sys/netinet/ip_icmp.c

   1 /*
   2  * Copyright (c) 1982, 1986, 1988, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. All advertising materials mentioning features or use of this software
  14  *    must display the following acknowledgement:
  15  *      This product includes software developed by the University of
  16  *      California, Berkeley and its contributors.
  17  * 4. Neither the name of the University nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  *
  33  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
  34  * $FreeBSD: src/sys/netinet/ip_icmp.c,v 1.39.2.19 2003/01/24 05:11:34 sam Exp $
  35  * $DragonFly: src/sys/netinet/ip_icmp.c,v 1.32 2008/10/27 02:56:30 sephe Exp $
  36  */
  37
  38 #include "opt_ipsec.h"
  39
  40 #include <sys/param.h>
  41 #include <sys/systm.h>
  42 #include <sys/mbuf.h>
  43 #include <sys/protosw.h>
  44 #include <sys/socket.h>
  45 #include <sys/socketops.h>
  46 #include <sys/time.h>
  47 #include <sys/kernel.h>
  48 #include <sys/sysctl.h>
  49 #include <sys/in_cksum.h>
  50
  51 #include <machine/stdarg.h>
  52
  53 #include <net/if.h>
  54 #include <net/if_types.h>
  55 #include <net/route.h>
  56
  57 #define _IP_VHL
  58 #include <netinet/in.h>
  59 #include <netinet/in_systm.h>
  60 #include <netinet/in_var.h>
  61 #include <netinet/ip.h>
  62 #include <netinet/ip_icmp.h>
  63 #include <netinet/ip_var.h>
  64 #include <netinet/icmp_var.h>
  65
  66 #ifdef IPSEC
  67 #include <netinet6/ipsec.h>
  68 #include <netproto/key/key.h>
  69 #endif
  70
  71 #ifdef FAST_IPSEC
  72 #include <netproto/ipsec/ipsec.h>
  73 #include <netproto/ipsec/key.h>
  74 #define IPSEC
  75 #endif
  76
  77 /*
  78  * ICMP routines: error generation, receive packet processing, and
  79  * routines to turnaround packets back to the originator, and
  80  * host table maintenance routines.
  81  */
  82
  83 struct icmpstat icmpstat;
  84 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
  85         &icmpstat, icmpstat, "");
  86
  87 static int      icmpmaskrepl = 0;
  88 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
  89         &icmpmaskrepl, 0, "");
  90
  91 static int      drop_redirect = 0;
  92 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
  93         &drop_redirect, 0, "");
  94
  95 static int      log_redirect = 0;
  96 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
  97         &log_redirect, 0, "");
  98
  99 #ifdef ICMP_BANDLIM
 100
 101 /*
 102  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
 103  *      variable content is -1 and read-only.
 104  */
 105
 106 static int      icmplim = 200;
 107 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
 108         &icmplim, 0, "");
 109 #else
 110
 111 static int      icmplim = -1;
 112 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
 113         &icmplim, 0, "");
 114
 115 #endif
 116
 117 static int      icmplim_output = 1;
 118 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
 119         &icmplim_output, 0, "");
 120
 121 /*
 122  * ICMP broadcast echo sysctl
 123  */
 124
 125 static int      icmpbmcastecho = 0;
 126 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
 127         &icmpbmcastecho, 0, "");
 128
 129 static char     icmp_reply_src[IFNAMSIZ];
 130 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
 131         icmp_reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
 132
 133 static int      icmp_rfi;
 134 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
 135         &icmp_rfi, 0, "ICMP reply from incoming interface for "
 136         "non-local packets");
 137
 138 #ifdef ICMPPRINTFS
 139 int     icmpprintfs = 0;
 140 #endif
 141
 142 static void     icmp_reflect (struct mbuf *);
 143 static void     icmp_send (struct mbuf *, struct mbuf *, struct route *);
 144
 145 extern  struct protosw inetsw[];
 146
 147 /*
 148  * Generate an error packet of type error
 149  * in response to bad packet ip.
 150  */
 151 void
 152 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
 153 {
 154         struct ip *oip = mtod(n, struct ip *), *nip;
 155         unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
 156         struct icmp *icp;
 157         struct mbuf *m;
 158         unsigned icmplen;
 159
 160 #ifdef ICMPPRINTFS
 161         if (icmpprintfs)
 162                 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code);
 163 #endif
 164         if (type != ICMP_REDIRECT)
 165                 icmpstat.icps_error++;
 166         /*
 167          * Don't send error if the original packet was encrypted.
 168          * Don't send error if not the first fragment of message.
 169          * Don't error if the old packet protocol was ICMP
 170          * error message, only known informational types.
 171          */
 172         if (n->m_flags & M_DECRYPTED)
 173                 goto freeit;
 174         if (oip->ip_off &~ (IP_MF|IP_DF))
 175                 goto freeit;
 176         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
 177           n->m_len >= oiplen + ICMP_MINLEN &&
 178           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
 179                 icmpstat.icps_oldicmp++;
 180                 goto freeit;
 181         }
 182         /* Don't send error in response to a multicast or broadcast packet */
 183         if (n->m_flags & (M_BCAST|M_MCAST))
 184                 goto freeit;
 185         /*
 186          * First, formulate icmp message
 187          */
 188         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
 189         if (m == NULL)
 190                 goto freeit;
 191         icmplen = min(oiplen + 8, oip->ip_len);
 192         if (icmplen < sizeof(struct ip))
 193                 panic("icmp_error: bad length");
 194         m->m_len = icmplen + ICMP_MINLEN;
 195         MH_ALIGN(m, m->m_len);
 196         icp = mtod(m, struct icmp *);
 197         if ((u_int)type > ICMP_MAXTYPE)
 198                 panic("icmp_error");
 199         icmpstat.icps_outhist[type]++;
 200         icp->icmp_type = type;
 201         if (type == ICMP_REDIRECT)
 202                 icp->icmp_gwaddr.s_addr = dest;
 203         else {
 204                 icp->icmp_void = 0;
 205                 /*
 206                  * The following assignments assume an overlay with the
 207                  * zeroed icmp_void field.
 208                  */
 209                 if (type == ICMP_PARAMPROB) {
 210                         icp->icmp_pptr = code;
 211                         code = 0;
 212                 } else if (type == ICMP_UNREACH &&
 213                         code == ICMP_UNREACH_NEEDFRAG && destmtu) {
 214                         icp->icmp_nextmtu = htons(destmtu);
 215                 }
 216         }
 217
 218         icp->icmp_code = code;
 219         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
 220         nip = &icp->icmp_ip;
 221
 222         /*
 223          * Convert fields to network representation.
 224          */
 225         nip->ip_len = htons(nip->ip_len);
 226         nip->ip_off = htons(nip->ip_off);
 227
 228         /*
 229          * Now, copy old ip header (without options)
 230          * in front of icmp message.
 231          */
 232         if (m->m_data - sizeof(struct ip) < m->m_pktdat)
 233                 panic("icmp len");
 234         m->m_data -= sizeof(struct ip);
 235         m->m_len += sizeof(struct ip);
 236         m->m_pkthdr.len = m->m_len;
 237         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
 238         nip = mtod(m, struct ip *);
 239         bcopy(oip, nip, sizeof(struct ip));
 240         nip->ip_len = m->m_len;
 241         nip->ip_vhl = IP_VHL_BORING;
 242         nip->ip_p = IPPROTO_ICMP;
 243         nip->ip_tos = 0;
 244         m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED;
 245         icmp_reflect(m);
 246
 247 freeit:
 248         m_freem(n);
 249 }
 250
 251 /*
 252  * Process a received ICMP message.
 253  */
 254 void
 255 icmp_input(struct mbuf *m, ...)
 256 {
 257         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 258         struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
 259         struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
 260         struct icmp *icp;
 261         struct in_ifaddr *ia;
 262         struct ip *ip = mtod(m, struct ip *);
 263         int icmplen = ip->ip_len;
 264         int i, hlen;
 265         int code, off, proto;
 266         __va_list ap;
 267
 268         __va_start(ap, m);
 269         off = __va_arg(ap, int);
 270         proto = __va_arg(ap, int);
 271         __va_end(ap);
 272
 273         hlen = off;
 274
 275         /*
 276          * Locate icmp structure in mbuf, and check
 277          * that not corrupted and of at least minimum length.
 278          */
 279 #ifdef ICMPPRINTFS
 280         if (icmpprintfs) {
 281                 char buf[sizeof "aaa.bbb.ccc.ddd"];
 282
 283                 strcpy(buf, inet_ntoa(ip->ip_src));
 284                 kprintf("icmp_input from %s to %s, len %d\n",
 285                        buf, inet_ntoa(ip->ip_dst), icmplen);
 286         }
 287 #endif
 288         if (icmplen < ICMP_MINLEN) {
 289                 icmpstat.icps_tooshort++;
 290                 goto freeit;
 291         }
 292         i = hlen + min(icmplen, ICMP_ADVLENMIN);
 293         if (m->m_len < i && (m = m_pullup(m, i)) == 0)  {
 294                 icmpstat.icps_tooshort++;
 295                 return;
 296         }
 297         ip = mtod(m, struct ip *);
 298         m->m_len -= hlen;
 299         m->m_data += hlen;
 300         icp = mtod(m, struct icmp *);
 301         if (in_cksum(m, icmplen)) {
 302                 icmpstat.icps_checksum++;
 303                 goto freeit;
 304         }
 305         m->m_len += hlen;
 306         m->m_data -= hlen;
 307
 308         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
 309                 /*
 310                  * Deliver very specific ICMP type only.
 311                  */
 312                 switch (icp->icmp_type) {
 313                 case ICMP_UNREACH:
 314                 case ICMP_TIMXCEED:
 315                         break;
 316                 default:
 317                         goto freeit;
 318                 }
 319         }
 320
 321 #ifdef ICMPPRINTFS
 322         if (icmpprintfs)
 323                 kprintf("icmp_input, type %d code %d\n", icp->icmp_type,
 324                     icp->icmp_code);
 325 #endif
 326
 327         /*
 328          * Message type specific processing.
 329          */
 330         if (icp->icmp_type > ICMP_MAXTYPE)
 331                 goto raw;
 332         icmpstat.icps_inhist[icp->icmp_type]++;
 333         code = icp->icmp_code;
 334         switch (icp->icmp_type) {
 335
 336         case ICMP_UNREACH:
 337                 switch (code) {
 338                         case ICMP_UNREACH_NET:
 339                         case ICMP_UNREACH_HOST:
 340                         case ICMP_UNREACH_SRCFAIL:
 341                         case ICMP_UNREACH_NET_UNKNOWN:
 342                         case ICMP_UNREACH_HOST_UNKNOWN:
 343                         case ICMP_UNREACH_ISOLATED:
 344                         case ICMP_UNREACH_TOSNET:
 345                         case ICMP_UNREACH_TOSHOST:
 346                         case ICMP_UNREACH_HOST_PRECEDENCE:
 347                         case ICMP_UNREACH_PRECEDENCE_CUTOFF:
 348                                 code = PRC_UNREACH_NET;
 349                                 break;
 350
 351                         case ICMP_UNREACH_NEEDFRAG:
 352                                 code = PRC_MSGSIZE;
 353                                 break;
 354
 355                         /*
 356                          * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
 357                          * Treat subcodes 2,3 as immediate RST
 358                          */
 359                         case ICMP_UNREACH_PROTOCOL:
 360                         case ICMP_UNREACH_PORT:
 361                                 code = PRC_UNREACH_PORT;
 362                                 break;
 363
 364                         case ICMP_UNREACH_NET_PROHIB:
 365                         case ICMP_UNREACH_HOST_PROHIB:
 366                         case ICMP_UNREACH_FILTER_PROHIB:
 367                                 code = PRC_UNREACH_ADMIN_PROHIB;
 368                                 break;
 369
 370                         default:
 371                                 goto badcode;
 372                 }
 373                 goto deliver;
 374
 375         case ICMP_TIMXCEED:
 376                 if (code > 1)
 377                         goto badcode;
 378                 code += PRC_TIMXCEED_INTRANS;
 379                 goto deliver;
 380
 381         case ICMP_PARAMPROB:
 382                 if (code > 1)
 383                         goto badcode;
 384                 code = PRC_PARAMPROB;
 385                 goto deliver;
 386
 387         case ICMP_SOURCEQUENCH:
 388                 if (code)
 389                         goto badcode;
 390                 code = PRC_QUENCH;
 391 deliver:
 392                 /*
 393                  * Problem with datagram; advise higher level routines.
 394                  */
 395                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 396                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 397                         icmpstat.icps_badlen++;
 398                         goto freeit;
 399                 }
 400                 icp->icmp_ip.ip_len = ntohs(icp->icmp_ip.ip_len);
 401                 /* Discard ICMP's in response to multicast packets */
 402                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
 403                         goto badcode;
 404 #ifdef ICMPPRINTFS
 405                 if (icmpprintfs)
 406                         kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
 407 #endif
 408                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 409 #if 1
 410                 /*
 411                  * MTU discovery:
 412                  * If we got a needfrag and there is a host route to the
 413                  * original destination, and the MTU is not locked, then
 414                  * set the MTU in the route to the suggested new value
 415                  * (if given) and then notify as usual.  The ULPs will
 416                  * notice that the MTU has changed and adapt accordingly.
 417                  * If no new MTU was suggested, then we guess a new one
 418                  * less than the current value.  If the new MTU is
 419                  * unreasonably small (arbitrarily set at 296), then
 420                  * we reset the MTU to the interface value and enable the
 421                  * lock bit, indicating that we are no longer doing MTU
 422                  * discovery.
 423                  */
 424                 if (code == PRC_MSGSIZE) {
 425                         struct rtentry *rt;
 426                         int mtu;
 427
 428                         rt = rtpurelookup((struct sockaddr *)&icmpsrc);
 429                         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
 430                             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
 431                                 mtu = ntohs(icp->icmp_nextmtu);
 432                                 if (!mtu)
 433                                         mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu,
 434                                                           1);
 435 #ifdef DEBUG_MTUDISC
 436                                 kprintf("MTU for %s reduced to %d\n",
 437                                         inet_ntoa(icmpsrc.sin_addr), mtu);
 438 #endif
 439                                 if (mtu < 296) {
 440                                         /* rt->rt_rmx.rmx_mtu =
 441                                                 rt->rt_ifp->if_mtu; */
 442                                         rt->rt_rmx.rmx_locks |= RTV_MTU;
 443                                 } else if (rt->rt_rmx.rmx_mtu > mtu) {
 444                                         rt->rt_rmx.rmx_mtu = mtu;
 445                                 }
 446                         }
 447                         if (rt != NULL)
 448                                 --rt->rt_refcnt;
 449                 }
 450 #endif
 451                 /*
 452                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
 453                  * notification to TCP layer.
 454                  */
 455                 so_pru_ctlinput(
 456                         &inetsw[ip_protox[icp->icmp_ip.ip_p]],
 457                         code, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
 458                 break;
 459
 460 badcode:
 461                 icmpstat.icps_badcode++;
 462                 break;
 463
 464         case ICMP_ECHO:
 465                 if (!icmpbmcastecho
 466                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 467                         icmpstat.icps_bmcastecho++;
 468                         break;
 469                 }
 470                 icp->icmp_type = ICMP_ECHOREPLY;
 471 #ifdef ICMP_BANDLIM
 472                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 473                         goto freeit;
 474                 else
 475 #endif
 476                         goto reflect;
 477
 478         case ICMP_TSTAMP:
 479                 if (!icmpbmcastecho
 480                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 481                         icmpstat.icps_bmcasttstamp++;
 482                         break;
 483                 }
 484                 if (icmplen < ICMP_TSLEN) {
 485                         icmpstat.icps_badlen++;
 486                         break;
 487                 }
 488                 icp->icmp_type = ICMP_TSTAMPREPLY;
 489                 icp->icmp_rtime = iptime();
 490                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 491 #ifdef ICMP_BANDLIM
 492                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 493                         goto freeit;
 494                 else
 495 #endif
 496                         goto reflect;
 497
 498         case ICMP_MASKREQ:
 499                 if (icmpmaskrepl == 0)
 500                         break;
 501                 /*
 502                  * We are not able to respond with all ones broadcast
 503                  * unless we receive it over a point-to-point interface.
 504                  */
 505                 if (icmplen < ICMP_MASKLEN)
 506                         break;
 507                 switch (ip->ip_dst.s_addr) {
 508
 509                 case INADDR_BROADCAST:
 510                 case INADDR_ANY:
 511                         icmpdst.sin_addr = ip->ip_src;
 512                         break;
 513
 514                 default:
 515                         icmpdst.sin_addr = ip->ip_dst;
 516                 }
 517                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 518                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 519                 if (ia == 0)
 520                         break;
 521                 if (ia->ia_ifp == 0)
 522                         break;
 523                 icp->icmp_type = ICMP_MASKREPLY;
 524                 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 525                 if (ip->ip_src.s_addr == 0) {
 526                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 527                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 528                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 529                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 530                 }
 531 reflect:
 532                 ip->ip_len += hlen;     /* since ip_input deducts this */
 533                 icmpstat.icps_reflect++;
 534                 icmpstat.icps_outhist[icp->icmp_type]++;
 535                 icmp_reflect(m);
 536                 return;
 537
 538         case ICMP_REDIRECT:
 539                 if (log_redirect) {
 540                         u_long src, dst, gw;
 541
 542                         src = ntohl(ip->ip_src.s_addr);
 543                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
 544                         gw = ntohl(icp->icmp_gwaddr.s_addr);
 545                         kprintf("icmp redirect from %d.%d.%d.%d: "
 546                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
 547                                (int)(src >> 24), (int)((src >> 16) & 0xff),
 548                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
 549                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
 550                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
 551                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
 552                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
 553                 }
 554                 if (drop_redirect)
 555                         break;
 556                 if (code > 3)
 557                         goto badcode;
 558                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 559                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 560                         icmpstat.icps_badlen++;
 561                         break;
 562                 }
 563                 /*
 564                  * Short circuit routing redirects to force
 565                  * immediate change in the kernel's routing
 566                  * tables.  The message is also handed to anyone
 567                  * listening on a raw socket (e.g. the routing
 568                  * daemon for use in updating its tables).
 569                  */
 570                 icmpgw.sin_addr = ip->ip_src;
 571                 icmpdst.sin_addr = icp->icmp_gwaddr;
 572 #ifdef  ICMPPRINTFS
 573                 if (icmpprintfs) {
 574                         char buf[sizeof "aaa.bbb.ccc.ddd"];
 575
 576                         strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
 577                         kprintf("redirect dst %s to %s\n",
 578                                buf, inet_ntoa(icp->icmp_gwaddr));
 579                 }
 580 #endif
 581                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 582                 rtredirect((struct sockaddr *)&icmpsrc,
 583                   (struct sockaddr *)&icmpdst,
 584                   NULL, RTF_GATEWAY | RTF_HOST,
 585                   (struct sockaddr *)&icmpgw);
 586                 kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 587 #ifdef IPSEC
 588                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 589 #endif
 590                 break;
 591
 592         /*
 593          * No kernel processing for the following;
 594          * just fall through to send to raw listener.
 595          */
 596         case ICMP_ECHOREPLY:
 597         case ICMP_ROUTERADVERT:
 598         case ICMP_ROUTERSOLICIT:
 599         case ICMP_TSTAMPREPLY:
 600         case ICMP_IREQREPLY:
 601         case ICMP_MASKREPLY:
 602         default:
 603                 break;
 604         }
 605
 606 raw:
 607         rip_input(m, off, proto);
 608         return;
 609
 610 freeit:
 611         m_freem(m);
 612 }
 613
 614 /*
 615  * Reflect the ip packet back to the source
 616  */
 617 static void
 618 icmp_reflect(struct mbuf *m)
 619 {
 620         struct ip *ip = mtod(m, struct ip *);
 621         struct in_ifaddr *ia;
 622         struct in_ifaddr_container *iac;
 623         struct ifaddr_container *ifac;
 624         struct ifnet *ifp;
 625         struct in_addr t;
 626         struct mbuf *opts = 0;
 627         int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
 628         struct route *ro = NULL, rt;
 629
 630         if (!in_canforward(ip->ip_src) &&
 631             ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
 632              (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
 633                 m_freem(m);     /* Bad return address */
 634                 icmpstat.icps_badaddr++;
 635                 goto done;      /* Ip_output() will check for broadcast */
 636         }
 637         t = ip->ip_dst;
 638         ip->ip_dst = ip->ip_src;
 639
 640         ro = &rt;
 641         bzero(ro, sizeof *ro);
 642
 643         /*
 644          * If the incoming packet was addressed directly to us,
 645          * use dst as the src for the reply.  Otherwise (broadcast
 646          * or anonymous), use the address which corresponds
 647          * to the incoming interface.
 648          */
 649         ia = NULL;
 650         LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
 651                 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
 652                         ia = iac->ia;
 653                         goto match;
 654                 }
 655         }
 656         ifp = m->m_pkthdr.rcvif;
 657         if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) {
 658                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 659                         struct ifaddr *ifa = ifac->ifa;
 660
 661                         if (ifa->ifa_addr->sa_family != AF_INET)
 662                                 continue;
 663                         ia = ifatoia(ifa);
 664                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 665                             t.s_addr)
 666                                 goto match;
 667                 }
 668         }
 669         /*
 670          * If the packet was transiting through us, use the address of
 671          * the interface the packet came through in.  If that interface
 672          * doesn't have a suitable IP address, the normal selection
 673          * criteria apply.
 674          */
 675         if (icmp_rfi && ifp != NULL) {
 676                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 677                         struct ifaddr *ifa = ifac->ifa;
 678
 679                         if (ifa->ifa_addr->sa_family != AF_INET)
 680                                 continue;
 681                         ia = ifatoia(ifa);
 682                         goto match;
 683                 }
 684         }
 685         /*
 686          * If the incoming packet was not addressed directly to us, use
 687          * designated interface for icmp replies specified by sysctl
 688          * net.inet.icmp.reply_src (default not set). Otherwise continue
 689          * with normal source selection.
 690          */
 691         if (icmp_reply_src[0] != '\0' && (ifp = ifunit(icmp_reply_src))) {
 692                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 693                         struct ifaddr *ifa = ifac->ifa;
 694
 695                         if (ifa->ifa_addr->sa_family != AF_INET)
 696                                 continue;
 697                         ia = ifatoia(ifa);
 698                         goto match;
 699                 }
 700         }
 701         /*
 702          * If the packet was transiting through us, use the address of
 703          * the interface that is the closest to the packet source.
 704          * When we don't have a route back to the packet source, stop here
 705          * and drop the packet.
 706          */
 707         ia = ip_rtaddr(ip->ip_dst, ro);
 708         if (ia == NULL) {
 709                 m_freem(m);
 710                 icmpstat.icps_noroute++;
 711                 goto done;
 712         }
 713 match:
 714         t = IA_SIN(ia)->sin_addr;
 715         ip->ip_src = t;
 716         ip->ip_ttl = ip_defttl;
 717
 718         if (optlen > 0) {
 719                 u_char *cp;
 720                 int opt, cnt;
 721                 u_int len;
 722
 723                 /*
 724                  * Retrieve any source routing from the incoming packet;
 725                  * add on any record-route or timestamp options.
 726                  */
 727                 cp = (u_char *) (ip + 1);
 728                 if ((opts = ip_srcroute(m)) == 0 &&
 729                     (opts = m_gethdr(MB_DONTWAIT, MT_HEADER))) {
 730                         opts->m_len = sizeof(struct in_addr);
 731                         mtod(opts, struct in_addr *)->s_addr = 0;
 732                 }
 733                 if (opts) {
 734 #ifdef ICMPPRINTFS
 735                         if (icmpprintfs)
 736                                 kprintf("icmp_reflect optlen %d rt %d => ",
 737                                        optlen, opts->m_len);
 738 #endif
 739                         for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
 740                                 opt = cp[IPOPT_OPTVAL];
 741                                 if (opt == IPOPT_EOL)
 742                                         break;
 743                                 if (opt == IPOPT_NOP)
 744                                         len = 1;
 745                                 else {
 746                                         if (cnt < IPOPT_OLEN + sizeof *cp)
 747                                                 break;
 748                                         len = cp[IPOPT_OLEN];
 749                                         if (len < IPOPT_OLEN + sizeof *cp ||
 750                                             len > cnt)
 751                                         break;
 752                                 }
 753                                 /*
 754                                  * Should check for overflow, but it
 755                                  * "can't happen".
 756                                  */
 757                                 if (opt == IPOPT_RR || opt == IPOPT_TS ||
 758                                     opt == IPOPT_SECURITY) {
 759                                         bcopy(cp,
 760                                               mtod(opts, caddr_t) + opts->m_len,
 761                                               len);
 762                                         opts->m_len += len;
 763                                 }
 764                         }
 765                         /* Terminate & pad, if necessary */
 766                         cnt = opts->m_len % 4;
 767                         if (cnt) {
 768                                 for (; cnt < 4; cnt++) {
 769                                         *(mtod(opts, caddr_t) + opts->m_len) =
 770                                             IPOPT_EOL;
 771                                         opts->m_len++;
 772                                 }
 773                         }
 774 #ifdef ICMPPRINTFS
 775                         if (icmpprintfs)
 776                                 kprintf("%d\n", opts->m_len);
 777 #endif
 778                 }
 779                 /*
 780                  * Now strip out original options by copying rest of first
 781                  * mbuf's data back, and adjust the IP length.
 782                  */
 783                 ip->ip_len -= optlen;
 784                 ip->ip_vhl = IP_VHL_BORING;
 785                 m->m_len -= optlen;
 786                 if (m->m_flags & M_PKTHDR)
 787                         m->m_pkthdr.len -= optlen;
 788                 optlen += sizeof(struct ip);
 789                 bcopy((caddr_t)ip + optlen, ip + 1,
 790                       m->m_len - sizeof(struct ip));
 791         }
 792         m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
 793         m->m_flags &= ~(M_BCAST|M_MCAST);
 794         icmp_send(m, opts, ro);
 795 done:
 796         if (opts)
 797                 m_free(opts);
 798         if (ro && ro->ro_rt)
 799                 RTFREE(ro->ro_rt);
 800 }
 801
 802 /*
 803  * Send an icmp packet back to the ip level,
 804  * after supplying a checksum.
 805  */
 806 static void
 807 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
 808 {
 809         struct ip *ip = mtod(m, struct ip *);
 810         int hlen;
 811         struct icmp *icp;
 812
 813         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 814         m->m_data += hlen;
 815         m->m_len -= hlen;
 816         icp = mtod(m, struct icmp *);
 817         icp->icmp_cksum = 0;
 818         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
 819         m->m_data -= hlen;
 820         m->m_len += hlen;
 821         m->m_pkthdr.rcvif = NULL;
 822 #ifdef ICMPPRINTFS
 823         if (icmpprintfs) {
 824                 char buf[sizeof "aaa.bbb.ccc.ddd"];
 825
 826                 strcpy(buf, inet_ntoa(ip->ip_dst));
 827                 kprintf("icmp_send dst %s src %s\n", buf, inet_ntoa(ip->ip_src));
 828         }
 829 #endif
 830         ip_output(m, opts, rt, 0, NULL, NULL);
 831 }
 832
 833 n_time
 834 iptime(void)
 835 {
 836         struct timeval atv;
 837         u_long t;
 838
 839         getmicrotime(&atv);
 840         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 841         return (htonl(t));
 842 }
 843
 844 #if 1
 845 /*
 846  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 847  * given current value MTU.  If DIR is less than zero, a larger plateau
 848  * is returned; otherwise, a smaller value is returned.
 849  */
 850 int
 851 ip_next_mtu(int mtu, int dir)
 852 {
 853         static int mtutab[] = {
 854                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
 855                 68, 0
 856         };
 857         int i;
 858
 859         for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
 860                 if (mtu >= mtutab[i])
 861                         break;
 862         }
 863
 864         if (dir < 0) {
 865                 if (i == 0) {
 866                         return 0;
 867                 } else {
 868                         return mtutab[i - 1];
 869                 }
 870         } else {
 871                 if (mtutab[i] == 0) {
 872                         return 0;
 873                 } else if(mtu > mtutab[i]) {
 874                         return mtutab[i];
 875                 } else {
 876                         return mtutab[i + 1];
 877                 }
 878         }
 879 }
 880 #endif
 881
 882 #ifdef ICMP_BANDLIM
 883 /*
 884  * badport_bandlim() - check for ICMP bandwidth limit
 885  *
 886  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 887  *      hit our bandwidth limit and it is not ok.
 888  *
 889  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 890  *
 891  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 892  *      values.  We may eventually remove this separation (and simplify the
 893  *      code further).
 894  *
 895  *      Note that the printing of the error message is delayed so we can
 896  *      properly print the icmp error rate that the system was trying to do
 897  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 898  *      the 'final' error, but it doesn't make sense to solve the printing
 899  *      delay with more complex code.
 900  */
 901 int
 902 badport_bandlim(int which)
 903 {
 904         static int lticks[BANDLIM_MAX + 1];
 905         static int lpackets[BANDLIM_MAX + 1];
 906         int dticks;
 907         const char *bandlimittype[] = {
 908                 "Limiting icmp unreach response",
 909                 "Limiting icmp ping response",
 910                 "Limiting icmp tstamp response",
 911                 "Limiting closed port RST response",
 912                 "Limiting open port RST response"
 913                 };
 914
 915         /*
 916          * Return ok status if feature disabled or argument out of
 917          * ranage.
 918          */
 919
 920         if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
 921                 return(0);
 922         dticks = ticks - lticks[which];
 923
 924         /*
 925          * reset stats when cumulative dt exceeds one second.
 926          */
 927
 928         if ((unsigned int)dticks > hz) {
 929                 if (lpackets[which] > icmplim && icmplim_output) {
 930                         kprintf("%s from %d to %d packets per second\n",
 931                                 bandlimittype[which],
 932                                 lpackets[which],
 933                                 icmplim
 934                         );
 935                 }
 936                 lticks[which] = ticks;
 937                 lpackets[which] = 0;
 938         }
 939
 940         /*
 941          * bump packet count
 942          */
 943
 944         if (++lpackets[which] > icmplim) {
 945                 return(-1);
 946         }
 947         return(0);
 948 }
 949 #endif