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_divert.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 
   32 #include <sys/cdefs.h>
   33 __FBSDID("$FreeBSD$");
   34 
   35 #include "opt_inet.h"
   36 #include "opt_inet6.h"
   37 #include "opt_sctp.h"
   38 #ifndef INET
   39 #error "IPDIVERT requires INET"
   40 #endif
   41 
   42 #include <sys/param.h>
   43 #include <sys/eventhandler.h>
   44 #include <sys/kernel.h>
   45 #include <sys/lock.h>
   46 #include <sys/malloc.h>
   47 #include <sys/mbuf.h>
   48 #include <sys/module.h>
   49 #include <sys/kernel.h>
   50 #include <sys/priv.h>
   51 #include <sys/proc.h>
   52 #include <sys/protosw.h>
   53 #include <sys/socket.h>
   54 #include <sys/socketvar.h>
   55 #include <sys/sysctl.h>
   56 #include <net/vnet.h>
   57 
   58 #include <net/if.h>
   59 #include <net/if_var.h>
   60 #include <net/netisr.h> 
   61 
   62 #include <netinet/in.h>
   63 #include <netinet/in_pcb.h>
   64 #include <netinet/in_systm.h>
   65 #include <netinet/in_var.h>
   66 #include <netinet/ip.h>
   67 #include <netinet/ip_var.h>
   68 #ifdef INET6
   69 #include <netinet/ip6.h>
   70 #include <netinet6/ip6_var.h>
   71 #endif
   72 #if defined(SCTP) || defined(SCTP_SUPPORT)
   73 #include <netinet/sctp_crc32.h>
   74 #endif
   75 
   76 #include <security/mac/mac_framework.h>
   77 /*
   78  * Divert sockets
   79  */
   80 
   81 /*
   82  * Allocate enough space to hold a full IP packet
   83  */
   84 #define DIVSNDQ         (65536 + 100)
   85 #define DIVRCVQ         (65536 + 100)
   86 
   87 /*
   88  * Divert sockets work in conjunction with ipfw or other packet filters,
   89  * see the divert(4) manpage for features.
   90  * Packets are selected by the packet filter and tagged with an
   91  * MTAG_IPFW_RULE tag carrying the 'divert port' number (as set by
   92  * the packet filter) and information on the matching filter rule for
   93  * subsequent reinjection. The divert_port is used to put the packet
   94  * on the corresponding divert socket, while the rule number is passed
   95  * up (at least partially) as the sin_port in the struct sockaddr.
   96  *
   97  * Packets written to the divert socket carry in sin_addr a
   98  * destination address, and in sin_port the number of the filter rule
   99  * after which to continue processing.
  100  * If the destination address is INADDR_ANY, the packet is treated as
  101  * as outgoing and sent to ip_output(); otherwise it is treated as
  102  * incoming and sent to ip_input().
  103  * Further, sin_zero carries some information on the interface,
  104  * which can be used in the reinject -- see comments in the code.
  105  *
  106  * On reinjection, processing in ip_input() and ip_output()
  107  * will be exactly the same as for the original packet, except that
  108  * packet filter processing will start at the rule number after the one
  109  * written in the sin_port (ipfw does not allow a rule #0, so sin_port=0
  110  * will apply the entire ruleset to the packet).
  111  */
  112 
  113 /* Internal variables. */
  114 VNET_DEFINE_STATIC(struct inpcbhead, divcb);
  115 VNET_DEFINE_STATIC(struct inpcbinfo, divcbinfo);
  116 
  117 #define V_divcb                         VNET(divcb)
  118 #define V_divcbinfo                     VNET(divcbinfo)
  119 
  120 static u_long   div_sendspace = DIVSNDQ;        /* XXX sysctl ? */
  121 static u_long   div_recvspace = DIVRCVQ;        /* XXX sysctl ? */
  122 
  123 static eventhandler_tag ip_divert_event_tag;
  124 
  125 /*
  126  * Initialize divert connection block queue.
  127  */
  128 static void
  129 div_zone_change(void *tag)
  130 {
  131 
  132         uma_zone_set_max(V_divcbinfo.ipi_zone, maxsockets);
  133 }
  134 
  135 static int
  136 div_inpcb_init(void *mem, int size, int flags)
  137 {
  138         struct inpcb *inp = mem;
  139 
  140         INP_LOCK_INIT(inp, "inp", "divinp");
  141         return (0);
  142 }
  143 
  144 static void
  145 div_init(void)
  146 {
  147 
  148         /*
  149          * XXX We don't use the hash list for divert IP, but it's easier to
  150          * allocate one-entry hash lists than it is to check all over the
  151          * place for hashbase == NULL.
  152          */
  153         in_pcbinfo_init(&V_divcbinfo, "div", &V_divcb, 1, 1, "divcb",
  154             div_inpcb_init, IPI_HASHFIELDS_NONE);
  155 }
  156 
  157 static void
  158 div_destroy(void *unused __unused)
  159 {
  160 
  161         in_pcbinfo_destroy(&V_divcbinfo);
  162 }
  163 VNET_SYSUNINIT(divert, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY,
  164     div_destroy, NULL);
  165 
  166 /*
  167  * IPPROTO_DIVERT is not in the real IP protocol number space; this
  168  * function should never be called.  Just in case, drop any packets.
  169  */
  170 static int
  171 div_input(struct mbuf **mp, int *offp, int proto)
  172 {
  173         struct mbuf *m = *mp;
  174 
  175         KMOD_IPSTAT_INC(ips_noproto);
  176         m_freem(m);
  177         return (IPPROTO_DONE);
  178 }
  179 
  180 /*
  181  * Divert a packet by passing it up to the divert socket at port 'port'.
  182  *
  183  * Setup generic address and protocol structures for div_input routine,
  184  * then pass them along with mbuf chain.
  185  */
  186 static void
  187 divert_packet(struct mbuf *m, int incoming)
  188 {
  189         struct ip *ip;
  190         struct inpcb *inp;
  191         struct socket *sa;
  192         u_int16_t nport;
  193         struct sockaddr_in divsrc;
  194         struct m_tag *mtag;
  195         struct epoch_tracker et;
  196 
  197         mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL);
  198         if (mtag == NULL) {
  199                 m_freem(m);
  200                 return;
  201         }
  202         /* Assure header */
  203         if (m->m_len < sizeof(struct ip) &&
  204             (m = m_pullup(m, sizeof(struct ip))) == NULL)
  205                 return;
  206         ip = mtod(m, struct ip *);
  207 
  208         /* Delayed checksums are currently not compatible with divert. */
  209         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
  210                 in_delayed_cksum(m);
  211                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
  212         }
  213 #if defined(SCTP) || defined(SCTP_SUPPORT)
  214         if (m->m_pkthdr.csum_flags & CSUM_SCTP) {
  215                 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
  216                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
  217         }
  218 #endif
  219 #ifdef INET6
  220         if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
  221                 in6_delayed_cksum(m, m->m_pkthdr.len -
  222                     sizeof(struct ip6_hdr), sizeof(struct ip6_hdr));
  223                 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
  224         }
  225 #if defined(SCTP) || defined(SCTP_SUPPORT)
  226         if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
  227                 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
  228                 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
  229         }
  230 #endif
  231 #endif /* INET6 */
  232         bzero(&divsrc, sizeof(divsrc));
  233         divsrc.sin_len = sizeof(divsrc);
  234         divsrc.sin_family = AF_INET;
  235         /* record matching rule, in host format */
  236         divsrc.sin_port = ((struct ipfw_rule_ref *)(mtag+1))->rulenum;
  237         /*
  238          * Record receive interface address, if any.
  239          * But only for incoming packets.
  240          */
  241         if (incoming) {
  242                 struct ifaddr *ifa;
  243                 struct ifnet *ifp;
  244 
  245                 /* Sanity check */
  246                 M_ASSERTPKTHDR(m);
  247 
  248                 /* Find IP address for receive interface */
  249                 ifp = m->m_pkthdr.rcvif;
  250                 if_addr_rlock(ifp);
  251                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
  252                         if (ifa->ifa_addr->sa_family != AF_INET)
  253                                 continue;
  254                         divsrc.sin_addr =
  255                             ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
  256                         break;
  257                 }
  258                 if_addr_runlock(ifp);
  259         }
  260         /*
  261          * Record the incoming interface name whenever we have one.
  262          */
  263         if (m->m_pkthdr.rcvif) {
  264                 /*
  265                  * Hide the actual interface name in there in the 
  266                  * sin_zero array. XXX This needs to be moved to a
  267                  * different sockaddr type for divert, e.g.
  268                  * sockaddr_div with multiple fields like 
  269                  * sockaddr_dl. Presently we have only 7 bytes
  270                  * but that will do for now as most interfaces
  271                  * are 4 or less + 2 or less bytes for unit.
  272                  * There is probably a faster way of doing this,
  273                  * possibly taking it from the sockaddr_dl on the iface.
  274                  * This solves the problem of a P2P link and a LAN interface
  275                  * having the same address, which can result in the wrong
  276                  * interface being assigned to the packet when fed back
  277                  * into the divert socket. Theoretically if the daemon saves
  278                  * and re-uses the sockaddr_in as suggested in the man pages,
  279                  * this iface name will come along for the ride.
  280                  * (see div_output for the other half of this.)
  281                  */ 
  282                 strlcpy(divsrc.sin_zero, m->m_pkthdr.rcvif->if_xname,
  283                     sizeof(divsrc.sin_zero));
  284         }
  285 
  286         /* Put packet on socket queue, if any */
  287         sa = NULL;
  288         nport = htons((u_int16_t)(((struct ipfw_rule_ref *)(mtag+1))->info));
  289         INP_INFO_RLOCK_ET(&V_divcbinfo, et);
  290         CK_LIST_FOREACH(inp, &V_divcb, inp_list) {
  291                 /* XXX why does only one socket match? */
  292                 if (inp->inp_lport == nport) {
  293                         INP_RLOCK(inp);
  294                         if (__predict_false(inp->inp_flags2 & INP_FREED)) {
  295                                 INP_RUNLOCK(inp);
  296                                 continue;
  297                         }
  298                         sa = inp->inp_socket;
  299                         SOCKBUF_LOCK(&sa->so_rcv);
  300                         if (sbappendaddr_locked(&sa->so_rcv,
  301                             (struct sockaddr *)&divsrc, m,
  302                             (struct mbuf *)0) == 0) {
  303                                 soroverflow_locked(sa);
  304                                 sa = NULL;      /* force mbuf reclaim below */
  305                         } else
  306                                 sorwakeup_locked(sa);
  307                         INP_RUNLOCK(inp);
  308                         break;
  309                 }
  310         }
  311         INP_INFO_RUNLOCK_ET(&V_divcbinfo, et);
  312         if (sa == NULL) {
  313                 m_freem(m);
  314                 KMOD_IPSTAT_INC(ips_noproto);
  315                 KMOD_IPSTAT_DEC(ips_delivered);
  316         }
  317 }
  318 
  319 /*
  320  * Deliver packet back into the IP processing machinery.
  321  *
  322  * If no address specified, or address is 0.0.0.0, send to ip_output();
  323  * otherwise, send to ip_input() and mark as having been received on
  324  * the interface with that address.
  325  */
  326 static int
  327 div_output(struct socket *so, struct mbuf *m, struct sockaddr_in *sin,
  328     struct mbuf *control)
  329 {
  330         struct ip *const ip = mtod(m, struct ip *);
  331         struct m_tag *mtag;
  332         struct ipfw_rule_ref *dt;
  333         int error = 0;
  334 
  335         /*
  336          * An mbuf may hasn't come from userland, but we pretend
  337          * that it has.
  338          */
  339         m->m_pkthdr.rcvif = NULL;
  340         m->m_nextpkt = NULL;
  341         M_SETFIB(m, so->so_fibnum);
  342 
  343         if (control)
  344                 m_freem(control);               /* XXX */
  345 
  346         mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL);
  347         if (mtag == NULL) {
  348                 /* this should be normal */
  349                 mtag = m_tag_alloc(MTAG_IPFW_RULE, 0,
  350                     sizeof(struct ipfw_rule_ref), M_NOWAIT | M_ZERO);
  351                 if (mtag == NULL) {
  352                         error = ENOBUFS;
  353                         goto cantsend;
  354                 }
  355                 m_tag_prepend(m, mtag);
  356         }
  357         dt = (struct ipfw_rule_ref *)(mtag+1);
  358 
  359         /* Loopback avoidance and state recovery */
  360         if (sin) {
  361                 int i;
  362 
  363                 /* set the starting point. We provide a non-zero slot,
  364                  * but a non_matching chain_id to skip that info and use
  365                  * the rulenum/rule_id.
  366                  */
  367                 dt->slot = 1; /* dummy, chain_id is invalid */
  368                 dt->chain_id = 0;
  369                 dt->rulenum = sin->sin_port+1; /* host format ? */
  370                 dt->rule_id = 0;
  371                 /*
  372                  * Find receive interface with the given name, stuffed
  373                  * (if it exists) in the sin_zero[] field.
  374                  * The name is user supplied data so don't trust its size
  375                  * or that it is zero terminated.
  376                  */
  377                 for (i = 0; i < sizeof(sin->sin_zero) && sin->sin_zero[i]; i++)
  378                         ;
  379                 if ( i > 0 && i < sizeof(sin->sin_zero))
  380                         m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
  381         }
  382 
  383         /* Reinject packet into the system as incoming or outgoing */
  384         if (!sin || sin->sin_addr.s_addr == 0) {
  385                 struct mbuf *options = NULL;
  386                 struct inpcb *inp;
  387 
  388                 dt->info |= IPFW_IS_DIVERT | IPFW_INFO_OUT;
  389                 inp = sotoinpcb(so);
  390                 INP_RLOCK(inp);
  391                 switch (ip->ip_v) {
  392                 case IPVERSION:
  393                         /*
  394                          * Don't allow both user specified and setsockopt
  395                          * options, and don't allow packet length sizes that
  396                          * will crash.
  397                          */
  398                         if ((((ip->ip_hl << 2) != sizeof(struct ip)) &&
  399                             inp->inp_options != NULL) ||
  400                             ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
  401                                 error = EINVAL;
  402                                 INP_RUNLOCK(inp);
  403                                 goto cantsend;
  404                         }
  405                         break;
  406 #ifdef INET6
  407                 case IPV6_VERSION >> 4:
  408                     {
  409                         struct ip6_hdr *const ip6 = mtod(m, struct ip6_hdr *);
  410 
  411                         /* Don't allow packet length sizes that will crash */
  412                         if (((u_short)ntohs(ip6->ip6_plen) > m->m_pkthdr.len)) {
  413                                 error = EINVAL;
  414                                 INP_RUNLOCK(inp);
  415                                 goto cantsend;
  416                         }
  417                         break;
  418                     }
  419 #endif
  420                 default:
  421                         error = EINVAL;
  422                         INP_RUNLOCK(inp);
  423                         goto cantsend;
  424                 }
  425 
  426                 /* Send packet to output processing */
  427                 KMOD_IPSTAT_INC(ips_rawout);            /* XXX */
  428 
  429 #ifdef MAC
  430                 mac_inpcb_create_mbuf(inp, m);
  431 #endif
  432                 /*
  433                  * Get ready to inject the packet into ip_output().
  434                  * Just in case socket options were specified on the
  435                  * divert socket, we duplicate them.  This is done
  436                  * to avoid having to hold the PCB locks over the call
  437                  * to ip_output(), as doing this results in a number of
  438                  * lock ordering complexities.
  439                  *
  440                  * Note that we set the multicast options argument for
  441                  * ip_output() to NULL since it should be invariant that
  442                  * they are not present.
  443                  */
  444                 KASSERT(inp->inp_moptions == NULL,
  445                     ("multicast options set on a divert socket"));
  446                 /*
  447                  * XXXCSJP: It is unclear to me whether or not it makes
  448                  * sense for divert sockets to have options.  However,
  449                  * for now we will duplicate them with the INP locks
  450                  * held so we can use them in ip_output() without
  451                  * requring a reference to the pcb.
  452                  */
  453                 if (inp->inp_options != NULL) {
  454                         options = m_dup(inp->inp_options, M_NOWAIT);
  455                         if (options == NULL) {
  456                                 INP_RUNLOCK(inp);
  457                                 error = ENOBUFS;
  458                                 goto cantsend;
  459                         }
  460                 }
  461                 INP_RUNLOCK(inp);
  462 
  463                 switch (ip->ip_v) {
  464                 case IPVERSION:
  465                         error = ip_output(m, options, NULL,
  466                             ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0)
  467                             | IP_ALLOWBROADCAST | IP_RAWOUTPUT, NULL, NULL);
  468                         break;
  469 #ifdef INET6
  470                 case IPV6_VERSION >> 4:
  471                         error = ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
  472                         break;
  473 #endif
  474                 }
  475                 if (options != NULL)
  476                         m_freem(options);
  477         } else {
  478                 dt->info |= IPFW_IS_DIVERT | IPFW_INFO_IN;
  479                 if (m->m_pkthdr.rcvif == NULL) {
  480                         /*
  481                          * No luck with the name, check by IP address.
  482                          * Clear the port and the ifname to make sure
  483                          * there are no distractions for ifa_ifwithaddr.
  484                          */
  485                         struct  ifaddr *ifa;
  486 
  487                         bzero(sin->sin_zero, sizeof(sin->sin_zero));
  488                         sin->sin_port = 0;
  489                         NET_EPOCH_ENTER();
  490                         ifa = ifa_ifwithaddr((struct sockaddr *) sin);
  491                         if (ifa == NULL) {
  492                                 error = EADDRNOTAVAIL;
  493                                 NET_EPOCH_EXIT();
  494                                 goto cantsend;
  495                         }
  496                         m->m_pkthdr.rcvif = ifa->ifa_ifp;
  497                         NET_EPOCH_EXIT();
  498                 }
  499 #ifdef MAC
  500                 mac_socket_create_mbuf(so, m);
  501 #endif
  502                 /* Send packet to input processing via netisr */
  503                 switch (ip->ip_v) {
  504                 case IPVERSION:
  505                         /*
  506                          * Restore M_BCAST flag when destination address is
  507                          * broadcast. It is expected by ip_tryforward().
  508                          */
  509                         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)))
  510                                 m->m_flags |= M_MCAST;
  511                         else if (in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))
  512                                 m->m_flags |= M_BCAST;
  513                         netisr_queue_src(NETISR_IP, (uintptr_t)so, m);
  514                         break;
  515 #ifdef INET6
  516                 case IPV6_VERSION >> 4:
  517                         netisr_queue_src(NETISR_IPV6, (uintptr_t)so, m);
  518                         break;
  519 #endif
  520                 default:
  521                         error = EINVAL;
  522                         goto cantsend;
  523                 }
  524         }
  525 
  526         return (error);
  527 
  528 cantsend:
  529         m_freem(m);
  530         return (error);
  531 }
  532 
  533 static int
  534 div_attach(struct socket *so, int proto, struct thread *td)
  535 {
  536         struct inpcb *inp;
  537         int error;
  538 
  539         inp  = sotoinpcb(so);
  540         KASSERT(inp == NULL, ("div_attach: inp != NULL"));
  541         if (td != NULL) {
  542                 error = priv_check(td, PRIV_NETINET_DIVERT);
  543                 if (error)
  544                         return (error);
  545         }
  546         error = soreserve(so, div_sendspace, div_recvspace);
  547         if (error)
  548                 return error;
  549         INP_INFO_WLOCK(&V_divcbinfo);
  550         error = in_pcballoc(so, &V_divcbinfo);
  551         if (error) {
  552                 INP_INFO_WUNLOCK(&V_divcbinfo);
  553                 return error;
  554         }
  555         inp = (struct inpcb *)so->so_pcb;
  556         INP_INFO_WUNLOCK(&V_divcbinfo);
  557         inp->inp_ip_p = proto;
  558         inp->inp_vflag |= INP_IPV4;
  559         inp->inp_flags |= INP_HDRINCL;
  560         INP_WUNLOCK(inp);
  561         return 0;
  562 }
  563 
  564 static void
  565 div_detach(struct socket *so)
  566 {
  567         struct inpcb *inp;
  568 
  569         inp = sotoinpcb(so);
  570         KASSERT(inp != NULL, ("div_detach: inp == NULL"));
  571         INP_INFO_WLOCK(&V_divcbinfo);
  572         INP_WLOCK(inp);
  573         in_pcbdetach(inp);
  574         in_pcbfree(inp);
  575         INP_INFO_WUNLOCK(&V_divcbinfo);
  576 }
  577 
  578 static int
  579 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
  580 {
  581         struct inpcb *inp;
  582         int error;
  583 
  584         inp = sotoinpcb(so);
  585         KASSERT(inp != NULL, ("div_bind: inp == NULL"));
  586         /* in_pcbbind assumes that nam is a sockaddr_in
  587          * and in_pcbbind requires a valid address. Since divert
  588          * sockets don't we need to make sure the address is
  589          * filled in properly.
  590          * XXX -- divert should not be abusing in_pcbind
  591          * and should probably have its own family.
  592          */
  593         if (nam->sa_family != AF_INET)
  594                 return EAFNOSUPPORT;
  595         ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
  596         INP_INFO_WLOCK(&V_divcbinfo);
  597         INP_WLOCK(inp);
  598         INP_HASH_WLOCK(&V_divcbinfo);
  599         error = in_pcbbind(inp, nam, td->td_ucred);
  600         INP_HASH_WUNLOCK(&V_divcbinfo);
  601         INP_WUNLOCK(inp);
  602         INP_INFO_WUNLOCK(&V_divcbinfo);
  603         return error;
  604 }
  605 
  606 static int
  607 div_shutdown(struct socket *so)
  608 {
  609         struct inpcb *inp;
  610 
  611         inp = sotoinpcb(so);
  612         KASSERT(inp != NULL, ("div_shutdown: inp == NULL"));
  613         INP_WLOCK(inp);
  614         socantsendmore(so);
  615         INP_WUNLOCK(inp);
  616         return 0;
  617 }
  618 
  619 static int
  620 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
  621     struct mbuf *control, struct thread *td)
  622 {
  623 
  624         /* Packet must have a header (but that's about it) */
  625         if (m->m_len < sizeof (struct ip) &&
  626             (m = m_pullup(m, sizeof (struct ip))) == NULL) {
  627                 KMOD_IPSTAT_INC(ips_toosmall);
  628                 m_freem(m);
  629                 return EINVAL;
  630         }
  631 
  632         /* Send packet */
  633         return div_output(so, m, (struct sockaddr_in *)nam, control);
  634 }
  635 
  636 static void
  637 div_ctlinput(int cmd, struct sockaddr *sa, void *vip)
  638 {
  639         struct in_addr faddr;
  640 
  641         faddr = ((struct sockaddr_in *)sa)->sin_addr;
  642         if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
  643                 return;
  644         if (PRC_IS_REDIRECT(cmd))
  645                 return;
  646 }
  647 
  648 static int
  649 div_pcblist(SYSCTL_HANDLER_ARGS)
  650 {
  651         int error, i, n;
  652         struct inpcb *inp, **inp_list;
  653         inp_gen_t gencnt;
  654         struct xinpgen xig;
  655         struct epoch_tracker et;
  656 
  657         /*
  658          * The process of preparing the TCB list is too time-consuming and
  659          * resource-intensive to repeat twice on every request.
  660          */
  661         if (req->oldptr == 0) {
  662                 n = V_divcbinfo.ipi_count;
  663                 n += imax(n / 8, 10);
  664                 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
  665                 return 0;
  666         }
  667 
  668         if (req->newptr != 0)
  669                 return EPERM;
  670 
  671         /*
  672          * OK, now we're committed to doing something.
  673          */
  674         INP_INFO_WLOCK(&V_divcbinfo);
  675         gencnt = V_divcbinfo.ipi_gencnt;
  676         n = V_divcbinfo.ipi_count;
  677         INP_INFO_WUNLOCK(&V_divcbinfo);
  678 
  679         error = sysctl_wire_old_buffer(req,
  680             2 * sizeof(xig) + n*sizeof(struct xinpcb));
  681         if (error != 0)
  682                 return (error);
  683 
  684         bzero(&xig, sizeof(xig));
  685         xig.xig_len = sizeof xig;
  686         xig.xig_count = n;
  687         xig.xig_gen = gencnt;
  688         xig.xig_sogen = so_gencnt;
  689         error = SYSCTL_OUT(req, &xig, sizeof xig);
  690         if (error)
  691                 return error;
  692 
  693         inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
  694         if (inp_list == NULL)
  695                 return ENOMEM;
  696         
  697         INP_INFO_RLOCK_ET(&V_divcbinfo, et);
  698         for (inp = CK_LIST_FIRST(V_divcbinfo.ipi_listhead), i = 0; inp && i < n;
  699              inp = CK_LIST_NEXT(inp, inp_list)) {
  700                 INP_WLOCK(inp);
  701                 if (inp->inp_gencnt <= gencnt &&
  702                     cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
  703                         in_pcbref(inp);
  704                         inp_list[i++] = inp;
  705                 }
  706                 INP_WUNLOCK(inp);
  707         }
  708         INP_INFO_RUNLOCK_ET(&V_divcbinfo, et);
  709         n = i;
  710 
  711         error = 0;
  712         for (i = 0; i < n; i++) {
  713                 inp = inp_list[i];
  714                 INP_RLOCK(inp);
  715                 if (inp->inp_gencnt <= gencnt) {
  716                         struct xinpcb xi;
  717 
  718                         in_pcbtoxinpcb(inp, &xi);
  719                         INP_RUNLOCK(inp);
  720                         error = SYSCTL_OUT(req, &xi, sizeof xi);
  721                 } else
  722                         INP_RUNLOCK(inp);
  723         }
  724         INP_INFO_WLOCK(&V_divcbinfo);
  725         for (i = 0; i < n; i++) {
  726                 inp = inp_list[i];
  727                 INP_RLOCK(inp);
  728                 if (!in_pcbrele_rlocked(inp))
  729                         INP_RUNLOCK(inp);
  730         }
  731         INP_INFO_WUNLOCK(&V_divcbinfo);
  732 
  733         if (!error) {
  734                 struct epoch_tracker et;
  735                 /*
  736                  * Give the user an updated idea of our state.
  737                  * If the generation differs from what we told
  738                  * her before, she knows that something happened
  739                  * while we were processing this request, and it
  740                  * might be necessary to retry.
  741                  */
  742                 INP_INFO_RLOCK_ET(&V_divcbinfo, et);
  743                 xig.xig_gen = V_divcbinfo.ipi_gencnt;
  744                 xig.xig_sogen = so_gencnt;
  745                 xig.xig_count = V_divcbinfo.ipi_count;
  746                 INP_INFO_RUNLOCK_ET(&V_divcbinfo, et);
  747                 error = SYSCTL_OUT(req, &xig, sizeof xig);
  748         }
  749         free(inp_list, M_TEMP);
  750         return error;
  751 }
  752 
  753 #ifdef SYSCTL_NODE
  754 static SYSCTL_NODE(_net_inet, IPPROTO_DIVERT, divert, CTLFLAG_RW, 0,
  755     "IPDIVERT");
  756 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLTYPE_OPAQUE | CTLFLAG_RD,
  757     NULL, 0, div_pcblist, "S,xinpcb", "List of active divert sockets");
  758 #endif
  759 
  760 struct pr_usrreqs div_usrreqs = {
  761         .pru_attach =           div_attach,
  762         .pru_bind =             div_bind,
  763         .pru_control =          in_control,
  764         .pru_detach =           div_detach,
  765         .pru_peeraddr =         in_getpeeraddr,
  766         .pru_send =             div_send,
  767         .pru_shutdown =         div_shutdown,
  768         .pru_sockaddr =         in_getsockaddr,
  769         .pru_sosetlabel =       in_pcbsosetlabel
  770 };
  771 
  772 struct protosw div_protosw = {
  773         .pr_type =              SOCK_RAW,
  774         .pr_protocol =          IPPROTO_DIVERT,
  775         .pr_flags =             PR_ATOMIC|PR_ADDR,
  776         .pr_input =             div_input,
  777         .pr_ctlinput =          div_ctlinput,
  778         .pr_ctloutput =         ip_ctloutput,
  779         .pr_init =              div_init,
  780         .pr_usrreqs =           &div_usrreqs
  781 };
  782 
  783 static int
  784 div_modevent(module_t mod, int type, void *unused)
  785 {
  786         int err = 0;
  787 
  788         switch (type) {
  789         case MOD_LOAD:
  790                 /*
  791                  * Protocol will be initialized by pf_proto_register().
  792                  * We don't have to register ip_protox because we are not
  793                  * a true IP protocol that goes over the wire.
  794                  */
  795                 err = pf_proto_register(PF_INET, &div_protosw);
  796                 if (err != 0)
  797                         return (err);
  798                 ip_divert_ptr = divert_packet;
  799                 ip_divert_event_tag = EVENTHANDLER_REGISTER(maxsockets_change,
  800                     div_zone_change, NULL, EVENTHANDLER_PRI_ANY);
  801                 break;
  802         case MOD_QUIESCE:
  803                 /*
  804                  * IPDIVERT may normally not be unloaded because of the
  805                  * potential race conditions.  Tell kldunload we can't be
  806                  * unloaded unless the unload is forced.
  807                  */
  808                 err = EPERM;
  809                 break;
  810         case MOD_UNLOAD:
  811                 /*
  812                  * Forced unload.
  813                  *
  814                  * Module ipdivert can only be unloaded if no sockets are
  815                  * connected.  Maybe this can be changed later to forcefully
  816                  * disconnect any open sockets.
  817                  *
  818                  * XXXRW: Note that there is a slight race here, as a new
  819                  * socket open request could be spinning on the lock and then
  820                  * we destroy the lock.
  821                  */
  822                 INP_INFO_WLOCK(&V_divcbinfo);
  823                 if (V_divcbinfo.ipi_count != 0) {
  824                         err = EBUSY;
  825                         INP_INFO_WUNLOCK(&V_divcbinfo);
  826                         break;
  827                 }
  828                 ip_divert_ptr = NULL;
  829                 err = pf_proto_unregister(PF_INET, IPPROTO_DIVERT, SOCK_RAW);
  830                 INP_INFO_WUNLOCK(&V_divcbinfo);
  831 #ifndef VIMAGE
  832                 div_destroy(NULL);
  833 #endif
  834                 EVENTHANDLER_DEREGISTER(maxsockets_change, ip_divert_event_tag);
  835                 break;
  836         default:
  837                 err = EOPNOTSUPP;
  838                 break;
  839         }
  840         return err;
  841 }
  842 
  843 static moduledata_t ipdivertmod = {
  844         "ipdivert",
  845         div_modevent,
  846         0
  847 };
  848 
  849 DECLARE_MODULE(ipdivert, ipdivertmod, SI_SUB_PROTO_FIREWALL, SI_ORDER_ANY);
  850 MODULE_DEPEND(ipdivert, ipfw, 3, 3, 3);
  851 MODULE_VERSION(ipdivert, 1);

Cache object: ba299cb09204541a8b99abf29f134bb6


[ 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.