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_inet.h"
  36 #include "opt_ipsec.h"
  37
  38 #include <sys/param.h>
  39 #include <sys/systm.h>
  40 #include <sys/mbuf.h>
  41 #include <sys/protosw.h>
  42 #include <sys/socket.h>
  43 #include <sys/time.h>
  44 #include <sys/kernel.h>
  45 #include <sys/sysctl.h>
  46 #include <sys/syslog.h>
  47
  48 #include <net/if.h>
  49 #include <net/if_var.h>
  50 #include <net/if_types.h>
  51 #include <net/route.h>
  52 #include <net/vnet.h>
  53
  54 #include <netinet/in.h>
  55 #include <netinet/in_pcb.h>
  56 #include <netinet/in_systm.h>
  57 #include <netinet/in_var.h>
  58 #include <netinet/ip.h>
  59 #include <netinet/ip_icmp.h>
  60 #include <netinet/ip_var.h>
  61 #include <netinet/ip_options.h>
  62 #include <netinet/tcp.h>
  63 #include <netinet/tcp_var.h>
  64 #include <netinet/tcpip.h>
  65 #include <netinet/icmp_var.h>
  66
  67 #ifdef INET
  68 #ifdef IPSEC
  69 #include <netipsec/ipsec.h>
  70 #include <netipsec/key.h>
  71 #endif
  72
  73 #include <machine/in_cksum.h>
  74
  75 #include <security/mac/mac_framework.h>
  76 #endif /* INET */
  77
  78 /*
  79  * ICMP routines: error generation, receive packet processing, and
  80  * routines to turnaround packets back to the originator, and
  81  * host table maintenance routines.
  82  */
  83 static VNET_DEFINE(int, icmplim) = 200;
  84 #define V_icmplim                       VNET(icmplim)
  85 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
  86         &VNET_NAME(icmplim), 0,
  87         "Maximum number of ICMP responses per second");
  88
  89 static VNET_DEFINE(int, icmplim_output) = 1;
  90 #define V_icmplim_output                VNET(icmplim_output)
  91 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
  92         &VNET_NAME(icmplim_output), 0,
  93         "Enable logging of ICMP response rate limiting");
  94
  95 #ifdef INET
  96 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
  97 VNET_PCPUSTAT_SYSINIT(icmpstat);
  98 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
  99     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
 100
 101 #ifdef VIMAGE
 102 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
 103 #endif /* VIMAGE */
 104
 105 static VNET_DEFINE(int, icmpmaskrepl) = 0;
 106 #define V_icmpmaskrepl                  VNET(icmpmaskrepl)
 107 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
 108         &VNET_NAME(icmpmaskrepl), 0,
 109         "Reply to ICMP Address Mask Request packets.");
 110
 111 static VNET_DEFINE(u_int, icmpmaskfake) = 0;
 112 #define V_icmpmaskfake                  VNET(icmpmaskfake)
 113 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
 114         &VNET_NAME(icmpmaskfake), 0,
 115         "Fake reply to ICMP Address Mask Request packets.");
 116
 117 VNET_DEFINE(int, drop_redirect) = 0;
 118
 119 static VNET_DEFINE(int, log_redirect) = 0;
 120 #define V_log_redirect                  VNET(log_redirect)
 121 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
 122         &VNET_NAME(log_redirect), 0,
 123         "Log ICMP redirects to the console");
 124
 125 static VNET_DEFINE(char, reply_src[IFNAMSIZ]);
 126 #define V_reply_src                     VNET(reply_src)
 127 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
 128         &VNET_NAME(reply_src), IFNAMSIZ,
 129         "icmp reply source for non-local packets.");
 130
 131 static VNET_DEFINE(int, icmp_rfi) = 0;
 132 #define V_icmp_rfi                      VNET(icmp_rfi)
 133 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
 134         &VNET_NAME(icmp_rfi), 0,
 135         "ICMP reply from incoming interface for non-local packets");
 136
 137 static VNET_DEFINE(int, icmp_quotelen) = 8;
 138 #define V_icmp_quotelen                 VNET(icmp_quotelen)
 139 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
 140         &VNET_NAME(icmp_quotelen), 0,
 141         "Number of bytes from original packet to quote in ICMP reply");
 142
 143 /*
 144  * ICMP broadcast echo sysctl
 145  */
 146 static VNET_DEFINE(int, icmpbmcastecho) = 0;
 147 #define V_icmpbmcastecho                VNET(icmpbmcastecho)
 148 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
 149         &VNET_NAME(icmpbmcastecho), 0,
 150         "");
 151
 152 static VNET_DEFINE(int, icmptstamprepl) = 1;
 153 #define V_icmptstamprepl                VNET(icmptstamprepl)
 154 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_RW,
 155         &VNET_NAME(icmptstamprepl), 0, "Respond to ICMP Timestamp packets");
 156
 157 #ifdef ICMPPRINTFS
 158 int     icmpprintfs = 0;
 159 #endif
 160
 161 static void     icmp_reflect(struct mbuf *);
 162 static void     icmp_send(struct mbuf *, struct mbuf *);
 163
 164 extern  struct protosw inetsw[];
 165
 166 static int
 167 sysctl_net_icmp_drop_redir(SYSCTL_HANDLER_ARGS)
 168 {
 169         int error, new;
 170         int i;
 171         struct radix_node_head *rnh;
 172
 173         new = V_drop_redirect;
 174         error = sysctl_handle_int(oidp, &new, 0, req);
 175         if (error == 0 && req->newptr) {
 176                 new = (new != 0) ? 1 : 0;
 177
 178                 if (new == V_drop_redirect)
 179                         return (0);
 180
 181                 for (i = 0; i < rt_numfibs; i++) {
 182                         if ((rnh = rt_tables_get_rnh(i, AF_INET)) == NULL)
 183                                 continue;
 184                         RADIX_NODE_HEAD_LOCK(rnh);
 185                         in_setmatchfunc(rnh, new);
 186                         RADIX_NODE_HEAD_UNLOCK(rnh);
 187                 }
 188
 189                 V_drop_redirect = new;
 190         }
 191
 192         return (error);
 193 }
 194
 195 SYSCTL_PROC(_net_inet_icmp, OID_AUTO, drop_redirect,
 196     CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 0, 0,
 197     sysctl_net_icmp_drop_redir, "I", "Ignore ICMP redirects");
 198
 199 /*
 200  * Kernel module interface for updating icmpstat.  The argument is an index
 201  * into icmpstat treated as an array of u_long.  While this encodes the
 202  * general layout of icmpstat into the caller, it doesn't encode its
 203  * location, so that future changes to add, for example, per-CPU stats
 204  * support won't cause binary compatibility problems for kernel modules.
 205  */
 206 void
 207 kmod_icmpstat_inc(int statnum)
 208 {
 209
 210         counter_u64_add(VNET(icmpstat)[statnum], 1);
 211 }
 212
 213 /*
 214  * Generate an error packet of type error
 215  * in response to bad packet ip.
 216  */
 217 void
 218 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
 219 {
 220         register struct ip *oip = mtod(n, struct ip *), *nip;
 221         register unsigned oiphlen = oip->ip_hl << 2;
 222         register struct icmp *icp;
 223         register struct mbuf *m;
 224         unsigned icmplen, icmpelen, nlen;
 225
 226         KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type", __func__));
 227 #ifdef ICMPPRINTFS
 228         if (icmpprintfs)
 229                 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
 230 #endif
 231         if (type != ICMP_REDIRECT)
 232                 ICMPSTAT_INC(icps_error);
 233         /*
 234          * Don't send error:
 235          *  if the original packet was encrypted.
 236          *  if not the first fragment of message.
 237          *  in response to a multicast or broadcast packet.
 238          *  if the old packet protocol was an ICMP error message.
 239          */
 240         if (n->m_flags & M_DECRYPTED)
 241                 goto freeit;
 242         if (oip->ip_off & htons(~(IP_MF|IP_DF)))
 243                 goto freeit;
 244         if (n->m_flags & (M_BCAST|M_MCAST))
 245                 goto freeit;
 246         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
 247           n->m_len >= oiphlen + ICMP_MINLEN &&
 248           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiphlen))->icmp_type)) {
 249                 ICMPSTAT_INC(icps_oldicmp);
 250                 goto freeit;
 251         }
 252         /* Drop if IP header plus 8 bytes is not contignous in first mbuf. */
 253         if (oiphlen + 8 > n->m_len)
 254                 goto freeit;
 255         /*
 256          * Calculate length to quote from original packet and
 257          * prevent the ICMP mbuf from overflowing.
 258          * Unfortunatly this is non-trivial since ip_forward()
 259          * sends us truncated packets.
 260          */
 261         nlen = m_length(n, NULL);
 262         if (oip->ip_p == IPPROTO_TCP) {
 263                 struct tcphdr *th;
 264                 int tcphlen;
 265
 266                 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
 267                     n->m_next == NULL)
 268                         goto stdreply;
 269                 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
 270                     ((n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL))
 271                         goto freeit;
 272                 th = (struct tcphdr *)((caddr_t)oip + oiphlen);
 273                 tcphlen = th->th_off << 2;
 274                 if (tcphlen < sizeof(struct tcphdr))
 275                         goto freeit;
 276                 if (ntohs(oip->ip_len) < oiphlen + tcphlen)
 277                         goto freeit;
 278                 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
 279                         goto stdreply;
 280                 if (n->m_len < oiphlen + tcphlen &&
 281                     ((n = m_pullup(n, oiphlen + tcphlen)) == NULL))
 282                         goto freeit;
 283                 icmpelen = max(tcphlen, min(V_icmp_quotelen,
 284                     ntohs(oip->ip_len) - oiphlen));
 285         } else
 286 stdreply:       icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
 287
 288         icmplen = min(oiphlen + icmpelen, nlen);
 289         if (icmplen < sizeof(struct ip))
 290                 goto freeit;
 291
 292         if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
 293                 m = m_gethdr(M_NOWAIT, MT_DATA);
 294         else
 295                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
 296         if (m == NULL)
 297                 goto freeit;
 298 #ifdef MAC
 299         mac_netinet_icmp_reply(n, m);
 300 #endif
 301         icmplen = min(icmplen, M_TRAILINGSPACE(m) - sizeof(struct ip) - ICMP_MINLEN);
 302         m_align(m, ICMP_MINLEN + icmplen);
 303         m->m_len = ICMP_MINLEN + icmplen;
 304
 305         /* XXX MRT  make the outgoing packet use the same FIB
 306          * that was associated with the incoming packet
 307          */
 308         M_SETFIB(m, M_GETFIB(n));
 309         icp = mtod(m, struct icmp *);
 310         ICMPSTAT_INC(icps_outhist[type]);
 311         icp->icmp_type = type;
 312         if (type == ICMP_REDIRECT)
 313                 icp->icmp_gwaddr.s_addr = dest;
 314         else {
 315                 icp->icmp_void = 0;
 316                 /*
 317                  * The following assignments assume an overlay with the
 318                  * just zeroed icmp_void field.
 319                  */
 320                 if (type == ICMP_PARAMPROB) {
 321                         icp->icmp_pptr = code;
 322                         code = 0;
 323                 } else if (type == ICMP_UNREACH &&
 324                         code == ICMP_UNREACH_NEEDFRAG && mtu) {
 325                         icp->icmp_nextmtu = htons(mtu);
 326                 }
 327         }
 328         icp->icmp_code = code;
 329
 330         /*
 331          * Copy the quotation into ICMP message and
 332          * convert quoted IP header back to network representation.
 333          */
 334         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
 335         nip = &icp->icmp_ip;
 336
 337         /*
 338          * Set up ICMP message mbuf and copy old IP header (without options
 339          * in front of ICMP message.
 340          * If the original mbuf was meant to bypass the firewall, the error
 341          * reply should bypass as well.
 342          */
 343         m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
 344         m->m_data -= sizeof(struct ip);
 345         m->m_len += sizeof(struct ip);
 346         m->m_pkthdr.len = m->m_len;
 347         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
 348         nip = mtod(m, struct ip *);
 349         bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
 350         nip->ip_len = htons(m->m_len);
 351         nip->ip_v = IPVERSION;
 352         nip->ip_hl = 5;
 353         nip->ip_p = IPPROTO_ICMP;
 354         nip->ip_tos = 0;
 355         nip->ip_off = 0;
 356         icmp_reflect(m);
 357
 358 freeit:
 359         m_freem(n);
 360 }
 361
 362 /*
 363  * Process a received ICMP message.
 364  */
 365 int
 366 icmp_input(struct mbuf **mp, int *offp, int proto)
 367 {
 368         struct icmp *icp;
 369         struct in_ifaddr *ia;
 370         struct mbuf *m = *mp;
 371         struct ip *ip = mtod(m, struct ip *);
 372         struct sockaddr_in icmpsrc, icmpdst, icmpgw;
 373         int hlen = *offp;
 374         int icmplen = ntohs(ip->ip_len) - *offp;
 375         int i, code;
 376         void (*ctlfunc)(int, struct sockaddr *, void *);
 377         int fibnum;
 378
 379         *mp = NULL;
 380
 381         /*
 382          * Locate icmp structure in mbuf, and check
 383          * that not corrupted and of at least minimum length.
 384          */
 385 #ifdef ICMPPRINTFS
 386         if (icmpprintfs) {
 387                 char buf[4 * sizeof "123"];
 388                 strcpy(buf, inet_ntoa(ip->ip_src));
 389                 printf("icmp_input from %s to %s, len %d\n",
 390                        buf, inet_ntoa(ip->ip_dst), icmplen);
 391         }
 392 #endif
 393         if (icmplen < ICMP_MINLEN) {
 394                 ICMPSTAT_INC(icps_tooshort);
 395                 goto freeit;
 396         }
 397         i = hlen + min(icmplen, ICMP_ADVLENMIN);
 398         if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
 399                 ICMPSTAT_INC(icps_tooshort);
 400                 return (IPPROTO_DONE);
 401         }
 402         ip = mtod(m, struct ip *);
 403         m->m_len -= hlen;
 404         m->m_data += hlen;
 405         icp = mtod(m, struct icmp *);
 406         if (in_cksum(m, icmplen)) {
 407                 ICMPSTAT_INC(icps_checksum);
 408                 goto freeit;
 409         }
 410         m->m_len += hlen;
 411         m->m_data -= hlen;
 412
 413 #ifdef ICMPPRINTFS
 414         if (icmpprintfs)
 415                 printf("icmp_input, type %d code %d\n", icp->icmp_type,
 416                     icp->icmp_code);
 417 #endif
 418
 419         /*
 420          * Message type specific processing.
 421          */
 422         if (icp->icmp_type > ICMP_MAXTYPE)
 423                 goto raw;
 424
 425         /* Initialize */
 426         bzero(&icmpsrc, sizeof(icmpsrc));
 427         icmpsrc.sin_len = sizeof(struct sockaddr_in);
 428         icmpsrc.sin_family = AF_INET;
 429         bzero(&icmpdst, sizeof(icmpdst));
 430         icmpdst.sin_len = sizeof(struct sockaddr_in);
 431         icmpdst.sin_family = AF_INET;
 432         bzero(&icmpgw, sizeof(icmpgw));
 433         icmpgw.sin_len = sizeof(struct sockaddr_in);
 434         icmpgw.sin_family = AF_INET;
 435
 436         ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
 437         code = icp->icmp_code;
 438         switch (icp->icmp_type) {
 439
 440         case ICMP_UNREACH:
 441                 switch (code) {
 442                         case ICMP_UNREACH_NET:
 443                         case ICMP_UNREACH_HOST:
 444                         case ICMP_UNREACH_SRCFAIL:
 445                         case ICMP_UNREACH_NET_UNKNOWN:
 446                         case ICMP_UNREACH_HOST_UNKNOWN:
 447                         case ICMP_UNREACH_ISOLATED:
 448                         case ICMP_UNREACH_TOSNET:
 449                         case ICMP_UNREACH_TOSHOST:
 450                         case ICMP_UNREACH_HOST_PRECEDENCE:
 451                         case ICMP_UNREACH_PRECEDENCE_CUTOFF:
 452                                 code = PRC_UNREACH_NET;
 453                                 break;
 454
 455                         case ICMP_UNREACH_NEEDFRAG:
 456                                 code = PRC_MSGSIZE;
 457                                 break;
 458
 459                         /*
 460                          * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
 461                          * Treat subcodes 2,3 as immediate RST
 462                          */
 463                         case ICMP_UNREACH_PROTOCOL:
 464                         case ICMP_UNREACH_PORT:
 465                                 code = PRC_UNREACH_PORT;
 466                                 break;
 467
 468                         case ICMP_UNREACH_NET_PROHIB:
 469                         case ICMP_UNREACH_HOST_PROHIB:
 470                         case ICMP_UNREACH_FILTER_PROHIB:
 471                                 code = PRC_UNREACH_ADMIN_PROHIB;
 472                                 break;
 473
 474                         default:
 475                                 goto badcode;
 476                 }
 477                 goto deliver;
 478
 479         case ICMP_TIMXCEED:
 480                 if (code > 1)
 481                         goto badcode;
 482                 code += PRC_TIMXCEED_INTRANS;
 483                 goto deliver;
 484
 485         case ICMP_PARAMPROB:
 486                 if (code > 1)
 487                         goto badcode;
 488                 code = PRC_PARAMPROB;
 489                 goto deliver;
 490
 491         case ICMP_SOURCEQUENCH:
 492                 if (code)
 493                         goto badcode;
 494                 code = PRC_QUENCH;
 495         deliver:
 496                 /*
 497                  * Problem with datagram; advise higher level routines.
 498                  */
 499                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 500                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
 501                         ICMPSTAT_INC(icps_badlen);
 502                         goto freeit;
 503                 }
 504                 /* Discard ICMP's in response to multicast packets */
 505                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
 506                         goto badcode;
 507 #ifdef ICMPPRINTFS
 508                 if (icmpprintfs)
 509                         printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
 510 #endif
 511                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 512                 /*
 513                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
 514                  * notification to TCP layer.
 515                  */
 516                 ctlfunc = inetsw[ip_protox[icp->icmp_ip.ip_p]].pr_ctlinput;
 517                 if (ctlfunc)
 518                         (*ctlfunc)(code, (struct sockaddr *)&icmpsrc,
 519                                    (void *)&icp->icmp_ip);
 520                 break;
 521
 522         badcode:
 523                 ICMPSTAT_INC(icps_badcode);
 524                 break;
 525
 526         case ICMP_ECHO:
 527                 if (!V_icmpbmcastecho
 528                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 529                         ICMPSTAT_INC(icps_bmcastecho);
 530                         break;
 531                 }
 532                 icp->icmp_type = ICMP_ECHOREPLY;
 533                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 534                         goto freeit;
 535                 else
 536                         goto reflect;
 537
 538         case ICMP_TSTAMP:
 539                 if (V_icmptstamprepl == 0)
 540                         break;
 541                 if (!V_icmpbmcastecho
 542                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 543                         ICMPSTAT_INC(icps_bmcasttstamp);
 544                         break;
 545                 }
 546                 if (icmplen < ICMP_TSLEN) {
 547                         ICMPSTAT_INC(icps_badlen);
 548                         break;
 549                 }
 550                 icp->icmp_type = ICMP_TSTAMPREPLY;
 551                 icp->icmp_rtime = iptime();
 552                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 553                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 554                         goto freeit;
 555                 else
 556                         goto reflect;
 557
 558         case ICMP_MASKREQ:
 559                 if (V_icmpmaskrepl == 0)
 560                         break;
 561                 /*
 562                  * We are not able to respond with all ones broadcast
 563                  * unless we receive it over a point-to-point interface.
 564                  */
 565                 if (icmplen < ICMP_MASKLEN)
 566                         break;
 567                 switch (ip->ip_dst.s_addr) {
 568
 569                 case INADDR_BROADCAST:
 570                 case INADDR_ANY:
 571                         icmpdst.sin_addr = ip->ip_src;
 572                         break;
 573
 574                 default:
 575                         icmpdst.sin_addr = ip->ip_dst;
 576                 }
 577                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 578                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 579                 if (ia == NULL)
 580                         break;
 581                 if (ia->ia_ifp == NULL) {
 582                         ifa_free(&ia->ia_ifa);
 583                         break;
 584                 }
 585                 icp->icmp_type = ICMP_MASKREPLY;
 586                 if (V_icmpmaskfake == 0)
 587                         icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 588                 else
 589                         icp->icmp_mask = V_icmpmaskfake;
 590                 if (ip->ip_src.s_addr == 0) {
 591                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 592                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 593                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 594                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 595                 }
 596                 ifa_free(&ia->ia_ifa);
 597 reflect:
 598                 ICMPSTAT_INC(icps_reflect);
 599                 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
 600                 icmp_reflect(m);
 601                 return (IPPROTO_DONE);
 602
 603         case ICMP_REDIRECT:
 604                 if (V_log_redirect) {
 605                         u_long src, dst, gw;
 606
 607                         src = ntohl(ip->ip_src.s_addr);
 608                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
 609                         gw = ntohl(icp->icmp_gwaddr.s_addr);
 610                         printf("icmp redirect from %d.%d.%d.%d: "
 611                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
 612                                (int)(src >> 24), (int)((src >> 16) & 0xff),
 613                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
 614                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
 615                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
 616                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
 617                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
 618                 }
 619                 /*
 620                  * RFC1812 says we must ignore ICMP redirects if we
 621                  * are acting as router.
 622                  */
 623                 if (V_drop_redirect || V_ipforwarding)
 624                         break;
 625                 if (code > 3)
 626                         goto badcode;
 627                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 628                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
 629                         ICMPSTAT_INC(icps_badlen);
 630                         break;
 631                 }
 632                 /*
 633                  * Short circuit routing redirects to force
 634                  * immediate change in the kernel's routing
 635                  * tables.  The message is also handed to anyone
 636                  * listening on a raw socket (e.g. the routing
 637                  * daemon for use in updating its tables).
 638                  */
 639                 icmpgw.sin_addr = ip->ip_src;
 640                 icmpdst.sin_addr = icp->icmp_gwaddr;
 641 #ifdef  ICMPPRINTFS
 642                 if (icmpprintfs) {
 643                         char buf[4 * sizeof "123"];
 644                         strcpy(buf, inet_ntoa(icp->icmp_ip.ip_dst));
 645
 646                         printf("redirect dst %s to %s\n",
 647                                buf, inet_ntoa(icp->icmp_gwaddr));
 648                 }
 649 #endif
 650                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 651                 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
 652                         in_rtredirect((struct sockaddr *)&icmpsrc,
 653                           (struct sockaddr *)&icmpdst,
 654                           (struct sockaddr *)0, RTF_GATEWAY | RTF_HOST,
 655                           (struct sockaddr *)&icmpgw, fibnum);
 656                 }
 657                 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 658 #ifdef IPSEC
 659                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 660 #endif
 661                 break;
 662
 663         /*
 664          * No kernel processing for the following;
 665          * just fall through to send to raw listener.
 666          */
 667         case ICMP_ECHOREPLY:
 668         case ICMP_ROUTERADVERT:
 669         case ICMP_ROUTERSOLICIT:
 670         case ICMP_TSTAMPREPLY:
 671         case ICMP_IREQREPLY:
 672         case ICMP_MASKREPLY:
 673         default:
 674                 break;
 675         }
 676
 677 raw:
 678         *mp = m;
 679         rip_input(mp, offp, proto);
 680         return (IPPROTO_DONE);
 681
 682 freeit:
 683         m_freem(m);
 684         return (IPPROTO_DONE);
 685 }
 686
 687 /*
 688  * Reflect the ip packet back to the source
 689  */
 690 static void
 691 icmp_reflect(struct mbuf *m)
 692 {
 693         struct ip *ip = mtod(m, struct ip *);
 694         struct ifaddr *ifa;
 695         struct ifnet *ifp;
 696         struct in_ifaddr *ia;
 697         struct in_addr t;
 698         struct mbuf *opts = 0;
 699         int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
 700
 701         if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
 702             IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) ||
 703             IN_ZERONET(ntohl(ip->ip_src.s_addr)) ) {
 704                 m_freem(m);     /* Bad return address */
 705                 ICMPSTAT_INC(icps_badaddr);
 706                 goto done;      /* Ip_output() will check for broadcast */
 707         }
 708
 709         t = ip->ip_dst;
 710         ip->ip_dst = ip->ip_src;
 711
 712         /*
 713          * Source selection for ICMP replies:
 714          *
 715          * If the incoming packet was addressed directly to one of our
 716          * own addresses, use dst as the src for the reply.
 717          */
 718         IN_IFADDR_RLOCK();
 719         LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
 720                 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
 721                         t = IA_SIN(ia)->sin_addr;
 722                         IN_IFADDR_RUNLOCK();
 723                         goto match;
 724                 }
 725         }
 726         IN_IFADDR_RUNLOCK();
 727
 728         /*
 729          * If the incoming packet was addressed to one of our broadcast
 730          * addresses, use the first non-broadcast address which corresponds
 731          * to the incoming interface.
 732          */
 733         ifp = m->m_pkthdr.rcvif;
 734         if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
 735                 IF_ADDR_RLOCK(ifp);
 736                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 737                         if (ifa->ifa_addr->sa_family != AF_INET)
 738                                 continue;
 739                         ia = ifatoia(ifa);
 740                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
 741                             t.s_addr) {
 742                                 t = IA_SIN(ia)->sin_addr;
 743                                 IF_ADDR_RUNLOCK(ifp);
 744                                 goto match;
 745                         }
 746                 }
 747                 IF_ADDR_RUNLOCK(ifp);
 748         }
 749         /*
 750          * If the packet was transiting through us, use the address of
 751          * the interface the packet came through in.  If that interface
 752          * doesn't have a suitable IP address, the normal selection
 753          * criteria apply.
 754          */
 755         if (V_icmp_rfi && ifp != NULL) {
 756                 IF_ADDR_RLOCK(ifp);
 757                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 758                         if (ifa->ifa_addr->sa_family != AF_INET)
 759                                 continue;
 760                         ia = ifatoia(ifa);
 761                         t = IA_SIN(ia)->sin_addr;
 762                         IF_ADDR_RUNLOCK(ifp);
 763                         goto match;
 764                 }
 765                 IF_ADDR_RUNLOCK(ifp);
 766         }
 767         /*
 768          * If the incoming packet was not addressed directly to us, use
 769          * designated interface for icmp replies specified by sysctl
 770          * net.inet.icmp.reply_src (default not set). Otherwise continue
 771          * with normal source selection.
 772          */
 773         if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
 774                 IF_ADDR_RLOCK(ifp);
 775                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 776                         if (ifa->ifa_addr->sa_family != AF_INET)
 777                                 continue;
 778                         ia = ifatoia(ifa);
 779                         t = IA_SIN(ia)->sin_addr;
 780                         IF_ADDR_RUNLOCK(ifp);
 781                         goto match;
 782                 }
 783                 IF_ADDR_RUNLOCK(ifp);
 784         }
 785         /*
 786          * If the packet was transiting through us, use the address of
 787          * the interface that is the closest to the packet source.
 788          * When we don't have a route back to the packet source, stop here
 789          * and drop the packet.
 790          */
 791         ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
 792         if (ia == NULL) {
 793                 m_freem(m);
 794                 ICMPSTAT_INC(icps_noroute);
 795                 goto done;
 796         }
 797         t = IA_SIN(ia)->sin_addr;
 798         ifa_free(&ia->ia_ifa);
 799 match:
 800 #ifdef MAC
 801         mac_netinet_icmp_replyinplace(m);
 802 #endif
 803         ip->ip_src = t;
 804         ip->ip_ttl = V_ip_defttl;
 805
 806         if (optlen > 0) {
 807                 register u_char *cp;
 808                 int opt, cnt;
 809                 u_int len;
 810
 811                 /*
 812                  * Retrieve any source routing from the incoming packet;
 813                  * add on any record-route or timestamp options.
 814                  */
 815                 cp = (u_char *) (ip + 1);
 816                 if ((opts = ip_srcroute(m)) == 0 &&
 817                     (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
 818                         opts->m_len = sizeof(struct in_addr);
 819                         mtod(opts, struct in_addr *)->s_addr = 0;
 820                 }
 821                 if (opts) {
 822 #ifdef ICMPPRINTFS
 823                     if (icmpprintfs)
 824                             printf("icmp_reflect optlen %d rt %d => ",
 825                                 optlen, opts->m_len);
 826 #endif
 827                     for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
 828                             opt = cp[IPOPT_OPTVAL];
 829                             if (opt == IPOPT_EOL)
 830                                     break;
 831                             if (opt == IPOPT_NOP)
 832                                     len = 1;
 833                             else {
 834                                     if (cnt < IPOPT_OLEN + sizeof(*cp))
 835                                             break;
 836                                     len = cp[IPOPT_OLEN];
 837                                     if (len < IPOPT_OLEN + sizeof(*cp) ||
 838                                         len > cnt)
 839                                             break;
 840                             }
 841                             /*
 842                              * Should check for overflow, but it "can't happen"
 843                              */
 844                             if (opt == IPOPT_RR || opt == IPOPT_TS ||
 845                                 opt == IPOPT_SECURITY) {
 846                                     bcopy((caddr_t)cp,
 847                                         mtod(opts, caddr_t) + opts->m_len, len);
 848                                     opts->m_len += len;
 849                             }
 850                     }
 851                     /* Terminate & pad, if necessary */
 852                     cnt = opts->m_len % 4;
 853                     if (cnt) {
 854                             for (; cnt < 4; cnt++) {
 855                                     *(mtod(opts, caddr_t) + opts->m_len) =
 856                                         IPOPT_EOL;
 857                                     opts->m_len++;
 858                             }
 859                     }
 860 #ifdef ICMPPRINTFS
 861                     if (icmpprintfs)
 862                             printf("%d\n", opts->m_len);
 863 #endif
 864                 }
 865                 ip_stripoptions(m);
 866         }
 867         m_tag_delete_nonpersistent(m);
 868         m->m_flags &= ~(M_BCAST|M_MCAST);
 869         icmp_send(m, opts);
 870 done:
 871         if (opts)
 872                 (void)m_free(opts);
 873 }
 874
 875 /*
 876  * Send an icmp packet back to the ip level,
 877  * after supplying a checksum.
 878  */
 879 static void
 880 icmp_send(struct mbuf *m, struct mbuf *opts)
 881 {
 882         register struct ip *ip = mtod(m, struct ip *);
 883         register int hlen;
 884         register struct icmp *icp;
 885
 886         hlen = ip->ip_hl << 2;
 887         m->m_data += hlen;
 888         m->m_len -= hlen;
 889         icp = mtod(m, struct icmp *);
 890         icp->icmp_cksum = 0;
 891         icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
 892         m->m_data -= hlen;
 893         m->m_len += hlen;
 894         m->m_pkthdr.rcvif = (struct ifnet *)0;
 895 #ifdef ICMPPRINTFS
 896         if (icmpprintfs) {
 897                 char buf[4 * sizeof "123"];
 898                 strcpy(buf, inet_ntoa(ip->ip_dst));
 899                 printf("icmp_send dst %s src %s\n",
 900                        buf, inet_ntoa(ip->ip_src));
 901         }
 902 #endif
 903         (void) ip_output(m, opts, NULL, 0, NULL, NULL);
 904 }
 905
 906 /*
 907  * Return milliseconds since 00:00 GMT in network format.
 908  */
 909 uint32_t
 910 iptime(void)
 911 {
 912         struct timeval atv;
 913         u_long t;
 914
 915         getmicrotime(&atv);
 916         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 917         return (htonl(t));
 918 }
 919
 920 /*
 921  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 922  * given current value MTU.  If DIR is less than zero, a larger plateau
 923  * is returned; otherwise, a smaller value is returned.
 924  */
 925 int
 926 ip_next_mtu(int mtu, int dir)
 927 {
 928         static int mtutab[] = {
 929                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
 930                 296, 68, 0
 931         };
 932         int i, size;
 933
 934         size = (sizeof mtutab) / (sizeof mtutab[0]);
 935         if (dir >= 0) {
 936                 for (i = 0; i < size; i++)
 937                         if (mtu > mtutab[i])
 938                                 return mtutab[i];
 939         } else {
 940                 for (i = size - 1; i >= 0; i--)
 941                         if (mtu < mtutab[i])
 942                                 return mtutab[i];
 943                 if (mtu == mtutab[0])
 944                         return mtutab[0];
 945         }
 946         return 0;
 947 }
 948 #endif /* INET */
 949
 950
 951 /*
 952  * badport_bandlim() - check for ICMP bandwidth limit
 953  *
 954  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 955  *      hit our bandwidth limit and it is not ok.
 956  *
 957  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 958  *
 959  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 960  *      values.  We may eventually remove this separation (and simplify the
 961  *      code further).
 962  *
 963  *      Note that the printing of the error message is delayed so we can
 964  *      properly print the icmp error rate that the system was trying to do
 965  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 966  *      the 'final' error, but it doesn't make sense to solve the printing
 967  *      delay with more complex code.
 968  */
 969
 970 int
 971 badport_bandlim(int which)
 972 {
 973
 974 #define N(a)    (sizeof (a) / sizeof (a[0]))
 975         static struct rate {
 976                 const char      *type;
 977                 struct timeval  lasttime;
 978                 int             curpps;
 979         } rates[BANDLIM_MAX+1] = {
 980                 { "icmp unreach response" },
 981                 { "icmp ping response" },
 982                 { "icmp tstamp response" },
 983                 { "closed port RST response" },
 984                 { "open port RST response" },
 985                 { "icmp6 unreach response" },
 986                 { "sctp ootb response" }
 987         };
 988
 989         /*
 990          * Return ok status if feature disabled or argument out of range.
 991          */
 992         if (V_icmplim > 0 && (u_int) which < N(rates)) {
 993                 struct rate *r = &rates[which];
 994                 int opps = r->curpps;
 995
 996                 if (!ppsratecheck(&r->lasttime, &r->curpps, V_icmplim))
 997                         return -1;      /* discard packet */
 998                 /*
 999                  * If we've dropped below the threshold after having
1000                  * rate-limited traffic print the message.  This preserves
1001                  * the previous behaviour at the expense of added complexity.
1002                  */
1003                 if (V_icmplim_output && opps > V_icmplim)
1004                         log(LOG_NOTICE, "Limiting %s from %d to %d packets/sec\n",
1005                                 r->type, opps, V_icmplim);
1006         }
1007         return 0;                       /* okay to send packet */
1008 #undef N
1009 }