sys/netinet/ip_divert.c

   1 /*
   2  * Copyright (c) 1982, 1986, 1988, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. All advertising materials mentioning features or use of this software
  14  *    must display the following acknowledgement:
  15  *      This product includes software developed by the University of
  16  *      California, Berkeley and its contributors.
  17  * 4. Neither the name of the University nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  *
  33  * $FreeBSD: src/sys/netinet/ip_divert.c,v 1.42.2.6 2003/01/23 21:06:45 sam Exp $
  34  * $DragonFly: src/sys/netinet/ip_divert.c,v 1.28 2007/04/22 01:13:14 dillon Exp $
  35  */
  36
  37 #include "opt_inet.h"
  38 #include "opt_ipfw.h"
  39 #include "opt_ipdivert.h"
  40 #include "opt_ipsec.h"
  41
  42 #ifndef INET
  43 #error "IPDIVERT requires INET."
  44 #endif
  45
  46 #include <sys/param.h>
  47 #include <sys/kernel.h>
  48 #include <sys/malloc.h>
  49 #include <sys/mbuf.h>
  50 #include <sys/socket.h>
  51 #include <sys/protosw.h>
  52 #include <sys/socketvar.h>
  53 #include <sys/sysctl.h>
  54 #include <sys/systm.h>
  55 #include <sys/proc.h>
  56 #include <sys/thread2.h>
  57
  58 #include <vm/vm_zone.h>
  59
  60 #include <net/if.h>
  61 #include <net/route.h>
  62
  63 #include <netinet/in.h>
  64 #include <netinet/in_systm.h>
  65 #include <netinet/ip.h>
  66 #include <netinet/in_pcb.h>
  67 #include <netinet/in_var.h>
  68 #include <netinet/ip_var.h>
  69
  70 /*
  71  * Divert sockets
  72  */
  73
  74 /*
  75  * Allocate enough space to hold a full IP packet
  76  */
  77 #define DIVSNDQ         (65536 + 100)
  78 #define DIVRCVQ         (65536 + 100)
  79
  80 /*
  81  * Divert sockets work in conjunction with ipfw, see the divert(4)
  82  * manpage for features.
  83  * Internally, packets selected by ipfw in ip_input() or ip_output(),
  84  * and never diverted before, are passed to the input queue of the
  85  * divert socket with a given 'divert_port' number (as specified in
  86  * the matching ipfw rule), and they are tagged with a 16 bit cookie
  87  * (representing the rule number of the matching ipfw rule), which
  88  * is passed to process reading from the socket.
  89  *
  90  * Packets written to the divert socket are again tagged with a cookie
  91  * (usually the same as above) and a destination address.
  92  * If the destination address is INADDR_ANY then the packet is
  93  * treated as outgoing and sent to ip_output(), otherwise it is
  94  * treated as incoming and sent to ip_input().
  95  * In both cases, the packet is tagged with the cookie.
  96  *
  97  * On reinjection, processing in ip_input() and ip_output()
  98  * will be exactly the same as for the original packet, except that
  99  * ipfw processing will start at the rule number after the one
 100  * written in the cookie (so, tagging a packet with a cookie of 0
 101  * will cause it to be effectively considered as a standard packet).
 102  */
 103
 104 /* Internal variables */
 105 static struct inpcbinfo divcbinfo;
 106
 107 static u_long   div_sendspace = DIVSNDQ;        /* XXX sysctl ? */
 108 static u_long   div_recvspace = DIVRCVQ;        /* XXX sysctl ? */
 109
 110 /*
 111  * Initialize divert connection block queue.
 112  */
 113 void
 114 div_init(void)
 115 {
 116         in_pcbinfo_init(&divcbinfo);
 117         /*
 118          * XXX We don't use the hash list for divert IP, but it's easier
 119          * to allocate a one entry hash list than it is to check all
 120          * over the place for hashbase == NULL.
 121          */
 122         divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
 123         divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
 124         divcbinfo.wildcardhashbase = hashinit(1, M_PCB,
 125                                               &divcbinfo.wildcardhashmask);
 126         divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb),
 127                                    maxsockets, ZONE_INTERRUPT, 0);
 128 }
 129
 130 /*
 131  * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
 132  * with that protocol number to enter the system from the outside.
 133  */
 134 void
 135 div_input(struct mbuf *m, ...)
 136 {
 137         ipstat.ips_noproto++;
 138         m_freem(m);
 139 }
 140
 141 /*
 142  * Divert a packet by passing it up to the divert socket at port 'port'.
 143  *
 144  * Setup generic address and protocol structures for div_input routine,
 145  * then pass them along with mbuf chain.
 146  */
 147 void
 148 divert_packet(struct mbuf *m, int incoming, int port)
 149 {
 150         struct sockaddr_in divsrc = { sizeof divsrc, AF_INET };
 151         struct ip *ip;
 152         struct inpcb *inp;
 153         struct socket *sa;
 154         struct m_tag *mtag;
 155         u_int16_t nport;
 156
 157         /* Sanity check */
 158         KASSERT(port != 0, ("%s: port=0", __func__));
 159
 160         if ((mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
 161                 divsrc.sin_port = *(u_int16_t *)m_tag_data(mtag);
 162         else
 163                 divsrc.sin_port = 0;
 164
 165         /* Assure header */
 166         if (m->m_len < sizeof(struct ip) &&
 167             (m = m_pullup(m, sizeof(struct ip))) == NULL)
 168                 return;
 169         ip = mtod(m, struct ip *);
 170
 171         /*
 172          * Record receive interface address, if any.
 173          * But only for incoming packets.
 174          */
 175         divsrc.sin_addr.s_addr = 0;
 176         if (incoming) {
 177                 struct ifaddr *ifa;
 178
 179                 /* Sanity check */
 180                 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __func__));
 181
 182                 /* Find IP address for receive interface */
 183                 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
 184                         if (ifa->ifa_addr == NULL)
 185                                 continue;
 186                         if (ifa->ifa_addr->sa_family != AF_INET)
 187                                 continue;
 188                         divsrc.sin_addr =
 189                             ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
 190                         break;
 191                 }
 192         }
 193         /*
 194          * Record the incoming interface name whenever we have one.
 195          */
 196         if (m->m_pkthdr.rcvif) {
 197                 /*
 198                  * Hide the actual interface name in there in the
 199                  * sin_zero array. XXX This needs to be moved to a
 200                  * different sockaddr type for divert, e.g.
 201                  * sockaddr_div with multiple fields like
 202                  * sockaddr_dl. Presently we have only 7 bytes
 203                  * but that will do for now as most interfaces
 204                  * are 4 or less + 2 or less bytes for unit.
 205                  * There is probably a faster way of doing this,
 206                  * possibly taking it from the sockaddr_dl on the iface.
 207                  * This solves the problem of a P2P link and a LAN interface
 208                  * having the same address, which can result in the wrong
 209                  * interface being assigned to the packet when fed back
 210                  * into the divert socket. Theoretically if the daemon saves
 211                  * and re-uses the sockaddr_in as suggested in the man pages,
 212                  * this iface name will come along for the ride.
 213                  * (see div_output for the other half of this.)
 214                  */
 215                 ksnprintf(divsrc.sin_zero, sizeof divsrc.sin_zero,
 216                          m->m_pkthdr.rcvif->if_xname);
 217         }
 218
 219         /* Put packet on socket queue, if any */
 220         sa = NULL;
 221         nport = htons((u_int16_t)port);
 222         LIST_FOREACH(inp, &divcbinfo.pcblisthead, inp_list) {
 223                 if (inp->inp_flags & INP_PLACEMARKER)
 224                         continue;
 225                 if (inp->inp_lport == nport)
 226                         sa = inp->inp_socket;
 227         }
 228         if (sa) {
 229                 if (ssb_appendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, m,
 230                                  (struct mbuf *)NULL) == 0)
 231                         m_freem(m);
 232                 else
 233                         sorwakeup(sa);
 234         } else {
 235                 m_freem(m);
 236                 ipstat.ips_noproto++;
 237                 ipstat.ips_delivered--;
 238         }
 239 }
 240
 241 /*
 242  * Deliver packet back into the IP processing machinery.
 243  *
 244  * If no address specified, or address is 0.0.0.0, send to ip_output();
 245  * otherwise, send to ip_input() and mark as having been received on
 246  * the interface with that address.
 247  */
 248 static int
 249 div_output(struct socket *so, struct mbuf *m,
 250         struct sockaddr_in *sin, struct mbuf *control)
 251 {
 252         int error = 0;
 253         struct m_tag *mtag;
 254
 255         /*
 256          * Prepare the tag for divert info. Note that a packet
 257          * with a 0 tag in mh_data is effectively untagged,
 258          * so we could optimize that case.
 259          */
 260         mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT, sizeof(u_int16_t), M_NOWAIT);
 261         if (mtag == NULL) {
 262                 error = ENOBUFS;
 263                 goto cantsend;
 264         }
 265         m_tag_prepend(m, mtag);
 266         m->m_pkthdr.rcvif = NULL;       /* XXX is it necessary ? */
 267
 268         if (control)
 269                 m_freem(control);               /* XXX */
 270
 271         /* Loopback avoidance and state recovery */
 272         if (sin) {
 273                 int i;
 274
 275                 *(u_int16_t *)m_tag_data(mtag) = sin->sin_port;
 276                 /*
 277                  * Find receive interface with the given name, stuffed
 278                  * (if it exists) in the sin_zero[] field.
 279                  * The name is user supplied data so don't trust its size
 280                  * or that it is zero terminated.
 281                  */
 282                 for (i = 0; sin->sin_zero[i] && i < sizeof sin->sin_zero; i++)
 283                         ;
 284                 if ( i > 0 && i < sizeof sin->sin_zero)
 285                         m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
 286         } else {
 287                 *(u_int16_t *)m_tag_data(mtag) = 0;
 288         }
 289
 290         /* Reinject packet into the system as incoming or outgoing */
 291         if (!sin || sin->sin_addr.s_addr == 0) {
 292                 struct inpcb *const inp = so->so_pcb;
 293                 struct ip *const ip = mtod(m, struct ip *);
 294
 295                 /*
 296                  * Don't allow both user specified and setsockopt options,
 297                  * and don't allow packet length sizes that will crash
 298                  */
 299                 if (((ip->ip_hl != (sizeof *ip) >> 2) && inp->inp_options) ||
 300                      ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
 301                         error = EINVAL;
 302                         goto cantsend;
 303                 }
 304
 305                 /* Convert fields to host order for ip_output() */
 306                 ip->ip_len = ntohs(ip->ip_len);
 307                 ip->ip_off = ntohs(ip->ip_off);
 308
 309                 /* Send packet to output processing */
 310                 ipstat.ips_rawout++;                    /* XXX */
 311                 error = ip_output(m,
 312                             inp->inp_options, &inp->inp_route,
 313                             (so->so_options & SO_DONTROUTE) |
 314                             IP_ALLOWBROADCAST | IP_RAWOUTPUT,
 315                             inp->inp_moptions, NULL);
 316         } else {
 317                 if (m->m_pkthdr.rcvif == NULL) {
 318                         /*
 319                          * No luck with the name, check by IP address.
 320                          * Clear the port and the ifname to make sure
 321                          * there are no distractions for ifa_ifwithaddr.
 322                          */
 323                         struct  ifaddr *ifa;
 324
 325                         bzero(sin->sin_zero, sizeof sin->sin_zero);
 326                         sin->sin_port = 0;
 327                         ifa = ifa_ifwithaddr((struct sockaddr *) sin);
 328                         if (ifa == NULL) {
 329                                 error = EADDRNOTAVAIL;
 330                                 goto cantsend;
 331                         }
 332                         m->m_pkthdr.rcvif = ifa->ifa_ifp;
 333                 }
 334                 ip_input(m);
 335         }
 336
 337         return error;
 338
 339 cantsend:
 340         m_freem(m);
 341         return error;
 342 }
 343
 344 static int
 345 div_attach(struct socket *so, int proto, struct pru_attach_info *ai)
 346 {
 347         struct inpcb *inp;
 348         int error;
 349
 350         inp  = so->so_pcb;
 351         if (inp)
 352                 panic("div_attach");
 353         if ((error = suser_cred(ai->p_ucred, NULL_CRED_OKAY)) != 0)
 354                 return error;
 355
 356         error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit);
 357         if (error)
 358                 return error;
 359         crit_enter();
 360         error = in_pcballoc(so, &divcbinfo);
 361         crit_exit();
 362         if (error)
 363                 return error;
 364         inp = (struct inpcb *)so->so_pcb;
 365         inp->inp_ip_p = proto;
 366         inp->inp_vflag |= INP_IPV4;
 367         inp->inp_flags |= INP_HDRINCL;
 368         /*
 369          * The socket is always "connected" because
 370          * we always know "where" to send the packet.
 371          */
 372         so->so_state |= SS_ISCONNECTED;
 373         return 0;
 374 }
 375
 376 static int
 377 div_detach(struct socket *so)
 378 {
 379         struct inpcb *inp;
 380
 381         inp = so->so_pcb;
 382         if (inp == NULL)
 383                 panic("div_detach");
 384         in_pcbdetach(inp);
 385         return 0;
 386 }
 387
 388 static int
 389 div_abort(struct socket *so)
 390 {
 391         soisdisconnected(so);
 392         return div_detach(so);
 393 }
 394
 395 static int
 396 div_disconnect(struct socket *so)
 397 {
 398         if (!(so->so_state & SS_ISCONNECTED))
 399                 return ENOTCONN;
 400         return div_abort(so);
 401 }
 402
 403 static int
 404 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 405 {
 406         struct inpcb *inp;
 407         int error;
 408
 409         crit_enter();
 410         inp = so->so_pcb;
 411         /* in_pcbbind assumes that nam is a sockaddr_in
 412          * and in_pcbbind requires a valid address. Since divert
 413          * sockets don't we need to make sure the address is
 414          * filled in properly.
 415          * XXX -- divert should not be abusing in_pcbind
 416          * and should probably have its own family.
 417          */
 418         if (nam->sa_family != AF_INET)
 419                 error = EAFNOSUPPORT;
 420         else {
 421                 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
 422                 error = in_pcbbind(inp, nam, td);
 423         }
 424         crit_exit();
 425         return error;
 426 }
 427
 428 static int
 429 div_shutdown(struct socket *so)
 430 {
 431         socantsendmore(so);
 432         return 0;
 433 }
 434
 435 static int
 436 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 437          struct mbuf *control, struct thread *td)
 438 {
 439         /* Packet must have a header (but that's about it) */
 440         if (m->m_len < sizeof(struct ip) &&
 441             (m = m_pullup(m, sizeof(struct ip))) == NULL) {
 442                 ipstat.ips_toosmall++;
 443                 m_freem(m);
 444                 return EINVAL;
 445         }
 446
 447         /* Send packet */
 448         return div_output(so, m, (struct sockaddr_in *)nam, control);
 449 }
 450
 451 SYSCTL_DECL(_net_inet_divert);
 452 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, &divcbinfo, 0,
 453             in_pcblist_global, "S,xinpcb", "List of active divert sockets");
 454
 455 struct pr_usrreqs div_usrreqs = {
 456         .pru_abort = div_abort,
 457         .pru_accept = pru_accept_notsupp,
 458         .pru_attach = div_attach,
 459         .pru_bind = div_bind,
 460         .pru_connect = pru_connect_notsupp,
 461         .pru_connect2 = pru_connect2_notsupp,
 462         .pru_control = in_control,
 463         .pru_detach = div_detach,
 464         .pru_disconnect = div_disconnect,
 465         .pru_listen = pru_listen_notsupp,
 466         .pru_peeraddr = in_setpeeraddr,
 467         .pru_rcvd = pru_rcvd_notsupp,
 468         .pru_rcvoob = pru_rcvoob_notsupp,
 469         .pru_send = div_send,
 470         .pru_sense = pru_sense_null,
 471         .pru_shutdown = div_shutdown,
 472         .pru_sockaddr = in_setsockaddr,
 473         .pru_sosend = sosend,
 474         .pru_soreceive = soreceive,
 475         .pru_sopoll = sopoll
 476 };