 }
 608 #endif
 609                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 610                 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
 611                         in_rtredirect((struct sockaddr *)&icmpsrc,
 612                           (struct sockaddr *)&icmpdst,
 613                           (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
 614                           (struct sockaddr *)&icmpgw, fibnum);
 615                 }
 616                 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 617 #ifdef IPSEC
 618                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 619 #endif
 620                 break;
 621
 622         /*
 623          * No kernel processing for the following;
 624          * just fall through to send to raw listener.
 625          */
 626         case ICMP_ECHOREPLY:
 627         case ICMP_ROUTERADVERT:
 628         case ICMP_ROUTERSOLICIT:
 629         case ICMP_TSTAMPREPLY:
 630         case ICMP_IREQREPLY:
 631         case ICMP_MASKREPLY:
 632         default:
 633                 break;
 634         }
 635
 636 raw:
 637         rip_input(m, off);
 638         return;
 639
 640 freeit:
 641         m_freem(m);
 642 }
 643
 644 /*
 645  * Reflect the ip packet back to the source
 646  */
 647 static void
 648 icmp_reflect(struct mbuf *m)
 649 {
 650         INIT_VNET_INET(curvnet);
 651         struct ip *ip = mtod(m, struct ip *);
 652         struct ifaddr *ifa;
 653         struct ifnet *ifn;
 654         struct in_ifaddr *ia;
 655         struct in_addr t;
 656         struct mbuf *opts = 0;
 657         int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
 658
 659         if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
 660             IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
 661             IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
 662                 m_freem(m);     /* Bad return address */
 663                 V_icmpstat.icps_badaddr++;
 664                 goto done;      /* Ip_output() will check for broadcast */
 665         }
 666
 667         t = ip->ip_dst;
 668         ip->ip_dst = ip->ip_src;
 669
 670         /*
 671          * Source selection for ICMP replies:
 672          *
 673          * If the incoming packet was addressed directly to one of our
 674          * own addresses, use dst as the src for the reply.
 675          */
 676         LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash)
 677                 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr)
 678                         goto match;
 679         /*
 680          * If the incoming packet was addressed to one of our broadcast
 681          * addresses, use the first non-broadcast address which corresponds
 682          * to the incoming interface.
 683          */
 684         if (m->m_pkthdr.rcvif != NULL &&
 685             m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
 686                 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
 687                         if (ifa->ifa_addr->sa_family != AF_INET)
 688                                 continue;
 689                         ia = ifatoia(ifa);
 690                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 691                             t.s_addr)
 692                                 goto match;
 693                 }
 694         }
 695         /*
 696          * If the packet was transiting through us, use the address of
 697          * the interface the packet came through in.  If that interface
 698          * doesn't have a suitable IP address, the normal selection
 699          * criteria apply.
 700          */
 701         if (V_icmp_rfi && m->m_pkthdr.rcvif != NULL) {
 702                 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
 703                         if (ifa->ifa_addr->sa_family != AF_INET)
 704                                 continue;
 705                         ia = ifatoia(ifa);
 706                         goto match;
 707                 }
 708         }
 709         /*
 710          * If the incoming packet was not addressed directly to us, use
 711          * designated interface for icmp replies specified by sysctl
 712          * net.inet.icmp.reply_src (default not set). Otherwise continue
 713          * with normal source selection.
 714          */
 715         if (V_reply_src[0] != '\0' && (ifn = ifunit(V_reply_src))) {
 716                 TAILQ_FOREACH(ifa, &ifn->if_addrhead, ifa_link) {
 717                         if (ifa->ifa_addr->sa_family != AF_INET)
 718                                 continue;
 719                         ia = ifatoia(ifa);
 720                         goto match;
 721                 }
 722         }
 723         /*
 724          * If the packet was transiting through us, use the address of
 725          * the interface that is the closest to the packet source.
 726          * When we don't have a route back to the packet source, stop here
 727          * and drop the packet.
 728          */
 729         ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
 730         if (ia == NULL) {
 731                 m_freem(m);
 732                 V_icmpstat.icps_noroute++;
 733                 goto done;
 734         }
 735 match:
 736 #ifdef MAC
 737         mac_netinet_icmp_replyinplace(m);
 738 #endif
 739         t = IA_SIN(ia)->sin_addr;
 740         ip->ip_src = t;
 741         ip->ip_ttl = V_ip_defttl;
 742
 743         if (optlen > 0) {
 744                 register u_char *cp;
 745                 int opt, cnt;
 746                 u_int len;
 747
 748                 /*
 749                  * Retrieve any source routing from the incoming packet;
 750                  * add on any record-route or timestamp options.
 751                  */
 752                 cp = (u_char *) (ip + 1);
 753                 if ((opts = ip_srcroute(m)) == 0 &&
 754                     (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
 755                         opts->m_len = sizeof(struct in_addr);
 756                         mtod(opts, struct in_addr *)->s_addr = 0;
 757                 }
 758                 if (opts) {
 759 #ifdef ICMPPRINTFS
 760                     if (icmpprintfs)
 761                             printf("icmp_reflect optlen %d rt %d => ",
 762                                 optlen, opts->m_len);
 763 #endif
 764                     for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
 765                             opt = cp[IPOPT_OPTVAL];
 766                             if (opt == IPOPT_EOL)
 767                                     break;
 768                             if (opt == IPOPT_NOP)
 769                                     len = 1;
 770                             else {
 771                                     if (cnt < IPOPT_OLEN + sizeof(*cp))
 772                                             break;
 773                                     len = cp[IPOPT_OLEN];
 774                                     if (len < IPOPT_OLEN + sizeof(*cp) ||
 775                                         len > cnt)
 776                                             break;
 777                             }
 778                             /*
 779                              * Should check for overflow, but it "can't happen"
 780                              */
 781                             if (opt == IPOPT_RR || opt == IPOPT_TS ||
 782                                 opt == IPOPT_SECURITY) {
 783                                     bcopy((caddr_t)cp,
 784                                         mtod(opts, caddr_t) + opts->m_len, len);
 785                                     opts->m_len += len;
 786                             }
 787                     }
 788                     /* Terminate & pad, if necessary */
 789                     cnt = opts->m_len % 4;
 790                     if (cnt) {
 791                             for (; cnt < 4; cnt++) {
 792                                     *(mtod(opts, caddr_t) + opts->m_len) =
 793                                         IPOPT_EOL;
 794                                     opts->m_len++;
 795                             }
 796                     }
 797 #ifdef ICMPPRINTFS
 798                     if (icmpprintfs)
 799                             printf("%d\n", opts->m_len);
 800 #endif
 801                 }
 802                 /*
 803                  * Now strip out original options by copying rest of first
 804                  * mbuf's data back, and adjust the IP length.
 805                  */
 806                 ip->ip_len -= optlen;
 807                 ip->ip_v = IPVERSION;
 808                 ip->ip_hl = 5;
 809                 m->m_len -= optlen;
 810                 if (m->m_flags & M_PKTHDR)
 811                         m->m_pkthdr.len -= optlen;
 812                 optlen += sizeof(struct ip);
 813                 bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
 814                          (unsigned)(m->m_len - sizeof(struct ip)));
 815         }
 816         m_tag_delete_nonpersistent(m);
 817         m->m_flags &= ~(M_BCAST|M_MCAST);
 818         icmp_send(m, opts);
 819 done:
 820         if (opts)
 821                 (void)m_free(opts);
 822 }
 823
 824 /*
 825  * Send an icmp packet back to the ip level,
 826  * after supplying a checksum.
 827  */
 828 static void
 829 icmp_send(struct mbuf *m, struct mbuf *opts)
 830 {
 831         register struct ip *ip = mtod(m, struct ip *);
 832         register int hlen;
 833         register struct icmp *icp;
 834
 835         hlen = ip->ip_hl << 2;
 836         m->m_data += hlen;
 837         m->m_len -= hlen;
 838         icp = mtod(m, struct icmp *);
 839         icp->icmp_cksum = 0;
 840         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
 841         m->m_data -= hlen;
 842         m->m_len += hlen;
 843         m->m_pkthdr.rcvif = (struct ifnet *)0;
 844 #ifdef ICMPPRINTFS
 845         if (icmpprintfs) {
 846                 char buf[4 * sizeof "123"];
 847                 strcpy(buf, inet_ntoa(ip->ip_dst));
 848                 printf("icmp_send dst %s src %s\n",
 849                        buf, inet_ntoa(ip->ip_src));
 850         }
 851 #endif
 852         (void) ip_output(m, opts, NULL, 0, NULL, NULL);
 853 }
 854
 855 /*
 856  * Return milliseconds since 00:00 GMT in network format.
 857  */
 858 uint32_t
 859 iptime(void)
 860 {
 861         struct timeval atv;
 862         u_long t;
 863
 864         getmicrotime(&atv);
 865         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 866         return (htonl(t));
 867 }
 868
 869 /*
 870  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 871  * given current value MTU.  If DIR is less than zero, a larger plateau
 872  * is returned; otherwise, a smaller value is returned.
 873  */
 874 int
 875 ip_next_mtu(int mtu, int dir)
 876 {
 877         static int mtutab[] = {
 878                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
 879                 296, 68, 0
 880         };
 881         int i, size;
 882
 883         size = (sizeof mtutab) / (sizeof mtutab[0]);
 884         if (dir >= 0) {
 885                 for (i = 0; i < size; i++)
 886                         if (mtu > mtutab[i])
 887                                 return mtutab[i];
 888         } else {
 889                 for (i = size - 1; i >= 0; i--)
 890                         if (mtu < mtutab[i])
 891                                 return mtutab[i];
 892                 if (mtu == mtutab[0])
 893                         return mtutab[0];
 894         }
 895         return 0;
 896 }
 897
 898
 899 /*
 900  * badport_bandlim() - check for ICMP bandwidth limit
 901  *
 902  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 903  *      hit our bandwidth limit and it is not ok.
 904  *
 905  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 906  *
 907  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 908  *      values.  We may eventually remove this separation (and simplify the
 909  *      code further).
 910  *
 911  *      Note that the printing of the error message is delayed so we can
 912  *      properly print the icmp error rate that the system was trying to do
 913  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 914  *      the 'final' error, but it doesn't make sense to solve the printing
 915  *      delay with more complex code.
 916  */
 917
 918 int
 919 badport_bandlim(int which)
 920 {
 921         INIT_VNET_INET(curvnet);
 922
 923 #define N(a)    (sizeof (a) / sizeof (a[0]))
 924         static struct rate {
 925                 const char      *type;
 926                 struct timeval  lasttime;
 927                 int             curpps;
 928         } rates[BANDLIM_MAX+1] = {
 929                 { "icmp unreach response" },
 930                 { "icmp ping response" },
 931                 { "icmp tstamp response" },
 932                 { "closed port RST response" },
 933                 { "open port RST response" },
 934                 { "icmp6 unreach response" }
 935         };
 936
 937         /*
 938          * Return ok status if feature disabled or argument out of range.
 939          */
 940         if (V_icmplim > 0 && (u_int) which < N(rates)) {
 941                 struct rate *r = &rates[which];
 942                 int opps = r->curpps;
 943
 944                 if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
 945                         return -1;      /* discard packet */
 946                 /*
 947                  * If we've dropped below the threshold after having
 948                  * rate-limited traffic print the message.  This preserves
 949                  * the previous behaviour at the expense of added complexity.
 950                  */
 951                 if (V_icmplim_output && opps > V_icmplim)
 952                         printf("Limiting %s from %d to %d packets/sec\n",
 953                                 r->type, opps, V_icmplim);
 954         }
 955         return 0;                       /* okay to send packet */
 956 #undef N
 957 }