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_input.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  * Copyright (c) 1982, 1986, 1988, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    8  * 1. Redistributions of source code must retain the above copyright
    9  *    notice, this list of conditions and the following disclaimer.
   10  * 2. Redistributions in binary form must reproduce the above copyright
   11  *    notice, this list of conditions and the following disclaimer in the
   12  *    documentation and/or other materials provided with the distribution.
   13  * 4. Neither the name of the University nor the names of its contributors
   14  *    may be used to endorse or promote products derived from this software
   15  *    without specific prior written permission.
   16  *
   17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   27  * SUCH DAMAGE.
   28  *
   29  *      @(#)ip_input.c  8.2 (Berkeley) 1/4/94
   30  */
   31 
   32 #include <sys/cdefs.h>
   33 __FBSDID("$FreeBSD: releng/7.4/sys/netinet/ip_input.c 216125 2010-12-02 16:27:28Z bz $");
   34 
   35 #include "opt_bootp.h"
   36 #include "opt_ipfw.h"
   37 #include "opt_ipstealth.h"
   38 #include "opt_ipsec.h"
   39 #include "opt_mac.h"
   40 #include "opt_carp.h"
   41 
   42 #include <sys/param.h>
   43 #include <sys/systm.h>
   44 #include <sys/callout.h>
   45 #include <sys/mbuf.h>
   46 #include <sys/malloc.h>
   47 #include <sys/domain.h>
   48 #include <sys/protosw.h>
   49 #include <sys/socket.h>
   50 #include <sys/time.h>
   51 #include <sys/kernel.h>
   52 #include <sys/syslog.h>
   53 #include <sys/sysctl.h>
   54 
   55 #include <net/pfil.h>
   56 #include <net/if.h>
   57 #include <net/if_types.h>
   58 #include <net/if_var.h>
   59 #include <net/if_dl.h>
   60 #include <net/route.h>
   61 #include <net/netisr.h>
   62 
   63 #include <netinet/in.h>
   64 #include <netinet/in_var.h>
   65 #include <netinet/ip.h>
   66 #include <netinet/in_pcb.h>
   67 #include <netinet/ip_var.h>
   68 #include <netinet/ip_icmp.h>
   69 #include <netinet/ip_options.h>
   70 #include <machine/in_cksum.h>
   71 #ifdef DEV_CARP
   72 #include <netinet/ip_carp.h>
   73 #endif
   74 #ifdef IPSEC
   75 #include <netinet/ip_ipsec.h>
   76 #endif /* IPSEC */
   77 
   78 #include <sys/socketvar.h>
   79 
   80 /* XXX: Temporary until ipfw_ether and ipfw_bridge are converted. */
   81 #include <netinet/ip_fw.h>
   82 #include <netinet/ip_dummynet.h>
   83 
   84 #include <security/mac/mac_framework.h>
   85 
   86 int rsvp_on = 0;
   87 
   88 int     ipforwarding = 0;
   89 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_RW,
   90     &ipforwarding, 0, "Enable IP forwarding between interfaces");
   91 
   92 static int      ipsendredirects = 1; /* XXX */
   93 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_RW,
   94     &ipsendredirects, 0, "Enable sending IP redirects");
   95 
   96 int     ip_defttl = IPDEFTTL;
   97 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW,
   98     &ip_defttl, 0, "Maximum TTL on IP packets");
   99 
  100 static int      ip_keepfaith = 0;
  101 SYSCTL_INT(_net_inet_ip, IPCTL_KEEPFAITH, keepfaith, CTLFLAG_RW,
  102     &ip_keepfaith,      0,
  103     "Enable packet capture for FAITH IPv4->IPv6 translater daemon");
  104 
  105 static int      ip_sendsourcequench = 0;
  106 SYSCTL_INT(_net_inet_ip, OID_AUTO, sendsourcequench, CTLFLAG_RW,
  107     &ip_sendsourcequench, 0,
  108     "Enable the transmission of source quench packets");
  109 
  110 int     ip_do_randomid = 0;
  111 SYSCTL_INT(_net_inet_ip, OID_AUTO, random_id, CTLFLAG_RW,
  112     &ip_do_randomid, 0,
  113     "Assign random ip_id values");
  114 
  115 /*
  116  * XXX - Setting ip_checkinterface mostly implements the receive side of
  117  * the Strong ES model described in RFC 1122, but since the routing table
  118  * and transmit implementation do not implement the Strong ES model,
  119  * setting this to 1 results in an odd hybrid.
  120  *
  121  * XXX - ip_checkinterface currently must be disabled if you use ipnat
  122  * to translate the destination address to another local interface.
  123  *
  124  * XXX - ip_checkinterface must be disabled if you add IP aliases
  125  * to the loopback interface instead of the interface where the
  126  * packets for those addresses are received.
  127  */
  128 static int      ip_checkinterface = 0;
  129 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_RW,
  130     &ip_checkinterface, 0, "Verify packet arrives on correct interface");
  131 
  132 struct pfil_head inet_pfil_hook;        /* Packet filter hooks */
  133 
  134 static struct   ifqueue ipintrq;
  135 
  136 extern  struct domain inetdomain;
  137 extern  struct protosw inetsw[];
  138 u_char  ip_protox[IPPROTO_MAX];
  139 struct  in_ifaddrhead in_ifaddrhead;            /* first inet address */
  140 struct  in_ifaddrhashhead *in_ifaddrhashtbl;    /* inet addr hash table  */
  141 u_long  in_ifaddrhmask;                         /* mask for hash table */
  142 
  143 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, CTLFLAG_RW,
  144     &ipintrq.ifq_maxlen, 0, "Maximum size of the IP input queue");
  145 SYSCTL_INT(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, CTLFLAG_RD,
  146     &ipintrq.ifq_drops, 0,
  147     "Number of packets dropped from the IP input queue");
  148 
  149 struct ipstat ipstat;
  150 SYSCTL_STRUCT(_net_inet_ip, IPCTL_STATS, stats, CTLFLAG_RW,
  151     &ipstat, ipstat, "IP statistics (struct ipstat, netinet/ip_var.h)");
  152 
  153 /*
  154  * IP datagram reassembly.
  155  */
  156 #define IPREASS_NHASH_LOG2      6
  157 #define IPREASS_NHASH           (1 << IPREASS_NHASH_LOG2)
  158 #define IPREASS_HMASK           (IPREASS_NHASH - 1)
  159 #define IPREASS_HASH(x,y) \
  160         (((((x) & 0xF) | ((((x) >> 8) & 0xF) << 4)) ^ (y)) & IPREASS_HMASK)
  161 
  162 static uma_zone_t ipq_zone;
  163 static TAILQ_HEAD(ipqhead, ipq) ipq[IPREASS_NHASH];
  164 static struct mtx ipqlock;
  165 
  166 #define IPQ_LOCK()      mtx_lock(&ipqlock)
  167 #define IPQ_UNLOCK()    mtx_unlock(&ipqlock)
  168 #define IPQ_LOCK_INIT() mtx_init(&ipqlock, "ipqlock", NULL, MTX_DEF)
  169 #define IPQ_LOCK_ASSERT()       mtx_assert(&ipqlock, MA_OWNED)
  170 
  171 static void     maxnipq_update(void);
  172 static void     ipq_zone_change(void *);
  173 
  174 static int      maxnipq;        /* Administrative limit on # reass queues. */
  175 static int      nipq = 0;       /* Total # of reass queues */
  176 SYSCTL_INT(_net_inet_ip, OID_AUTO, fragpackets, CTLFLAG_RD,
  177     &nipq, 0, "Current number of IPv4 fragment reassembly queue entries");
  178 
  179 static int      maxfragsperpacket;
  180 SYSCTL_INT(_net_inet_ip, OID_AUTO, maxfragsperpacket, CTLFLAG_RW,
  181     &maxfragsperpacket, 0,
  182     "Maximum number of IPv4 fragments allowed per packet");
  183 
  184 struct callout  ipport_tick_callout;
  185 
  186 #ifdef IPCTL_DEFMTU
  187 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
  188     &ip_mtu, 0, "Default MTU");
  189 #endif
  190 
  191 #ifdef IPSTEALTH
  192 int     ipstealth = 0;
  193 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_RW,
  194     &ipstealth, 0, "IP stealth mode, no TTL decrementation on forwarding");
  195 #endif
  196 
  197 /*
  198  * ipfw_ether and ipfw_bridge hooks.
  199  * XXX: Temporary until those are converted to pfil_hooks as well.
  200  */
  201 ip_fw_chk_t *ip_fw_chk_ptr = NULL;
  202 ip_dn_io_t *ip_dn_io_ptr = NULL;
  203 int fw_one_pass = 1;
  204 
  205 static void     ip_freef(struct ipqhead *, struct ipq *);
  206 
  207 /*
  208  * IP initialization: fill in IP protocol switch table.
  209  * All protocols not implemented in kernel go to raw IP protocol handler.
  210  */
  211 void
  212 ip_init(void)
  213 {
  214         struct protosw *pr;
  215         int i;
  216 
  217         TAILQ_INIT(&in_ifaddrhead);
  218         in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &in_ifaddrhmask);
  219         pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
  220         if (pr == NULL)
  221                 panic("ip_init: PF_INET not found");
  222 
  223         /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
  224         for (i = 0; i < IPPROTO_MAX; i++)
  225                 ip_protox[i] = pr - inetsw;
  226         /*
  227          * Cycle through IP protocols and put them into the appropriate place
  228          * in ip_protox[].
  229          */
  230         for (pr = inetdomain.dom_protosw;
  231             pr < inetdomain.dom_protoswNPROTOSW; pr++)
  232                 if (pr->pr_domain->dom_family == PF_INET &&
  233                     pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
  234                         /* Be careful to only index valid IP protocols. */
  235                         if (pr->pr_protocol < IPPROTO_MAX)
  236                                 ip_protox[pr->pr_protocol] = pr - inetsw;
  237                 }
  238 
  239         /* Initialize packet filter hooks. */
  240         inet_pfil_hook.ph_type = PFIL_TYPE_AF;
  241         inet_pfil_hook.ph_af = AF_INET;
  242         if ((i = pfil_head_register(&inet_pfil_hook)) != 0)
  243                 printf("%s: WARNING: unable to register pfil hook, "
  244                         "error %d\n", __func__, i);
  245 
  246         /* Initialize IP reassembly queue. */
  247         IPQ_LOCK_INIT();
  248         for (i = 0; i < IPREASS_NHASH; i++)
  249             TAILQ_INIT(&ipq[i]);
  250         maxnipq = nmbclusters / 32;
  251         maxfragsperpacket = 16;
  252         ipq_zone = uma_zcreate("ipq", sizeof(struct ipq), NULL, NULL, NULL,
  253             NULL, UMA_ALIGN_PTR, 0);
  254         maxnipq_update();
  255 
  256         /* Start ipport_tick. */
  257         callout_init(&ipport_tick_callout, CALLOUT_MPSAFE);
  258         ipport_tick(NULL);
  259         EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL,
  260                 SHUTDOWN_PRI_DEFAULT);
  261         EVENTHANDLER_REGISTER(nmbclusters_change, ipq_zone_change,
  262                 NULL, EVENTHANDLER_PRI_ANY);
  263 
  264         /* Initialize various other remaining things. */
  265         ip_id = time_second & 0xffff;
  266         ipintrq.ifq_maxlen = ifqmaxlen;
  267         mtx_init(&ipintrq.ifq_mtx, "ip_inq", NULL, MTX_DEF);
  268         netisr_register(NETISR_IP, ip_input, &ipintrq, NETISR_MPSAFE);
  269 }
  270 
  271 void
  272 ip_fini(void *xtp)
  273 {
  274 
  275         callout_stop(&ipport_tick_callout);
  276 }
  277 
  278 /*
  279  * Ip input routine.  Checksum and byte swap header.  If fragmented
  280  * try to reassemble.  Process options.  Pass to next level.
  281  */
  282 void
  283 ip_input(struct mbuf *m)
  284 {
  285         struct ip *ip = NULL;
  286         struct in_ifaddr *ia = NULL;
  287         struct ifaddr *ifa;
  288         int    checkif, hlen = 0;
  289         u_short sum;
  290         int dchg = 0;                           /* dest changed after fw */
  291         struct in_addr odst;                    /* original dst address */
  292 
  293         M_ASSERTPKTHDR(m);
  294 
  295         if (m->m_flags & M_FASTFWD_OURS) {
  296                 /*
  297                  * Firewall or NAT changed destination to local.
  298                  * We expect ip_len and ip_off to be in host byte order.
  299                  */
  300                 m->m_flags &= ~M_FASTFWD_OURS;
  301                 /* Set up some basics that will be used later. */
  302                 ip = mtod(m, struct ip *);
  303                 hlen = ip->ip_hl << 2;
  304                 goto ours;
  305         }
  306 
  307         ipstat.ips_total++;
  308 
  309         if (m->m_pkthdr.len < sizeof(struct ip))
  310                 goto tooshort;
  311 
  312         if (m->m_len < sizeof (struct ip) &&
  313             (m = m_pullup(m, sizeof (struct ip))) == NULL) {
  314                 ipstat.ips_toosmall++;
  315                 return;
  316         }
  317         ip = mtod(m, struct ip *);
  318 
  319         if (ip->ip_v != IPVERSION) {
  320                 ipstat.ips_badvers++;
  321                 goto bad;
  322         }
  323 
  324         hlen = ip->ip_hl << 2;
  325         if (hlen < sizeof(struct ip)) { /* minimum header length */
  326                 ipstat.ips_badhlen++;
  327                 goto bad;
  328         }
  329         if (hlen > m->m_len) {
  330                 if ((m = m_pullup(m, hlen)) == NULL) {
  331                         ipstat.ips_badhlen++;
  332                         return;
  333                 }
  334                 ip = mtod(m, struct ip *);
  335         }
  336 
  337         /* 127/8 must not appear on wire - RFC1122 */
  338         if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET ||
  339             (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
  340                 if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
  341                         ipstat.ips_badaddr++;
  342                         goto bad;
  343                 }
  344         }
  345 
  346         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
  347                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
  348         } else {
  349                 if (hlen == sizeof(struct ip)) {
  350                         sum = in_cksum_hdr(ip);
  351                 } else {
  352                         sum = in_cksum(m, hlen);
  353                 }
  354         }
  355         if (sum) {
  356                 ipstat.ips_badsum++;
  357                 goto bad;
  358         }
  359 
  360 #ifdef ALTQ
  361         if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0)
  362                 /* packet is dropped by traffic conditioner */
  363                 return;
  364 #endif
  365 
  366         /*
  367          * Convert fields to host representation.
  368          */
  369         ip->ip_len = ntohs(ip->ip_len);
  370         if (ip->ip_len < hlen) {
  371                 ipstat.ips_badlen++;
  372                 goto bad;
  373         }
  374         ip->ip_off = ntohs(ip->ip_off);
  375 
  376         /*
  377          * Check that the amount of data in the buffers
  378          * is as at least much as the IP header would have us expect.
  379          * Trim mbufs if longer than we expect.
  380          * Drop packet if shorter than we expect.
  381          */
  382         if (m->m_pkthdr.len < ip->ip_len) {
  383 tooshort:
  384                 ipstat.ips_tooshort++;
  385                 goto bad;
  386         }
  387         if (m->m_pkthdr.len > ip->ip_len) {
  388                 if (m->m_len == m->m_pkthdr.len) {
  389                         m->m_len = ip->ip_len;
  390                         m->m_pkthdr.len = ip->ip_len;
  391                 } else
  392                         m_adj(m, ip->ip_len - m->m_pkthdr.len);
  393         }
  394 #ifdef IPSEC
  395         /*
  396          * Bypass packet filtering for packets from a tunnel (gif).
  397          */
  398         if (ip_ipsec_filtertunnel(m))
  399                 goto passin;
  400 #endif /* IPSEC */
  401 
  402         /*
  403          * Run through list of hooks for input packets.
  404          *
  405          * NB: Beware of the destination address changing (e.g.
  406          *     by NAT rewriting).  When this happens, tell
  407          *     ip_forward to do the right thing.
  408          */
  409 
  410         /* Jump over all PFIL processing if hooks are not active. */
  411         if (!PFIL_HOOKED(&inet_pfil_hook))
  412                 goto passin;
  413 
  414         odst = ip->ip_dst;
  415         if (pfil_run_hooks(&inet_pfil_hook, &m, m->m_pkthdr.rcvif,
  416             PFIL_IN, NULL) != 0)
  417                 return;
  418         if (m == NULL)                  /* consumed by filter */
  419                 return;
  420 
  421         ip = mtod(m, struct ip *);
  422         dchg = (odst.s_addr != ip->ip_dst.s_addr);
  423 
  424 #ifdef IPFIREWALL_FORWARD
  425         if (m->m_flags & M_FASTFWD_OURS) {
  426                 m->m_flags &= ~M_FASTFWD_OURS;
  427                 goto ours;
  428         }
  429         if ((dchg = (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL)) != 0) {
  430                 /*
  431                  * Directly ship on the packet.  This allows to forward packets
  432                  * that were destined for us to some other directly connected
  433                  * host.
  434                  */
  435                 ip_forward(m, dchg);
  436                 return;
  437         }
  438 #endif /* IPFIREWALL_FORWARD */
  439 
  440 passin:
  441         /*
  442          * Process options and, if not destined for us,
  443          * ship it on.  ip_dooptions returns 1 when an
  444          * error was detected (causing an icmp message
  445          * to be sent and the original packet to be freed).
  446          */
  447         if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
  448                 return;
  449 
  450         /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
  451          * matter if it is destined to another node, or whether it is 
  452          * a multicast one, RSVP wants it! and prevents it from being forwarded
  453          * anywhere else. Also checks if the rsvp daemon is running before
  454          * grabbing the packet.
  455          */
  456         if (rsvp_on && ip->ip_p==IPPROTO_RSVP) 
  457                 goto ours;
  458 
  459         /*
  460          * Check our list of addresses, to see if the packet is for us.
  461          * If we don't have any addresses, assume any unicast packet
  462          * we receive might be for us (and let the upper layers deal
  463          * with it).
  464          */
  465         if (TAILQ_EMPTY(&in_ifaddrhead) &&
  466             (m->m_flags & (M_MCAST|M_BCAST)) == 0)
  467                 goto ours;
  468 
  469         /*
  470          * Enable a consistency check between the destination address
  471          * and the arrival interface for a unicast packet (the RFC 1122
  472          * strong ES model) if IP forwarding is disabled and the packet
  473          * is not locally generated and the packet is not subject to
  474          * 'ipfw fwd'.
  475          *
  476          * XXX - Checking also should be disabled if the destination
  477          * address is ipnat'ed to a different interface.
  478          *
  479          * XXX - Checking is incompatible with IP aliases added
  480          * to the loopback interface instead of the interface where
  481          * the packets are received.
  482          *
  483          * XXX - This is the case for carp vhost IPs as well so we
  484          * insert a workaround. If the packet got here, we already
  485          * checked with carp_iamatch() and carp_forus().
  486          */
  487         checkif = ip_checkinterface && (ipforwarding == 0) && 
  488             m->m_pkthdr.rcvif != NULL &&
  489             ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) &&
  490 #ifdef DEV_CARP
  491             !m->m_pkthdr.rcvif->if_carp &&
  492 #endif
  493             (dchg == 0);
  494 
  495         /*
  496          * Check for exact addresses in the hash bucket.
  497          */
  498         LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
  499                 /*
  500                  * If the address matches, verify that the packet
  501                  * arrived via the correct interface if checking is
  502                  * enabled.
  503                  */
  504                 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 
  505                     (!checkif || ia->ia_ifp == m->m_pkthdr.rcvif))
  506                         goto ours;
  507         }
  508         /*
  509          * Check for broadcast addresses.
  510          *
  511          * Only accept broadcast packets that arrive via the matching
  512          * interface.  Reception of forwarded directed broadcasts would
  513          * be handled via ip_forward() and ether_output() with the loopback
  514          * into the stack for SIMPLEX interfaces handled by ether_output().
  515          */
  516         if (m->m_pkthdr.rcvif != NULL &&
  517             m->m_pkthdr.rcvif->if_flags & IFF_BROADCAST) {
  518                 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
  519                         if (ifa->ifa_addr->sa_family != AF_INET)
  520                                 continue;
  521                         ia = ifatoia(ifa);
  522                         if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
  523                             ip->ip_dst.s_addr)
  524                                 goto ours;
  525                         if (ia->ia_netbroadcast.s_addr == ip->ip_dst.s_addr)
  526                                 goto ours;
  527 #ifdef BOOTP_COMPAT
  528                         if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY)
  529                                 goto ours;
  530 #endif
  531                 }
  532         }
  533         /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */
  534         if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) {
  535                 ipstat.ips_cantforward++;
  536                 m_freem(m);
  537                 return;
  538         }
  539         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
  540                 struct in_multi *inm;
  541                 if (ip_mrouter) {
  542                         /*
  543                          * If we are acting as a multicast router, all
  544                          * incoming multicast packets are passed to the
  545                          * kernel-level multicast forwarding function.
  546                          * The packet is returned (relatively) intact; if
  547                          * ip_mforward() returns a non-zero value, the packet
  548                          * must be discarded, else it may be accepted below.
  549                          */
  550                         if (ip_mforward &&
  551                             ip_mforward(ip, m->m_pkthdr.rcvif, m, 0) != 0) {
  552                                 ipstat.ips_cantforward++;
  553                                 m_freem(m);
  554                                 return;
  555                         }
  556 
  557                         /*
  558                          * The process-level routing daemon needs to receive
  559                          * all multicast IGMP packets, whether or not this
  560                          * host belongs to their destination groups.
  561                          */
  562                         if (ip->ip_p == IPPROTO_IGMP)
  563                                 goto ours;
  564                         ipstat.ips_forward++;
  565                 }
  566                 /*
  567                  * See if we belong to the destination multicast group on the
  568                  * arrival interface.
  569                  */
  570                 IN_MULTI_LOCK();
  571                 IN_LOOKUP_MULTI(ip->ip_dst, m->m_pkthdr.rcvif, inm);
  572                 IN_MULTI_UNLOCK();
  573                 if (inm == NULL) {
  574                         ipstat.ips_notmember++;
  575                         m_freem(m);
  576                         return;
  577                 }
  578                 goto ours;
  579         }
  580         if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
  581                 goto ours;
  582         if (ip->ip_dst.s_addr == INADDR_ANY)
  583                 goto ours;
  584 
  585         /*
  586          * FAITH(Firewall Aided Internet Translator)
  587          */
  588         if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
  589                 if (ip_keepfaith) {
  590                         if (ip->ip_p == IPPROTO_TCP || ip->ip_p == IPPROTO_ICMP) 
  591                                 goto ours;
  592                 }
  593                 m_freem(m);
  594                 return;
  595         }
  596 
  597         /*
  598          * Not for us; forward if possible and desirable.
  599          */
  600         if (ipforwarding == 0) {
  601                 ipstat.ips_cantforward++;
  602                 m_freem(m);
  603         } else {
  604 #ifdef IPSEC
  605                 if (ip_ipsec_fwd(m))
  606                         goto bad;
  607 #endif /* IPSEC */
  608                 ip_forward(m, dchg);
  609         }
  610         return;
  611 
  612 ours:
  613 #ifdef IPSTEALTH
  614         /*
  615          * IPSTEALTH: Process non-routing options only
  616          * if the packet is destined for us.
  617          */
  618         if (ipstealth && hlen > sizeof (struct ip) &&
  619             ip_dooptions(m, 1))
  620                 return;
  621 #endif /* IPSTEALTH */
  622 
  623         /* Count the packet in the ip address stats */
  624         if (ia != NULL) {
  625                 ia->ia_ifa.if_ipackets++;
  626                 ia->ia_ifa.if_ibytes += m->m_pkthdr.len;
  627         }
  628 
  629         /*
  630          * Attempt reassembly; if it succeeds, proceed.
  631          * ip_reass() will return a different mbuf.
  632          */
  633         if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
  634                 m = ip_reass(m);
  635                 if (m == NULL)
  636                         return;
  637                 ip = mtod(m, struct ip *);
  638                 /* Get the header length of the reassembled packet */
  639                 hlen = ip->ip_hl << 2;
  640         }
  641 
  642         /*
  643          * Further protocols expect the packet length to be w/o the
  644          * IP header.
  645          */
  646         ip->ip_len -= hlen;
  647 
  648 #ifdef IPSEC
  649         /*
  650          * enforce IPsec policy checking if we are seeing last header.
  651          * note that we do not visit this with protocols with pcb layer
  652          * code - like udp/tcp/raw ip.
  653          */
  654         if (ip_ipsec_input(m))
  655                 goto bad;
  656 #endif /* IPSEC */
  657 
  658         /*
  659          * Switch out to protocol's input routine.
  660          */
  661         ipstat.ips_delivered++;
  662 
  663         (*inetsw[ip_protox[ip->ip_p]].pr_input)(m, hlen);
  664         return;
  665 bad:
  666         m_freem(m);
  667 }
  668 
  669 /*
  670  * After maxnipq has been updated, propagate the change to UMA.  The UMA zone
  671  * max has slightly different semantics than the sysctl, for historical
  672  * reasons.
  673  */
  674 static void
  675 maxnipq_update(void)
  676 {
  677 
  678         /*
  679          * -1 for unlimited allocation.
  680          */
  681         if (maxnipq < 0)
  682                 uma_zone_set_max(ipq_zone, 0);
  683         /*
  684          * Positive number for specific bound.
  685          */
  686         if (maxnipq > 0)
  687                 uma_zone_set_max(ipq_zone, maxnipq);
  688         /*
  689          * Zero specifies no further fragment queue allocation -- set the
  690          * bound very low, but rely on implementation elsewhere to actually
  691          * prevent allocation and reclaim current queues.
  692          */
  693         if (maxnipq == 0)
  694                 uma_zone_set_max(ipq_zone, 1);
  695 }
  696 
  697 static void
  698 ipq_zone_change(void *tag)
  699 {
  700 
  701         if (maxnipq > 0 && maxnipq < (nmbclusters / 32)) {
  702                 maxnipq = nmbclusters / 32;
  703                 maxnipq_update();
  704         }
  705 }
  706 
  707 static int
  708 sysctl_maxnipq(SYSCTL_HANDLER_ARGS)
  709 {
  710         int error, i;
  711 
  712         i = maxnipq;
  713         error = sysctl_handle_int(oidp, &i, 0, req);
  714         if (error || !req->newptr)
  715                 return (error);
  716 
  717         /*
  718          * XXXRW: Might be a good idea to sanity check the argument and place
  719          * an extreme upper bound.
  720          */
  721         if (i < -1)
  722                 return (EINVAL);
  723         maxnipq = i;
  724         maxnipq_update();
  725         return (0);
  726 }
  727 
  728 SYSCTL_PROC(_net_inet_ip, OID_AUTO, maxfragpackets, CTLTYPE_INT|CTLFLAG_RW,
  729     NULL, 0, sysctl_maxnipq, "I",
  730     "Maximum number of IPv4 fragment reassembly queue entries");
  731 
  732 /*
  733  * Take incoming datagram fragment and try to reassemble it into
  734  * whole datagram.  If the argument is the first fragment or one
  735  * in between the function will return NULL and store the mbuf
  736  * in the fragment chain.  If the argument is the last fragment
  737  * the packet will be reassembled and the pointer to the new
  738  * mbuf returned for further processing.  Only m_tags attached
  739  * to the first packet/fragment are preserved.
  740  * The IP header is *NOT* adjusted out of iplen.
  741  */
  742 struct mbuf *
  743 ip_reass(struct mbuf *m)
  744 {
  745         struct ip *ip;
  746         struct mbuf *p, *q, *nq, *t;
  747         struct ipq *fp = NULL;
  748         struct ipqhead *head;
  749         int i, hlen, next;
  750         u_int8_t ecn, ecn0;
  751         u_short hash;
  752 
  753         /* If maxnipq or maxfragsperpacket are 0, never accept fragments. */
  754         if (maxnipq == 0 || maxfragsperpacket == 0) {
  755                 ipstat.ips_fragments++;
  756                 ipstat.ips_fragdropped++;
  757                 m_freem(m);
  758                 return (NULL);
  759         }
  760 
  761         ip = mtod(m, struct ip *);
  762         hlen = ip->ip_hl << 2;
  763 
  764         hash = IPREASS_HASH(ip->ip_src.s_addr, ip->ip_id);
  765         head = &ipq[hash];
  766         IPQ_LOCK();
  767 
  768         /*
  769          * Look for queue of fragments
  770          * of this datagram.
  771          */
  772         TAILQ_FOREACH(fp, head, ipq_list)
  773                 if (ip->ip_id == fp->ipq_id &&
  774                     ip->ip_src.s_addr == fp->ipq_src.s_addr &&
  775                     ip->ip_dst.s_addr == fp->ipq_dst.s_addr &&
  776 #ifdef MAC
  777                     mac_fragment_match(m, fp) &&
  778 #endif
  779                     ip->ip_p == fp->ipq_p)
  780                         goto found;
  781 
  782         fp = NULL;
  783 
  784         /*
  785          * Attempt to trim the number of allocated fragment queues if it
  786          * exceeds the administrative limit.
  787          */
  788         if ((nipq > maxnipq) && (maxnipq > 0)) {
  789                 /*
  790                  * drop something from the tail of the current queue
  791                  * before proceeding further
  792                  */
  793                 struct ipq *q = TAILQ_LAST(head, ipqhead);
  794                 if (q == NULL) {   /* gak */
  795                         for (i = 0; i < IPREASS_NHASH; i++) {
  796                                 struct ipq *r = TAILQ_LAST(&ipq[i], ipqhead);
  797                                 if (r) {
  798                                         ipstat.ips_fragtimeout += r->ipq_nfrags;
  799                                         ip_freef(&ipq[i], r);
  800                                         break;
  801                                 }
  802                         }
  803                 } else {
  804                         ipstat.ips_fragtimeout += q->ipq_nfrags;
  805                         ip_freef(head, q);
  806                 }
  807         }
  808 
  809 found:
  810         /*
  811          * Adjust ip_len to not reflect header,
  812          * convert offset of this to bytes.
  813          */
  814         ip->ip_len -= hlen;
  815         if (ip->ip_off & IP_MF) {
  816                 /*
  817                  * Make sure that fragments have a data length
  818                  * that's a non-zero multiple of 8 bytes.
  819                  */
  820                 if (ip->ip_len == 0 || (ip->ip_len & 0x7) != 0) {
  821                         ipstat.ips_toosmall++; /* XXX */
  822                         goto dropfrag;
  823                 }
  824                 m->m_flags |= M_FRAG;
  825         } else
  826                 m->m_flags &= ~M_FRAG;
  827         ip->ip_off <<= 3;
  828 
  829 
  830         /*
  831          * Attempt reassembly; if it succeeds, proceed.
  832          * ip_reass() will return a different mbuf.
  833          */
  834         ipstat.ips_fragments++;
  835         m->m_pkthdr.header = ip;
  836 
  837         /* Previous ip_reass() started here. */
  838         /*
  839          * Presence of header sizes in mbufs
  840          * would confuse code below.
  841          */
  842         m->m_data += hlen;
  843         m->m_len -= hlen;
  844 
  845         /*
  846          * If first fragment to arrive, create a reassembly queue.
  847          */
  848         if (fp == NULL) {
  849                 fp = uma_zalloc(ipq_zone, M_NOWAIT);
  850                 if (fp == NULL)
  851                         goto dropfrag;
  852 #ifdef MAC
  853                 if (mac_init_ipq(fp, M_NOWAIT) != 0) {
  854                         uma_zfree(ipq_zone, fp);
  855                         fp = NULL;
  856                         goto dropfrag;
  857                 }
  858                 mac_create_ipq(m, fp);
  859 #endif
  860                 TAILQ_INSERT_HEAD(head, fp, ipq_list);
  861                 nipq++;
  862                 fp->ipq_nfrags = 1;
  863                 fp->ipq_ttl = IPFRAGTTL;
  864                 fp->ipq_p = ip->ip_p;
  865                 fp->ipq_id = ip->ip_id;
  866                 fp->ipq_src = ip->ip_src;
  867                 fp->ipq_dst = ip->ip_dst;
  868                 fp->ipq_frags = m;
  869                 m->m_nextpkt = NULL;
  870                 goto done;
  871         } else {
  872                 fp->ipq_nfrags++;
  873 #ifdef MAC
  874                 mac_update_ipq(m, fp);
  875 #endif
  876         }
  877 
  878 #define GETIP(m)        ((struct ip*)((m)->m_pkthdr.header))
  879 
  880         /*
  881          * Handle ECN by comparing this segment with the first one;
  882          * if CE is set, do not lose CE.
  883          * drop if CE and not-ECT are mixed for the same packet.
  884          */
  885         ecn = ip->ip_tos & IPTOS_ECN_MASK;
  886         ecn0 = GETIP(fp->ipq_frags)->ip_tos & IPTOS_ECN_MASK;
  887         if (ecn == IPTOS_ECN_CE) {
  888                 if (ecn0 == IPTOS_ECN_NOTECT)
  889                         goto dropfrag;
  890                 if (ecn0 != IPTOS_ECN_CE)
  891                         GETIP(fp->ipq_frags)->ip_tos |= IPTOS_ECN_CE;
  892         }
  893         if (ecn == IPTOS_ECN_NOTECT && ecn0 != IPTOS_ECN_NOTECT)
  894                 goto dropfrag;
  895 
  896         /*
  897          * Find a segment which begins after this one does.
  898          */
  899         for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt)
  900                 if (GETIP(q)->ip_off > ip->ip_off)
  901                         break;
  902 
  903         /*
  904          * If there is a preceding segment, it may provide some of
  905          * our data already.  If so, drop the data from the incoming
  906          * segment.  If it provides all of our data, drop us, otherwise
  907          * stick new segment in the proper place.
  908          *
  909          * If some of the data is dropped from the the preceding
  910          * segment, then it's checksum is invalidated.
  911          */
  912         if (p) {
  913                 i = GETIP(p)->ip_off + GETIP(p)->ip_len - ip->ip_off;
  914                 if (i > 0) {
  915                         if (i >= ip->ip_len)
  916                                 goto dropfrag;
  917                         m_adj(m, i);
  918                         m->m_pkthdr.csum_flags = 0;
  919                         ip->ip_off += i;
  920                         ip->ip_len -= i;
  921                 }
  922                 m->m_nextpkt = p->m_nextpkt;
  923                 p->m_nextpkt = m;
  924         } else {
  925                 m->m_nextpkt = fp->ipq_frags;
  926                 fp->ipq_frags = m;
  927         }
  928 
  929         /*
  930          * While we overlap succeeding segments trim them or,
  931          * if they are completely covered, dequeue them.
  932          */
  933         for (; q != NULL && ip->ip_off + ip->ip_len > GETIP(q)->ip_off;
  934              q = nq) {
  935                 i = (ip->ip_off + ip->ip_len) - GETIP(q)->ip_off;
  936                 if (i < GETIP(q)->ip_len) {
  937                         GETIP(q)->ip_len -= i;
  938                         GETIP(q)->ip_off += i;
  939                         m_adj(q, i);
  940                         q->m_pkthdr.csum_flags = 0;
  941                         break;
  942                 }
  943                 nq = q->m_nextpkt;
  944                 m->m_nextpkt = nq;
  945                 ipstat.ips_fragdropped++;
  946                 fp->ipq_nfrags--;
  947                 m_freem(q);
  948         }
  949 
  950         /*
  951          * Check for complete reassembly and perform frag per packet
  952          * limiting.
  953          *
  954          * Frag limiting is performed here so that the nth frag has
  955          * a chance to complete the packet before we drop the packet.
  956          * As a result, n+1 frags are actually allowed per packet, but
  957          * only n will ever be stored. (n = maxfragsperpacket.)
  958          *
  959          */
  960         next = 0;
  961         for (p = NULL, q = fp->ipq_frags; q; p = q, q = q->m_nextpkt) {
  962                 if (GETIP(q)->ip_off != next) {
  963                         if (fp->ipq_nfrags > maxfragsperpacket) {
  964                                 ipstat.ips_fragdropped += fp->ipq_nfrags;
  965                                 ip_freef(head, fp);
  966                         }
  967                         goto done;
  968                 }
  969                 next += GETIP(q)->ip_len;
  970         }
  971         /* Make sure the last packet didn't have the IP_MF flag */
  972         if (p->m_flags & M_FRAG) {
  973                 if (fp->ipq_nfrags > maxfragsperpacket) {
  974                         ipstat.ips_fragdropped += fp->ipq_nfrags;
  975                         ip_freef(head, fp);
  976                 }
  977                 goto done;
  978         }
  979 
  980         /*
  981          * Reassembly is complete.  Make sure the packet is a sane size.
  982          */
  983         q = fp->ipq_frags;
  984         ip = GETIP(q);
  985         if (next + (ip->ip_hl << 2) > IP_MAXPACKET) {
  986                 ipstat.ips_toolong++;
  987                 ipstat.ips_fragdropped += fp->ipq_nfrags;
  988                 ip_freef(head, fp);
  989                 goto done;
  990         }
  991 
  992         /*
  993          * Concatenate fragments.
  994          */
  995         m = q;
  996         t = m->m_next;
  997         m->m_next = NULL;
  998         m_cat(m, t);
  999         nq = q->m_nextpkt;
 1000         q->m_nextpkt = NULL;
 1001         for (q = nq; q != NULL; q = nq) {
 1002                 nq = q->m_nextpkt;
 1003                 q->m_nextpkt = NULL;
 1004                 m->m_pkthdr.csum_flags &= q->m_pkthdr.csum_flags;
 1005                 m->m_pkthdr.csum_data += q->m_pkthdr.csum_data;
 1006                 m_cat(m, q);
 1007         }
 1008         /*
 1009          * In order to do checksumming faster we do 'end-around carry' here
 1010          * (and not in for{} loop), though it implies we are not going to
 1011          * reassemble more than 64k fragments.
 1012          */
 1013         m->m_pkthdr.csum_data =
 1014             (m->m_pkthdr.csum_data & 0xffff) + (m->m_pkthdr.csum_data >> 16);
 1015 #ifdef MAC
 1016         mac_create_datagram_from_ipq(fp, m);
 1017         mac_destroy_ipq(fp);
 1018 #endif
 1019 
 1020         /*
 1021          * Create header for new ip packet by modifying header of first
 1022          * packet;  dequeue and discard fragment reassembly header.
 1023          * Make header visible.
 1024          */
 1025         ip->ip_len = (ip->ip_hl << 2) + next;
 1026         ip->ip_src = fp->ipq_src;
 1027         ip->ip_dst = fp->ipq_dst;
 1028         TAILQ_REMOVE(head, fp, ipq_list);
 1029         nipq--;
 1030         uma_zfree(ipq_zone, fp);
 1031         m->m_len += (ip->ip_hl << 2);
 1032         m->m_data -= (ip->ip_hl << 2);
 1033         /* some debugging cruft by sklower, below, will go away soon */
 1034         if (m->m_flags & M_PKTHDR)      /* XXX this should be done elsewhere */
 1035                 m_fixhdr(m);
 1036         ipstat.ips_reassembled++;
 1037         IPQ_UNLOCK();
 1038         return (m);
 1039 
 1040 dropfrag:
 1041         ipstat.ips_fragdropped++;
 1042         if (fp != NULL)
 1043                 fp->ipq_nfrags--;
 1044         m_freem(m);
 1045 done:
 1046         IPQ_UNLOCK();
 1047         return (NULL);
 1048 
 1049 #undef GETIP
 1050 }
 1051 
 1052 /*
 1053  * Free a fragment reassembly header and all
 1054  * associated datagrams.
 1055  */
 1056 static void
 1057 ip_freef(struct ipqhead *fhp, struct ipq *fp)
 1058 {
 1059         struct mbuf *q;
 1060 
 1061         IPQ_LOCK_ASSERT();
 1062 
 1063         while (fp->ipq_frags) {
 1064                 q = fp->ipq_frags;
 1065                 fp->ipq_frags = q->m_nextpkt;
 1066                 m_freem(q);
 1067         }
 1068         TAILQ_REMOVE(fhp, fp, ipq_list);
 1069         uma_zfree(ipq_zone, fp);
 1070         nipq--;
 1071 }
 1072 
 1073 /*
 1074  * IP timer processing;
 1075  * if a timer expires on a reassembly
 1076  * queue, discard it.
 1077  */
 1078 void
 1079 ip_slowtimo(void)
 1080 {
 1081         struct ipq *fp;
 1082         int i;
 1083 
 1084         IPQ_LOCK();
 1085         for (i = 0; i < IPREASS_NHASH; i++) {
 1086                 for(fp = TAILQ_FIRST(&ipq[i]); fp;) {
 1087                         struct ipq *fpp;
 1088 
 1089                         fpp = fp;
 1090                         fp = TAILQ_NEXT(fp, ipq_list);
 1091                         if(--fpp->ipq_ttl == 0) {
 1092                                 ipstat.ips_fragtimeout += fpp->ipq_nfrags;
 1093                                 ip_freef(&ipq[i], fpp);
 1094                         }
 1095                 }
 1096         }
 1097         /*
 1098          * If we are over the maximum number of fragments
 1099          * (due to the limit being lowered), drain off
 1100          * enough to get down to the new limit.
 1101          */
 1102         if (maxnipq >= 0 && nipq > maxnipq) {
 1103                 for (i = 0; i < IPREASS_NHASH; i++) {
 1104                         while (nipq > maxnipq && !TAILQ_EMPTY(&ipq[i])) {
 1105                                 ipstat.ips_fragdropped +=
 1106                                     TAILQ_FIRST(&ipq[i])->ipq_nfrags;
 1107                                 ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
 1108                         }
 1109                 }
 1110         }
 1111         IPQ_UNLOCK();
 1112 }
 1113 
 1114 /*
 1115  * Drain off all datagram fragments.
 1116  */
 1117 void
 1118 ip_drain(void)
 1119 {
 1120         int     i;
 1121 
 1122         IPQ_LOCK();
 1123         for (i = 0; i < IPREASS_NHASH; i++) {
 1124                 while(!TAILQ_EMPTY(&ipq[i])) {
 1125                         ipstat.ips_fragdropped +=
 1126                             TAILQ_FIRST(&ipq[i])->ipq_nfrags;
 1127                         ip_freef(&ipq[i], TAILQ_FIRST(&ipq[i]));
 1128                 }
 1129         }
 1130         IPQ_UNLOCK();
 1131         in_rtqdrain();
 1132 }
 1133 
 1134 /*
 1135  * The protocol to be inserted into ip_protox[] must be already registered
 1136  * in inetsw[], either statically or through pf_proto_register().
 1137  */
 1138 int
 1139 ipproto_register(short ipproto)
 1140 {
 1141         struct protosw *pr;
 1142 
 1143         /* Sanity checks. */
 1144         if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
 1145                 return (EPROTONOSUPPORT);
 1146 
 1147         /*
 1148          * The protocol slot must not be occupied by another protocol
 1149          * already.  An index pointing to IPPROTO_RAW is unused.
 1150          */
 1151         pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
 1152         if (pr == NULL)
 1153                 return (EPFNOSUPPORT);
 1154         if (ip_protox[ipproto] != pr - inetsw)  /* IPPROTO_RAW */
 1155                 return (EEXIST);
 1156 
 1157         /* Find the protocol position in inetsw[] and set the index. */
 1158         for (pr = inetdomain.dom_protosw;
 1159              pr < inetdomain.dom_protoswNPROTOSW; pr++) {
 1160                 if (pr->pr_domain->dom_family == PF_INET &&
 1161                     pr->pr_protocol && pr->pr_protocol == ipproto) {
 1162                         ip_protox[pr->pr_protocol] = pr - inetsw;
 1163                         return (0);
 1164                 }
 1165         }
 1166         return (EPROTONOSUPPORT);
 1167 }
 1168 
 1169 int
 1170 ipproto_unregister(short ipproto)
 1171 {
 1172         struct protosw *pr;
 1173 
 1174         /* Sanity checks. */
 1175         if (ipproto <= 0 || ipproto >= IPPROTO_MAX)
 1176                 return (EPROTONOSUPPORT);
 1177 
 1178         /* Check if the protocol was indeed registered. */
 1179         pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW);
 1180         if (pr == NULL)
 1181                 return (EPFNOSUPPORT);
 1182         if (ip_protox[ipproto] == pr - inetsw)  /* IPPROTO_RAW */
 1183                 return (ENOENT);
 1184 
 1185         /* Reset the protocol slot to IPPROTO_RAW. */
 1186         ip_protox[ipproto] = pr - inetsw;
 1187         return (0);
 1188 }
 1189 
 1190 /*
 1191  * Given address of next destination (final or next hop),
 1192  * return internet address info of interface to be used to get there.
 1193  */
 1194 struct in_ifaddr *
 1195 ip_rtaddr(struct in_addr dst, u_int fibnum)
 1196 {
 1197         struct route sro;
 1198         struct sockaddr_in *sin;
 1199         struct in_ifaddr *ifa;
 1200 
 1201         bzero(&sro, sizeof(sro));
 1202         sin = (struct sockaddr_in *)&sro.ro_dst;
 1203         sin->sin_family = AF_INET;
 1204         sin->sin_len = sizeof(*sin);
 1205         sin->sin_addr = dst;
 1206         in_rtalloc_ign(&sro, RTF_CLONING, fibnum);
 1207 
 1208         if (sro.ro_rt == NULL)
 1209                 return (NULL);
 1210 
 1211         ifa = ifatoia(sro.ro_rt->rt_ifa);
 1212         RTFREE(sro.ro_rt);
 1213         return (ifa);
 1214 }
 1215 
 1216 u_char inetctlerrmap[PRC_NCMDS] = {
 1217         0,              0,              0,              0,
 1218         0,              EMSGSIZE,       EHOSTDOWN,      EHOSTUNREACH,
 1219         EHOSTUNREACH,   EHOSTUNREACH,   ECONNREFUSED,   ECONNREFUSED,
 1220         EMSGSIZE,       EHOSTUNREACH,   0,              0,
 1221         0,              0,              EHOSTUNREACH,   0,
 1222         ENOPROTOOPT,    ECONNREFUSED
 1223 };
 1224 
 1225 /*
 1226  * Forward a packet.  If some error occurs return the sender
 1227  * an icmp packet.  Note we can't always generate a meaningful
 1228  * icmp message because icmp doesn't have a large enough repertoire
 1229  * of codes and types.
 1230  *
 1231  * If not forwarding, just drop the packet.  This could be confusing
 1232  * if ipforwarding was zero but some routing protocol was advancing
 1233  * us as a gateway to somewhere.  However, we must let the routing
 1234  * protocol deal with that.
 1235  *
 1236  * The srcrt parameter indicates whether the packet is being forwarded
 1237  * via a source route.
 1238  */
 1239 void
 1240 ip_forward(struct mbuf *m, int srcrt)
 1241 {
 1242         struct ip *ip = mtod(m, struct ip *);
 1243         struct in_ifaddr *ia = NULL;
 1244         struct mbuf *mcopy;
 1245         struct in_addr dest;
 1246         struct route ro;
 1247         int error, type = 0, code = 0, mtu = 0;
 1248 
 1249         if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
 1250                 ipstat.ips_cantforward++;
 1251                 m_freem(m);
 1252                 return;
 1253         }
 1254 #ifdef IPSTEALTH
 1255         if (!ipstealth) {
 1256 #endif
 1257                 if (ip->ip_ttl <= IPTTLDEC) {
 1258                         icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS,
 1259                             0, 0);
 1260                         return;
 1261                 }
 1262 #ifdef IPSTEALTH
 1263         }
 1264 #endif
 1265 
 1266         ia = ip_rtaddr(ip->ip_dst, M_GETFIB(m));
 1267         if (!srcrt && ia == NULL) {
 1268                 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0);
 1269                 return;
 1270         }
 1271 
 1272         /*
 1273          * Save the IP header and at most 8 bytes of the payload,
 1274          * in case we need to generate an ICMP message to the src.
 1275          *
 1276          * XXX this can be optimized a lot by saving the data in a local
 1277          * buffer on the stack (72 bytes at most), and only allocating the
 1278          * mbuf if really necessary. The vast majority of the packets
 1279          * are forwarded without having to send an ICMP back (either
 1280          * because unnecessary, or because rate limited), so we are
 1281          * really we are wasting a lot of work here.
 1282          *
 1283          * We don't use m_copy() because it might return a reference
 1284          * to a shared cluster. Both this function and ip_output()
 1285          * assume exclusive access to the IP header in `m', so any
 1286          * data in a cluster may change before we reach icmp_error().
 1287          */
 1288         MGETHDR(mcopy, M_DONTWAIT, m->m_type);
 1289         if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_DONTWAIT)) {
 1290                 /*
 1291                  * It's probably ok if the pkthdr dup fails (because
 1292                  * the deep copy of the tag chain failed), but for now
 1293                  * be conservative and just discard the copy since
 1294                  * code below may some day want the tags.
 1295                  */
 1296                 m_free(mcopy);
 1297                 mcopy = NULL;
 1298         }
 1299         if (mcopy != NULL) {
 1300                 mcopy->m_len = min(ip->ip_len, M_TRAILINGSPACE(mcopy));
 1301                 mcopy->m_pkthdr.len = mcopy->m_len;
 1302                 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
 1303         }
 1304 
 1305 #ifdef IPSTEALTH
 1306         if (!ipstealth) {
 1307 #endif
 1308                 ip->ip_ttl -= IPTTLDEC;
 1309 #ifdef IPSTEALTH
 1310         }
 1311 #endif
 1312 
 1313         /*
 1314          * If forwarding packet using same interface that it came in on,
 1315          * perhaps should send a redirect to sender to shortcut a hop.
 1316          * Only send redirect if source is sending directly to us,
 1317          * and if packet was not source routed (or has any options).
 1318          * Also, don't send redirect if forwarding using a default route
 1319          * or a route modified by a redirect.
 1320          */
 1321         dest.s_addr = 0;
 1322         if (!srcrt && ipsendredirects && ia->ia_ifp == m->m_pkthdr.rcvif) {
 1323                 struct sockaddr_in *sin;
 1324                 struct rtentry *rt;
 1325 
 1326                 bzero(&ro, sizeof(ro));
 1327                 sin = (struct sockaddr_in *)&ro.ro_dst;
 1328                 sin->sin_family = AF_INET;
 1329                 sin->sin_len = sizeof(*sin);
 1330                 sin->sin_addr = ip->ip_dst;
 1331                 in_rtalloc_ign(&ro, RTF_CLONING, M_GETFIB(m));
 1332 
 1333                 rt = ro.ro_rt;
 1334 
 1335                 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 &&
 1336                     satosin(rt_key(rt))->sin_addr.s_addr != 0) {
 1337 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa))
 1338                         u_long src = ntohl(ip->ip_src.s_addr);
 1339 
 1340                         if (RTA(rt) &&
 1341                             (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) {
 1342                                 if (rt->rt_flags & RTF_GATEWAY)
 1343                                         dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr;
 1344                                 else
 1345                                         dest.s_addr = ip->ip_dst.s_addr;
 1346                                 /* Router requirements says to only send host redirects */
 1347                                 type = ICMP_REDIRECT;
 1348                                 code = ICMP_REDIRECT_HOST;
 1349                         }
 1350                 }
 1351                 if (rt)
 1352                         RTFREE(rt);
 1353         }
 1354 
 1355         /*
 1356          * Try to cache the route MTU from ip_output so we can consider it for
 1357          * the ICMP_UNREACH_NEEDFRAG "Next-Hop MTU" field described in RFC1191.
 1358          */
 1359         bzero(&ro, sizeof(ro));
 1360 
 1361         error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
 1362 
 1363         if (error == EMSGSIZE && ro.ro_rt)
 1364                 mtu = ro.ro_rt->rt_rmx.rmx_mtu;
 1365         if (ro.ro_rt)
 1366                 RTFREE(ro.ro_rt);
 1367 
 1368         if (error)
 1369                 ipstat.ips_cantforward++;
 1370         else {
 1371                 ipstat.ips_forward++;
 1372                 if (type)
 1373                         ipstat.ips_redirectsent++;
 1374                 else {
 1375                         if (mcopy)
 1376                                 m_freem(mcopy);
 1377                         return;
 1378                 }
 1379         }
 1380         if (mcopy == NULL)
 1381                 return;
 1382 
 1383         switch (error) {
 1384 
 1385         case 0:                         /* forwarded, but need redirect */
 1386                 /* type, code set above */
 1387                 break;
 1388 
 1389         case ENETUNREACH:               /* shouldn't happen, checked above */
 1390         case EHOSTUNREACH:
 1391         case ENETDOWN:
 1392         case EHOSTDOWN:
 1393         default:
 1394                 type = ICMP_UNREACH;
 1395                 code = ICMP_UNREACH_HOST;
 1396                 break;
 1397 
 1398         case EMSGSIZE:
 1399                 type = ICMP_UNREACH;
 1400                 code = ICMP_UNREACH_NEEDFRAG;
 1401 
 1402 #ifdef IPSEC
 1403                 /* 
 1404                  * If IPsec is configured for this path,
 1405                  * override any possibly mtu value set by ip_output.
 1406                  */ 
 1407                 mtu = ip_ipsec_mtu(mcopy, mtu);
 1408 #endif /* IPSEC */
 1409                 /*
 1410                  * If the MTU was set before make sure we are below the
 1411                  * interface MTU.
 1412                  * If the MTU wasn't set before use the interface mtu or
 1413                  * fall back to the next smaller mtu step compared to the
 1414                  * current packet size.
 1415                  */
 1416                 if (mtu != 0) {
 1417                         if (ia != NULL)
 1418                                 mtu = min(mtu, ia->ia_ifp->if_mtu);
 1419                 } else {
 1420                         if (ia != NULL)
 1421                                 mtu = ia->ia_ifp->if_mtu;
 1422                         else
 1423                                 mtu = ip_next_mtu(ip->ip_len, 0);
 1424                 }
 1425                 ipstat.ips_cantfrag++;
 1426                 break;
 1427 
 1428         case ENOBUFS:
 1429                 /*
 1430                  * A router should not generate ICMP_SOURCEQUENCH as
 1431                  * required in RFC1812 Requirements for IP Version 4 Routers.
 1432                  * Source quench could be a big problem under DoS attacks,
 1433                  * or if the underlying interface is rate-limited.
 1434                  * Those who need source quench packets may re-enable them
 1435                  * via the net.inet.ip.sendsourcequench sysctl.
 1436                  */
 1437                 if (ip_sendsourcequench == 0) {
 1438                         m_freem(mcopy);
 1439                         return;
 1440                 } else {
 1441                         type = ICMP_SOURCEQUENCH;
 1442                         code = 0;
 1443                 }
 1444                 break;
 1445 
 1446         case EACCES:                    /* ipfw denied packet */
 1447                 m_freem(mcopy);
 1448                 return;
 1449         }
 1450         icmp_error(mcopy, type, code, dest.s_addr, mtu);
 1451 }
 1452 
 1453 void
 1454 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
 1455     struct mbuf *m)
 1456 {
 1457         if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) {
 1458                 struct bintime bt;
 1459 
 1460                 bintime(&bt);
 1461                 if (inp->inp_socket->so_options & SO_BINTIME) {
 1462                         *mp = sbcreatecontrol((caddr_t) &bt, sizeof(bt),
 1463                         SCM_BINTIME, SOL_SOCKET);
 1464                         if (*mp)
 1465                                 mp = &(*mp)->m_next;
 1466                 }
 1467                 if (inp->inp_socket->so_options & SO_TIMESTAMP) {
 1468                         struct timeval tv;
 1469 
 1470                         bintime2timeval(&bt, &tv);
 1471                         *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
 1472                                 SCM_TIMESTAMP, SOL_SOCKET);
 1473                         if (*mp)
 1474                                 mp = &(*mp)->m_next;
 1475                 }
 1476         }
 1477         if (inp->inp_flags & INP_RECVDSTADDR) {
 1478                 *mp = sbcreatecontrol((caddr_t) &ip->ip_dst,
 1479                     sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP);
 1480                 if (*mp)
 1481                         mp = &(*mp)->m_next;
 1482         }
 1483         if (inp->inp_flags & INP_RECVTTL) {
 1484                 *mp = sbcreatecontrol((caddr_t) &ip->ip_ttl,
 1485                     sizeof(u_char), IP_RECVTTL, IPPROTO_IP);
 1486                 if (*mp)
 1487                         mp = &(*mp)->m_next;
 1488         }
 1489 #ifdef notyet
 1490         /* XXX
 1491          * Moving these out of udp_input() made them even more broken
 1492          * than they already were.
 1493          */
 1494         /* options were tossed already */
 1495         if (inp->inp_flags & INP_RECVOPTS) {
 1496                 *mp = sbcreatecontrol((caddr_t) opts_deleted_above,
 1497                     sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP);
 1498                 if (*mp)
 1499                         mp = &(*mp)->m_next;
 1500         }
 1501         /* ip_srcroute doesn't do what we want here, need to fix */
 1502         if (inp->inp_flags & INP_RECVRETOPTS) {
 1503                 *mp = sbcreatecontrol((caddr_t) ip_srcroute(m),
 1504                     sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP);
 1505                 if (*mp)
 1506                         mp = &(*mp)->m_next;
 1507         }
 1508 #endif
 1509         if (inp->inp_flags & INP_RECVIF) {
 1510                 struct ifnet *ifp;
 1511                 struct sdlbuf {
 1512                         struct sockaddr_dl sdl;
 1513                         u_char  pad[32];
 1514                 } sdlbuf;
 1515                 struct sockaddr_dl *sdp;
 1516                 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
 1517 
 1518                 if (((ifp = m->m_pkthdr.rcvif)) 
 1519                 && ( ifp->if_index && (ifp->if_index <= if_index))) {
 1520                         sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
 1521                         /*
 1522                          * Change our mind and don't try copy.
 1523                          */
 1524                         if ((sdp->sdl_family != AF_LINK)
 1525                         || (sdp->sdl_len > sizeof(sdlbuf))) {
 1526                                 goto makedummy;
 1527                         }
 1528                         bcopy(sdp, sdl2, sdp->sdl_len);
 1529                 } else {
 1530 makedummy:      
 1531                         sdl2->sdl_len
 1532                                 = offsetof(struct sockaddr_dl, sdl_data[0]);
 1533                         sdl2->sdl_family = AF_LINK;
 1534                         sdl2->sdl_index = 0;
 1535                         sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
 1536                 }
 1537                 *mp = sbcreatecontrol((caddr_t) sdl2, sdl2->sdl_len,
 1538                         IP_RECVIF, IPPROTO_IP);
 1539                 if (*mp)
 1540                         mp = &(*mp)->m_next;
 1541         }
 1542 }
 1543 
 1544 /*
 1545  * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
 1546  * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
 1547  * locking.  This code remains in ip_input.c as ip_mroute.c is optionally
 1548  * compiled.
 1549  */
 1550 static int ip_rsvp_on;
 1551 struct socket *ip_rsvpd;
 1552 int
 1553 ip_rsvp_init(struct socket *so)
 1554 {
 1555         if (so->so_type != SOCK_RAW ||
 1556             so->so_proto->pr_protocol != IPPROTO_RSVP)
 1557                 return EOPNOTSUPP;
 1558 
 1559         if (ip_rsvpd != NULL)
 1560                 return EADDRINUSE;
 1561 
 1562         ip_rsvpd = so;
 1563         /*
 1564          * This may seem silly, but we need to be sure we don't over-increment
 1565          * the RSVP counter, in case something slips up.
 1566          */
 1567         if (!ip_rsvp_on) {
 1568                 ip_rsvp_on = 1;
 1569                 rsvp_on++;
 1570         }
 1571 
 1572         return 0;
 1573 }
 1574 
 1575 int
 1576 ip_rsvp_done(void)
 1577 {
 1578         ip_rsvpd = NULL;
 1579         /*
 1580          * This may seem silly, but we need to be sure we don't over-decrement
 1581          * the RSVP counter, in case something slips up.
 1582          */
 1583         if (ip_rsvp_on) {
 1584                 ip_rsvp_on = 0;
 1585                 rsvp_on--;
 1586         }
 1587         return 0;
 1588 }
 1589 
 1590 void
 1591 rsvp_input(struct mbuf *m, int off)     /* XXX must fixup manually */
 1592 {
 1593         if (rsvp_input_p) { /* call the real one if loaded */
 1594                 rsvp_input_p(m, off);
 1595                 return;
 1596         }
 1597 
 1598         /* Can still get packets with rsvp_on = 0 if there is a local member
 1599          * of the group to which the RSVP packet is addressed.  But in this
 1600          * case we want to throw the packet away.
 1601          */
 1602         
 1603         if (!rsvp_on) {
 1604                 m_freem(m);
 1605                 return;
 1606         }
 1607 
 1608         if (ip_rsvpd != NULL) { 
 1609                 rip_input(m, off);
 1610                 return;
 1611         }
 1612         /* Drop the packet */
 1613         m_freem(m);
 1614 }

Cache object: 51834f23a19d1458a6e7ee581649c8bc


[ 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.