bsd/netinet/ip_divert.c

   1 /*
   2  * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
   7  *
   8  * This file contains Original Code and/or Modifications of Original Code
   9  * as defined in and that are subject to the Apple Public Source License
  10  * Version 2.0 (the 'License'). You may not use this file except in
  11  * compliance with the License. Please obtain a copy of the License at
  12  * http://www.opensource.apple.com/apsl/ and read it before using this
  13  * file.
  14  *
  15  * The Original Code and all software distributed under the License are
  16  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  17  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  18  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  20  * Please see the License for the specific language governing rights and
  21  * limitations under the License.
  22  *
  23  * @APPLE_LICENSE_HEADER_END@
  24  */
  25 /*
  26  * Copyright (c) 1982, 1986, 1988, 1993
  27  *      The Regents of the University of California.  All rights reserved.
  28  *
  29  * Redistribution and use in source and binary forms, with or without
  30  * modification, are permitted provided that the following conditions
  31  * are met:
  32  * 1. Redistributions of source code must retain the above copyright
  33  *    notice, this list of conditions and the following disclaimer.
  34  * 2. Redistributions in binary form must reproduce the above copyright
  35  *    notice, this list of conditions and the following disclaimer in the
  36  *    documentation and/or other materials provided with the distribution.
  37  * 3. All advertising materials mentioning features or use of this software
  38  *    must display the following acknowledgement:
  39  *      This product includes software developed by the University of
  40  *      California, Berkeley and its contributors.
  41  * 4. Neither the name of the University nor the names of its contributors
  42  *    may be used to endorse or promote products derived from this software
  43  *    without specific prior written permission.
  44  *
  45  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  46  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  47  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  48  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  49  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  50  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  51  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  52  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  53  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  54  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  55  * SUCH DAMAGE.
  56  *
  57  * $FreeBSD: src/sys/netinet/ip_divert.c,v 1.42.2.4 2001/07/29 19:32:40 ume Exp $
  58  */
  59
  60
  61 #ifndef INET
  62 #error "IPDIVERT requires INET."
  63 #endif
  64
  65 #include <sys/param.h>
  66 #include <sys/kernel.h>
  67 #include <sys/malloc.h>
  68 #include <sys/mbuf.h>
  69 #include <sys/socket.h>
  70 #include <sys/protosw.h>
  71 #include <sys/socketvar.h>
  72 #include <sys/sysctl.h>
  73 #include <sys/systm.h>
  74 #include <sys/proc.h>
  75
  76
  77 #include <net/if.h>
  78 #include <net/route.h>
  79
  80 #include <netinet/in.h>
  81 #include <netinet/in_systm.h>
  82 #include <netinet/ip.h>
  83 #include <netinet/in_pcb.h>
  84 #include <netinet/in_var.h>
  85 #include <netinet/ip_var.h>
  86
  87 /*
  88  * Divert sockets
  89  */
  90
  91 /*
  92  * Allocate enough space to hold a full IP packet
  93  */
  94 #define DIVSNDQ         (65536 + 100)
  95 #define DIVRCVQ         (65536 + 100)
  96
  97 /*
  98  * A 16 bit cookie is passed to and from the user process.
  99  * The user process can send it back to help the caller know
 100  * something about where the packet originally came from.
 101  *
 102  * In the case of ipfw, then the cookie is the rule that sent
 103  * us here. On reinjection is is the rule after which processing
 104  * should continue. Leaving it the same will make processing start
 105  * at the rule number after that which sent it here. Setting it to
 106  * 0 will restart processing at the beginning.
 107  *
 108  * For divert_packet(), ip_divert_cookie is an input value only.
 109  * For div_output(), ip_divert_cookie is an output value only.
 110  */
 111 u_int16_t ip_divert_cookie;
 112
 113 /* Internal variables */
 114 static struct inpcbhead divcb;
 115 static struct inpcbinfo divcbinfo;
 116
 117 static u_long   div_sendspace = DIVSNDQ;        /* XXX sysctl ? */
 118 static u_long   div_recvspace = DIVRCVQ;        /* XXX sysctl ? */
 119
 120 /* Optimization: have this preinitialized */
 121 static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET };
 122
 123 /* Internal functions */
 124 static int div_output(struct socket *so,
 125                 struct mbuf *m, struct sockaddr *addr, struct mbuf *control);
 126
 127 /*
 128  * Initialize divert connection block queue.
 129  */
 130 void
 131 div_init(void)
 132 {
 133         LIST_INIT(&divcb);
 134         divcbinfo.listhead = &divcb;
 135         /*
 136          * XXX We don't use the hash list for divert IP, but it's easier
 137          * to allocate a one entry hash list than it is to check all
 138          * over the place for hashbase == NULL.
 139          */
 140         divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
 141         divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
 142         divcbinfo.ipi_zone = (void *) zinit(sizeof(struct inpcb),(maxsockets * sizeof(struct inpcb)),
 143                                    4096, "divzone");
 144
 145 /*
 146  * ### LD 08/03: init IP forwarding at this point [ipfw is not a module yet]
 147  */
 148 #if !IPFIREWALL_KEXT
 149         ip_fw_init();
 150 #endif
 151 }
 152
 153 /*
 154  * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
 155  * with that protocol number to enter the system from the outside.
 156  */
 157 void
 158 div_input(struct mbuf *m, int off)
 159 {
 160         ipstat.ips_noproto++;
 161         m_freem(m);
 162 }
 163
 164 /*
 165  * Divert a packet by passing it up to the divert socket at port 'port'.
 166  *
 167  * Setup generic address and protocol structures for div_input routine,
 168  * then pass them along with mbuf chain.
 169  */
 170 void
 171 divert_packet(struct mbuf *m, int incoming, int port)
 172 {
 173         struct ip *ip;
 174         struct inpcb *inp;
 175         struct socket *sa;
 176         u_int16_t nport;
 177
 178         /* Sanity check */
 179         KASSERT(port != 0, ("%s: port=0", __FUNCTION__));
 180
 181         /* Record and reset divert cookie */
 182         divsrc.sin_port = ip_divert_cookie;
 183         ip_divert_cookie = 0;
 184
 185         /* Assure header */
 186         if (m->m_len < sizeof(struct ip) &&
 187             (m = m_pullup(m, sizeof(struct ip))) == 0) {
 188                 return;
 189         }
 190         ip = mtod(m, struct ip *);
 191
 192         /*
 193          * Record receive interface address, if any.
 194          * But only for incoming packets.
 195          */
 196         divsrc.sin_addr.s_addr = 0;
 197         if (incoming) {
 198                 struct ifaddr *ifa;
 199
 200                 /* Sanity check */
 201                 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __FUNCTION__));
 202
 203                 /* Find IP address for receive interface */
 204                 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
 205                         if (ifa->ifa_addr == NULL)
 206                                 continue;
 207                         if (ifa->ifa_addr->sa_family != AF_INET)
 208                                 continue;
 209                         divsrc.sin_addr =
 210                             ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
 211                         break;
 212                 }
 213         }
 214         /*
 215          * Record the incoming interface name whenever we have one.
 216          */
 217         bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero));
 218         if (m->m_pkthdr.rcvif) {
 219                 /*
 220                  * Hide the actual interface name in there in the
 221                  * sin_zero array. XXX This needs to be moved to a
 222                  * different sockaddr type for divert, e.g.
 223                  * sockaddr_div with multiple fields like
 224                  * sockaddr_dl. Presently we have only 7 bytes
 225                  * but that will do for now as most interfaces
 226                  * are 4 or less + 2 or less bytes for unit.
 227                  * There is probably a faster way of doing this,
 228                  * possibly taking it from the sockaddr_dl on the iface.
 229                  * This solves the problem of a P2P link and a LAN interface
 230                  * having the same address, which can result in the wrong
 231                  * interface being assigned to the packet when fed back
 232                  * into the divert socket. Theoretically if the daemon saves
 233                  * and re-uses the sockaddr_in as suggested in the man pages,
 234                  * this iface name will come along for the ride.
 235                  * (see div_output for the other half of this.)
 236                  */
 237                 snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero),
 238                         "%s%d", m->m_pkthdr.rcvif->if_name,
 239                         m->m_pkthdr.rcvif->if_unit);
 240         }
 241
 242         /* Put packet on socket queue, if any */
 243         sa = NULL;
 244         nport = htons((u_int16_t)port);
 245         LIST_FOREACH(inp, &divcb, inp_list) {
 246                 if (inp->inp_lport == nport)
 247                         sa = inp->inp_socket;
 248         }
 249         if (sa) {
 250                 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc,
 251                                 m, (struct mbuf *)0) == 0)
 252                         m_freem(m);
 253                 else
 254                         sorwakeup(sa);
 255         } else {
 256                 m_freem(m);
 257                 ipstat.ips_noproto++;
 258                 ipstat.ips_delivered--;
 259         }
 260 }
 261
 262 /*
 263  * Deliver packet back into the IP processing machinery.
 264  *
 265  * If no address specified, or address is 0.0.0.0, send to ip_output();
 266  * otherwise, send to ip_input() and mark as having been received on
 267  * the interface with that address.
 268  */
 269 static int
 270 div_output(so, m, addr, control)
 271         struct socket *so;
 272         register struct mbuf *m;
 273         struct sockaddr *addr;
 274         struct mbuf *control;
 275 {
 276         register struct inpcb *const inp = sotoinpcb(so);
 277         register struct ip *const ip = mtod(m, struct ip *);
 278         struct sockaddr_in *sin = (struct sockaddr_in *)addr;
 279         int error = 0;
 280
 281         if (control)
 282                 m_freem(control);               /* XXX */
 283
 284         /* Loopback avoidance and state recovery */
 285         if (sin) {
 286                 int     len = 0;
 287                 char    *c = sin->sin_zero;
 288
 289                 ip_divert_cookie = sin->sin_port;
 290
 291                 /*
 292                  * Find receive interface with the given name or IP address.
 293                  * The name is user supplied data so don't trust it's size or
 294                  * that it is zero terminated. The name has priority.
 295                  * We are presently assuming that the sockaddr_in
 296                  * has not been replaced by a sockaddr_div, so we limit it
 297                  * to 16 bytes in total. the name is stuffed (if it exists)
 298                  * in the sin_zero[] field.
 299                  */
 300                 while (*c++ && (len++ < sizeof(sin->sin_zero)));
 301                 if ((len > 0) && (len < sizeof(sin->sin_zero)))
 302                         m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
 303         } else {
 304                 ip_divert_cookie = 0;
 305         }
 306
 307         /* Reinject packet into the system as incoming or outgoing */
 308         if (!sin || sin->sin_addr.s_addr == 0) {
 309                 /*
 310                  * Don't allow both user specified and setsockopt options,
 311                  * and don't allow packet length sizes that will crash
 312                  */
 313                 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) ||
 314                      ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
 315                         error = EINVAL;
 316                         goto cantsend;
 317                 }
 318
 319                 /* Convert fields to host order for ip_output() */
 320                 NTOHS(ip->ip_len);
 321                 NTOHS(ip->ip_off);
 322
 323                 /* Send packet to output processing */
 324                 ipstat.ips_rawout++;                    /* XXX */
 325                 error = ip_output(m, inp->inp_options, &inp->inp_route,
 326                         (so->so_options & SO_DONTROUTE) |
 327                         IP_ALLOWBROADCAST | IP_RAWOUTPUT,
 328                         inp->inp_moptions);
 329         } else {
 330                 struct  ifaddr *ifa;
 331
 332                 /* If no luck with the name above. check by IP address.  */
 333                 if (m->m_pkthdr.rcvif == NULL) {
 334                         /*
 335                          * Make sure there are no distractions
 336                          * for ifa_ifwithaddr. Clear the port and the ifname.
 337                          * Maybe zap all 8 bytes at once using a 64bit write?
 338                          */
 339                         bzero(sin->sin_zero, sizeof(sin->sin_zero));
 340                         /* *((u_int64_t *)sin->sin_zero) = 0; */ /* XXX ?? */
 341                         sin->sin_port = 0;
 342                         if (!(ifa = ifa_ifwithaddr((struct sockaddr *) sin))) {
 343                                 error = EADDRNOTAVAIL;
 344                                 goto cantsend;
 345                         }
 346                         m->m_pkthdr.rcvif = ifa->ifa_ifp;
 347                 }
 348
 349                 /* Send packet to input processing */
 350                 ip_input(m);
 351         }
 352
 353         /* paranoid: Reset for next time (and other packets) */
 354         /* almost definitly already done in the ipfw filter but.. */
 355         ip_divert_cookie = 0;
 356         return error;
 357
 358 cantsend:
 359         m_freem(m);
 360         ip_divert_cookie = 0;
 361         return error;
 362 }
 363
 364 static int
 365 div_attach(struct socket *so, int proto, struct proc *p)
 366 {
 367         struct inpcb *inp;
 368         int error, s;
 369
 370         inp  = sotoinpcb(so);
 371         if (inp)
 372                 panic("div_attach");
 373         if (p && (error = suser(p->p_ucred, &p->p_acflag)) != 0)
 374                 return error;
 375
 376         error = soreserve(so, div_sendspace, div_recvspace);
 377         if (error)
 378                 return error;
 379         s = splnet();
 380         error = in_pcballoc(so, &divcbinfo, p);
 381         splx(s);
 382         if (error)
 383                 return error;
 384         inp = (struct inpcb *)so->so_pcb;
 385         inp->inp_ip_p = proto;
 386         inp->inp_vflag |= INP_IPV4;
 387         inp->inp_flags |= INP_HDRINCL;
 388         /* The socket is always "connected" because
 389            we always know "where" to send the packet */
 390         so->so_state |= SS_ISCONNECTED;
 391         return 0;
 392 }
 393
 394 static int
 395 div_detach(struct socket *so)
 396 {
 397         struct inpcb *inp;
 398
 399         inp = sotoinpcb(so);
 400         if (inp == 0)
 401                 panic("div_detach");
 402         in_pcbdetach(inp);
 403         return 0;
 404 }
 405
 406 static int
 407 div_abort(struct socket *so)
 408 {
 409         soisdisconnected(so);
 410         return div_detach(so);
 411 }
 412
 413 static int
 414 div_disconnect(struct socket *so)
 415 {
 416         if ((so->so_state & SS_ISCONNECTED) == 0)
 417                 return ENOTCONN;
 418         return div_abort(so);
 419 }
 420
 421 static int
 422 div_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
 423 {
 424         struct inpcb *inp;
 425         int s;
 426         int error;
 427
 428         s = splnet();
 429         inp = sotoinpcb(so);
 430         /* in_pcbbind assumes that the socket is a sockaddr_in
 431         * and in_pcbbind requires a valid address. Since divert
 432         * sockets don't we need to make sure the address is
 433         * filled in properly.
 434         * XXX -- divert should not be abusing in_pcbind
 435         * and should probably have its own family.
 436         */
 437         if (nam->sa_family != AF_INET) {
 438                 error = EAFNOSUPPORT;
 439         } else {
 440                ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
 441                 error = in_pcbbind(inp, nam, p);
 442         }
 443         splx(s);
 444         return error;
 445 }
 446
 447 static int
 448 div_shutdown(struct socket *so)
 449 {
 450         socantsendmore(so);
 451         return 0;
 452 }
 453
 454 static int
 455 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 456          struct mbuf *control, struct proc *p)
 457 {
 458         /* Packet must have a header (but that's about it) */
 459         if (m->m_len < sizeof (struct ip) &&
 460             (m = m_pullup(m, sizeof (struct ip))) == 0) {
 461                 ipstat.ips_toosmall++;
 462                 m_freem(m);
 463                 return EINVAL;
 464         }
 465
 466         /* Send packet */
 467         return div_output(so, m, nam, control);
 468 }
 469
 470 static int
 471 div_pcblist SYSCTL_HANDLER_ARGS
 472 {
 473         int error, i, n, s;
 474         struct inpcb *inp, **inp_list;
 475         inp_gen_t gencnt;
 476         struct xinpgen xig;
 477
 478         /*
 479          * The process of preparing the TCB list is too time-consuming and
 480          * resource-intensive to repeat twice on every request.
 481          */
 482         if (req->oldptr == 0) {
 483                 n = divcbinfo.ipi_count;
 484                 req->oldidx = 2 * (sizeof xig)
 485                         + (n + n/8) * sizeof(struct xinpcb);
 486                 return 0;
 487         }
 488
 489         if (req->newptr != 0)
 490                 return EPERM;
 491
 492         /*
 493          * OK, now we're committed to doing something.
 494          */
 495         s = splnet();
 496         gencnt = divcbinfo.ipi_gencnt;
 497         n = divcbinfo.ipi_count;
 498         splx(s);
 499
 500         xig.xig_len = sizeof xig;
 501         xig.xig_count = n;
 502         xig.xig_gen = gencnt;
 503         xig.xig_sogen = so_gencnt;
 504         error = SYSCTL_OUT(req, &xig, sizeof xig);
 505         if (error)
 506                 return error;
 507
 508         inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
 509         if (inp_list == 0)
 510                 return ENOMEM;
 511
 512         s = splnet();
 513         for (inp = LIST_FIRST(divcbinfo.listhead), i = 0; inp && i < n;
 514              inp = LIST_NEXT(inp, inp_list)) {
 515 #ifdef __APPLE__
 516                 if (inp->inp_gencnt <= gencnt)
 517 #else
 518                 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->p, inp))
 519 #endif
 520                         inp_list[i++] = inp;
 521         }
 522         splx(s);
 523         n = i;
 524
 525         error = 0;
 526         for (i = 0; i < n; i++) {
 527                 inp = inp_list[i];
 528                 if (inp->inp_gencnt <= gencnt) {
 529                         struct xinpcb xi;
 530                         xi.xi_len = sizeof xi;
 531                         /* XXX should avoid extra copy */
 532                         bcopy(inp, &xi.xi_inp, sizeof *inp);
 533                         if (inp->inp_socket)
 534                                 sotoxsocket(inp->inp_socket, &xi.xi_socket);
 535                         error = SYSCTL_OUT(req, &xi, sizeof xi);
 536                 }
 537         }
 538         if (!error) {
 539                 /*
 540                  * Give the user an updated idea of our state.
 541                  * If the generation differs from what we told
 542                  * her before, she knows that something happened
 543                  * while we were processing this request, and it
 544                  * might be necessary to retry.
 545                  */
 546                 s = splnet();
 547                 xig.xig_gen = divcbinfo.ipi_gencnt;
 548                 xig.xig_sogen = so_gencnt;
 549                 xig.xig_count = divcbinfo.ipi_count;
 550                 splx(s);
 551                 error = SYSCTL_OUT(req, &xig, sizeof xig);
 552         }
 553         FREE(inp_list, M_TEMP);
 554         return error;
 555 }
 556
 557 #warning Fix SYSCTL net_inet_divert
 558 #if 0
 559 SYSCTL_DECL(_net_inet_divert);
 560 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0,
 561             div_pcblist, "S,xinpcb", "List of active divert sockets");
 562 #endif
 563
 564 struct pr_usrreqs div_usrreqs = {
 565         div_abort, pru_accept_notsupp, div_attach, div_bind,
 566         pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach,
 567         div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
 568         pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown,
 569         in_setsockaddr, sosend, soreceive, sopoll
 570 };