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 1.36 2008/03/07 11:34:20 sephe 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 #ifdef SMP
  58 #include <sys/msgport.h>
  59 #endif
  60
  61 #include <vm/vm_zone.h>
  62
  63 #include <net/if.h>
  64 #include <net/route.h>
  65 #ifdef SMP
  66 #include <net/netmsg2.h>
  67 #endif
  68
  69 #include <netinet/in.h>
  70 #include <netinet/in_systm.h>
  71 #include <netinet/ip.h>
  72 #include <netinet/in_pcb.h>
  73 #include <netinet/in_var.h>
  74 #include <netinet/ip_var.h>
  75
  76 /*
  77  * Divert sockets
  78  */
  79
  80 /*
  81  * Allocate enough space to hold a full IP packet
  82  */
  83 #define DIVSNDQ         (65536 + 100)
  84 #define DIVRCVQ         (65536 + 100)
  85
  86 #define DIV_IS_OUTPUT(sin)      ((sin) == NULL || (sin)->sin_addr.s_addr == 0)
  87
  88 #define DIV_OUTPUT      0x10000
  89 #define DIV_INPUT       0x20000
  90
  91 /*
  92  * Divert sockets work in conjunction with ipfw, see the divert(4)
  93  * manpage for features.
  94  * Internally, packets selected by ipfw in ip_input() or ip_output(),
  95  * and never diverted before, are passed to the input queue of the
  96  * divert socket with a given 'divert_port' number (as specified in
  97  * the matching ipfw rule), and they are tagged with a 16 bit cookie
  98  * (representing the rule number of the matching ipfw rule), which
  99  * is passed to process reading from the socket.
 100  *
 101  * Packets written to the divert socket are again tagged with a cookie
 102  * (usually the same as above) and a destination address.
 103  * If the destination address is INADDR_ANY then the packet is
 104  * treated as outgoing and sent to ip_output(), otherwise it is
 105  * treated as incoming and sent to ip_input().
 106  * In both cases, the packet is tagged with the cookie.
 107  *
 108  * On reinjection, processing in ip_input() and ip_output()
 109  * will be exactly the same as for the original packet, except that
 110  * ipfw processing will start at the rule number after the one
 111  * written in the cookie (so, tagging a packet with a cookie of 0
 112  * will cause it to be effectively considered as a standard packet).
 113  */
 114
 115 /* Internal variables */
 116 static struct inpcbinfo divcbinfo;
 117
 118 static u_long   div_sendspace = DIVSNDQ;        /* XXX sysctl ? */
 119 static u_long   div_recvspace = DIVRCVQ;        /* XXX sysctl ? */
 120
 121 /*
 122  * Initialize divert connection block queue.
 123  */
 124 void
 125 div_init(void)
 126 {
 127         in_pcbinfo_init(&divcbinfo);
 128         /*
 129          * XXX We don't use the hash list for divert IP, but it's easier
 130          * to allocate a one entry hash list than it is to check all
 131          * over the place for hashbase == NULL.
 132          */
 133         divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
 134         divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
 135         divcbinfo.wildcardhashbase = hashinit(1, M_PCB,
 136                                               &divcbinfo.wildcardhashmask);
 137         divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb),
 138                                    maxsockets, ZONE_INTERRUPT, 0);
 139 }
 140
 141 /*
 142  * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
 143  * with that protocol number to enter the system from the outside.
 144  */
 145 void
 146 div_input(struct mbuf *m, ...)
 147 {
 148         ipstat.ips_noproto++;
 149         m_freem(m);
 150 }
 151
 152 struct lwkt_port *
 153 div_soport(struct socket *so, struct sockaddr *nam,
 154            struct mbuf **mptr, int req)
 155 {
 156         struct sockaddr_in *sin;
 157         struct mbuf *m;
 158         int dir;
 159
 160         /* Except for send(), everything happens on CPU0 */
 161         if (req != PRU_SEND)
 162                 return cpu0_soport(so, nam, mptr, req);
 163
 164         sin = (struct sockaddr_in *)nam;
 165         m = *mptr;
 166         M_ASSERTPKTHDR(m);
 167
 168         m->m_pkthdr.rcvif = NULL;
 169         dir = DIV_IS_OUTPUT(sin) ? IP_MPORT_OUT : IP_MPORT_IN;
 170
 171         if (sin != NULL) {
 172                 int i;
 173
 174                 /*
 175                  * Try locating the interface, if we originally had one.
 176                  * This is done even for outgoing packets, since for a
 177                  * forwarded packet, there must be an interface attached.
 178                  *
 179                  * Find receive interface with the given name, stuffed
 180                  * (if it exists) in the sin_zero[] field.
 181                  * The name is user supplied data so don't trust its size
 182                  * or that it is zero terminated.
 183                  */
 184                 for (i = 0; sin->sin_zero[i] && i < sizeof(sin->sin_zero); i++)
 185                         ;
 186                 if (i > 0 && i < sizeof(sin->sin_zero))
 187                         m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
 188         }
 189
 190         if (dir == IP_MPORT_IN && m->m_pkthdr.rcvif == NULL) {
 191                 /*
 192                  * No luck with the name, check by IP address.
 193                  * Clear the port and the ifname to make sure
 194                  * there are no distractions for ifa_ifwithaddr.
 195                  *
 196                  * Be careful not to trash sin->sin_port; it will
 197                  * be used later in div_output().
 198                  */
 199                 struct ifaddr *ifa;
 200                 u_short sin_port;
 201
 202                 bzero(sin->sin_zero, sizeof(sin->sin_zero));
 203                 sin_port = sin->sin_port; /* save */
 204                 sin->sin_port = 0;
 205                 ifa = ifa_ifwithaddr((struct sockaddr *)sin);
 206                 if (ifa == NULL) {
 207                         m_freem(m);
 208                         *mptr = NULL;
 209                         return NULL;
 210                 }
 211                 sin->sin_port = sin_port; /* restore */
 212                 m->m_pkthdr.rcvif = ifa->ifa_ifp;
 213         }
 214
 215         return ip_mport(mptr, dir);
 216 }
 217
 218 /*
 219  * Divert a packet by passing it up to the divert socket at port 'port'.
 220  *
 221  * Setup generic address and protocol structures for div_input routine,
 222  * then pass them along with mbuf chain.
 223  */
 224 static void
 225 div_packet(struct mbuf *m, int incoming, int port)
 226 {
 227         struct sockaddr_in divsrc = { sizeof divsrc, AF_INET };
 228         struct inpcb *inp;
 229         struct socket *sa;
 230         struct m_tag *mtag;
 231         u_int16_t nport;
 232
 233         /* Locate the divert tag */
 234         mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
 235         divsrc.sin_port = *(u_int16_t *)m_tag_data(mtag);
 236
 237         /*
 238          * Record receive interface address, if any.
 239          * But only for incoming packets.
 240          */
 241         divsrc.sin_addr.s_addr = 0;
 242         if (incoming) {
 243                 struct ifaddr_container *ifac;
 244
 245                 /* Find IP address for receive interface */
 246                 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
 247                               ifa_link) {
 248                         struct ifaddr *ifa = ifac->ifa;
 249
 250                         if (ifa->ifa_addr == NULL)
 251                                 continue;
 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         }
 259         /*
 260          * Record the incoming interface name whenever we have one.
 261          */
 262         if (m->m_pkthdr.rcvif) {
 263                 /*
 264                  * Hide the actual interface name in there in the
 265                  * sin_zero array. XXX This needs to be moved to a
 266                  * different sockaddr type for divert, e.g.
 267                  * sockaddr_div with multiple fields like
 268                  * sockaddr_dl. Presently we have only 7 bytes
 269                  * but that will do for now as most interfaces
 270                  * are 4 or less + 2 or less bytes for unit.
 271                  * There is probably a faster way of doing this,
 272                  * possibly taking it from the sockaddr_dl on the iface.
 273                  * This solves the problem of a P2P link and a LAN interface
 274                  * having the same address, which can result in the wrong
 275                  * interface being assigned to the packet when fed back
 276                  * into the divert socket. Theoretically if the daemon saves
 277                  * and re-uses the sockaddr_in as suggested in the man pages,
 278                  * this iface name will come along for the ride.
 279                  * (see div_output for the other half of this.)
 280                  */
 281                 ksnprintf(divsrc.sin_zero, sizeof divsrc.sin_zero,
 282                          m->m_pkthdr.rcvif->if_xname);
 283         }
 284
 285         /* Put packet on socket queue, if any */
 286         sa = NULL;
 287         nport = htons((u_int16_t)port);
 288         LIST_FOREACH(inp, &divcbinfo.pcblisthead, inp_list) {
 289                 if (inp->inp_flags & INP_PLACEMARKER)
 290                         continue;
 291                 if (inp->inp_lport == nport)
 292                         sa = inp->inp_socket;
 293         }
 294         if (sa) {
 295                 if (ssb_appendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, m,
 296                                  (struct mbuf *)NULL) == 0)
 297                         m_freem(m);
 298                 else
 299                         sorwakeup(sa);
 300         } else {
 301                 m_freem(m);
 302                 ipstat.ips_noproto++;
 303                 ipstat.ips_delivered--;
 304         }
 305 }
 306
 307 #ifdef SMP
 308 static void
 309 div_packet_handler(struct netmsg *nmsg)
 310 {
 311         struct netmsg_packet *nmp;
 312         struct lwkt_msg *msg;
 313         struct mbuf *m;
 314         int port, incoming = 0;
 315
 316         nmp = (struct netmsg_packet *)nmsg;
 317         m = nmp->nm_packet;
 318
 319         msg = &nmsg->nm_lmsg;
 320         port = msg->u.ms_result32 & 0xffff;
 321         if (msg->u.ms_result32 & DIV_INPUT)
 322                 incoming = 1;
 323
 324         div_packet(m, incoming, port);
 325 }
 326 #endif  /* SMP */
 327
 328 void
 329 divert_packet(struct mbuf *m, int incoming, int port)
 330 {
 331         /* Sanity check */
 332         KASSERT(port != 0, ("%s: port=0", __func__));
 333         M_ASSERTPKTHDR(m);
 334
 335         /* Assure header */
 336         if (m->m_len < sizeof(struct ip) &&
 337             (m = m_pullup(m, sizeof(struct ip))) == NULL)
 338                 return;
 339
 340         /* Sanity check */
 341         KASSERT(m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL) != NULL,
 342                 ("%s no divert tag!", __func__));
 343
 344 #ifdef SMP
 345         if (mycpuid != 0) {
 346                 struct netmsg_packet *nmp;
 347                 struct lwkt_msg *msg;
 348
 349                 nmp = &m->m_hdr.mh_netmsg;
 350                 netmsg_init(&nmp->nm_netmsg, &netisr_apanic_rport, 0,
 351                             div_packet_handler);
 352                 nmp->nm_packet = m;
 353
 354                 msg = &nmp->nm_netmsg.nm_lmsg;
 355                 msg->u.ms_result32 = port; /* port is 16bits */
 356                 if (incoming)
 357                         msg->u.ms_result32 |= DIV_INPUT;
 358                 else
 359                         msg->u.ms_result32 |= DIV_OUTPUT;
 360
 361                 lwkt_sendmsg(cpu_portfn(0), &nmp->nm_netmsg.nm_lmsg);
 362         } else
 363 #endif
 364         div_packet(m, incoming, port);
 365 }
 366
 367 /*
 368  * Deliver packet back into the IP processing machinery.
 369  *
 370  * If no address specified, or address is 0.0.0.0, send to ip_output();
 371  * otherwise, send to ip_input() and mark as having been received on
 372  * the interface with that address.
 373  */
 374 static int
 375 div_output(struct socket *so, struct mbuf *m,
 376         struct sockaddr_in *sin, struct mbuf *control)
 377 {
 378         int error = 0;
 379         struct m_tag *mtag;
 380
 381         if (control)
 382                 m_freem(control);               /* XXX */
 383
 384         /*
 385          * Prepare the tag for divert info. Note that a packet
 386          * with a 0 tag in mh_data is effectively untagged,
 387          * so we could optimize that case.
 388          */
 389         mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT, sizeof(u_int16_t), MB_DONTWAIT);
 390         if (mtag == NULL) {
 391                 error = ENOBUFS;
 392                 goto cantsend;
 393         }
 394         m_tag_prepend(m, mtag);
 395
 396         /* Loopback avoidance and state recovery */
 397         if (sin)
 398                 *(u_int16_t *)m_tag_data(mtag) = sin->sin_port;
 399         else
 400                 *(u_int16_t *)m_tag_data(mtag) = 0;
 401
 402         /* Reinject packet into the system as incoming or outgoing */
 403         if (DIV_IS_OUTPUT(sin)) {
 404                 struct ip *const ip = mtod(m, struct ip *);
 405
 406                 /* Don't allow packet length sizes that will crash */
 407                 if ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len) {
 408                         error = EINVAL;
 409                         goto cantsend;
 410                 }
 411
 412                 /* Convert fields to host order for ip_output() */
 413                 ip->ip_len = ntohs(ip->ip_len);
 414                 ip->ip_off = ntohs(ip->ip_off);
 415
 416                 /* Send packet to output processing */
 417                 ipstat.ips_rawout++;                    /* XXX */
 418                 error = ip_output(m, NULL, NULL,
 419                             (so->so_options & SO_DONTROUTE) |
 420                             IP_ALLOWBROADCAST | IP_RAWOUTPUT,
 421                             NULL, NULL);
 422         } else {
 423                 ip_input(m);
 424         }
 425         return error;
 426
 427 cantsend:
 428         m_freem(m);
 429         return error;
 430 }
 431
 432 static int
 433 div_attach(struct socket *so, int proto, struct pru_attach_info *ai)
 434 {
 435         struct inpcb *inp;
 436         int error;
 437
 438         inp  = so->so_pcb;
 439         if (inp)
 440                 panic("div_attach");
 441         if ((error = suser_cred(ai->p_ucred, NULL_CRED_OKAY)) != 0)
 442                 return error;
 443
 444         error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit);
 445         if (error)
 446                 return error;
 447         error = in_pcballoc(so, &divcbinfo);
 448         if (error)
 449                 return error;
 450         inp = (struct inpcb *)so->so_pcb;
 451         inp->inp_ip_p = proto;
 452         inp->inp_vflag |= INP_IPV4;
 453         inp->inp_flags |= INP_HDRINCL;
 454         /*
 455          * The socket is always "connected" because
 456          * we always know "where" to send the packet.
 457          */
 458         so->so_state |= SS_ISCONNECTED;
 459         return 0;
 460 }
 461
 462 static int
 463 div_detach(struct socket *so)
 464 {
 465         struct inpcb *inp;
 466
 467         inp = so->so_pcb;
 468         if (inp == NULL)
 469                 panic("div_detach");
 470         in_pcbdetach(inp);
 471         return 0;
 472 }
 473
 474 static int
 475 div_abort(struct socket *so)
 476 {
 477         soisdisconnected(so);
 478         return div_detach(so);
 479 }
 480
 481 static int
 482 div_disconnect(struct socket *so)
 483 {
 484         if (!(so->so_state & SS_ISCONNECTED))
 485                 return ENOTCONN;
 486         return div_abort(so);
 487 }
 488
 489 static int
 490 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 491 {
 492         int error;
 493
 494         /*
 495          * in_pcbbind assumes that nam is a sockaddr_in
 496          * and in_pcbbind requires a valid address. Since divert
 497          * sockets don't we need to make sure the address is
 498          * filled in properly.
 499          * XXX -- divert should not be abusing in_pcbind
 500          * and should probably have its own family.
 501          */
 502         if (nam->sa_family != AF_INET) {
 503                 error = EAFNOSUPPORT;
 504         } else {
 505                 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
 506                 error = in_pcbbind(so->so_pcb, nam, td);
 507         }
 508         return error;
 509 }
 510
 511 static int
 512 div_shutdown(struct socket *so)
 513 {
 514         socantsendmore(so);
 515         return 0;
 516 }
 517
 518 static int
 519 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 520          struct mbuf *control, struct thread *td)
 521 {
 522         /* Length check already done in ip_mport() */
 523         KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
 524
 525         /* Send packet */
 526         return div_output(so, m, (struct sockaddr_in *)nam, control);
 527 }
 528
 529 SYSCTL_DECL(_net_inet_divert);
 530 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, &divcbinfo, 0,
 531             in_pcblist_global, "S,xinpcb", "List of active divert sockets");
 532
 533 struct pr_usrreqs div_usrreqs = {
 534         .pru_abort = div_abort,
 535         .pru_accept = pru_accept_notsupp,
 536         .pru_attach = div_attach,
 537         .pru_bind = div_bind,
 538         .pru_connect = pru_connect_notsupp,
 539         .pru_connect2 = pru_connect2_notsupp,
 540         .pru_control = in_control,
 541         .pru_detach = div_detach,
 542         .pru_disconnect = div_disconnect,
 543         .pru_listen = pru_listen_notsupp,
 544         .pru_peeraddr = in_setpeeraddr,
 545         .pru_rcvd = pru_rcvd_notsupp,
 546         .pru_rcvoob = pru_rcvoob_notsupp,
 547         .pru_send = div_send,
 548         .pru_sense = pru_sense_null,
 549         .pru_shutdown = div_shutdown,
 550         .pru_sockaddr = in_setsockaddr,
 551         .pru_sosend = sosend,
 552         .pru_soreceive = soreceive,
 553         .pru_sopoll = sopoll
 554 };