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
  37 #include <sys/param.h>
  38 #include <sys/systm.h>
  39 #include <sys/mbuf.h>
  40 #include <sys/protosw.h>
  41 #include <sys/socket.h>
  42 #include <sys/time.h>
  43 #include <sys/kernel.h>
  44 #include <sys/sysctl.h>
  45
  46 #include <net/if.h>
  47 #include <net/if_types.h>
  48 #include <net/route.h>
  49 #include <net/vnet.h>
  50
  51 #include <netinet/in.h>
  52 #include <netinet/in_pcb.h>
  53 #include <netinet/in_systm.h>
  54 #include <netinet/in_var.h>
  55 #include <netinet/ip.h>
  56 #include <netinet/ip_icmp.h>
  57 #include <netinet/ip_var.h>
  58 #include <netinet/ip_options.h>
  59 #include <netinet/tcp.h>
  60 #include <netinet/tcp_var.h>
  61 #include <netinet/tcpip.h>
  62 #include <netinet/icmp_var.h>
  63
  64 #ifdef IPSEC
  65 #include <netipsec/ipsec.h>
  66 #include <netipsec/key.h>
  67 #endif
  68
  69 #include <machine/in_cksum.h>
  70
  71 #include <security/mac/mac_framework.h>
  72
  73 /*
  74  * ICMP routines: error generation, receive packet processing, and
  75  * routines to turnaround packets back to the originator, and
  76  * host table maintenance routines.
  77  */
  78
  79 VNET_DEFINE(struct icmpstat, icmpstat);
  80 static VNET_DEFINE(int, icmpmaskrepl);
  81 static VNET_DEFINE(u_int, icmpmaskfake);
  82 static VNET_DEFINE(int, drop_redirect);
  83 static VNET_DEFINE(int, log_redirect);
  84 static VNET_DEFINE(int, icmplim);
  85 static VNET_DEFINE(int, icmplim_output);
  86 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
  87 static VNET_DEFINE(int, icmp_rfi);
  88 static VNET_DEFINE(int, icmp_quotelen);
  89 static VNET_DEFINE(int, icmpbmcastecho);
  90
  91 #define V_icmpmaskrepl                  VNET(icmpmaskrepl)
  92 #define V_icmpmaskfake                  VNET(icmpmaskfake)
  93 #define V_drop_redirect                 VNET(drop_redirect)
  94 #define V_log_redirect                  VNET(log_redirect)
  95 #define V_icmplim                       VNET(icmplim)
  96 #define V_icmplim_output                VNET(icmplim_output)
  97 #define V_reply_src                     VNET(reply_src)
  98 #define V_icmp_rfi                      VNET(icmp_rfi)
  99 #define V_icmp_quotelen                 VNET(icmp_quotelen)
 100 #define V_icmpbmcastecho                VNET(icmpbmcastecho)
 101
 102 SYSCTL_VNET_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
 103         &VNET_NAME(icmpstat), icmpstat, "");
 104
 105 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
 106         &VNET_NAME(icmpmaskrepl), 0,
 107         "Reply to ICMP Address Mask Request packets.");
 108
 109 SYSCTL_VNET_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
 110         &VNET_NAME(icmpmaskfake), 0,
 111         "Fake reply to ICMP Address Mask Request packets.");
 112
 113 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
 114         &VNET_NAME(drop_redirect), 0,
 115         "Ignore ICMP redirects");
 116
 117 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
 118         &VNET_NAME(log_redirect), 0,
 119         "Log ICMP redirects to the console");
 120
 121 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
 122         &VNET_NAME(icmplim), 0,
 123         "Maximum number of ICMP responses per second");
 124
 125 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
 126         &VNET_NAME(icmplim_output), 0,
 127         "Enable rate limiting of ICMP responses");
 128
 129 SYSCTL_VNET_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
 130         &VNET_NAME(reply_src), IFNAMSIZ,
 131         "icmp reply source for non-local packets.");
 132
 133 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
 134         &VNET_NAME(icmp_rfi), 0,
 135         "ICMP reply from incoming interface for non-local packets");
 136
 137 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
 138         &VNET_NAME(icmp_quotelen), 0,
 139         "Number of bytes from original packet to quote in ICMP reply");
 140
 141 /*
 142  * ICMP broadcast echo sysctl
 143  */
 144
 145 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
 146         &VNET_NAME(icmpbmcastecho), 0,
 147         "");
 148
 149
 150 #ifdef ICMPPRINTFS
 151 int     icmpprintfs = 0;
 152 #endif
 153
 154 static void     icmp_reflect(struct mbuf *);
 155 static void     icmp_send(struct mbuf *, struct mbuf *);
 156
 157 extern  struct protosw inetsw[];
 158
 159 void
 160 icmp_init(void)
 161 {
 162
 163         V_icmpmaskrepl = 0;
 164         V_icmpmaskfake = 0;
 165         V_drop_redirect = 0;
 166         V_log_redirect = 0;
 167         V_icmplim = 200;
 168         V_icmplim_output = 1;
 169         V_icmp_rfi = 0;
 170         V_icmp_quotelen = 8;
 171         V_icmpbmcastecho = 0;
 172 }
 173
 174 /*
 175  * Kernel module interface for updating icmpstat.  The argument is an index
 176  * into icmpstat treated as an array of u_long.  While this encodes the
 177  * general layout of icmpstat into the caller, it doesn't encode its
 178  * location, so that future changes to add, for example, per-CPU stats
 179  * support won't cause binary compatibility problems for kernel modules.
 180  */
 181 void
 182 kmod_icmpstat_inc(int statnum)
 183 {
 184
 185         (*((u_long *)&V_icmpstat + statnum))++;
 186 }
 187
 188 /*
 189  * Generate an error packet of type error
 190  * in response to bad packet ip.
 191  */
 192 void
 193 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
 194 {
 195         register struct ip *oip = mtod(n, struct ip *), *nip;
 196         register unsigned oiphlen = oip->ip_hl << 2;
 197         register struct icmp *icp;
 198         register struct mbuf *m;
 199         unsigned icmplen, icmpelen, nlen;
 200
 201         KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
 202 #ifdef ICMPPRINTFS
 203         if (icmpprintfs)
 204                 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
 205 #endif
 206         if (type != ICMP_REDIRECT)
 207                 ICMPSTAT_INC(icps_error);
 208         /*
 209          * Don't send error:
 210          *  if the original packet was encrypted.
 211          *  if not the first fragment of message.
 212          *  in response to a multicast or broadcast packet.
 213          *  if the old packet protocol was an ICMP error message.
 214          */
 215         if (n->m_flags & M_DECRYPTED)
 216                 goto freeit;
 217         if (oip->ip_off & ~(IP_MF|IP_DF))
 218                 goto freeit;
 219         if (n->m_flags & (M_BCAST|M_MCAST))
 220                 goto freeit;
 221         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
 222           n->m_len >= oiphlen + ICMP_MINLEN &&
 223           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
 224                 ICMPSTAT_INC(icps_oldicmp);
 225                 goto freeit;
 226         }
 227         /* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
 228         if (oiphlen + 8 > n->m_len)
 229                 goto freeit;
 230         /*
 231          * Calculate length to quote from original packet and
 232          * prevent the ICMP mbuf from overflowing.
 233          * Unfortunatly this is non-trivial since ip_forward()
 234          * sends us truncated packets.
 235          */
 236         nlen = m_length(n, NULL);
 237         if (oip->ip_p == IPPROTO_TCP) {
 238                 struct tcphdr *th;
 239                 int tcphlen;
 240
 241                 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
 242                     n->m_next == NULL)
 243                         goto stdreply;
 244                 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
 245                     ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
 246                         goto freeit;
 247                 th = (struct tcphdr *)((caddr_t)oip + oiphlen);
 248                 tcphlen = th->th_off << 2;
 249                 if (tcphlen < sizeof(struct tcphdr))
 250                         goto freeit;
 251                 if (oip->ip_len < oiphlen + tcphlen)
 252                         goto freeit;
 253                 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
 254                         goto stdreply;
 255                 if (n->m_len < oiphlen + tcphlen &&
 256                     ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
 257                         goto freeit;
 258                 icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
 259         } else
 260 stdreply:       icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
 261
 262         icmplen = min(oiphlen + icmpelen, nlen);
 263         if (icmplen < sizeof(struct ip))
 264                 goto freeit;
 265
 266         if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
 267                 m = m_gethdr(M_DONTWAIT, MT_DATA);
 268         else
 269                 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
 270         if (m == NULL)
 271                 goto freeit;
 272 #ifdef MAC
 273         mac_netinet_icmp_reply(n, m);
 274 #endif
 275         icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
 276         m_align(m, ICMP_MINLEN + icmplen);
 277         m->m_len = ICMP_MINLEN + icmplen;
 278
 279         /* XXX MRT  make the outgoing packet use the same FIB
 280          * that was associated with the incoming packet
 281          */
 282         M_SETFIB(m, M_GETFIB(n));
 283         icp = mtod(m, struct icmp *);
 284         ICMPSTAT_INC(icps_outhist[type]);
 285         icp->icmp_type = type;
 286         if (type == ICMP_REDIRECT)
 287                 icp->icmp_gwaddr.s_addr = dest;
 288         else {
 289                 icp->icmp_void = 0;
 290                 /*
 291                  * The following assignments assume an overlay with the
 292                  * just zeroed icmp_void field.
 293                  */
 294                 if (type == ICMP_PARAMPROB) {
 295                         icp->icmp_pptr = code;
 296                         code = 0;
 297                 } else if (type == ICMP_UNREACH &&
 298                         code == ICMP_UNREACH_NEEDFRAG && mtu) {
 299                         icp->icmp_nextmtu = htons(mtu);
 300                 }
 301         }
 302         icp->icmp_code = code;
 303
 304         /*
 305          * Copy the quotation into ICMP message and
 306          * convert quoted IP header back to network representation.
 307          */
 308         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
 309         nip = &icp->icmp_ip;
 310         nip->ip_len = htons(nip->ip_len);
 311         nip->ip_off = htons(nip->ip_off);
 312
 313         /*
 314          * Set up ICMP message mbuf and copy old IP header (without options
 315          * in front of ICMP message.
 316          * If the original mbuf was meant to bypass the firewall, the error
 317          * reply should bypass as well.
 318          */
 319         m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
 320         m->m_data -= sizeof(struct ip);
 321         m->m_len += sizeof(struct ip);
 322         m->m_pkthdr.len = m->m_len;
 323         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
 324         nip = mtod(m, struct ip *);
 325         bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
 326         nip->ip_len = m->m_len;
 327         nip->ip_v = IPVERSION;
 328         nip->ip_hl = 5;
 329         nip->ip_p = IPPROTO_ICMP;
 330         nip->ip_tos = 0;
 331         icmp_reflect(m);
 332
 333 freeit:
 334         m_freem(n);
 335 }
 336
 337 /*
 338  * Process a received ICMP message.
 339  */
 340 void
 341 icmp_input(struct mbuf *m, int off)
 342 {
 343         struct icmp *icp;
 344         struct in_ifaddr *ia;
 345         struct ip *ip = mtod(m, struct ip *);
 346         struct sockaddr_in icmpsrc, icmpdst, icmpgw;
 347         int hlen = off;
 348         int icmplen = ip->ip_len;
 349         int i, code;
 350         void (*ctlfunc)(int, struct sockaddr *, void *);
 351         int fibnum;
 352
 353         /*
 354          * Locate icmp structure in mbuf, and check
 355          * that not corrupted and of at least minimum length.
 356          */
 357 #ifdef ICMPPRINTFS
 358         if (icmpprintfs) {
 359                 char buf[4 * sizeof "123"];
 360                 strcpy(buf, inet_ntoa(ip->ip_src));
 361                 printf("icmp_input from %s to %s, len %d\n",
 362                        buf, inet_ntoa(ip->ip_dst), icmplen);
 363         }
 364 #endif
 365         if (icmplen < ICMP_MINLEN) {
 366                 ICMPSTAT_INC(icps_tooshort);
 367                 goto freeit;
 368         }
 369         i = hlen + min(icmplen, ICMP_ADVLENMIN);
 370         if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
 371                 ICMPSTAT_INC(icps_tooshort);
 372                 return;
 373         }
 374         ip = mtod(m, struct ip *);
 375         m->m_len -= hlen;
 376         m->m_data += hlen;
 377         icp = mtod(m, struct icmp *);
 378         if (in_cksum(m, icmplen)) {
 379                 ICMPSTAT_INC(icps_checksum);
 380                 goto freeit;
 381         }
 382         m->m_len += hlen;
 383         m->m_data -= hlen;
 384
 385         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
 386                 /*
 387                  * Deliver very specific ICMP type only.
 388                  */
 389                 switch (icp->icmp_type) {
 390                 case ICMP_UNREACH:
 391                 case ICMP_TIMXCEED:
 392                         break;
 393                 default:
 394                         goto freeit;
 395                 }
 396         }
 397
 398 #ifdef ICMPPRINTFS
 399         if (icmpprintfs)
 400                 printf("icmp_input, type %d code %d\n", icp->icmp_type,
 401                     icp->icmp_code);
 402 #endif
 403
 404         /*
 405          * Message type specific processing.
 406          */
 407         if (icp->icmp_type > ICMP_MAXTYPE)
 408                 goto raw;
 409
 410         /* Initialize */
 411         bzero(&icmpsrc, sizeof(icmpsrc));
 412         icmpsrc.sin_len = sizeof(struct sockaddr_in);
 413         icmpsrc.sin_family = AF_INET;
 414         bzero(&icmpdst, sizeof(icmpdst));
 415         icmpdst.sin_len = sizeof(struct sockaddr_in);
 416         icmpdst.sin_family = AF_INET;
 417         bzero(&icmpgw, sizeof(icmpgw));
 418         icmpgw.sin_len = sizeof(struct sockaddr_in);
 419         icmpgw.sin_family = AF_INET;
 420
 421         ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
 422         code = icp->icmp_code;
 423         switch (icp->icmp_type) {
 424
 425         case ICMP_UNREACH:
 426                 switch (code) {
 427                         case ICMP_UNREACH_NET:
 428                         case ICMP_UNREACH_HOST:
 429                         case ICMP_UNREACH_SRCFAIL:
 430                         case ICMP_UNREACH_NET_UNKNOWN:
 431                         case ICMP_UNREACH_HOST_UNKNOWN:
 432                         case ICMP_UNREACH_ISOLATED:
 433                         case ICMP_UNREACH_TOSNET:
 434                         case ICMP_UNREACH_TOSHOST:
 435                         case ICMP_UNREACH_HOST_PRECEDENCE:
 436                         case ICMP_UNREACH_PRECEDENCE_CUTOFF:
 437                                 code = PRC_UNREACH_NET;
 438                                 break;
 439
 440                         case ICMP_UNREACH_NEEDFRAG:
 441                                 code = PRC_MSGSIZE;
 442                                 break;
 443
 444                         /*
 445                          * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
 446                          * Treat subcodes 2,3 as immediate RST
 447                          */
 448                         case ICMP_UNREACH_PROTOCOL:
 449                         case ICMP_UNREACH_PORT:
 450                                 code = PRC_UNREACH_PORT;
 451                                 break;
 452
 453                         case ICMP_UNREACH_NET_PROHIB:
 454                         case ICMP_UNREACH_HOST_PROHIB:
 455                         case ICMP_UNREACH_FILTER_PROHIB:
 456                                 code = PRC_UNREACH_ADMIN_PROHIB;
 457                                 break;
 458
 459                         default:
 460                                 goto badcode;
 461                 }
 462                 goto deliver;
 463
 464         case ICMP_TIMXCEED:
 465                 if (code > 1)
 466                         goto badcode;
 467                 code += PRC_TIMXCEED_INTRANS;
 468                 goto deliver;
 469
 470         case ICMP_PARAMPROB:
 471                 if (code > 1)
 472                         goto badcode;
 473                 code = PRC_PARAMPROB;
 474                 goto deliver;
 475
 476         case ICMP_SOURCEQUENCH:
 477                 if (code)
 478                         goto badcode;
 479                 code = PRC_QUENCH;
 480         deliver:
 481                 /*
 482                  * Problem with datagram; advise higher level routines.
 483                  */
 484                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 485                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
 486                         ICMPSTAT_INC(icps_badlen);
 487                         goto freeit;
 488                 }
 489                 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
 490                 /* Discard ICMP's in response to multicast packets */
 491                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
 492                         goto badcode;
 493 #ifdef ICMPPRINTFS
 494                 if (icmpprintfs)
 495                         printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
 496 #endif
 497                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 498                 /*
 499                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
 500                  * notification to TCP layer.
 501                  */
 502                 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
 503                 if (ctlfunc)
 504                         (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
 505                                    (void *)&icp->icmp_ip);
 506                 break;
 507
 508         badcode:
 509                 ICMPSTAT_INC(icps_badcode);
 510                 break;
 511
 512         case ICMP_ECHO:
 513                 if (!V_icmpbmcastecho
 514                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 515                         ICMPSTAT_INC(icps_bmcastecho);
 516                         break;
 517                 }
 518                 icp->icmp_type = ICMP_ECHOREPLY;
 519                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 520                         goto freeit;
 521                 else
 522                         goto reflect;
 523
 524         case ICMP_TSTAMP:
 525                 if (!V_icmpbmcastecho
 526                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 527                         ICMPSTAT_INC(icps_bmcasttstamp);
 528                         break;
 529                 }
 530                 if (icmplen < ICMP_TSLEN) {
 531                         ICMPSTAT_INC(icps_badlen);
 532                         break;
 533                 }
 534                 icp->icmp_type = ICMP_TSTAMPREPLY;
 535                 icp->icmp_rtime = iptime();
 536                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 537                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 538                         goto freeit;
 539                 else
 540                         goto reflect;
 541
 542         case ICMP_MASKREQ:
 543                 if (V_icmpmaskrepl == 0)
 544                         break;
 545                 /*
 546                  * We are not able to respond with all ones broadcast
 547                  * unless we receive it over a point-to-point interface.
 548                  */
 549                 if (icmplen < ICMP_MASKLEN)
 550                         break;
 551                 switch (ip->ip_dst.s_addr) {
 552
 553                 case INADDR_BROADCAST:
 554                 case INADDR_ANY:
 555                         icmpdst.sin_addr = ip->ip_src;
 556                         break;
 557
 558                 default:
 559                         icmpdst.sin_addr = ip->ip_dst;
 560                 }
 561                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 562                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 563                 if (ia == NULL)
 564                         break;
 565                 if (ia->ia_ifp == NULL) {
 566                         ifa_free(&ia->ia_ifa);
 567                         break;
 568                 }
 569                 icp->icmp_type = ICMP_MASKREPLY;
 570                 if (V_icmpmaskfake == 0)
 571                         icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 572                 else
 573                         icp->icmp_mask = V_icmpmaskfake;
 574                 if (ip->ip_src.s_addr == 0) {
 575                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 576                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 577                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 578                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 579                 }
 580                 ifa_free(&ia->ia_ifa);
 581 reflect:
 582                 ip->ip_len += hlen;     /* since ip_input deducts this */
 583                 ICMPSTAT_INC(icps_reflect);
 584                 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
 585                 icmp_reflect(m);
 586                 return;
 587
 588         case ICMP_REDIRECT:
 589                 if (V_log_redirect) {
 590                         u_long src, dst, gw;
 591
 592                         src = ntohl(ip->ip_src.s_addr);
 593                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
 594                         gw = ntohl(icp->icmp_gwaddr.s_addr);
 595                         printf("icmp redirect from %d.%d.%d.%d: "
 596                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
 597                                (int)(src >> 24), (int)((src >> 16) & 0xff),
 598                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
 599                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
 600                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
 601                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
 602                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
 603                 }
 604                 /*
 605                  * RFC1812 says we must ignore ICMP redirects if we
 606                  * are acting as router.
 607                  */
 608                 if (V_drop_redirect || V_ipforwarding)
 609                         break;
 610                 if (code > 3)
 611                         goto badcode;
 612                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 613                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
 614                         ICMPSTAT_INC(icps_badlen);
 615                         break;
 616                 }
 617                 /*
 618                  * Short circuit routing redirects to force
 619                  * immediate change in the kernel's routing
 620                  * tables.  The message is also handed to anyone
 621                  * listening on a raw socket (e.g. the routing
 622                  * daemon for use in updating its tables).
 623                  */
 624                 icmpgw.sin_addr = ip->ip_src;
 625                 icmpdst.sin_addr = icp->icmp_gwaddr;
 626 #ifdef  ICMPPRINTFS
 627                 if (icmpprintfs) {
 628                         char buf[4 * sizeof "123"];
 629                         strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
 630
 631                         printf("redirect dst %s to %s\n",
 632                                buf, inet_ntoa(icp->icmp_gwaddr));
 633                 }
 634 #endif
 635                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 636                 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
 637                         in_rtredirect((struct sockaddr *)&icmpsrc,
 638                           (struct sockaddr *)&icmpdst,
 639                           (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
 640                           (struct sockaddr *)&icmpgw, fibnum);
 641                 }
 642                 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 643 #ifdef IPSEC
 644                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 645 #endif
 646                 break;
 647
 648         /*
 649          * No kernel processing for the following;
 650          * just fall through to send to raw listener.
 651          */
 652         case ICMP_ECHOREPLY:
 653         case ICMP_ROUTERADVERT:
 654         case ICMP_ROUTERSOLICIT:
 655         case ICMP_TSTAMPREPLY:
 656         case ICMP_IREQREPLY:
 657         case ICMP_MASKREPLY:
 658         default:
 659                 break;
 660         }
 661
 662 raw:
 663         rip_input(m, off);
 664         return;
 665
 666 freeit:
 667         m_freem(m);
 668 }
 669
 670 /*
 671  * Reflect the ip packet back to the source
 672  */
 673 static void
 674 icmp_reflect(struct mbuf *m)
 675 {
 676         struct ip *ip = mtod(m, struct ip *);
 677         struct ifaddr *ifa;
 678         struct ifnet *ifp;
 679         struct in_ifaddr *ia;
 680         struct in_addr t;
 681         struct mbuf *opts = 0;
 682         int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
 683
 684         if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
 685             IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
 686             IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
 687                 m_freem(m);     /* Bad return address */
 688                 ICMPSTAT_INC(icps_badaddr);
 689                 goto done;      /* Ip_output() will check for broadcast */
 690         }
 691
 692         t = ip->ip_dst;
 693         ip->ip_dst = ip->ip_src;
 694
 695         /*
 696          * Source selection for ICMP replies:
 697          *
 698          * If the incoming packet was addressed directly to one of our
 699          * own addresses, use dst as the src for the reply.
 700          */
 701         IN_IFADDR_RLOCK();
 702         LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
 703                 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
 704                         t = IA_SIN(ia)->sin_addr;
 705                         IN_IFADDR_RUNLOCK();
 706                         goto match;
 707                 }
 708         }
 709         IN_IFADDR_RUNLOCK();
 710
 711         /*
 712          * If the incoming packet was addressed to one of our broadcast
 713          * addresses, use the first non-broadcast address which corresponds
 714          * to the incoming interface.
 715          */
 716         ifp = m->m_pkthdr.rcvif;
 717         if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
 718                 IF_ADDR_LOCK(ifp);
 719                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 720                         if (ifa->ifa_addr->sa_family != AF_INET)
 721                                 continue;
 722                         ia = ifatoia(ifa);
 723                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 724                             t.s_addr) {
 725                                 t = IA_SIN(ia)->sin_addr;
 726                                 IF_ADDR_UNLOCK(ifp);
 727                                 goto match;
 728                         }
 729                 }
 730                 IF_ADDR_UNLOCK(ifp);
 731         }
 732         /*
 733          * If the packet was transiting through us, use the address of
 734          * the interface the packet came through in.  If that interface
 735          * doesn't have a suitable IP address, the normal selection
 736          * criteria apply.
 737          */
 738         if (V_icmp_rfi && ifp != NULL) {
 739                 IF_ADDR_LOCK(ifp);
 740                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 741                         if (ifa->ifa_addr->sa_family != AF_INET)
 742                                 continue;
 743                         ia = ifatoia(ifa);
 744                         t = IA_SIN(ia)->sin_addr;
 745                         IF_ADDR_UNLOCK(ifp);
 746                         goto match;
 747                 }
 748                 IF_ADDR_UNLOCK(ifp);
 749         }
 750         /*
 751          * If the incoming packet was not addressed directly to us, use
 752          * designated interface for icmp replies specified by sysctl
 753          * net.inet.icmp.reply_src (default not set). Otherwise continue
 754          * with normal source selection.
 755          */
 756         if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
 757                 IF_ADDR_LOCK(ifp);
 758                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 759                         if (ifa->ifa_addr->sa_family != AF_INET)
 760                                 continue;
 761                         ia = ifatoia(ifa);
 762                         t = IA_SIN(ia)->sin_addr;
 763                         IF_ADDR_UNLOCK(ifp);
 764                         goto match;
 765                 }
 766                 IF_ADDR_UNLOCK(ifp);
 767         }
 768         /*
 769          * If the packet was transiting through us, use the address of
 770          * the interface that is the closest to the packet source.
 771          * When we don't have a route back to the packet source, stop here
 772          * and drop the packet.
 773          */
 774         ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
 775         if (ia == NULL) {
 776                 m_freem(m);
 777                 ICMPSTAT_INC(icps_noroute);
 778                 goto done;
 779         }
 780         t = IA_SIN(ia)->sin_addr;
 781         ifa_free(&ia->ia_ifa);
 782 match:
 783 #ifdef MAC
 784         mac_netinet_icmp_replyinplace(m);
 785 #endif
 786         ip->ip_src = t;
 787         ip->ip_ttl = V_ip_defttl;
 788
 789         if (optlen > 0) {
 790                 register u_char *cp;
 791                 int opt, cnt;
 792                 u_int len;
 793
 794                 /*
 795                  * Retrieve any source routing from the incoming packet;
 796                  * add on any record-route or timestamp options.
 797                  */
 798                 cp = (u_char *) (ip + 1);
 799                 if ((opts = ip_srcroute(m)) == 0 &&
 800                     (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
 801                         opts->m_len = sizeof(struct in_addr);
 802                         mtod(opts, struct in_addr *)->s_addr = 0;
 803                 }
 804                 if (opts) {
 805 #ifdef ICMPPRINTFS
 806                     if (icmpprintfs)
 807                             printf("icmp_reflect optlen %d rt %d => ",
 808                                 optlen, opts->m_len);
 809 #endif
 810                     for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
 811                             opt = cp[IPOPT_OPTVAL];
 812                             if (opt == IPOPT_EOL)
 813                                     break;
 814                             if (opt == IPOPT_NOP)
 815                                     len = 1;
 816                             else {
 817                                     if (cnt < IPOPT_OLEN + sizeof(*cp))
 818                                             break;
 819                                     len = cp[IPOPT_OLEN];
 820                                     if (len < IPOPT_OLEN + sizeof(*cp) ||
 821                                         len > cnt)
 822                                             break;
 823                             }
 824                             /*
 825                              * Should check for overflow, but it "can't happen"
 826                              */
 827                             if (opt == IPOPT_RR || opt == IPOPT_TS ||
 828                                 opt == IPOPT_SECURITY) {
 829                                     bcopy((caddr_t)cp,
 830                                         mtod(opts, caddr_t) + opts->m_len, len);
 831                                     opts->m_len += len;
 832                             }
 833                     }
 834                     /* Terminate & pad, if necessary */
 835                     cnt = opts->m_len % 4;
 836                     if (cnt) {
 837                             for (; cnt < 4; cnt++) {
 838                                     *(mtod(opts, caddr_t) + opts->m_len) =
 839                                         IPOPT_EOL;
 840                                     opts->m_len++;
 841                             }
 842                     }
 843 #ifdef ICMPPRINTFS
 844                     if (icmpprintfs)
 845                             printf("%d\n", opts->m_len);
 846 #endif
 847                 }
 848                 /*
 849                  * Now strip out original options by copying rest of first
 850                  * mbuf's data back, and adjust the IP length.
 851                  */
 852                 ip->ip_len -= optlen;
 853                 ip->ip_v = IPVERSION;
 854                 ip->ip_hl = 5;
 855                 m->m_len -= optlen;
 856                 if (m->m_flags & M_PKTHDR)
 857                         m->m_pkthdr.len -= optlen;
 858                 optlen += sizeof(struct ip);
 859                 bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
 860                          (unsigned)(m->m_len - sizeof(struct ip)));
 861         }
 862         m_tag_delete_nonpersistent(m);
 863         m->m_flags &= ~(M_BCAST|M_MCAST);
 864         icmp_send(m, opts);
 865 done:
 866         if (opts)
 867                 (void)m_free(opts);
 868 }
 869
 870 /*
 871  * Send an icmp packet back to the ip level,
 872  * after supplying a checksum.
 873  */
 874 static void
 875 icmp_send(struct mbuf *m, struct mbuf *opts)
 876 {
 877         register struct ip *ip = mtod(m, struct ip *);
 878         register int hlen;
 879         register struct icmp *icp;
 880
 881         hlen = ip->ip_hl << 2;
 882         m->m_data += hlen;
 883         m->m_len -= hlen;
 884         icp = mtod(m, struct icmp *);
 885         icp->icmp_cksum = 0;
 886         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
 887         m->m_data -= hlen;
 888         m->m_len += hlen;
 889         m->m_pkthdr.rcvif = (struct ifnet *)0;
 890 #ifdef ICMPPRINTFS
 891         if (icmpprintfs) {
 892                 char buf[4 * sizeof "123"];
 893                 strcpy(buf, inet_ntoa(ip->ip_dst));
 894                 printf("icmp_send dst %s src %s\n",
 895                        buf, inet_ntoa(ip->ip_src));
 896         }
 897 #endif
 898         (void) ip_output(m, opts, NULL, 0, NULL, NULL);
 899 }
 900
 901 /*
 902  * Return milliseconds since 00:00 GMT in network format.
 903  */
 904 uint32_t
 905 iptime(void)
 906 {
 907         struct timeval atv;
 908         u_long t;
 909
 910         getmicrotime(&atv);
 911         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 912         return (htonl(t));
 913 }
 914
 915 /*
 916  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 917  * given current value MTU.  If DIR is less than zero, a larger plateau
 918  * is returned; otherwise, a smaller value is returned.
 919  */
 920 int
 921 ip_next_mtu(int mtu, int dir)
 922 {
 923         static int mtutab[] = {
 924                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
 925                 296, 68, 0
 926         };
 927         int i, size;
 928
 929         size = (sizeof mtutab) / (sizeof mtutab[0]);
 930         if (dir >= 0) {
 931                 for (i = 0; i < size; i++)
 932                         if (mtu > mtutab[i])
 933                                 return mtutab[i];
 934         } else {
 935                 for (i = size - 1; i >= 0; i--)
 936                         if (mtu < mtutab[i])
 937                                 return mtutab[i];
 938                 if (mtu == mtutab[0])
 939                         return mtutab[0];
 940         }
 941         return 0;
 942 }
 943
 944
 945 /*
 946  * badport_bandlim() - check for ICMP bandwidth limit
 947  *
 948  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 949  *      hit our bandwidth limit and it is not ok.
 950  *
 951  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 952  *
 953  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 954  *      values.  We may eventually remove this separation (and simplify the
 955  *      code further).
 956  *
 957  *      Note that the printing of the error message is delayed so we can
 958  *      properly print the icmp error rate that the system was trying to do
 959  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 960  *      the 'final' error, but it doesn't make sense to solve the printing
 961  *      delay with more complex code.
 962  */
 963
 964 int
 965 badport_bandlim(int which)
 966 {
 967
 968 #define N(a)    (sizeof (a) / sizeof (a[0]))
 969         static struct rate {
 970                 const char      *type;
 971                 struct timeval  lasttime;
 972                 int             curpps;
 973         } rates[BANDLIM_MAX+1] = {
 974                 { "icmp unreach response" },
 975                 { "icmp ping response" },
 976                 { "icmp tstamp response" },
 977                 { "closed port RST response" },
 978                 { "open port RST response" },
 979                 { "icmp6 unreach response" }
 980         };
 981
 982         /*
 983          * Return ok status if feature disabled or argument out of range.
 984          */
 985         if (V_icmplim > 0 && (u_int) which < N(rates)) {
 986                 struct rate *r = &rates[which];
 987                 int opps = r->curpps;
 988
 989                 if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
 990                         return -1;      /* discard packet */
 991                 /*
 992                  * If we've dropped below the threshold after having
 993                  * rate-limited traffic print the message.  This preserves
 994                  * the previous behaviour at the expense of added complexity.
 995                  */
 996                 if (V_icmplim_output && opps > V_icmplim)
 997                         printf("Limiting %s from %d to %d packets/sec\n",
 998                                 r->type, opps, V_icmplim);
 999         }
1000         return 0;                       /* okay to send packet */
1001 #undef N
1002 }