The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/netinet/ip_icmp.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * SPDX-License-Identifier: BSD-3-Clause
    3  *
    4  * Copyright (c) 1982, 1986, 1988, 1993
    5  *      The Regents of the University of California.  All rights reserved.
    6  *
    7  * Redistribution and use in source and binary forms, with or without
    8  * modification, are permitted provided that the following conditions
    9  * are met:
   10  * 1. Redistributions of source code must retain the above copyright
   11  *    notice, this list of conditions and the following disclaimer.
   12  * 2. Redistributions in binary form must reproduce the above copyright
   13  *    notice, this list of conditions and the following disclaimer in the
   14  *    documentation and/or other materials provided with the distribution.
   15  * 3. Neither the name of the University nor the names of its contributors
   16  *    may be used to endorse or promote products derived from this software
   17  *    without specific prior written permission.
   18  *
   19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   29  * SUCH DAMAGE.
   30  *
   31  *      @(#)ip_icmp.c   8.2 (Berkeley) 1/4/94
   32  */
   33 
   34 #include <sys/cdefs.h>
   35 __FBSDID("$FreeBSD$");
   36 
   37 #include "opt_inet.h"
   38 
   39 #include <sys/param.h>
   40 #include <sys/systm.h>
   41 #include <sys/mbuf.h>
   42 #include <sys/protosw.h>
   43 #include <sys/socket.h>
   44 #include <sys/time.h>
   45 #include <sys/kernel.h>
   46 #include <sys/lock.h>
   47 #include <sys/sysctl.h>
   48 #include <sys/syslog.h>
   49 
   50 #include <net/if.h>
   51 #include <net/if_var.h>
   52 #include <net/if_types.h>
   53 #include <net/route.h>
   54 #include <net/route/route_ctl.h>
   55 #include <net/route/nhop.h>
   56 #include <net/vnet.h>
   57 
   58 #include <netinet/in.h>
   59 #include <netinet/in_fib.h>
   60 #include <netinet/in_pcb.h>
   61 #include <netinet/in_systm.h>
   62 #include <netinet/in_var.h>
   63 #include <netinet/ip.h>
   64 #include <netinet/ip_icmp.h>
   65 #include <netinet/ip_var.h>
   66 #include <netinet/ip_options.h>
   67 #include <netinet/sctp.h>
   68 #include <netinet/tcp.h>
   69 #include <netinet/tcpip.h>
   70 #include <netinet/icmp_var.h>
   71 
   72 #ifdef INET
   73 
   74 #include <machine/in_cksum.h>
   75 
   76 #include <security/mac/mac_framework.h>
   77 #endif /* INET */
   78 
   79 extern ipproto_ctlinput_t       *ip_ctlprotox[];
   80 
   81 /*
   82  * ICMP routines: error generation, receive packet processing, and
   83  * routines to turnaround packets back to the originator, and
   84  * host table maintenance routines.
   85  */
   86 VNET_DEFINE_STATIC(int, icmplim) = 200;
   87 #define V_icmplim                       VNET(icmplim)
   88 SYSCTL_INT(_net_inet_icmp, ICMPCTL_ICMPLIM, icmplim, CTLFLAG_VNET | CTLFLAG_RW,
   89         &VNET_NAME(icmplim), 0,
   90         "Maximum number of ICMP responses per second");
   91 
   92 VNET_DEFINE_STATIC(int, icmplim_curr_jitter) = 0;
   93 #define V_icmplim_curr_jitter           VNET(icmplim_curr_jitter)
   94 VNET_DEFINE_STATIC(int, icmplim_jitter) = 16;
   95 #define V_icmplim_jitter                VNET(icmplim_jitter)
   96 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_jitter, CTLFLAG_VNET | CTLFLAG_RW,
   97         &VNET_NAME(icmplim_jitter), 0,
   98         "Random icmplim jitter adjustment limit");
   99 
  100 VNET_DEFINE_STATIC(int, icmplim_output) = 1;
  101 #define V_icmplim_output                VNET(icmplim_output)
  102 SYSCTL_INT(_net_inet_icmp, OID_AUTO, icmplim_output, CTLFLAG_VNET | CTLFLAG_RW,
  103         &VNET_NAME(icmplim_output), 0,
  104         "Enable logging of ICMP response rate limiting");
  105 
  106 #ifdef INET
  107 VNET_PCPUSTAT_DEFINE(struct icmpstat, icmpstat);
  108 VNET_PCPUSTAT_SYSINIT(icmpstat);
  109 SYSCTL_VNET_PCPUSTAT(_net_inet_icmp, ICMPCTL_STATS, stats, struct icmpstat,
  110     icmpstat, "ICMP statistics (struct icmpstat, netinet/icmp_var.h)");
  111 
  112 #ifdef VIMAGE
  113 VNET_PCPUSTAT_SYSUNINIT(icmpstat);
  114 #endif /* VIMAGE */
  115 
  116 VNET_DEFINE_STATIC(int, icmpmaskrepl) = 0;
  117 #define V_icmpmaskrepl                  VNET(icmpmaskrepl)
  118 SYSCTL_INT(_net_inet_icmp, ICMPCTL_MASKREPL, maskrepl, CTLFLAG_VNET | CTLFLAG_RW,
  119         &VNET_NAME(icmpmaskrepl), 0,
  120         "Reply to ICMP Address Mask Request packets");
  121 
  122 VNET_DEFINE_STATIC(u_int, icmpmaskfake) = 0;
  123 #define V_icmpmaskfake                  VNET(icmpmaskfake)
  124 SYSCTL_UINT(_net_inet_icmp, OID_AUTO, maskfake, CTLFLAG_VNET | CTLFLAG_RW,
  125         &VNET_NAME(icmpmaskfake), 0,
  126         "Fake reply to ICMP Address Mask Request packets");
  127 
  128 VNET_DEFINE(int, drop_redirect) = 0;
  129 #define V_drop_redirect                 VNET(drop_redirect)
  130 SYSCTL_INT(_net_inet_icmp, OID_AUTO, drop_redirect, CTLFLAG_VNET | CTLFLAG_RW,
  131         &VNET_NAME(drop_redirect), 0,
  132         "Ignore ICMP redirects");
  133 
  134 VNET_DEFINE_STATIC(int, log_redirect) = 0;
  135 #define V_log_redirect                  VNET(log_redirect)
  136 SYSCTL_INT(_net_inet_icmp, OID_AUTO, log_redirect, CTLFLAG_VNET | CTLFLAG_RW,
  137         &VNET_NAME(log_redirect), 0,
  138         "Log ICMP redirects to the console");
  139 
  140 VNET_DEFINE_STATIC(int, redirtimeout) = 60 * 10; /* 10 minutes */
  141 #define V_redirtimeout                  VNET(redirtimeout)
  142 SYSCTL_INT(_net_inet_icmp, OID_AUTO, redirtimeout, CTLFLAG_VNET | CTLFLAG_RW,
  143         &VNET_NAME(redirtimeout), 0,
  144         "Delay in seconds before expiring redirect route");
  145 
  146 VNET_DEFINE_STATIC(char, reply_src[IFNAMSIZ]);
  147 #define V_reply_src                     VNET(reply_src)
  148 SYSCTL_STRING(_net_inet_icmp, OID_AUTO, reply_src, CTLFLAG_VNET | CTLFLAG_RW,
  149         &VNET_NAME(reply_src), IFNAMSIZ,
  150         "ICMP reply source for non-local packets");
  151 
  152 VNET_DEFINE_STATIC(int, icmp_rfi) = 0;
  153 #define V_icmp_rfi                      VNET(icmp_rfi)
  154 SYSCTL_INT(_net_inet_icmp, OID_AUTO, reply_from_interface, CTLFLAG_VNET | CTLFLAG_RW,
  155         &VNET_NAME(icmp_rfi), 0,
  156         "ICMP reply from incoming interface for non-local packets");
  157 /* Router requirements RFC 1812 section 4.3.2.3 requires 576 - 28. */
  158 VNET_DEFINE_STATIC(int, icmp_quotelen) = 548;
  159 #define V_icmp_quotelen                 VNET(icmp_quotelen)
  160 SYSCTL_INT(_net_inet_icmp, OID_AUTO, quotelen, CTLFLAG_VNET | CTLFLAG_RW,
  161         &VNET_NAME(icmp_quotelen), 0,
  162         "Number of bytes from original packet to quote in ICMP reply");
  163 
  164 VNET_DEFINE_STATIC(int, icmpbmcastecho) = 0;
  165 #define V_icmpbmcastecho                VNET(icmpbmcastecho)
  166 SYSCTL_INT(_net_inet_icmp, OID_AUTO, bmcastecho, CTLFLAG_VNET | CTLFLAG_RW,
  167         &VNET_NAME(icmpbmcastecho), 0,
  168         "Reply to multicast ICMP Echo Request and Timestamp packets");
  169 
  170 VNET_DEFINE_STATIC(int, icmptstamprepl) = 1;
  171 #define V_icmptstamprepl                VNET(icmptstamprepl)
  172 SYSCTL_INT(_net_inet_icmp, OID_AUTO, tstamprepl, CTLFLAG_VNET | CTLFLAG_RW,
  173         &VNET_NAME(icmptstamprepl), 0,
  174         "Respond to ICMP Timestamp packets");
  175 
  176 VNET_DEFINE_STATIC(int, error_keeptags) = 0;
  177 #define V_error_keeptags                VNET(error_keeptags)
  178 SYSCTL_INT(_net_inet_icmp, OID_AUTO, error_keeptags, CTLFLAG_VNET | CTLFLAG_RW,
  179         &VNET_NAME(error_keeptags), 0,
  180         "ICMP error response keeps copy of mbuf_tags of original packet");
  181 
  182 #ifdef ICMPPRINTFS
  183 int     icmpprintfs = 0;
  184 #endif
  185 
  186 static void     icmp_reflect(struct mbuf *);
  187 static void     icmp_send(struct mbuf *, struct mbuf *);
  188 static int      icmp_verify_redirect_gateway(struct sockaddr_in *,
  189     struct sockaddr_in *, struct sockaddr_in *, u_int);
  190 
  191 /*
  192  * Kernel module interface for updating icmpstat.  The argument is an index
  193  * into icmpstat treated as an array of u_long.  While this encodes the
  194  * general layout of icmpstat into the caller, it doesn't encode its
  195  * location, so that future changes to add, for example, per-CPU stats
  196  * support won't cause binary compatibility problems for kernel modules.
  197  */
  198 void
  199 kmod_icmpstat_inc(int statnum)
  200 {
  201 
  202         counter_u64_add(VNET(icmpstat)[statnum], 1);
  203 }
  204 
  205 /*
  206  * Generate an error packet of type error
  207  * in response to bad packet ip.
  208  */
  209 void
  210 icmp_error(struct mbuf *n, int type, int code, uint32_t dest, int mtu)
  211 {
  212         struct ip *oip, *nip;
  213         struct icmp *icp;
  214         struct mbuf *m;
  215         unsigned icmplen, icmpelen, nlen, oiphlen;
  216 
  217         KASSERT((u_int)type <= ICMP_MAXTYPE, ("%s: illegal ICMP type",
  218             __func__));
  219 
  220         if (type != ICMP_REDIRECT)
  221                 ICMPSTAT_INC(icps_error);
  222         /*
  223          * Don't send error:
  224          *  if the original packet was encrypted.
  225          *  if not the first fragment of message.
  226          *  in response to a multicast or broadcast packet.
  227          *  if the old packet protocol was an ICMP error message.
  228          */
  229         if (n->m_flags & M_DECRYPTED)
  230                 goto freeit;
  231         if (n->m_flags & (M_BCAST|M_MCAST))
  232                 goto freeit;
  233 
  234         /* Drop if IP header plus 8 bytes is not contiguous in first mbuf. */
  235         if (n->m_len < sizeof(struct ip) + ICMP_MINLEN)
  236                 goto freeit;
  237         oip = mtod(n, struct ip *);
  238         oiphlen = oip->ip_hl << 2;
  239         if (n->m_len < oiphlen + ICMP_MINLEN)
  240                 goto freeit;
  241 #ifdef ICMPPRINTFS
  242         if (icmpprintfs)
  243                 printf("icmp_error(%p, %x, %d)\n", oip, type, code);
  244 #endif
  245         if (oip->ip_off & htons(~(IP_MF|IP_DF)))
  246                 goto freeit;
  247         if (oip->ip_p == IPPROTO_ICMP && type != ICMP_REDIRECT &&
  248             !ICMP_INFOTYPE(((struct icmp *)((caddr_t)oip +
  249                 oiphlen))->icmp_type)) {
  250                 ICMPSTAT_INC(icps_oldicmp);
  251                 goto freeit;
  252         }
  253         /*
  254          * Calculate length to quote from original packet and
  255          * prevent the ICMP mbuf from overflowing.
  256          * Unfortunately this is non-trivial since ip_forward()
  257          * sends us truncated packets.
  258          */
  259         nlen = m_length(n, NULL);
  260         if (oip->ip_p == IPPROTO_TCP) {
  261                 struct tcphdr *th;
  262                 int tcphlen;
  263 
  264                 if (oiphlen + sizeof(struct tcphdr) > n->m_len &&
  265                     n->m_next == NULL)
  266                         goto stdreply;
  267                 if (n->m_len < oiphlen + sizeof(struct tcphdr) &&
  268                     (n = m_pullup(n, oiphlen + sizeof(struct tcphdr))) == NULL)
  269                         goto freeit;
  270                 oip = mtod(n, struct ip *);
  271                 th = mtodo(n, oiphlen);
  272                 tcphlen = th->th_off << 2;
  273                 if (tcphlen < sizeof(struct tcphdr))
  274                         goto freeit;
  275                 if (ntohs(oip->ip_len) < oiphlen + tcphlen)
  276                         goto freeit;
  277                 if (oiphlen + tcphlen > n->m_len && n->m_next == NULL)
  278                         goto stdreply;
  279                 if (n->m_len < oiphlen + tcphlen &&
  280                     (n = m_pullup(n, oiphlen + tcphlen)) == NULL)
  281                         goto freeit;
  282                 oip = mtod(n, struct ip *);
  283                 icmpelen = max(tcphlen, min(V_icmp_quotelen,
  284                     ntohs(oip->ip_len) - oiphlen));
  285         } else if (oip->ip_p == IPPROTO_SCTP) {
  286                 struct sctphdr *sh;
  287                 struct sctp_chunkhdr *ch;
  288 
  289                 if (ntohs(oip->ip_len) < oiphlen + sizeof(struct sctphdr))
  290                         goto stdreply;
  291                 if (oiphlen + sizeof(struct sctphdr) > n->m_len &&
  292                     n->m_next == NULL)
  293                         goto stdreply;
  294                 if (n->m_len < oiphlen + sizeof(struct sctphdr) &&
  295                     (n = m_pullup(n, oiphlen + sizeof(struct sctphdr))) == NULL)
  296                         goto freeit;
  297                 oip = mtod(n, struct ip *);
  298                 icmpelen = max(sizeof(struct sctphdr),
  299                     min(V_icmp_quotelen, ntohs(oip->ip_len) - oiphlen));
  300                 sh = mtodo(n, oiphlen);
  301                 if (ntohl(sh->v_tag) == 0 &&
  302                     ntohs(oip->ip_len) >= oiphlen +
  303                     sizeof(struct sctphdr) + 8 &&
  304                     (n->m_len >= oiphlen + sizeof(struct sctphdr) + 8 ||
  305                      n->m_next != NULL)) {
  306                         if (n->m_len < oiphlen + sizeof(struct sctphdr) + 8 &&
  307                             (n = m_pullup(n, oiphlen +
  308                             sizeof(struct sctphdr) + 8)) == NULL)
  309                                 goto freeit;
  310                         oip = mtod(n, struct ip *);
  311                         sh = mtodo(n, oiphlen);
  312                         ch = (struct sctp_chunkhdr *)(sh + 1);
  313                         if (ch->chunk_type == SCTP_INITIATION) {
  314                                 icmpelen = max(sizeof(struct sctphdr) + 8,
  315                                     min(V_icmp_quotelen, ntohs(oip->ip_len) -
  316                                     oiphlen));
  317                         }
  318                 }
  319         } else
  320 stdreply:       icmpelen = max(8, min(V_icmp_quotelen, ntohs(oip->ip_len) -
  321                     oiphlen));
  322 
  323         icmplen = min(oiphlen + icmpelen, nlen);
  324         if (icmplen < sizeof(struct ip))
  325                 goto freeit;
  326 
  327         if (MHLEN > sizeof(struct ip) + ICMP_MINLEN + icmplen)
  328                 m = m_gethdr(M_NOWAIT, MT_DATA);
  329         else
  330                 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
  331         if (m == NULL)
  332                 goto freeit;
  333 #ifdef MAC
  334         mac_netinet_icmp_reply(n, m);
  335 #endif
  336         icmplen = min(icmplen, M_TRAILINGSPACE(m) -
  337             sizeof(struct ip) - ICMP_MINLEN);
  338         m_align(m, sizeof(struct ip) + ICMP_MINLEN + icmplen);
  339         m->m_data += sizeof(struct ip);
  340         m->m_len = ICMP_MINLEN + icmplen;
  341 
  342         /* XXX MRT  make the outgoing packet use the same FIB
  343          * that was associated with the incoming packet
  344          */
  345         M_SETFIB(m, M_GETFIB(n));
  346         icp = mtod(m, struct icmp *);
  347         ICMPSTAT_INC(icps_outhist[type]);
  348         icp->icmp_type = type;
  349         if (type == ICMP_REDIRECT)
  350                 icp->icmp_gwaddr.s_addr = dest;
  351         else {
  352                 icp->icmp_void = 0;
  353                 /*
  354                  * The following assignments assume an overlay with the
  355                  * just zeroed icmp_void field.
  356                  */
  357                 if (type == ICMP_PARAMPROB) {
  358                         icp->icmp_pptr = code;
  359                         code = 0;
  360                 } else if (type == ICMP_UNREACH &&
  361                         code == ICMP_UNREACH_NEEDFRAG && mtu) {
  362                         icp->icmp_nextmtu = htons(mtu);
  363                 }
  364         }
  365         icp->icmp_code = code;
  366 
  367         /*
  368          * Copy the quotation into ICMP message and
  369          * convert quoted IP header back to network representation.
  370          */
  371         m_copydata(n, 0, icmplen, (caddr_t)&icp->icmp_ip);
  372         nip = &icp->icmp_ip;
  373 
  374         /*
  375          * Set up ICMP message mbuf and copy old IP header (without options
  376          * in front of ICMP message.
  377          * If the original mbuf was meant to bypass the firewall, the error
  378          * reply should bypass as well.
  379          */
  380         m->m_flags |= n->m_flags & M_SKIP_FIREWALL;
  381         KASSERT(M_LEADINGSPACE(m) >= sizeof(struct ip),
  382             ("insufficient space for ip header"));
  383         m->m_data -= sizeof(struct ip);
  384         m->m_len += sizeof(struct ip);
  385         m->m_pkthdr.len = m->m_len;
  386         m->m_pkthdr.rcvif = n->m_pkthdr.rcvif;
  387         nip = mtod(m, struct ip *);
  388         bcopy((caddr_t)oip, (caddr_t)nip, sizeof(struct ip));
  389         nip->ip_len = htons(m->m_len);
  390         nip->ip_v = IPVERSION;
  391         nip->ip_hl = 5;
  392         nip->ip_p = IPPROTO_ICMP;
  393         nip->ip_tos = 0;
  394         nip->ip_off = 0;
  395 
  396         if (V_error_keeptags)
  397                 m_tag_copy_chain(m, n, M_NOWAIT);
  398 
  399         icmp_reflect(m);
  400 
  401 freeit:
  402         m_freem(n);
  403 }
  404 
  405 int
  406 icmp_errmap(const struct icmp *icp)
  407 {
  408 
  409         switch (icp->icmp_type) {
  410         case ICMP_UNREACH:
  411                 switch (icp->icmp_code) {
  412                 case ICMP_UNREACH_NET:
  413                 case ICMP_UNREACH_HOST:
  414                 case ICMP_UNREACH_SRCFAIL:
  415                 case ICMP_UNREACH_NET_UNKNOWN:
  416                 case ICMP_UNREACH_HOST_UNKNOWN:
  417                 case ICMP_UNREACH_ISOLATED:
  418                 case ICMP_UNREACH_TOSNET:
  419                 case ICMP_UNREACH_TOSHOST:
  420                 case ICMP_UNREACH_HOST_PRECEDENCE:
  421                 case ICMP_UNREACH_PRECEDENCE_CUTOFF:
  422                         return (EHOSTUNREACH);
  423                 case ICMP_UNREACH_NEEDFRAG:
  424                         return (EMSGSIZE);
  425                 case ICMP_UNREACH_PROTOCOL:
  426                 case ICMP_UNREACH_PORT:
  427                 case ICMP_UNREACH_NET_PROHIB:
  428                 case ICMP_UNREACH_HOST_PROHIB:
  429                 case ICMP_UNREACH_FILTER_PROHIB:
  430                         return (ECONNREFUSED);
  431                 default:
  432                         return (0);
  433                 }
  434         case ICMP_TIMXCEED:
  435                 switch (icp->icmp_code) {
  436                 case ICMP_TIMXCEED_INTRANS:
  437                         return (EHOSTUNREACH);
  438                 default:
  439                         return (0);
  440                 }
  441         case ICMP_PARAMPROB:
  442                 switch (icp->icmp_code) {
  443                 case ICMP_PARAMPROB_ERRATPTR:
  444                 case ICMP_PARAMPROB_OPTABSENT:
  445                         return (ENOPROTOOPT);
  446                 default:
  447                         return (0);
  448                 }
  449         default:
  450                 return (0);
  451         }
  452 }
  453 
  454 /*
  455  * Process a received ICMP message.
  456  */
  457 int
  458 icmp_input(struct mbuf **mp, int *offp, int proto)
  459 {
  460         struct icmp *icp;
  461         struct in_ifaddr *ia;
  462         struct mbuf *m = *mp;
  463         struct ip *ip = mtod(m, struct ip *);
  464         struct sockaddr_in icmpsrc, icmpdst, icmpgw;
  465         int hlen = *offp;
  466         int icmplen = ntohs(ip->ip_len) - *offp;
  467         int i, code;
  468         int fibnum;
  469 
  470         NET_EPOCH_ASSERT();
  471 
  472         *mp = NULL;
  473 
  474         /*
  475          * Locate icmp structure in mbuf, and check
  476          * that not corrupted and of at least minimum length.
  477          */
  478 #ifdef ICMPPRINTFS
  479         if (icmpprintfs) {
  480                 char srcbuf[INET_ADDRSTRLEN];
  481                 char dstbuf[INET_ADDRSTRLEN];
  482 
  483                 printf("icmp_input from %s to %s, len %d\n",
  484                     inet_ntoa_r(ip->ip_src, srcbuf),
  485                     inet_ntoa_r(ip->ip_dst, dstbuf), icmplen);
  486         }
  487 #endif
  488         if (icmplen < ICMP_MINLEN) {
  489                 ICMPSTAT_INC(icps_tooshort);
  490                 goto freeit;
  491         }
  492         i = hlen + min(icmplen, ICMP_ADVLENMIN);
  493         if (m->m_len < i && (m = m_pullup(m, i)) == NULL)  {
  494                 ICMPSTAT_INC(icps_tooshort);
  495                 return (IPPROTO_DONE);
  496         }
  497         ip = mtod(m, struct ip *);
  498         m->m_len -= hlen;
  499         m->m_data += hlen;
  500         icp = mtod(m, struct icmp *);
  501         if (in_cksum(m, icmplen)) {
  502                 ICMPSTAT_INC(icps_checksum);
  503                 goto freeit;
  504         }
  505         m->m_len += hlen;
  506         m->m_data -= hlen;
  507 
  508 #ifdef ICMPPRINTFS
  509         if (icmpprintfs)
  510                 printf("icmp_input, type %d code %d\n", icp->icmp_type,
  511                     icp->icmp_code);
  512 #endif
  513 
  514         /*
  515          * Message type specific processing.
  516          */
  517         if (icp->icmp_type > ICMP_MAXTYPE)
  518                 goto raw;
  519 
  520         /* Initialize */
  521         bzero(&icmpsrc, sizeof(icmpsrc));
  522         icmpsrc.sin_len = sizeof(struct sockaddr_in);
  523         icmpsrc.sin_family = AF_INET;
  524         bzero(&icmpdst, sizeof(icmpdst));
  525         icmpdst.sin_len = sizeof(struct sockaddr_in);
  526         icmpdst.sin_family = AF_INET;
  527         bzero(&icmpgw, sizeof(icmpgw));
  528         icmpgw.sin_len = sizeof(struct sockaddr_in);
  529         icmpgw.sin_family = AF_INET;
  530 
  531         ICMPSTAT_INC(icps_inhist[icp->icmp_type]);
  532         code = icp->icmp_code;
  533         switch (icp->icmp_type) {
  534         case ICMP_UNREACH:
  535                 if (code > ICMP_UNREACH_PRECEDENCE_CUTOFF)
  536                         goto badcode;
  537                 else
  538                         goto deliver;
  539 
  540         case ICMP_TIMXCEED:
  541                 if (code > ICMP_TIMXCEED_REASS)
  542                         goto badcode;
  543                 else
  544                         goto deliver;
  545 
  546         case ICMP_PARAMPROB:
  547                 if (code > ICMP_PARAMPROB_LENGTH)
  548                         goto badcode;
  549 
  550         deliver:
  551                 /*
  552                  * Problem with datagram; advise higher level routines.
  553                  */
  554                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
  555                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
  556                         ICMPSTAT_INC(icps_badlen);
  557                         goto freeit;
  558                 }
  559                 /* Discard ICMP's in response to multicast packets */
  560                 if (IN_MULTICAST(ntohl(icp->icmp_ip.ip_dst.s_addr)))
  561                         goto badcode;
  562                 /* Filter out responses to INADDR_ANY, protocols ignore it. */
  563                 if (icp->icmp_ip.ip_dst.s_addr == INADDR_ANY)
  564                         goto freeit;
  565 #ifdef ICMPPRINTFS
  566                 if (icmpprintfs)
  567                         printf("deliver to protocol %d\n", icp->icmp_ip.ip_p);
  568 #endif
  569                 /*
  570                  * XXX if the packet contains [IPv4 AH TCP], we can't make a
  571                  * notification to TCP layer.
  572                  */
  573                 i = sizeof(struct ip) + min(icmplen, ICMP_ADVLENPREF(icp));
  574                 ip_stripoptions(m);
  575                 if (m->m_len < i && (m = m_pullup(m, i)) == NULL) {
  576                         /* This should actually not happen */
  577                         ICMPSTAT_INC(icps_tooshort);
  578                         return (IPPROTO_DONE);
  579                 }
  580                 ip = mtod(m, struct ip *);
  581                 icp = (struct icmp *)(ip + 1);
  582                 /*
  583                  * The upper layer handler can rely on:
  584                  * - The outer IP header has no options.
  585                  * - The outer IP header, the ICMP header, the inner IP header,
  586                  *   and the first n bytes of the inner payload are contiguous.
  587                  *   n is at least 8, but might be larger based on
  588                  *   ICMP_ADVLENPREF. See its definition in ip_icmp.h.
  589                  */
  590                 if (ip_ctlprotox[icp->icmp_ip.ip_p] != NULL)
  591                         ip_ctlprotox[icp->icmp_ip.ip_p](icp);
  592                 break;
  593 
  594         badcode:
  595                 ICMPSTAT_INC(icps_badcode);
  596                 break;
  597 
  598         case ICMP_ECHO:
  599                 if (!V_icmpbmcastecho
  600                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
  601                         ICMPSTAT_INC(icps_bmcastecho);
  602                         break;
  603                 }
  604                 if (badport_bandlim(BANDLIM_ICMP_ECHO) < 0)
  605                         goto freeit;
  606                 icp->icmp_type = ICMP_ECHOREPLY;
  607                 goto reflect;
  608 
  609         case ICMP_TSTAMP:
  610                 if (V_icmptstamprepl == 0)
  611                         break;
  612                 if (!V_icmpbmcastecho
  613                     && (m->m_flags & (M_MCAST | M_BCAST)) != 0) {
  614                         ICMPSTAT_INC(icps_bmcasttstamp);
  615                         break;
  616                 }
  617                 if (icmplen < ICMP_TSLEN) {
  618                         ICMPSTAT_INC(icps_badlen);
  619                         break;
  620                 }
  621                 if (badport_bandlim(BANDLIM_ICMP_TSTAMP) < 0)
  622                         goto freeit;
  623                 icp->icmp_type = ICMP_TSTAMPREPLY;
  624                 icp->icmp_rtime = iptime();
  625                 icp->icmp_ttime = icp->icmp_rtime;      /* bogus, do later! */
  626                 goto reflect;
  627 
  628         case ICMP_MASKREQ:
  629                 if (V_icmpmaskrepl == 0)
  630                         break;
  631                 /*
  632                  * We are not able to respond with all ones broadcast
  633                  * unless we receive it over a point-to-point interface.
  634                  */
  635                 if (icmplen < ICMP_MASKLEN)
  636                         break;
  637                 switch (ip->ip_dst.s_addr) {
  638                 case INADDR_BROADCAST:
  639                 case INADDR_ANY:
  640                         icmpdst.sin_addr = ip->ip_src;
  641                         break;
  642 
  643                 default:
  644                         icmpdst.sin_addr = ip->ip_dst;
  645                 }
  646                 ia = (struct in_ifaddr *)ifaof_ifpforaddr(
  647                             (struct sockaddr *)&icmpdst, m->m_pkthdr.rcvif);
  648                 if (ia == NULL)
  649                         break;
  650                 if (ia->ia_ifp == NULL)
  651                         break;
  652                 icp->icmp_type = ICMP_MASKREPLY;
  653                 if (V_icmpmaskfake == 0)
  654                         icp->icmp_mask = ia->ia_sockmask.sin_addr.s_addr;
  655                 else
  656                         icp->icmp_mask = V_icmpmaskfake;
  657                 if (ip->ip_src.s_addr == 0) {
  658                         if (ia->ia_ifp->if_flags & IFF_BROADCAST)
  659                             ip->ip_src = satosin(&ia->ia_broadaddr)->sin_addr;
  660                         else if (ia->ia_ifp->if_flags & IFF_POINTOPOINT)
  661                             ip->ip_src = satosin(&ia->ia_dstaddr)->sin_addr;
  662                 }
  663 reflect:
  664                 ICMPSTAT_INC(icps_reflect);
  665                 ICMPSTAT_INC(icps_outhist[icp->icmp_type]);
  666                 icmp_reflect(m);
  667                 return (IPPROTO_DONE);
  668 
  669         case ICMP_REDIRECT:
  670                 if (V_log_redirect) {
  671                         u_long src, dst, gw;
  672 
  673                         src = ntohl(ip->ip_src.s_addr);
  674                         dst = ntohl(icp->icmp_ip.ip_dst.s_addr);
  675                         gw = ntohl(icp->icmp_gwaddr.s_addr);
  676                         printf("icmp redirect from %d.%d.%d.%d: "
  677                                "%d.%d.%d.%d => %d.%d.%d.%d\n",
  678                                (int)(src >> 24), (int)((src >> 16) & 0xff),
  679                                (int)((src >> 8) & 0xff), (int)(src & 0xff),
  680                                (int)(dst >> 24), (int)((dst >> 16) & 0xff),
  681                                (int)((dst >> 8) & 0xff), (int)(dst & 0xff),
  682                                (int)(gw >> 24), (int)((gw >> 16) & 0xff),
  683                                (int)((gw >> 8) & 0xff), (int)(gw & 0xff));
  684                 }
  685                 /*
  686                  * RFC1812 says we must ignore ICMP redirects if we
  687                  * are acting as router.
  688                  */
  689                 if (V_drop_redirect || V_ipforwarding)
  690                         break;
  691                 if (code > 3)
  692                         goto badcode;
  693                 if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp) ||
  694                     icp->icmp_ip.ip_hl < (sizeof(struct ip) >> 2)) {
  695                         ICMPSTAT_INC(icps_badlen);
  696                         break;
  697                 }
  698                 /*
  699                  * Short circuit routing redirects to force
  700                  * immediate change in the kernel's routing
  701                  * tables.  The message is also handed to anyone
  702                  * listening on a raw socket (e.g. the routing
  703                  * daemon for use in updating its tables).
  704                  */
  705                 icmpgw.sin_addr = ip->ip_src;
  706                 icmpdst.sin_addr = icp->icmp_gwaddr;
  707 #ifdef  ICMPPRINTFS
  708                 if (icmpprintfs) {
  709                         char dstbuf[INET_ADDRSTRLEN];
  710                         char gwbuf[INET_ADDRSTRLEN];
  711 
  712                         printf("redirect dst %s to %s\n",
  713                                inet_ntoa_r(icp->icmp_ip.ip_dst, dstbuf),
  714                                inet_ntoa_r(icp->icmp_gwaddr, gwbuf));
  715                 }
  716 #endif
  717                 icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
  718 
  719                 /*
  720                  * RFC 1122 says network (code 0,2) redirects SHOULD
  721                  * be treated identically to the host redirects.
  722                  * Given that, ignore network masks.
  723                  */
  724 
  725                 /*
  726                  * Variable values:
  727                  * icmpsrc: route destination
  728                  * icmpdst: route gateway
  729                  * icmpgw: message source
  730                  */
  731 
  732                 if (icmp_verify_redirect_gateway(&icmpgw, &icmpsrc, &icmpdst,
  733                     M_GETFIB(m)) != 0) {
  734                         /* TODO: increment bad redirects here */
  735                         break;
  736                 }
  737 
  738                 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
  739                         rib_add_redirect(fibnum, (struct sockaddr *)&icmpsrc,
  740                             (struct sockaddr *)&icmpdst,
  741                             (struct sockaddr *)&icmpgw, m->m_pkthdr.rcvif,
  742                             RTF_GATEWAY, V_redirtimeout);
  743                 }
  744                 break;
  745 
  746         /*
  747          * No kernel processing for the following;
  748          * just fall through to send to raw listener.
  749          */
  750         case ICMP_ECHOREPLY:
  751         case ICMP_ROUTERADVERT:
  752         case ICMP_ROUTERSOLICIT:
  753         case ICMP_TSTAMPREPLY:
  754         case ICMP_IREQREPLY:
  755         case ICMP_MASKREPLY:
  756         case ICMP_SOURCEQUENCH:
  757         default:
  758                 break;
  759         }
  760 
  761 raw:
  762         *mp = m;
  763         rip_input(mp, offp, proto);
  764         return (IPPROTO_DONE);
  765 
  766 freeit:
  767         m_freem(m);
  768         return (IPPROTO_DONE);
  769 }
  770 
  771 /*
  772  * Reflect the ip packet back to the source
  773  */
  774 static void
  775 icmp_reflect(struct mbuf *m)
  776 {
  777         struct ip *ip = mtod(m, struct ip *);
  778         struct ifaddr *ifa;
  779         struct ifnet *ifp;
  780         struct in_ifaddr *ia;
  781         struct in_addr t;
  782         struct nhop_object *nh;
  783         struct mbuf *opts = NULL;
  784         int optlen = (ip->ip_hl << 2) - sizeof(struct ip);
  785 
  786         NET_EPOCH_ASSERT();
  787 
  788         if (IN_MULTICAST(ntohl(ip->ip_src.s_addr)) ||
  789             (IN_EXPERIMENTAL(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net240) ||
  790             (IN_ZERONET(ntohl(ip->ip_src.s_addr)) && !V_ip_allow_net0) ) {
  791                 m_freem(m);     /* Bad return address */
  792                 ICMPSTAT_INC(icps_badaddr);
  793                 goto done;      /* Ip_output() will check for broadcast */
  794         }
  795 
  796         t = ip->ip_dst;
  797         ip->ip_dst = ip->ip_src;
  798 
  799         /*
  800          * Source selection for ICMP replies:
  801          *
  802          * If the incoming packet was addressed directly to one of our
  803          * own addresses, use dst as the src for the reply.
  804          */
  805         CK_LIST_FOREACH(ia, INADDR_HASH(t.s_addr), ia_hash) {
  806                 if (t.s_addr == IA_SIN(ia)->sin_addr.s_addr) {
  807                         t = IA_SIN(ia)->sin_addr;
  808                         goto match;
  809                 }
  810         }
  811 
  812         /*
  813          * If the incoming packet was addressed to one of our broadcast
  814          * addresses, use the first non-broadcast address which corresponds
  815          * to the incoming interface.
  816          */
  817         ifp = m->m_pkthdr.rcvif;
  818         if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) {
  819                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
  820                         if (ifa->ifa_addr->sa_family != AF_INET)
  821                                 continue;
  822                         ia = ifatoia(ifa);
  823                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
  824                             t.s_addr) {
  825                                 t = IA_SIN(ia)->sin_addr;
  826                                 goto match;
  827                         }
  828                 }
  829         }
  830         /*
  831          * If the packet was transiting through us, use the address of
  832          * the interface the packet came through in.  If that interface
  833          * doesn't have a suitable IP address, the normal selection
  834          * criteria apply.
  835          */
  836         if (V_icmp_rfi && ifp != NULL) {
  837                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
  838                         if (ifa->ifa_addr->sa_family != AF_INET)
  839                                 continue;
  840                         ia = ifatoia(ifa);
  841                         t = IA_SIN(ia)->sin_addr;
  842                         goto match;
  843                 }
  844         }
  845         /*
  846          * If the incoming packet was not addressed directly to us, use
  847          * designated interface for icmp replies specified by sysctl
  848          * net.inet.icmp.reply_src (default not set). Otherwise continue
  849          * with normal source selection.
  850          */
  851         if (V_reply_src[0] != '\0' && (ifp = ifunit(V_reply_src))) {
  852                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
  853                         if (ifa->ifa_addr->sa_family != AF_INET)
  854                                 continue;
  855                         ia = ifatoia(ifa);
  856                         t = IA_SIN(ia)->sin_addr;
  857                         goto match;
  858                 }
  859         }
  860         /*
  861          * If the packet was transiting through us, use the address of
  862          * the interface that is the closest to the packet source.
  863          * When we don't have a route back to the packet source, stop here
  864          * and drop the packet.
  865          */
  866         nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_NONE, 0);
  867         if (nh == NULL) {
  868                 m_freem(m);
  869                 ICMPSTAT_INC(icps_noroute);
  870                 goto done;
  871         }
  872         t = IA_SIN(ifatoia(nh->nh_ifa))->sin_addr;
  873 match:
  874 #ifdef MAC
  875         mac_netinet_icmp_replyinplace(m);
  876 #endif
  877         ip->ip_src = t;
  878         ip->ip_ttl = V_ip_defttl;
  879 
  880         if (optlen > 0) {
  881                 u_char *cp;
  882                 int opt, cnt;
  883                 u_int len;
  884 
  885                 /*
  886                  * Retrieve any source routing from the incoming packet;
  887                  * add on any record-route or timestamp options.
  888                  */
  889                 cp = (u_char *) (ip + 1);
  890                 if ((opts = ip_srcroute(m)) == NULL &&
  891                     (opts = m_gethdr(M_NOWAIT, MT_DATA))) {
  892                         opts->m_len = sizeof(struct in_addr);
  893                         mtod(opts, struct in_addr *)->s_addr = 0;
  894                 }
  895                 if (opts) {
  896 #ifdef ICMPPRINTFS
  897                     if (icmpprintfs)
  898                             printf("icmp_reflect optlen %d rt %d => ",
  899                                 optlen, opts->m_len);
  900 #endif
  901                     for (cnt = optlen; cnt > 0; cnt -= len, cp += len) {
  902                             opt = cp[IPOPT_OPTVAL];
  903                             if (opt == IPOPT_EOL)
  904                                     break;
  905                             if (opt == IPOPT_NOP)
  906                                     len = 1;
  907                             else {
  908                                     if (cnt < IPOPT_OLEN + sizeof(*cp))
  909                                             break;
  910                                     len = cp[IPOPT_OLEN];
  911                                     if (len < IPOPT_OLEN + sizeof(*cp) ||
  912                                         len > cnt)
  913                                             break;
  914                             }
  915                             /*
  916                              * Should check for overflow, but it "can't happen"
  917                              */
  918                             if (opt == IPOPT_RR || opt == IPOPT_TS ||
  919                                 opt == IPOPT_SECURITY) {
  920                                     bcopy((caddr_t)cp,
  921                                         mtod(opts, caddr_t) + opts->m_len, len);
  922                                     opts->m_len += len;
  923                             }
  924                     }
  925                     /* Terminate & pad, if necessary */
  926                     cnt = opts->m_len % 4;
  927                     if (cnt) {
  928                             for (; cnt < 4; cnt++) {
  929                                     *(mtod(opts, caddr_t) + opts->m_len) =
  930                                         IPOPT_EOL;
  931                                     opts->m_len++;
  932                             }
  933                     }
  934 #ifdef ICMPPRINTFS
  935                     if (icmpprintfs)
  936                             printf("%d\n", opts->m_len);
  937 #endif
  938                 }
  939                 ip_stripoptions(m);
  940         }
  941         m_tag_delete_nonpersistent(m);
  942         m->m_flags &= ~(M_BCAST|M_MCAST);
  943         icmp_send(m, opts);
  944 done:
  945         if (opts)
  946                 (void)m_free(opts);
  947 }
  948 
  949 /*
  950  * Verifies if redirect message is valid, according to RFC 1122
  951  *
  952  * @src: sockaddr with address of redirect originator
  953  * @dst: sockaddr with destination in question
  954  * @gateway: new proposed gateway
  955  *
  956  * Returns 0 on success.
  957  */
  958 static int
  959 icmp_verify_redirect_gateway(struct sockaddr_in *src, struct sockaddr_in *dst,
  960     struct sockaddr_in *gateway, u_int fibnum)
  961 {
  962         struct nhop_object *nh;
  963         struct ifaddr *ifa;
  964 
  965         NET_EPOCH_ASSERT();
  966 
  967         /* Verify the gateway is directly reachable. */
  968         if ((ifa = ifa_ifwithnet((struct sockaddr *)gateway, 0, fibnum))==NULL)
  969                 return (ENETUNREACH);
  970 
  971         /* TODO: fib-aware. */
  972         if (ifa_ifwithaddr_check((struct sockaddr *)gateway))
  973                 return (EHOSTUNREACH);
  974 
  975         nh = fib4_lookup(fibnum, dst->sin_addr, 0, NHR_NONE, 0);
  976         if (nh == NULL)
  977                 return (EINVAL);
  978 
  979         /*
  980          * If the redirect isn't from our current router for this dst,
  981          * it's either old or wrong.  If it redirects us to ourselves,
  982          * we have a routing loop, perhaps as a result of an interface
  983          * going down recently.
  984          */
  985         if (!sa_equal((struct sockaddr *)src, &nh->gw_sa))
  986                 return (EINVAL);
  987         if (nh->nh_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK)
  988                 return (EINVAL);
  989 
  990         /* If host route already exists, ignore redirect. */
  991         if (nh->nh_flags & NHF_HOST)
  992                 return (EEXIST);
  993 
  994         /* If the prefix is directly reachable, ignore redirect. */
  995         if (!(nh->nh_flags & NHF_GATEWAY))
  996                 return (EEXIST);
  997 
  998         return (0);
  999 }
 1000 
 1001 /*
 1002  * Send an icmp packet back to the ip level,
 1003  * after supplying a checksum.
 1004  */
 1005 static void
 1006 icmp_send(struct mbuf *m, struct mbuf *opts)
 1007 {
 1008         struct ip *ip = mtod(m, struct ip *);
 1009         int hlen;
 1010         struct icmp *icp;
 1011 
 1012         hlen = ip->ip_hl << 2;
 1013         m->m_data += hlen;
 1014         m->m_len -= hlen;
 1015         icp = mtod(m, struct icmp *);
 1016         icp->icmp_cksum = 0;
 1017         icp->icmp_cksum = in_cksum(m, ntohs(ip->ip_len) - hlen);
 1018         m->m_data -= hlen;
 1019         m->m_len += hlen;
 1020         m->m_pkthdr.rcvif = (struct ifnet *)0;
 1021 #ifdef ICMPPRINTFS
 1022         if (icmpprintfs) {
 1023                 char dstbuf[INET_ADDRSTRLEN];
 1024                 char srcbuf[INET_ADDRSTRLEN];
 1025 
 1026                 printf("icmp_send dst %s src %s\n",
 1027                     inet_ntoa_r(ip->ip_dst, dstbuf),
 1028                     inet_ntoa_r(ip->ip_src, srcbuf));
 1029         }
 1030 #endif
 1031         (void) ip_output(m, opts, NULL, 0, NULL, NULL);
 1032 }
 1033 
 1034 /*
 1035  * Return milliseconds since 00:00 UTC in network format.
 1036  */
 1037 uint32_t
 1038 iptime(void)
 1039 {
 1040         struct timeval atv;
 1041         u_long t;
 1042 
 1043         getmicrotime(&atv);
 1044         t = (atv.tv_sec % (24*60*60)) * 1000 + atv.tv_usec / 1000;
 1045         return (htonl(t));
 1046 }
 1047 
 1048 /*
 1049  * Return the next larger or smaller MTU plateau (table from RFC 1191)
 1050  * given current value MTU.  If DIR is less than zero, a larger plateau
 1051  * is returned; otherwise, a smaller value is returned.
 1052  */
 1053 int
 1054 ip_next_mtu(int mtu, int dir)
 1055 {
 1056         static int mtutab[] = {
 1057                 65535, 32000, 17914, 8166, 4352, 2002, 1492, 1280, 1006, 508,
 1058                 296, 68, 0
 1059         };
 1060         int i, size;
 1061 
 1062         size = (sizeof mtutab) / (sizeof mtutab[0]);
 1063         if (dir >= 0) {
 1064                 for (i = 0; i < size; i++)
 1065                         if (mtu > mtutab[i])
 1066                                 return mtutab[i];
 1067         } else {
 1068                 for (i = size - 1; i >= 0; i--)
 1069                         if (mtu < mtutab[i])
 1070                                 return mtutab[i];
 1071                 if (mtu == mtutab[0])
 1072                         return mtutab[0];
 1073         }
 1074         return 0;
 1075 }
 1076 #endif /* INET */
 1077 
 1078 /*
 1079  * badport_bandlim() - check for ICMP bandwidth limit
 1080  *
 1081  *      Return 0 if it is ok to send an ICMP error response, -1 if we have
 1082  *      hit our bandwidth limit and it is not ok.
 1083  *
 1084  *      If icmplim is <= 0, the feature is disabled and 0 is returned.
 1085  *
 1086  *      For now we separate the TCP and UDP subsystems w/ different 'which'
 1087  *      values.  We may eventually remove this separation (and simplify the
 1088  *      code further).
 1089  *
 1090  *      Note that the printing of the error message is delayed so we can
 1091  *      properly print the icmp error rate that the system was trying to do
 1092  *      (i.e. 22000/100 pps, etc...).  This can cause long delays in printing
 1093  *      the 'final' error, but it doesn't make sense to solve the printing
 1094  *      delay with more complex code.
 1095  */
 1096 struct icmp_rate {
 1097         const char *descr;
 1098         struct counter_rate cr;
 1099 };
 1100 VNET_DEFINE_STATIC(struct icmp_rate, icmp_rates[BANDLIM_MAX]) = {
 1101         { "icmp unreach response" },
 1102         { "icmp ping response" },
 1103         { "icmp tstamp response" },
 1104         { "closed port RST response" },
 1105         { "open port RST response" },
 1106         { "icmp6 unreach response" },
 1107         { "sctp ootb response" }
 1108 };
 1109 #define V_icmp_rates    VNET(icmp_rates)
 1110 
 1111 static void
 1112 icmp_bandlimit_init(void)
 1113 {
 1114 
 1115         for (int i = 0; i < BANDLIM_MAX; i++) {
 1116                 V_icmp_rates[i].cr.cr_rate = counter_u64_alloc(M_WAITOK);
 1117                 V_icmp_rates[i].cr.cr_ticks = ticks;
 1118         }
 1119 }
 1120 VNET_SYSINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY,
 1121     icmp_bandlimit_init, NULL);
 1122 
 1123 static void
 1124 icmp_bandlimit_uninit(void)
 1125 {
 1126 
 1127         for (int i = 0; i < BANDLIM_MAX; i++)
 1128                 counter_u64_free(V_icmp_rates[i].cr.cr_rate);
 1129 }
 1130 VNET_SYSUNINIT(icmp_bandlimit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
 1131     icmp_bandlimit_uninit, NULL);
 1132 
 1133 int
 1134 badport_bandlim(int which)
 1135 {
 1136         int64_t pps;
 1137 
 1138         if (V_icmplim == 0 || which == BANDLIM_UNLIMITED)
 1139                 return (0);
 1140 
 1141         KASSERT(which >= 0 && which < BANDLIM_MAX,
 1142             ("%s: which %d", __func__, which));
 1143 
 1144         if ((V_icmplim + V_icmplim_curr_jitter) <= 0)
 1145                 V_icmplim_curr_jitter = -V_icmplim + 1;
 1146 
 1147         pps = counter_ratecheck(&V_icmp_rates[which].cr, V_icmplim +
 1148             V_icmplim_curr_jitter);
 1149         if (pps > 0) {
 1150                 /*
 1151                  * Adjust limit +/- to jitter the measurement to deny a
 1152                  * side-channel port scan as in CVE-2020-25705
 1153                  */
 1154                 if (V_icmplim_jitter > 0) {
 1155                         int32_t inc =
 1156                             arc4random_uniform(V_icmplim_jitter * 2 +1)
 1157                             - V_icmplim_jitter;
 1158 
 1159                         V_icmplim_curr_jitter = inc;
 1160                 }
 1161         }
 1162         if (pps == -1)
 1163                 return (-1);
 1164         if (pps > 0 && V_icmplim_output)
 1165                 log(LOG_NOTICE, "Limiting %s from %jd to %d packets/sec\n",
 1166                     V_icmp_rates[which].descr, (intmax_t )pps, V_icmplim +
 1167                     V_icmplim_curr_jitter);
 1168         return (0);
 1169 }

Cache object: 019e6f136462b16e086910cd581a84d9


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.