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 #include <sys/syslog.h>
  46
  47 #include <net/if.h>
  48 #include <net/if_types.h>
  49 #include <net/route.h>
  50 #include <net/vnet.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
  65 #ifdef IPSEC
  66 #include <netipsec/ipsec.h>
  67 #include <netipsec/key.h>
  68 #endif
  69
  70 #include <machine/in_cksum.h>
  71
  72 #include <security/mac/mac_framework.h>
  73
  74 /*
  75  * ICMP routines: error generation, receive packet processing, and
  76  * routines to turnaround packets back to the originator, and
  77  * host table maintenance routines.
  78  */
  79 VNET_DEFINE(struct icmpstat, icmpstat);
  80 SYSCTL_VNET_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
  81         &VNET_NAME(icmpstat), icmpstat, "");
  82
  83 static VNET_DEFINE(int, icmpmaskrepl) = 0;
  84 #define V_icmpmaskrepl                  VNET(icmpmaskrepl)
  85 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
  86         &VNET_NAME(icmpmaskrepl), 0,
  87         "Reply to ICMP Address Mask Request packets.");
  88
  89 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
  90 #define V_icmpmaskfake                  VNET(icmpmaskfake)
  91 SYSCTL_VNET_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_RW,
  92         &VNET_NAME(icmpmaskfake), 0,
  93         "Fake reply to ICMP Address Mask Request packets.");
  94
  95 VNET_DEFINE(int, drop_redirect) = 0;
  96
  97 static VNET_DEFINE(int, log_redirect) = 0;
  98 #define V_log_redirect                  VNET(log_redirect)
  99 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
 100         &VNET_NAME(log_redirect), 0,
 101         "Log ICMP redirects to the console");
 102
 103 static VNET_DEFINE(int, icmplim) = 200;
 104 #define V_icmplim                       VNET(icmplim)
 105 SYSCTL_VNET_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
 106         &VNET_NAME(icmplim), 0,
 107         "Maximum number of ICMP responses per second");
 108
 109 static VNET_DEFINE(int, icmplim_output) = 1;
 110 #define V_icmplim_output                VNET(icmplim_output)
 111 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
 112         &VNET_NAME(icmplim_output), 0,
 113         "Enable rate limiting of ICMP responses");
 114
 115 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
 116 #define V_reply_src                     VNET(reply_src)
 117 SYSCTL_VNET_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
 118         &VNET_NAME(reply_src), IFNAMSIZ,
 119         "icmp reply source for non-local packets.");
 120
 121 static VNET_DEFINE(int, icmp_rfi) = 0;
 122 #define V_icmp_rfi                      VNET(icmp_rfi)
 123 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
 124         &VNET_NAME(icmp_rfi), 0,
 125         "ICMP reply from incoming interface for non-local packets");
 126
 127 static VNET_DEFINE(int, icmp_quotelen) = 8;
 128 #define V_icmp_quotelen                 VNET(icmp_quotelen)
 129 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_RW,
 130         &VNET_NAME(icmp_quotelen), 0,
 131         "Number of bytes from original packet to quote in ICMP reply");
 132
 133 /*
 134  * ICMP broadcast echo sysctl
 135  */
 136 static VNET_DEFINE(int, icmpbmcastecho) = 0;
 137 #define V_icmpbmcastecho                VNET(icmpbmcastecho)
 138 SYSCTL_VNET_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
 139         &VNET_NAME(icmpbmcastecho), 0,
 140         "");
 141
 142
 143 #ifdef ICMPPRINTFS
 144 int     icmpprintfs = 0;
 145 #endif
 146
 147 static void     icmp_reflect(struct mbuf *);
 148 static void     icmp_send(struct mbuf *, struct mbuf *);
 149
 150 extern  struct protosw inetsw[];
 151
 152 static int
 153 sysctl_net_icmp_drop_redir(SYSCTL_HANDLER_ARGS)
 154 {
 155         int error, new;
 156         int i;
 157         struct radix_node_head *rnh;
 158
 159         new = V_drop_redirect;
 160         error = sysctl_handle_int(oidp, &new, 0, req);
 161         if (error == 0 && req->newptr) {
 162                 new = (new != 0) ? 1 : 0;
 163
 164                 if (new == V_drop_redirect)
 165                         return (0);
 166
 167                 for (i = 0; i < rt_numfibs; i++) {
 168                         if ((rnh = rt_tables_get_rnh(i, AF_INET)) == NULL)
 169                                 continue;
 170                         RADIX_NODE_HEAD_LOCK(rnh);
 171                         in_setmatchfunc(rnh, new);
 172                         RADIX_NODE_HEAD_UNLOCK(rnh);
 173                 }
 174
 175                 V_drop_redirect = new;
 176         }
 177
 178         return (error);
 179 }
 180
 181 SYSCTL_VNET_PROC(_net_inet_icmp, OID_AUTO, drop_redirect,
 182     CTLTYPE_INT|CTLFLAG_RW, 0, 0,
 183     sysctl_net_icmp_drop_redir, "I", "Ignore ICMP redirects");
 184
 185 /*
 186  * Kernel module interface for updating icmpstat.  The argument is an index
 187  * into icmpstat treated as an array of u_long.  While this encodes the
 188  * general layout of icmpstat into the caller, it doesn't encode its
 189  * location, so that future changes to add, for example, per-CPU stats
 190  * support won't cause binary compatibility problems for kernel modules.
 191  */
 192 void
 193 kmod_icmpstat_inc(int statnum)
 194 {
 195
 196         (*((u_long *)&V_icmpstat + statnum))++;
 197 }
 198
 199 /*
 200  * Generate an error packet of type error
 201  * in response to bad packet ip.
 202  */
 203 void
 204 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
 205 {
 206         register struct ip *oip = mtod(n, struct ip *), *nip;
 207         register unsigned oiphlen = oip->ip_hl << 2;
 208         register struct icmp *icp;
 209         register struct mbuf *m;
 210         unsigned icmplen, icmpelen, nlen;
 211
 212         KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
 213 #ifdef ICMPPRINTFS
 214         if (icmpprintfs)
 215                 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
 216 #endif
 217         if (type != ICMP_REDIRECT)
 218                 ICMPSTAT_INC(icps_error);
 219         /*
 220          * Don't send error:
 221          *  if the original packet was encrypted.
 222          *  if not the first fragment of message.
 223          *  in response to a multicast or broadcast packet.
 224          *  if the old packet protocol was an ICMP error message.
 225          */
 226         if (n->m_flags & M_DECRYPTED)
 227                 goto freeit;
 228         if (oip->ip_off & ~(IP_MF|IP_DF))
 229                 goto freeit;
 230         if (n->m_flags & (M_BCAST|M_MCAST))
 231                 goto freeit;
 232         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
 233           n->m_len >= oiphlen + ICMP_MINLEN &&
 234           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
 235                 ICMPSTAT_INC(icps_oldicmp);
 236                 goto freeit;
 237         }
 238         /* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
 239         if (oiphlen + 8 > n->m_len)
 240                 goto freeit;
 241         /*
 242          * Calculate length to quote from original packet and
 243          * prevent the ICMP mbuf from overflowing.
 244          * Unfortunatly this is non-trivial since ip_forward()
 245          * sends us truncated packets.
 246          */
 247         nlen = m_length(n, NULL);
 248         if (oip->ip_p == IPPROTO_TCP) {
 249                 struct tcphdr *th;
 250                 int tcphlen;
 251
 252                 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
 253                     n->m_next == NULL)
 254                         goto stdreply;
 255                 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
 256                     ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
 257                         goto freeit;
 258                 th = (struct tcphdr *)((caddr_t)oip + oiphlen);
 259                 tcphlen = th->th_off << 2;
 260                 if (tcphlen < sizeof(struct tcphdr))
 261                         goto freeit;
 262                 if (oip->ip_len < oiphlen + tcphlen)
 263                         goto freeit;
 264                 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
 265                         goto stdreply;
 266                 if (n->m_len < oiphlen + tcphlen &&
 267                     ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
 268                         goto freeit;
 269                 icmpelen = max(tcphlen, min(V_icmp_quotelen, oip->ip_len - oiphlen));
 270         } else
 271 stdreply:       icmpelen = max(8, min(V_icmp_quotelen, oip->ip_len - oiphlen));
 272
 273         icmplen = min(oiphlen + icmpelen, nlen);
 274         if (icmplen < sizeof(struct ip))
 275                 goto freeit;
 276
 277         if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
 278                 m = m_gethdr(M_DONTWAIT, MT_DATA);
 279         else
 280                 m = m_getcl(M_DONTWAIT, MT_DATA, M_PKTHDR);
 281         if (m == NULL)
 282                 goto freeit;
 283 #ifdef MAC
 284         mac_netinet_icmp_reply(n, m);
 285 #endif
 286         icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
 287         m_align(m, ICMP_MINLEN + icmplen);
 288         m->m_len = ICMP_MINLEN + icmplen;
 289
 290         /* XXX MRT  make the outgoing packet use the same FIB
 291          * that was associated with the incoming packet
 292          */
 293         M_SETFIB(m, M_GETFIB(n));
 294         icp = mtod(m, struct icmp *);
 295         ICMPSTAT_INC(icps_outhist[type]);
 296         icp->icmp_type = type;
 297         if (type == ICMP_REDIRECT)
 298                 icp->icmp_gwaddr.s_addr = dest;
 299         else {
 300                 icp->icmp_void = 0;
 301                 /*
 302                  * The following assignments assume an overlay with the
 303                  * just zeroed icmp_void field.
 304                  */
 305                 if (type == ICMP_PARAMPROB) {
 306                         icp->icmp_pptr = code;
 307                         code = 0;
 308                 } else if (type == ICMP_UNREACH &&
 309                         code == ICMP_UNREACH_NEEDFRAG && mtu) {
 310                         icp->icmp_nextmtu = htons(mtu);
 311                 }
 312         }
 313         icp->icmp_code = code;
 314
 315         /*
 316          * Copy the quotation into ICMP message and
 317          * convert quoted IP header back to network representation.
 318          */
 319         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
 320         nip = &icp->icmp_ip;
 321         nip->ip_len = htons(nip->ip_len);
 322         nip->ip_off = htons(nip->ip_off);
 323
 324         /*
 325          * Set up ICMP message mbuf and copy old IP header (without options
 326          * in front of ICMP message.
 327          * If the original mbuf was meant to bypass the firewall, the error
 328          * reply should bypass as well.
 329          */
 330         m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
 331         m->m_data -= sizeof(struct ip);
 332         m->m_len += sizeof(struct ip);
 333         m->m_pkthdr.len = m->m_len;
 334         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
 335         nip = mtod(m, struct ip *);
 336         bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
 337         nip->ip_len = m->m_len;
 338         nip->ip_v = IPVERSION;
 339         nip->ip_hl = 5;
 340         nip->ip_p = IPPROTO_ICMP;
 341         nip->ip_tos = 0;
 342         icmp_reflect(m);
 343
 344 freeit:
 345         m_freem(n);
 346 }
 347
 348 /*
 349  * Process a received ICMP message.
 350  */
 351 void
 352 icmp_input(struct mbuf *m, int off)
 353 {
 354         struct icmp *icp;
 355         struct in_ifaddr *ia;
 356         struct ip *ip = mtod(m, struct ip *);
 357         struct sockaddr_in icmpsrc, icmpdst, icmpgw;
 358         int hlen = off;
 359         int icmplen = ip->ip_len;
 360         int i, code;
 361         void (*ctlfunc)(int, struct sockaddr *, void *);
 362         int fibnum;
 363
 364         /*
 365          * Locate icmp structure in mbuf, and check
 366          * that not corrupted and of at least minimum length.
 367          */
 368 #ifdef ICMPPRINTFS
 369         if (icmpprintfs) {
 370                 char buf[4 * sizeof "123"];
 371                 strcpy(buf, inet_ntoa(ip->ip_src));
 372                 printf("icmp_input from %s to %s, len %d\n",
 373                        buf, inet_ntoa(ip->ip_dst), icmplen);
 374         }
 375 #endif
 376         if (icmplen < ICMP_MINLEN) {
 377                 ICMPSTAT_INC(icps_tooshort);
 378                 goto freeit;
 379         }
 380         i = hlen + min(icmplen, ICMP_ADVLENMIN);
 381         if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
 382                 ICMPSTAT_INC(icps_tooshort);
 383                 return;
 384         }
 385         ip = mtod(m, struct ip *);
 386         m->m_len -= hlen;
 387         m->m_data += hlen;
 388         icp = mtod(m, struct icmp *);
 389         if (in_cksum(m, icmplen)) {
 390                 ICMPSTAT_INC(icps_checksum);
 391                 goto freeit;
 392         }
 393         m->m_len += hlen;
 394         m->m_data -= hlen;
 395
 396         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
 397                 /*
 398                  * Deliver very specific ICMP type only.
 399                  */
 400                 switch (icp->icmp_type) {
 401                 case ICMP_UNREACH:
 402                 case ICMP_TIMXCEED:
 403                         break;
 404                 default:
 405                         goto freeit;
 406                 }
 407         }
 408
 409 #ifdef ICMPPRINTFS
 410         if (icmpprintfs)
 411                 printf("icmp_input, type %d code %d\n", icp->icmp_type,
 412                     icp->icmp_code);
 413 #endif
 414
 415         /*
 416          * Message type specific processing.
 417          */
 418         if (icp->icmp_type > ICMP_MAXTYPE)
 419                 goto raw;
 420
 421         /* Initialize */
 422         bzero(&icmpsrc, sizeof(icmpsrc));
 423         icmpsrc.sin_len = sizeof(struct sockaddr_in);
 424         icmpsrc.sin_family = AF_INET;
 425         bzero(&icmpdst, sizeof(icmpdst));
 426         icmpdst.sin_len = sizeof(struct sockaddr_in);
 427         icmpdst.sin_family = AF_INET;
 428         bzero(&icmpgw, sizeof(icmpgw));
 429         icmpgw.sin_len = sizeof(struct sockaddr_in);
 430         icmpgw.sin_family = AF_INET;
 431
 432         ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
 433         code = icp->icmp_code;
 434         switch (icp->icmp_type) {
 435
 436         case ICMP_UNREACH:
 437                 switch (code) {
 438                         case ICMP_UNREACH_NET:
 439                         case ICMP_UNREACH_HOST:
 440                         case ICMP_UNREACH_SRCFAIL:
 441                         case ICMP_UNREACH_NET_UNKNOWN:
 442                         case ICMP_UNREACH_HOST_UNKNOWN:
 443                         case ICMP_UNREACH_ISOLATED:
 444                         case ICMP_UNREACH_TOSNET:
 445                         case ICMP_UNREACH_TOSHOST:
 446                         case ICMP_UNREACH_HOST_PRECEDENCE:
 447                         case ICMP_UNREACH_PRECEDENCE_CUTOFF:
 448                                 code = PRC_UNREACH_NET;
 449                                 break;
 450
 451                         case ICMP_UNREACH_NEEDFRAG:
 452                                 code = PRC_MSGSIZE;
 453                                 break;
 454
 455                         /*
 456                          * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
 457                          * Treat subcodes 2,3 as immediate RST
 458                          */
 459                         case ICMP_UNREACH_PROTOCOL:
 460                         case ICMP_UNREACH_PORT:
 461                                 code = PRC_UNREACH_PORT;
 462                                 break;
 463
 464                         case ICMP_UNREACH_NET_PROHIB:
 465                         case ICMP_UNREACH_HOST_PROHIB:
 466                         case ICMP_UNREACH_FILTER_PROHIB:
 467                                 code = PRC_UNREACH_ADMIN_PROHIB;
 468                                 break;
 469
 470                         default:
 471                                 goto badcode;
 472                 }
 473                 goto deliver;
 474
 475         case ICMP_TIMXCEED:
 476                 if (code > 1)
 477                         goto badcode;
 478                 code += PRC_TIMXCEED_INTRANS;
 479                 goto deliver;
 480
 481         case ICMP_PARAMPROB:
 482                 if (code > 1)
 483                         goto badcode;
 484                 code = PRC_PARAMPROB;
 485                 goto deliver;
 486
 487         case ICMP_SOURCEQUENCH:
 488                 if (code)
 489                         goto badcode;
 490                 code = PRC_QUENCH;
 491         deliver:
 492                 /*
 493                  * Problem with datagram; advise higher level routines.
 494                  */
 495                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 496                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
 497                         ICMPSTAT_INC(icps_badlen);
 498                         goto freeit;
 499                 }
 500                 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
 501                 /* Discard ICMP's in response to multicast packets */
 502                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
 503                         goto badcode;
 504 #ifdef ICMPPRINTFS
 505                 if (icmpprintfs)
 506                         printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
 507 #endif
 508                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 509                 /*
 510                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
 511                  * notification to TCP layer.
 512                  */
 513                 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
 514                 if (ctlfunc)
 515                         (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
 516                                    (void *)&icp->icmp_ip);
 517                 break;
 518
 519         badcode:
 520                 ICMPSTAT_INC(icps_badcode);
 521                 break;
 522
 523         case ICMP_ECHO:
 524                 if (!V_icmpbmcastecho
 525                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 526                         ICMPSTAT_INC(icps_bmcastecho);
 527                         break;
 528                 }
 529                 icp->icmp_type = ICMP_ECHOREPLY;
 530                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 531                         goto freeit;
 532                 else
 533                         goto reflect;
 534
 535         case ICMP_TSTAMP:
 536                 if (!V_icmpbmcastecho
 537                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 538                         ICMPSTAT_INC(icps_bmcasttstamp);
 539                         break;
 540                 }
 541                 if (icmplen < ICMP_TSLEN) {
 542                         ICMPSTAT_INC(icps_badlen);
 543                         break;
 544                 }
 545                 icp->icmp_type = ICMP_TSTAMPREPLY;
 546                 icp->icmp_rtime = iptime();
 547                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 548                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 549                         goto freeit;
 550                 else
 551                         goto reflect;
 552
 553         case ICMP_MASKREQ:
 554                 if (V_icmpmaskrepl == 0)
 555                         break;
 556                 /*
 557                  * We are not able to respond with all ones broadcast
 558                  * unless we receive it over a point-to-point interface.
 559                  */
 560                 if (icmplen < ICMP_MASKLEN)
 561                         break;
 562                 switch (ip->ip_dst.s_addr) {
 563
 564                 case INADDR_BROADCAST:
 565                 case INADDR_ANY:
 566                         icmpdst.sin_addr = ip->ip_src;
 567                         break;
 568
 569                 default:
 570                         icmpdst.sin_addr = ip->ip_dst;
 571                 }
 572                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 573                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 574                 if (ia == NULL)
 575                         break;
 576                 if (ia->ia_ifp == NULL) {
 577                         ifa_free(&ia->ia_ifa);
 578                         break;
 579                 }
 580                 icp->icmp_type = ICMP_MASKREPLY;
 581                 if (V_icmpmaskfake == 0)
 582                         icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 583                 else
 584                         icp->icmp_mask = V_icmpmaskfake;
 585                 if (ip->ip_src.s_addr == 0) {
 586                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 587                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 588                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 589                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 590                 }
 591                 ifa_free(&ia->ia_ifa);
 592 reflect:
 593                 ip->ip_len += hlen;     /* since ip_input deducts this */
 594                 ICMPSTAT_INC(icps_reflect);
 595                 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
 596                 icmp_reflect(m);
 597                 return;
 598
 599         case ICMP_REDIRECT:
 600                 if (V_log_redirect) {
 601                         u_long src, dst, gw;
 602
 603                         src = ntohl(ip->ip_src.s_addr);
 604                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
 605                         gw = ntohl(icp->icmp_gwaddr.s_addr);
 606                         printf("icmp redirect from %d.%d.%d.%d: "
 607                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
 608                                (int)(src >> 24), (int)((src >> 16) & 0xff),
 609                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
 610                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
 611                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
 612                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
 613                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
 614                 }
 615                 /*
 616                  * RFC1812 says we must ignore ICMP redirects if we
 617                  * are acting as router.
 618                  */
 619                 if (V_drop_redirect || V_ipforwarding)
 620                         break;
 621                 if (code > 3)
 622                         goto badcode;
 623                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 624                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
 625                         ICMPSTAT_INC(icps_badlen);
 626                         break;
 627                 }
 628                 /*
 629                  * Short circuit routing redirects to force
 630                  * immediate change in the kernel's routing
 631                  * tables.  The message is also handed to anyone
 632                  * listening on a raw socket (e.g. the routing
 633                  * daemon for use in updating its tables).
 634                  */
 635                 icmpgw.sin_addr = ip->ip_src;
 636                 icmpdst.sin_addr = icp->icmp_gwaddr;
 637 #ifdef  ICMPPRINTFS
 638                 if (icmpprintfs) {
 639                         char buf[4 * sizeof "123"];
 640                         strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
 641
 642                         printf("redirect dst %s to %s\n",
 643                                buf, inet_ntoa(icp->icmp_gwaddr));
 644                 }
 645 #endif
 646                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 647                 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
 648                         in_rtredirect((struct sockaddr *)&icmpsrc,
 649                           (struct sockaddr *)&icmpdst,
 650                           (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
 651                           (struct sockaddr *)&icmpgw, fibnum);
 652                 }
 653                 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 654 #ifdef IPSEC
 655                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 656 #endif
 657                 break;
 658
 659         /*
 660          * No kernel processing for the following;
 661          * just fall through to send to raw listener.
 662          */
 663         case ICMP_ECHOREPLY:
 664         case ICMP_ROUTERADVERT:
 665         case ICMP_ROUTERSOLICIT:
 666         case ICMP_TSTAMPREPLY:
 667         case ICMP_IREQREPLY:
 668         case ICMP_MASKREPLY:
 669         default:
 670                 break;
 671         }
 672
 673 raw:
 674         rip_input(m, off);
 675         return;
 676
 677 freeit:
 678         m_freem(m);
 679 }
 680
 681 /*
 682  * Reflect the ip packet back to the source
 683  */
 684 static void
 685 icmp_reflect(struct mbuf *m)
 686 {
 687         struct ip *ip = mtod(m, struct ip *);
 688         struct ifaddr *ifa;
 689         struct ifnet *ifp;
 690         struct in_ifaddr *ia;
 691         struct in_addr t;
 692         struct mbuf *opts = 0;
 693         int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
 694
 695         if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
 696             IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
 697             IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
 698                 m_freem(m);     /* Bad return address */
 699                 ICMPSTAT_INC(icps_badaddr);
 700                 goto done;      /* Ip_output() will check for broadcast */
 701         }
 702
 703         t = ip->ip_dst;
 704         ip->ip_dst = ip->ip_src;
 705
 706         /*
 707          * Source selection for ICMP replies:
 708          *
 709          * If the incoming packet was addressed directly to one of our
 710          * own addresses, use dst as the src for the reply.
 711          */
 712         IN_IFADDR_RLOCK();
 713         LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
 714                 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
 715                         t = IA_SIN(ia)->sin_addr;
 716                         IN_IFADDR_RUNLOCK();
 717                         goto match;
 718                 }
 719         }
 720         IN_IFADDR_RUNLOCK();
 721
 722         /*
 723          * If the incoming packet was addressed to one of our broadcast
 724          * addresses, use the first non-broadcast address which corresponds
 725          * to the incoming interface.
 726          */
 727         ifp = m->m_pkthdr.rcvif;
 728         if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
 729                 IF_ADDR_RLOCK(ifp);
 730                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 731                         if (ifa->ifa_addr->sa_family != AF_INET)
 732                                 continue;
 733                         ia = ifatoia(ifa);
 734                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 735                             t.s_addr) {
 736                                 t = IA_SIN(ia)->sin_addr;
 737                                 IF_ADDR_RUNLOCK(ifp);
 738                                 goto match;
 739                         }
 740                 }
 741                 IF_ADDR_RUNLOCK(ifp);
 742         }
 743         /*
 744          * If the packet was transiting through us, use the address of
 745          * the interface the packet came through in.  If that interface
 746          * doesn't have a suitable IP address, the normal selection
 747          * criteria apply.
 748          */
 749         if (V_icmp_rfi && ifp != NULL) {
 750                 IF_ADDR_RLOCK(ifp);
 751                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 752                         if (ifa->ifa_addr->sa_family != AF_INET)
 753                                 continue;
 754                         ia = ifatoia(ifa);
 755                         t = IA_SIN(ia)->sin_addr;
 756                         IF_ADDR_RUNLOCK(ifp);
 757                         goto match;
 758                 }
 759                 IF_ADDR_RUNLOCK(ifp);
 760         }
 761         /*
 762          * If the incoming packet was not addressed directly to us, use
 763          * designated interface for icmp replies specified by sysctl
 764          * net.inet.icmp.reply_src (default not set). Otherwise continue
 765          * with normal source selection.
 766          */
 767         if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
 768                 IF_ADDR_RLOCK(ifp);
 769                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 770                         if (ifa->ifa_addr->sa_family != AF_INET)
 771                                 continue;
 772                         ia = ifatoia(ifa);
 773                         t = IA_SIN(ia)->sin_addr;
 774                         IF_ADDR_RUNLOCK(ifp);
 775                         goto match;
 776                 }
 777                 IF_ADDR_RUNLOCK(ifp);
 778         }
 779         /*
 780          * If the packet was transiting through us, use the address of
 781          * the interface that is the closest to the packet source.
 782          * When we don't have a route back to the packet source, stop here
 783          * and drop the packet.
 784          */
 785         ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
 786         if (ia == NULL) {
 787                 m_freem(m);
 788                 ICMPSTAT_INC(icps_noroute);
 789                 goto done;
 790         }
 791         t = IA_SIN(ia)->sin_addr;
 792         ifa_free(&ia->ia_ifa);
 793 match:
 794 #ifdef MAC
 795         mac_netinet_icmp_replyinplace(m);
 796 #endif
 797         ip->ip_src = t;
 798         ip->ip_ttl = V_ip_defttl;
 799
 800         if (optlen > 0) {
 801                 register u_char *cp;
 802                 int opt, cnt;
 803                 u_int len;
 804
 805                 /*
 806                  * Retrieve any source routing from the incoming packet;
 807                  * add on any record-route or timestamp options.
 808                  */
 809                 cp = (u_char *) (ip + 1);
 810                 if ((opts = ip_srcroute(m)) == 0 &&
 811                     (opts = m_gethdr(M_DONTWAIT, MT_DATA))) {
 812                         opts->m_len = sizeof(struct in_addr);
 813                         mtod(opts, struct in_addr *)->s_addr = 0;
 814                 }
 815                 if (opts) {
 816 #ifdef ICMPPRINTFS
 817                     if (icmpprintfs)
 818                             printf("icmp_reflect optlen %d rt %d => ",
 819                                 optlen, opts->m_len);
 820 #endif
 821                     for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
 822                             opt = cp[IPOPT_OPTVAL];
 823                             if (opt == IPOPT_EOL)
 824                                     break;
 825                             if (opt == IPOPT_NOP)
 826                                     len = 1;
 827                             else {
 828                                     if (cnt < IPOPT_OLEN + sizeof(*cp))
 829                                             break;
 830                                     len = cp[IPOPT_OLEN];
 831                                     if (len < IPOPT_OLEN + sizeof(*cp) ||
 832                                         len > cnt)
 833                                             break;
 834                             }
 835                             /*
 836                              * Should check for overflow, but it "can't happen"
 837                              */
 838                             if (opt == IPOPT_RR || opt == IPOPT_TS ||
 839                                 opt == IPOPT_SECURITY) {
 840                                     bcopy((caddr_t)cp,
 841                                         mtod(opts, caddr_t) + opts->m_len, len);
 842                                     opts->m_len += len;
 843                             }
 844                     }
 845                     /* Terminate & pad, if necessary */
 846                     cnt = opts->m_len % 4;
 847                     if (cnt) {
 848                             for (; cnt < 4; cnt++) {
 849                                     *(mtod(opts, caddr_t) + opts->m_len) =
 850                                         IPOPT_EOL;
 851                                     opts->m_len++;
 852                             }
 853                     }
 854 #ifdef ICMPPRINTFS
 855                     if (icmpprintfs)
 856                             printf("%d\n", opts->m_len);
 857 #endif
 858                 }
 859                 /*
 860                  * Now strip out original options by copying rest of first
 861                  * mbuf's data back, and adjust the IP length.
 862                  */
 863                 ip->ip_len -= optlen;
 864                 ip->ip_v = IPVERSION;
 865                 ip->ip_hl = 5;
 866                 m->m_len -= optlen;
 867                 if (m->m_flags & M_PKTHDR)
 868                         m->m_pkthdr.len -= optlen;
 869                 optlen += sizeof(struct ip);
 870                 bcopy((caddr_t)ip + optlen, (caddr_t)(ip + 1),
 871                          (unsigned)(m->m_len - sizeof(struct ip)));
 872         }
 873         m_tag_delete_nonpersistent(m);
 874         m->m_flags &= ~(M_BCAST|M_MCAST);
 875         icmp_send(m, opts);
 876 done:
 877         if (opts)
 878                 (void)m_free(opts);
 879 }
 880
 881 /*
 882  * Send an icmp packet back to the ip level,
 883  * after supplying a checksum.
 884  */
 885 static void
 886 icmp_send(struct mbuf *m, struct mbuf *opts)
 887 {
 888         register struct ip *ip = mtod(m, struct ip *);
 889         register int hlen;
 890         register struct icmp *icp;
 891
 892         hlen = ip->ip_hl << 2;
 893         m->m_data += hlen;
 894         m->m_len -= hlen;
 895         icp = mtod(m, struct icmp *);
 896         icp->icmp_cksum = 0;
 897         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
 898         m->m_data -= hlen;
 899         m->m_len += hlen;
 900         m->m_pkthdr.rcvif = (struct ifnet *)0;
 901 #ifdef ICMPPRINTFS
 902         if (icmpprintfs) {
 903                 char buf[4 * sizeof "123"];
 904                 strcpy(buf, inet_ntoa(ip->ip_dst));
 905                 printf("icmp_send dst %s src %s\n",
 906                        buf, inet_ntoa(ip->ip_src));
 907         }
 908 #endif
 909         (void) ip_output(m, opts, NULL, 0, NULL, NULL);
 910 }
 911
 912 /*
 913  * Return milliseconds since 00:00 GMT in network format.
 914  */
 915 uint32_t
 916 iptime(void)
 917 {
 918         struct timeval atv;
 919         u_long t;
 920
 921         getmicrotime(&atv);
 922         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 923         return (htonl(t));
 924 }
 925
 926 /*
 927  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 928  * given current value MTU.  If DIR is less than zero, a larger plateau
 929  * is returned; otherwise, a smaller value is returned.
 930  */
 931 int
 932 ip_next_mtu(int mtu, int dir)
 933 {
 934         static int mtutab[] = {
 935                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
 936                 296, 68, 0
 937         };
 938         int i, size;
 939
 940         size = (sizeof mtutab) / (sizeof mtutab[0]);
 941         if (dir >= 0) {
 942                 for (i = 0; i < size; i++)
 943                         if (mtu > mtutab[i])
 944                                 return mtutab[i];
 945         } else {
 946                 for (i = size - 1; i >= 0; i--)
 947                         if (mtu < mtutab[i])
 948                                 return mtutab[i];
 949                 if (mtu == mtutab[0])
 950                         return mtutab[0];
 951         }
 952         return 0;
 953 }
 954
 955
 956 /*
 957  * badport_bandlim() - check for ICMP bandwidth limit
 958  *
 959  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 960  *      hit our bandwidth limit and it is not ok.
 961  *
 962  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 963  *
 964  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 965  *      values.  We may eventually remove this separation (and simplify the
 966  *      code further).
 967  *
 968  *      Note that the printing of the error message is delayed so we can
 969  *      properly print the icmp error rate that the system was trying to do
 970  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 971  *      the 'final' error, but it doesn't make sense to solve the printing
 972  *      delay with more complex code.
 973  */
 974
 975 int
 976 badport_bandlim(int which)
 977 {
 978
 979 #define N(a)    (sizeof (a) / sizeof (a[0]))
 980         static struct rate {
 981                 const char      *type;
 982                 struct timeval  lasttime;
 983                 int             curpps;
 984         } rates[BANDLIM_MAX+1] = {
 985                 { "icmp unreach response" },
 986                 { "icmp ping response" },
 987                 { "icmp tstamp response" },
 988                 { "closed port RST response" },
 989                 { "open port RST response" },
 990                 { "icmp6 unreach response" }
 991         };
 992
 993         /*
 994          * Return ok status if feature disabled or argument out of range.
 995          */
 996         if (V_icmplim > 0 && (u_int) which < N(rates)) {
 997                 struct rate *r = &rates[which];
 998                 int opps = r->curpps;
 999
1000                 if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
1001                         return -1;      /* discard packet */
1002                 /*
1003                  * If we've dropped below the threshold after having
1004                  * rate-limited traffic print the message.  This preserves
1005                  * the previous behaviour at the expense of added complexity.
1006                  */
1007                 if (V_icmplim_output && opps > V_icmplim)
1008                         log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
1009                                 r->type, opps, V_icmplim);
1010         }
1011         return 0;                       /* okay to send packet */
1012 #undef N
1013 }