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. 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  * $FreeBSD: src/sys/netinet/ip_icmp.c,v 1.39.2.19 2003/01/24 05:11:34 sam Exp $
  31  */
  32
  33 #include "opt_ipsec.h"
  34
  35 #include <sys/param.h>
  36 #include <sys/systm.h>
  37 #include <sys/mbuf.h>
  38 #include <sys/protosw.h>
  39 #include <sys/socket.h>
  40 #include <sys/socketops.h>
  41 #include <sys/time.h>
  42 #include <sys/kernel.h>
  43 #include <sys/sysctl.h>
  44 #include <sys/in_cksum.h>
  45
  46 #include <machine/stdarg.h>
  47
  48 #include <net/if.h>
  49 #include <net/if_types.h>
  50 #include <net/netisr2.h>
  51 #include <net/netmsg2.h>
  52 #include <net/route.h>
  53
  54 #define _IP_VHL
  55 #include <netinet/in.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/icmp_var.h>
  62
  63 #ifdef IPSEC
  64 #include <netinet6/ipsec.h>
  65 #include <netproto/key/key.h>
  66 #endif
  67
  68 #ifdef FAST_IPSEC
  69 #include <netproto/ipsec/ipsec.h>
  70 #include <netproto/ipsec/key.h>
  71 #define IPSEC
  72 #endif
  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
  80 struct icmpstat icmpstat;
  81 SYSCTL_STRUCT(_net_inet_icmp, ICMPCTL_STATS, stats, CTLFLAG_RW,
  82         &icmpstat, icmpstat, "ICMP statistics");
  83
  84 static int      icmpmaskrepl = 0;
  85 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_RW,
  86         &icmpmaskrepl, 0, "Allow replies to netmask requests");
  87
  88 static int      drop_redirect = 0;
  89 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_RW,
  90         &drop_redirect, 0, "Ignore ICMP redirects");
  91
  92 static int      log_redirect = 0;
  93 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_RW,
  94         &log_redirect, 0, "Enable output about ICMP redirects");
  95
  96 static int      discard_sourcequench = 1;
  97 SYSCTL_INT(_net_inet_icmp, OID_AUTO, discard_sourcequench, CTLFLAG_RW,
  98         &discard_sourcequench, 0, "Discard ICMP Source Quench");
  99
 100 #ifdef ICMP_BANDLIM
 101
 102 /*
 103  * ICMP error-response bandwidth limiting sysctl.  If not enabled, sysctl
 104  *      variable content is -1 and read-only.
 105  */
 106
 107 static int      icmplim = 200;
 108 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RW,
 109         &icmplim, 0, "ICMP bandwidth limit");
 110 #else
 111
 112 static int      icmplim = -1;
 113 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_RD,
 114         &icmplim, 0, "ICMP bandwidth limit");
 115
 116 #endif
 117
 118 static int      icmplim_output = 0;
 119 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_RW,
 120         &icmplim_output, 0, "Enable output about ICMP bandwidth limits");
 121
 122 /*
 123  * ICMP broadcast echo sysctl
 124  */
 125
 126 static int      icmpbmcastecho = 0;
 127 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_RW,
 128     &icmpbmcastecho, 0, "");
 129
 130 static char     icmp_reply_src[IFNAMSIZ];
 131 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_RW,
 132         icmp_reply_src, IFNAMSIZ, "icmp reply source for non-local packets.");
 133
 134 static int      icmp_rfi;
 135 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_RW,
 136         &icmp_rfi, 0, "ICMP reply from incoming interface for "
 137         "non-local packets");
 138
 139 #ifdef ICMPPRINTFS
 140 static int      icmpprintfs = 0;
 141 SYSCTL_INT(_net_inet_icmp, OID_AUTO, debug_prints, CTLFLAG_RW,
 142         &icmpprintfs, 0, "extra ICMP debug prints");
 143 #endif
 144
 145 static void     icmp_reflect (struct mbuf *);
 146 static void     icmp_send (struct mbuf *, struct mbuf *, struct route *);
 147
 148 extern  struct protosw inetsw[];
 149
 150 /*
 151  * Generate an error packet of type error
 152  * in response to bad packet ip.
 153  */
 154 void
 155 icmp_error(struct mbuf *n, int type, int code, n_long dest, int destmtu)
 156 {
 157         struct ip *oip = mtod(n, struct ip *), *nip;
 158         unsigned oiplen = IP_VHL_HL(oip->ip_vhl) << 2;
 159         struct icmp *icp;
 160         struct mbuf *m;
 161         unsigned icmplen;
 162
 163 #ifdef ICMPPRINTFS
 164         if (icmpprintfs)
 165                 kprintf("icmp_error(%p, %d, %d)\n", oip, type, code);
 166 #endif
 167         if (type != ICMP_REDIRECT)
 168                 icmpstat.icps_error++;
 169         /*
 170          * Don't send error if the original packet was encrypted.
 171          * Don't send error if not the first fragment of message.
 172          * Don't error if the old packet protocol was ICMP
 173          * error message, only known informational types.
 174          */
 175         if (n->m_flags & M_DECRYPTED)
 176                 goto freeit;
 177         if (oip->ip_off &~ (IP_MF|IP_DF))
 178                 goto freeit;
 179         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
 180           n->m_len >= oiplen + ICMP_MINLEN &&
 181           !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip + oiplen))->icmp_type)) {
 182                 icmpstat.icps_oldicmp++;
 183                 goto freeit;
 184         }
 185         /* Don't send error in response to a multicast or broadcast packet */
 186         if (n->m_flags & (M_BCAST|M_MCAST))
 187                 goto freeit;
 188         /*
 189          * First, formulate icmp message
 190          */
 191         m = m_gethdr(M_NOWAIT, MT_HEADER);
 192         if (m == NULL)
 193                 goto freeit;
 194         icmplen = min(oiplen + 8, oip->ip_len);
 195         if (icmplen < sizeof(struct ip))
 196                 panic("icmp_error: bad length");
 197         m->m_len = icmplen + ICMP_MINLEN;
 198         MH_ALIGN(m, m->m_len);
 199         icp = mtod(m, struct icmp *);
 200         if ((u_int)type > ICMP_MAXTYPE)
 201                 panic("icmp_error");
 202         icmpstat.icps_outhist[type]++;
 203         icp->icmp_type = type;
 204         if (type == ICMP_REDIRECT)
 205                 icp->icmp_gwaddr.s_addr = dest;
 206         else {
 207                 icp->icmp_void = 0;
 208                 /*
 209                  * The following assignments assume an overlay with the
 210                  * zeroed icmp_void field.
 211                  */
 212                 if (type == ICMP_PARAMPROB) {
 213                         icp->icmp_pptr = code;
 214                         code = 0;
 215                 } else if (type == ICMP_UNREACH &&
 216                         code == ICMP_UNREACH_NEEDFRAG && destmtu) {
 217                         icp->icmp_nextmtu = htons(destmtu);
 218                 }
 219         }
 220
 221         icp->icmp_code = code;
 222         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
 223         nip = &icp->icmp_ip;
 224
 225         /*
 226          * Convert fields to network representation.
 227          */
 228         nip->ip_len = htons(nip->ip_len);
 229         nip->ip_off = htons(nip->ip_off);
 230
 231         /*
 232          * Now, copy old ip header (without options)
 233          * in front of icmp message.
 234          */
 235         if (m->m_data - sizeof(struct ip) < m->m_pktdat)
 236                 panic("icmp len");
 237         m->m_data -= sizeof(struct ip);
 238         m->m_len += sizeof(struct ip);
 239         m->m_pkthdr.len = m->m_len;
 240         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
 241         nip = mtod(m, struct ip *);
 242         bcopy(oip, nip, sizeof(struct ip));
 243         nip->ip_len = m->m_len;
 244         nip->ip_vhl = IP_VHL_BORING;
 245         nip->ip_p = IPPROTO_ICMP;
 246         nip->ip_tos = 0;
 247         m->m_pkthdr.fw_flags |= n->m_pkthdr.fw_flags & FW_MBUF_GENERATED;
 248         icmp_reflect(m);
 249
 250 freeit:
 251         m_freem(n);
 252 }
 253
 254 static void
 255 icmp_ctlinput_done_handler(netmsg_t nmsg)
 256 {
 257         struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
 258         struct mbuf *m = msg->m;
 259         int hlen = msg->hlen;
 260
 261         rip_input(&m, &hlen, msg->proto);
 262 }
 263
 264 static void
 265 icmp_ctlinput_done(struct mbuf *m)
 266 {
 267         struct netmsg_ctlinput *msg = &m->m_hdr.mh_ctlmsg;
 268
 269         netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
 270             icmp_ctlinput_done_handler);
 271         lwkt_sendmsg(netisr_cpuport(0), &msg->base.lmsg);
 272 }
 273
 274 static void
 275 icmp_mtudisc(struct mbuf *m, int hlen)
 276 {
 277         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 278         struct rtentry *rt;
 279         struct icmp *icp;
 280
 281         KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
 282
 283         icp = mtodoff(m, struct icmp *, hlen);
 284         icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 285
 286         /*
 287          * MTU discovery:
 288          * If we got a needfrag and there is a host route to the original
 289          * destination, and the MTU is not locked, then set the MTU in the
 290          * route to the suggested new value (if given) and then notify as
 291          * usual.  The ULPs will notice that the MTU has changed and adapt
 292          * accordingly.  If no new MTU was suggested, then we guess a new
 293          * one less than the current value.  If the new MTU is unreasonably
 294          * small (arbitrarily set at 296), then we reset the MTU to the
 295          * interface value and enable the lock bit, indicating that we are
 296          * no longer doing MTU discovery.
 297          */
 298         rt = rtpurelookup((struct sockaddr *)&icmpsrc);
 299         if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
 300             !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
 301 #ifdef DEBUG_MTUDISC
 302                 char src_buf[INET_ADDRSTRLEN];
 303 #endif
 304                 int mtu;
 305
 306                 mtu = ntohs(icp->icmp_nextmtu);
 307                 if (!mtu)
 308                         mtu = ip_next_mtu(rt->rt_rmx.rmx_mtu, 1);
 309 #ifdef DEBUG_MTUDISC
 310                 kprintf("MTU for %s reduced to %d\n",
 311                     inet_ntop(AF_INET, &icmpsrc.sin_addr,
 312                         src_buf, INET_ADDRSTRLEN), mtu);
 313 #endif
 314                 if (mtu < 296) {
 315                         /* rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu; */
 316                         rt->rt_rmx.rmx_locks |= RTV_MTU;
 317                 } else if (rt->rt_rmx.rmx_mtu > mtu) {
 318                         rt->rt_rmx.rmx_mtu = mtu;
 319                 }
 320         }
 321         if (rt != NULL)
 322                 --rt->rt_refcnt;
 323
 324         /*
 325          * XXX if the packet contains [IPv4 AH TCP], we can't make a
 326          * notification to TCP layer.
 327          */
 328         so_pr_ctlinput_direct(&inetsw[ip_protox[icp->icmp_ip.ip_p]],
 329             PRC_MSGSIZE, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
 330 }
 331
 332 static void
 333 icmp_mtudisc_handler(netmsg_t nmsg)
 334 {
 335         struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
 336         int nextcpu;
 337
 338         icmp_mtudisc(msg->m, msg->hlen);
 339
 340         nextcpu = mycpuid + 1;
 341         if (nextcpu < ncpus)
 342                 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg);
 343         else
 344                 icmp_ctlinput_done(msg->m);
 345 }
 346
 347 static boolean_t
 348 icmp_mtudisc_start(struct mbuf *m, int hlen, int proto)
 349 {
 350         struct netmsg_ctlinput *msg;
 351
 352         ASSERT_IN_NETISR(0);
 353
 354         icmp_mtudisc(m, hlen);
 355
 356         if (ncpus == 1) {
 357                 /* There is only one netisr; done */
 358                 return FALSE;
 359         }
 360
 361         msg = &m->m_hdr.mh_ctlmsg;
 362         netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
 363             icmp_mtudisc_handler);
 364         msg->m = m;
 365         msg->cmd = PRC_MSGSIZE;
 366         msg->hlen = hlen;
 367         msg->proto = proto;
 368
 369         lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg);
 370         return TRUE;
 371 }
 372
 373 static void
 374 icmp_ctlinput(struct mbuf *m, int cmd, int hlen)
 375 {
 376         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 377         struct icmp *icp;
 378
 379         KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
 380
 381         icp = mtodoff(m, struct icmp *, hlen);
 382         icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 383
 384         /*
 385          * XXX if the packet contains [IPv4 AH TCP], we can't make a
 386          * notification to TCP layer.
 387          */
 388         so_pr_ctlinput_direct(&inetsw[ip_protox[icp->icmp_ip.ip_p]],
 389             cmd, (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
 390 }
 391
 392 static void
 393 icmp_ctlinput_handler(netmsg_t nmsg)
 394 {
 395         struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
 396
 397         icmp_ctlinput(msg->m, msg->cmd, msg->hlen);
 398         icmp_ctlinput_done(msg->m);
 399 }
 400
 401 static void
 402 icmp_ctlinput_start(struct mbuf *m, struct lwkt_port *port,
 403     int cmd, int hlen, int proto)
 404 {
 405         struct netmsg_ctlinput *msg;
 406
 407         KASSERT(&curthread->td_msgport != port,
 408             ("send icmp ctlinput to the current netisr"));
 409
 410         msg = &m->m_hdr.mh_ctlmsg;
 411         netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
 412             icmp_ctlinput_handler);
 413         msg->m = m;
 414         msg->cmd = cmd;
 415         msg->hlen = hlen;
 416         msg->proto = proto;
 417
 418         lwkt_sendmsg(port, &msg->base.lmsg);
 419 }
 420
 421 static void
 422 icmp_ctlinput_global_handler(netmsg_t nmsg)
 423 {
 424         struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
 425         int nextcpu;
 426
 427         icmp_ctlinput(msg->m, msg->cmd, msg->hlen);
 428
 429         nextcpu = mycpuid + 1;
 430         if (nextcpu < ncpus)
 431                 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg);
 432         else
 433                 icmp_ctlinput_done(msg->m);
 434 }
 435
 436 static void
 437 icmp_ctlinput_global_start(struct mbuf *m, int cmd, int hlen, int proto)
 438 {
 439         struct netmsg_ctlinput *msg;
 440
 441         ASSERT_IN_NETISR(0);
 442         KASSERT(ncpus > 1, ("there is only 1 cpu"));
 443
 444         icmp_ctlinput(m, cmd, hlen);
 445
 446         msg = &m->m_hdr.mh_ctlmsg;
 447         netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
 448             icmp_ctlinput_global_handler);
 449         msg->m = m;
 450         msg->cmd = cmd;
 451         msg->hlen = hlen;
 452         msg->proto = proto;
 453
 454         lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg);
 455 }
 456
 457 #define ICMP_RTREDIRECT_FLAGS   (RTF_GATEWAY | RTF_HOST)
 458
 459 static void
 460 icmp_redirect(struct mbuf *m, int hlen, boolean_t prt)
 461 {
 462         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 463         struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
 464         struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
 465         struct icmp *icp;
 466         struct ip *ip;
 467
 468         KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
 469
 470         ip = mtod(m, struct ip *);
 471         icp = mtodoff(m, struct icmp *, hlen);
 472
 473         /*
 474          * Short circuit routing redirects to force immediate change
 475          * in the kernel's routing tables.  The message is also handed
 476          * to anyone listening on a raw socket (e.g. the routing daemon
 477          * for use in updating its tables).
 478          */
 479 #ifdef ICMPPRINTFS
 480         if (icmpprintfs && prt) {
 481                 char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN];
 482
 483                 kprintf("redirect dst %s to %s\n",
 484                     inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
 485                         dst_buf, INET_ADDRSTRLEN),
 486                     inet_ntop(AF_INET, &icp->icmp_gwaddr,
 487                         gw_buf, INET_ADDRSTRLEN));
 488         }
 489 #endif
 490         icmpgw.sin_addr = ip->ip_src;
 491         icmpdst.sin_addr = icp->icmp_gwaddr;
 492         icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 493         rtredirect_oncpu((struct sockaddr *)&icmpsrc,
 494             (struct sockaddr *)&icmpdst, NULL, ICMP_RTREDIRECT_FLAGS,
 495             (struct sockaddr *)&icmpgw);
 496         kpfctlinput_direct(PRC_REDIRECT_HOST, (struct sockaddr *)&icmpsrc);
 497 }
 498
 499 static void
 500 icmp_redirect_done_handler(netmsg_t nmsg)
 501 {
 502         struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
 503         struct mbuf *m = msg->m;
 504         int hlen = msg->hlen;
 505 #ifdef IPSEC
 506         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 507         struct icmp *icp = mtodoff(m, struct icmp *, hlen);;
 508
 509         icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 510         key_sa_routechange((struct sockaddr *)&icmpsrc);
 511 #endif
 512         rip_input(&m, &hlen, msg->proto);
 513 }
 514
 515 static void
 516 icmp_redirect_done(struct mbuf *m, int hlen, boolean_t dispatch_rip)
 517 {
 518         struct rt_addrinfo rtinfo;
 519         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 520         struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
 521         struct sockaddr_in icmpgw = { sizeof(struct sockaddr_in), AF_INET };
 522         struct icmp *icp;
 523         struct ip *ip;
 524
 525         ip = mtod(m, struct ip *);
 526         icp = mtodoff(m, struct icmp *, hlen);
 527
 528         icmpgw.sin_addr = ip->ip_src;
 529         icmpdst.sin_addr = icp->icmp_gwaddr;
 530         icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 531
 532         bzero(&rtinfo, sizeof(struct rt_addrinfo));
 533         rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&icmpsrc;
 534         rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&icmpdst;
 535         rtinfo.rti_info[RTAX_NETMASK] = NULL;
 536         rtinfo.rti_info[RTAX_AUTHOR] = (struct sockaddr *)&icmpgw;
 537         rt_missmsg(RTM_REDIRECT, &rtinfo, ICMP_RTREDIRECT_FLAGS, 0);
 538
 539         if (dispatch_rip) {
 540                 struct netmsg_ctlinput *msg = &m->m_hdr.mh_ctlmsg;
 541
 542                 netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
 543                     icmp_redirect_done_handler);
 544                 lwkt_sendmsg(netisr_cpuport(0), &msg->base.lmsg);
 545         }
 546 }
 547
 548 static void
 549 icmp_redirect_handler(netmsg_t nmsg)
 550 {
 551         struct netmsg_ctlinput *msg = (struct netmsg_ctlinput *)nmsg;
 552         int nextcpu;
 553
 554         icmp_redirect(msg->m, msg->hlen, FALSE);
 555
 556         nextcpu = mycpuid + 1;
 557         if (nextcpu < ncpus)
 558                 lwkt_forwardmsg(netisr_cpuport(nextcpu), &msg->base.lmsg);
 559         else
 560                 icmp_redirect_done(msg->m, msg->hlen, TRUE);
 561 }
 562
 563 static boolean_t
 564 icmp_redirect_start(struct mbuf *m, int hlen, int proto)
 565 {
 566         struct netmsg_ctlinput *msg;
 567
 568         ASSERT_IN_NETISR(0);
 569
 570         icmp_redirect(m, hlen, TRUE);
 571
 572         if (ncpus == 1) {
 573                 /* There is only one netisr; done */
 574                 icmp_redirect_done(m, hlen, FALSE);
 575                 return FALSE;
 576         }
 577
 578         msg = &m->m_hdr.mh_ctlmsg;
 579         netmsg_init(&msg->base, NULL, &netisr_apanic_rport, 0,
 580             icmp_redirect_handler);
 581         msg->m = m;
 582         msg->cmd = PRC_REDIRECT_HOST;
 583         msg->hlen = hlen;
 584         msg->proto = proto;
 585
 586         lwkt_sendmsg(netisr_cpuport(1), &msg->base.lmsg);
 587         return TRUE;
 588 }
 589
 590 /*
 591  * Process a received ICMP message.
 592  */
 593 int
 594 icmp_input(struct mbuf **mp, int *offp, int proto)
 595 {
 596         struct sockaddr_in icmpsrc = { sizeof(struct sockaddr_in), AF_INET };
 597         struct sockaddr_in icmpdst = { sizeof(struct sockaddr_in), AF_INET };
 598         struct icmp *icp;
 599         struct in_ifaddr *ia;
 600         struct mbuf *m = *mp;
 601         struct ip *ip = mtod(m, struct ip *);
 602         int icmplen = ip->ip_len;
 603         int i, hlen;
 604         int code;
 605
 606         ASSERT_IN_NETISR(0);
 607
 608         *mp = NULL;
 609         hlen = *offp;
 610
 611         /*
 612          * Locate icmp structure in mbuf, and check
 613          * that not corrupted and of at least minimum length.
 614          */
 615 #ifdef ICMPPRINTFS
 616         if (icmpprintfs) {
 617                 char src_buf[INET_ADDRSTRLEN], dst_buf[INET_ADDRSTRLEN];
 618
 619                 kprintf("icmp_input from %s to %s, len %d\n",
 620                     inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN),
 621                     inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN),
 622                     icmplen);
 623         }
 624 #endif
 625         if (icmplen < ICMP_MINLEN) {
 626                 icmpstat.icps_tooshort++;
 627                 goto freeit;
 628         }
 629         i = hlen + min(icmplen, ICMP_ADVLENMIN);
 630         if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
 631                 icmpstat.icps_tooshort++;
 632                 return(IPPROTO_DONE);
 633         }
 634         ip = mtod(m, struct ip *);
 635
 636         if (in_cksum_skip(m, hlen + icmplen, hlen)) {
 637                 icmpstat.icps_checksum++;
 638                 goto freeit;
 639         }
 640         icp = (struct icmp *)((caddr_t)ip + hlen);
 641
 642         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
 643                 /*
 644                  * Deliver very specific ICMP type only.
 645                  */
 646                 switch (icp->icmp_type) {
 647                 case ICMP_UNREACH:
 648                 case ICMP_TIMXCEED:
 649                         break;
 650                 default:
 651                         goto freeit;
 652                 }
 653         }
 654
 655 #ifdef ICMPPRINTFS
 656         if (icmpprintfs)
 657                 kprintf("icmp_input, type %d code %d\n", icp->icmp_type,
 658                     icp->icmp_code);
 659 #endif
 660
 661         /*
 662          * Message type specific processing.
 663          */
 664         if (icp->icmp_type > ICMP_MAXTYPE)
 665                 goto raw;
 666         icmpstat.icps_inhist[icp->icmp_type]++;
 667         code = icp->icmp_code;
 668         switch (icp->icmp_type) {
 669
 670         case ICMP_UNREACH:
 671                 switch (code) {
 672                         case ICMP_UNREACH_NET:
 673                         case ICMP_UNREACH_HOST:
 674                         case ICMP_UNREACH_SRCFAIL:
 675                         case ICMP_UNREACH_NET_UNKNOWN:
 676                         case ICMP_UNREACH_HOST_UNKNOWN:
 677                         case ICMP_UNREACH_ISOLATED:
 678                         case ICMP_UNREACH_TOSNET:
 679                         case ICMP_UNREACH_TOSHOST:
 680                         case ICMP_UNREACH_HOST_PRECEDENCE:
 681                         case ICMP_UNREACH_PRECEDENCE_CUTOFF:
 682                                 code = PRC_UNREACH_NET;
 683                                 break;
 684
 685                         case ICMP_UNREACH_NEEDFRAG:
 686                                 code = PRC_MSGSIZE;
 687                                 break;
 688
 689                         /*
 690                          * RFC 1122, Sections 3.2.2.1 and 4.2.3.9.
 691                          * Treat subcodes 2,3 as immediate RST
 692                          */
 693                         case ICMP_UNREACH_PROTOCOL:
 694                         case ICMP_UNREACH_PORT:
 695                                 code = PRC_UNREACH_PORT;
 696                                 break;
 697
 698                         case ICMP_UNREACH_NET_PROHIB:
 699                         case ICMP_UNREACH_HOST_PROHIB:
 700                         case ICMP_UNREACH_FILTER_PROHIB:
 701                                 code = PRC_UNREACH_ADMIN_PROHIB;
 702                                 break;
 703
 704                         default:
 705                                 goto badcode;
 706                 }
 707                 goto deliver;
 708
 709         case ICMP_TIMXCEED:
 710                 if (code > 1)
 711                         goto badcode;
 712                 code += PRC_TIMXCEED_INTRANS;
 713                 goto deliver;
 714
 715         case ICMP_PARAMPROB:
 716                 if (code > 1)
 717                         goto badcode;
 718                 code = PRC_PARAMPROB;
 719                 goto deliver;
 720
 721         case ICMP_SOURCEQUENCH:
 722                 if (code)
 723                         goto badcode;
 724                 if (discard_sourcequench)
 725                         break;
 726                 code = PRC_QUENCH;
 727 deliver:
 728                 /*
 729                  * Problem with datagram; advise higher level routines.
 730                  */
 731                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 732                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 733                         icmpstat.icps_badlen++;
 734                         goto freeit;
 735                 }
 736                 /* Discard ICMP's in response to multicast packets */
 737                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
 738                         goto badcode;
 739 #ifdef ICMPPRINTFS
 740                 if (icmpprintfs)
 741                         kprintf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
 742 #endif
 743                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 744
 745                 /*
 746                  * MTU discovery
 747                  */
 748                 if (code == PRC_MSGSIZE) {
 749                         /* Run MTU discovery in all netisrs */
 750                         if (icmp_mtudisc_start(m, hlen, proto)) {
 751                                 /* Forwarded; done */
 752                                 return IPPROTO_DONE;
 753                         }
 754                         /* Move on; run rip_input() directly */
 755                 } else {
 756                         struct protosw *pr;
 757                         struct lwkt_port *port;
 758                         int cpu;
 759
 760                         pr = &inetsw[ip_protox[icp->icmp_ip.ip_p]];
 761                         port = so_pr_ctlport(pr, code,
 762                             (struct sockaddr *)&icmpsrc, &icp->icmp_ip, &cpu);
 763                         if (port != NULL) {
 764                                 if (cpu == ncpus) {
 765                                         if (ncpus > 1) {
 766                                                 /*
 767                                                  * Run pr_ctlinput in all
 768                                                  * netisrs
 769                                                  */
 770                                                 icmp_ctlinput_global_start(m,
 771                                                     code, hlen, proto);
 772                                                 return IPPROTO_DONE;
 773                                         }
 774                                         /*
 775                                          * There is only one netisr; run
 776                                          * pr_ctlinput directly.
 777                                          */
 778                                 } else if (cpu != mycpuid) {
 779                                         /*
 780                                          * Send to the target netisr to run
 781                                          * pr_ctlinput.
 782                                          */
 783                                         icmp_ctlinput_start(m, port,
 784                                             code, hlen, proto);
 785                                         return IPPROTO_DONE;
 786                                 }
 787
 788                                 /*
 789                                  * The target netisr is this netisr.
 790                                  *
 791                                  * XXX if the packet contains [IPv4 AH TCP],
 792                                  * we can't make a notification to TCP layer.
 793                                  */
 794                                 so_pr_ctlinput_direct(pr, code,
 795                                     (struct sockaddr *)&icmpsrc, &icp->icmp_ip);
 796                         }
 797                         /* Move on; run rip_input() directly */
 798                 }
 799                 break;
 800 badcode:
 801                 icmpstat.icps_badcode++;
 802                 break;
 803
 804         case ICMP_ECHO:
 805                 if (!icmpbmcastecho
 806                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 807                         icmpstat.icps_bmcastecho++;
 808                         break;
 809                 }
 810                 icp->icmp_type = ICMP_ECHOREPLY;
 811 #ifdef ICMP_BANDLIM
 812                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
 813                         goto freeit;
 814                 else
 815 #endif
 816                         goto reflect;
 817
 818         case ICMP_TSTAMP:
 819                 if (!icmpbmcastecho
 820                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
 821                         icmpstat.icps_bmcasttstamp++;
 822                         break;
 823                 }
 824                 if (icmplen < ICMP_TSLEN) {
 825                         icmpstat.icps_badlen++;
 826                         break;
 827                 }
 828                 icp->icmp_type = ICMP_TSTAMPREPLY;
 829                 icp->icmp_rtime = iptime();
 830                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
 831 #ifdef ICMP_BANDLIM
 832                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
 833                         goto freeit;
 834                 else
 835 #endif
 836                         goto reflect;
 837
 838         case ICMP_MASKREQ:
 839                 if (icmpmaskrepl == 0)
 840                         break;
 841                 /*
 842                  * We are not able to respond with all ones broadcast
 843                  * unless we receive it over a point-to-point interface.
 844                  */
 845                 if (icmplen < ICMP_MASKLEN)
 846                         break;
 847                 switch (ip->ip_dst.s_addr) {
 848
 849                 case INADDR_BROADCAST:
 850                 case INADDR_ANY:
 851                         icmpdst.sin_addr = ip->ip_src;
 852                         break;
 853
 854                 default:
 855                         icmpdst.sin_addr = ip->ip_dst;
 856                 }
 857                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
 858                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
 859                 if (ia == NULL)
 860                         break;
 861                 if (ia->ia_ifp == 0)
 862                         break;
 863                 icp->icmp_type = ICMP_MASKREPLY;
 864                 icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
 865                 if (ip->ip_src.s_addr == 0) {
 866                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
 867                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
 868                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
 869                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
 870                 }
 871 reflect:
 872                 ip->ip_len += hlen;     /* since ip_input deducts this */
 873                 icmpstat.icps_reflect++;
 874                 icmpstat.icps_outhist[icp->icmp_type]++;
 875                 icmp_reflect(m);
 876                 return(IPPROTO_DONE);
 877
 878         case ICMP_REDIRECT:
 879                 if (log_redirect) {
 880                         char src_buf[INET_ADDRSTRLEN];
 881                         char dst_buf[INET_ADDRSTRLEN];
 882                         char gwy_buf[INET_ADDRSTRLEN];
 883
 884                         kprintf("icmp redirect from %s: %s => %s\n",
 885                             inet_ntop(AF_INET, &ip->ip_src,
 886                                 src_buf, INET_ADDRSTRLEN),
 887                             inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
 888                                 dst_buf, INET_ADDRSTRLEN),
 889                             inet_ntop(AF_INET, &icp->icmp_gwaddr,
 890                                 gwy_buf, INET_ADDRSTRLEN));
 891                 }
 892                 if (drop_redirect)
 893                         break;
 894                 if (code > 3)
 895                         goto badcode;
 896                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
 897                     IP_VHL_HL(icp->icmp_ip.ip_vhl) < (sizeof(struct ip) >> 2)) {
 898                         icmpstat.icps_badlen++;
 899                         break;
 900                 }
 901 #ifdef ICMPPRINTFS
 902                 if (icmpprintfs) {
 903                         char dst_buf[INET_ADDRSTRLEN], gw_buf[INET_ADDRSTRLEN];
 904
 905                         kprintf("redirect dst %s to %s\n",
 906                             inet_ntop(AF_INET, &icp->icmp_ip.ip_dst,
 907                                 dst_buf, INET_ADDRSTRLEN),
 908                             inet_ntop(AF_INET, &icp->icmp_gwaddr,
 909                                 gw_buf, INET_ADDRSTRLEN));
 910                 }
 911 #endif
 912                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
 913
 914                 /* Run redirect in all netisrs */
 915                 if (icmp_redirect_start(m, hlen, proto)) {
 916                         /* Forwarded; done */
 917                         return IPPROTO_DONE;
 918                 }
 919                 /* Move on; run rip_input() directly */
 920 #ifdef IPSEC
 921                 key_sa_routechange((struct sockaddr *)&icmpsrc);
 922 #endif
 923                 break;
 924
 925         /*
 926          * No kernel processing for the following;
 927          * just fall through to send to raw listener.
 928          */
 929         case ICMP_ECHOREPLY:
 930         case ICMP_ROUTERADVERT:
 931         case ICMP_ROUTERSOLICIT:
 932         case ICMP_TSTAMPREPLY:
 933         case ICMP_IREQREPLY:
 934         case ICMP_MASKREPLY:
 935         default:
 936                 break;
 937         }
 938
 939 raw:
 940         *mp = m;
 941         rip_input(mp, offp, proto);
 942         return(IPPROTO_DONE);
 943
 944 freeit:
 945         m_freem(m);
 946         return(IPPROTO_DONE);
 947 }
 948
 949 /*
 950  * Reflect the ip packet back to the source
 951  */
 952 static void
 953 icmp_reflect(struct mbuf *m)
 954 {
 955         struct ip *ip = mtod(m, struct ip *);
 956         struct in_ifaddr *ia;
 957         struct in_ifaddr_container *iac;
 958         struct ifaddr_container *ifac;
 959         struct ifnet *ifp;
 960         struct in_addr t;
 961         struct mbuf *opts = NULL;
 962         int optlen = (IP_VHL_HL(ip->ip_vhl) << 2) - sizeof(struct ip);
 963         struct route *ro = NULL, rt;
 964
 965         if (!in_canforward(ip->ip_src) &&
 966             ((ntohl(ip->ip_src.s_addr) & IN_CLASSA_NET) !=
 967              (IN_LOOPBACKNET << IN_CLASSA_NSHIFT))) {
 968                 m_freem(m);     /* Bad return address */
 969                 icmpstat.icps_badaddr++;
 970                 goto done;      /* Ip_output() will check for broadcast */
 971         }
 972         t = ip->ip_dst;
 973         ip->ip_dst = ip->ip_src;
 974
 975         ro = &rt;
 976         bzero(ro, sizeof *ro);
 977
 978         /*
 979          * If the incoming packet was addressed directly to us,
 980          * use dst as the src for the reply.  Otherwise (broadcast
 981          * or anonymous), use the address which corresponds
 982          * to the incoming interface.
 983          */
 984         ia = NULL;
 985         LIST_FOREACH(iac, INADDR_HASH(t.s_addr), ia_hash) {
 986                 if (t.s_addr == IA_SIN(iac->ia)->sin_addr.s_addr) {
 987                         ia = iac->ia;
 988                         goto match;
 989                 }
 990         }
 991         ifp = m->m_pkthdr.rcvif;
 992         if (ifp != NULL && (ifp->if_flags & IFF_BROADCAST)) {
 993                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
 994                         struct ifaddr *ifa = ifac->ifa;
 995
 996                         if (ifa->ifa_addr->sa_family != AF_INET)
 997                                 continue;
 998                         ia = ifatoia(ifa);
 999                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
1000                             t.s_addr)
1001                                 goto match;
1002                 }
1003         }
1004         /*
1005          * If the packet was transiting through us, use the address of
1006          * the interface the packet came through in.  If that interface
1007          * doesn't have a suitable IP address, the normal selection
1008          * criteria apply.
1009          */
1010         if (icmp_rfi && ifp != NULL) {
1011                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1012                         struct ifaddr *ifa = ifac->ifa;
1013
1014                         if (ifa->ifa_addr->sa_family != AF_INET)
1015                                 continue;
1016                         ia = ifatoia(ifa);
1017                         goto match;
1018                 }
1019         }
1020         /*
1021          * If the incoming packet was not addressed directly to us, use
1022          * designated interface for icmp replies specified by sysctl
1023          * net.inet.icmp.reply_src (default not set). Otherwise continue
1024          * with normal source selection.
1025          */
1026         if (icmp_reply_src[0] != '\0' &&
1027             (ifp = ifunit_netisr(icmp_reply_src))) {
1028                 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1029                         struct ifaddr *ifa = ifac->ifa;
1030
1031                         if (ifa->ifa_addr->sa_family != AF_INET)
1032                                 continue;
1033                         ia = ifatoia(ifa);
1034                         goto match;
1035                 }
1036         }
1037         /*
1038          * If the packet was transiting through us, use the address of
1039          * the interface that is the closest to the packet source.
1040          * When we don't have a route back to the packet source, stop here
1041          * and drop the packet.
1042          */
1043         ia = ip_rtaddr(ip->ip_dst, ro);
1044         if (ia == NULL) {
1045                 m_freem(m);
1046                 icmpstat.icps_noroute++;
1047                 goto done;
1048         }
1049 match:
1050         t = IA_SIN(ia)->sin_addr;
1051         ip->ip_src = t;
1052         ip->ip_ttl = ip_defttl;
1053
1054         if (optlen > 0) {
1055                 u_char *cp;
1056                 int opt, cnt;
1057                 u_int len;
1058
1059                 /*
1060                  * Retrieve any source routing from the incoming packet;
1061                  * add on any record-route or timestamp options.
1062                  */
1063                 cp = (u_char *) (ip + 1);
1064                 if ((opts = ip_srcroute(m)) == NULL &&
1065                     (opts = m_gethdr(M_NOWAIT, MT_HEADER))) {
1066                         opts->m_len = sizeof(struct in_addr);
1067                         mtod(opts, struct in_addr *)->s_addr = 0;
1068                 }
1069                 if (opts) {
1070 #ifdef ICMPPRINTFS
1071                         if (icmpprintfs)
1072                                 kprintf("icmp_reflect optlen %d rt %d => ",
1073                                        optlen, opts->m_len);
1074 #endif
1075                         for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
1076                                 opt = cp[IPOPT_OPTVAL];
1077                                 if (opt == IPOPT_EOL)
1078                                         break;
1079                                 if (opt == IPOPT_NOP)
1080                                         len = 1;
1081                                 else {
1082                                         if (cnt < IPOPT_OLEN + sizeof *cp)
1083                                                 break;
1084                                         len = cp[IPOPT_OLEN];
1085                                         if (len < IPOPT_OLEN + sizeof *cp ||
1086                                             len > cnt)
1087                                         break;
1088                                 }
1089                                 /*
1090                                  * Should check for overflow, but it
1091                                  * "can't happen".
1092                                  */
1093                                 if (opt == IPOPT_RR || opt == IPOPT_TS ||
1094                                     opt == IPOPT_SECURITY) {
1095                                         bcopy(cp,
1096                                               mtod(opts, caddr_t) + opts->m_len,
1097                                               len);
1098                                         opts->m_len += len;
1099                                 }
1100                         }
1101                         /* Terminate & pad, if necessary */
1102                         cnt = opts->m_len % 4;
1103                         if (cnt) {
1104                                 for (; cnt < 4; cnt++) {
1105                                         *(mtod(opts, caddr_t) + opts->m_len) =
1106                                             IPOPT_EOL;
1107                                         opts->m_len++;
1108                                 }
1109                         }
1110 #ifdef ICMPPRINTFS
1111                         if (icmpprintfs)
1112                                 kprintf("%d\n", opts->m_len);
1113 #endif
1114                 }
1115                 /*
1116                  * Now strip out original options by copying rest of first
1117                  * mbuf's data back, and adjust the IP length.
1118                  */
1119                 ip->ip_len -= optlen;
1120                 ip->ip_vhl = IP_VHL_BORING;
1121                 m->m_len -= optlen;
1122                 if (m->m_flags & M_PKTHDR)
1123                         m->m_pkthdr.len -= optlen;
1124                 optlen += sizeof(struct ip);
1125                 bcopy((caddr_t)ip + optlen, ip + 1,
1126                       m->m_len - sizeof(struct ip));
1127         }
1128         m->m_pkthdr.fw_flags &= FW_MBUF_GENERATED;
1129         m->m_flags &= ~(M_BCAST|M_MCAST);
1130         icmp_send(m, opts, ro);
1131 done:
1132         if (opts)
1133                 m_free(opts);
1134         if (ro && ro->ro_rt)
1135                 RTFREE(ro->ro_rt);
1136 }
1137
1138 /*
1139  * Send an icmp packet back to the ip level,
1140  * after supplying a checksum.
1141  */
1142 static void
1143 icmp_send(struct mbuf *m, struct mbuf *opts, struct route *rt)
1144 {
1145         struct ip *ip = mtod(m, struct ip *);
1146         struct icmp *icp;
1147         int hlen;
1148
1149         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
1150         m->m_data += hlen;
1151         m->m_len -= hlen;
1152         icp = mtod(m, struct icmp *);
1153         icp->icmp_cksum = 0;
1154         icp->icmp_cksum = in_cksum(m, ip->ip_len - hlen);
1155         m->m_data -= hlen;
1156         m->m_len += hlen;
1157         m->m_pkthdr.rcvif = NULL;
1158 #ifdef ICMPPRINTFS
1159         if (icmpprintfs) {
1160                 char dst_buf[INET_ADDRSTRLEN], src_buf[INET_ADDRSTRLEN];
1161
1162                 kprintf("icmp_send dst %s src %s\n",
1163                     inet_ntop(AF_INET, &ip->ip_dst, dst_buf, INET_ADDRSTRLEN),
1164                     inet_ntop(AF_INET, &ip->ip_src, src_buf, INET_ADDRSTRLEN));
1165         }
1166 #endif
1167         ip_output(m, opts, rt, 0, NULL, NULL);
1168 }
1169
1170 n_time
1171 iptime(void)
1172 {
1173         struct timeval atv;
1174         u_long t;
1175
1176         getmicrotime(&atv);
1177         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
1178         return (htonl(t));
1179 }
1180
1181 #if 1
1182 /*
1183  * Return the next larger or smaller MTU plateau (table from RFC 1191)
1184  * given current value MTU.  If DIR is less than zero, a larger plateau
1185  * is returned; otherwise, a smaller value is returned.
1186  */
1187 int
1188 ip_next_mtu(int mtu, int dir)
1189 {
1190         static int mtutab[] = {
1191                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1006, 508, 296,
1192                 68, 0
1193         };
1194         int i;
1195
1196         for (i = 0; i < (sizeof mtutab) / (sizeof mtutab[0]); i++) {
1197                 if (mtu >= mtutab[i])
1198                         break;
1199         }
1200
1201         if (dir < 0) {
1202                 if (i == 0) {
1203                         return 0;
1204                 } else {
1205                         return mtutab[i - 1];
1206                 }
1207         } else {
1208                 if (mtutab[i] == 0) {
1209                         return 0;
1210                 } else if(mtu > mtutab[i]) {
1211                         return mtutab[i];
1212                 } else {
1213                         return mtutab[i + 1];
1214                 }
1215         }
1216 }
1217 #endif
1218
1219 #ifdef ICMP_BANDLIM
1220 /*
1221  * badport_bandlim() - check for ICMP bandwidth limit
1222  *
1223  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
1224  *      hit our bandwidth limit and it is not ok.
1225  *
1226  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
1227  *
1228  *      For now we separate the TCP and UDP subsystems w/ different 'which'
1229  *      values.  We may eventually remove this separation (and simplify the
1230  *      code further).
1231  *
1232  *      Note that the printing of the error message is delayed so we can
1233  *      properly print the icmp error rate that the system was trying to do
1234  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
1235  *      the 'final' error, but it doesn't make sense to solve the printing
1236  *      delay with more complex code.
1237  */
1238 int
1239 badport_bandlim(int which)
1240 {
1241         static int lticks[BANDLIM_MAX + 1];
1242         static int lpackets[BANDLIM_MAX + 1];
1243         int dticks;
1244         const char *bandlimittype[] = {
1245                 "Limiting icmp unreach response",
1246                 "Limiting icmp ping response",
1247                 "Limiting icmp tstamp response",
1248                 "Limiting closed port RST response",
1249                 "Limiting open port RST response"
1250                 };
1251
1252         /*
1253          * Return ok status if feature disabled or argument out of
1254          * ranage.
1255          */
1256
1257         if (icmplim <= 0 || which > BANDLIM_MAX || which < 0)
1258                 return(0);
1259         dticks = ticks - lticks[which];
1260
1261         /*
1262          * reset stats when cumulative dt exceeds one second.
1263          */
1264
1265         if ((unsigned int)dticks > hz) {
1266                 if (lpackets[which] > icmplim && icmplim_output) {
1267                         kprintf("%s from %d to %d packets per second\n",
1268                                 bandlimittype[which],
1269                                 lpackets[which],
1270                                 icmplim
1271                         );
1272                 }
1273                 lticks[which] = ticks;
1274                 lpackets[which] = 0;
1275         }
1276
1277         /*
1278          * bump packet count
1279          */
1280
1281         if (++lpackets[which] > icmplim) {
1282                 return(-1);
1283         }
1284         return(0);
1285 }
1286 #endif