sys/netinet/raw_ip.c

   1 /*-
   2  * Copyright (c) 1982, 1986, 1988, 1993
   3  *      The Regents of the University of California.
   4  * All rights reserved.
   5  *
   6  * Redistribution and use in source and binary forms, with or without
   7  * modification, are permitted provided that the following conditions
   8  * are met:
   9  * 1. Redistributions of source code must retain the above copyright
  10  *    notice, this list of conditions and the following disclaimer.
  11  * 2. Redistributions in binary form must reproduce the above copyright
  12  *    notice, this list of conditions and the following disclaimer in the
  13  *    documentation and/or other materials provided with the distribution.
  14  * 4. Neither the name of the University nor the names of its contributors
  15  *    may be used to endorse or promote products derived from this software
  16  *    without specific prior written permission.
  17  *
  18  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  28  * SUCH DAMAGE.
  29  *
  30  *      @(#)raw_ip.c    8.7 (Berkeley) 5/15/95
  31  */
  32
  33 #include <sys/cdefs.h>
  34 __FBSDID("$FreeBSD$");
  35
  36 #include "opt_inet6.h"
  37 #include "opt_ipsec.h"
  38 #include "opt_mac.h"
  39
  40 #include <sys/param.h>
  41 #include <sys/jail.h>
  42 #include <sys/kernel.h>
  43 #include <sys/lock.h>
  44 #include <sys/malloc.h>
  45 #include <sys/mbuf.h>
  46 #include <sys/priv.h>
  47 #include <sys/proc.h>
  48 #include <sys/protosw.h>
  49 #include <sys/signalvar.h>
  50 #include <sys/socket.h>
  51 #include <sys/socketvar.h>
  52 #include <sys/sx.h>
  53 #include <sys/sysctl.h>
  54 #include <sys/systm.h>
  55 #include <sys/vimage.h>
  56
  57 #include <vm/uma.h>
  58
  59 #include <net/if.h>
  60 #include <net/route.h>
  61
  62 #include <netinet/in.h>
  63 #include <netinet/in_systm.h>
  64 #include <netinet/in_pcb.h>
  65 #include <netinet/in_var.h>
  66 #include <netinet/ip.h>
  67 #include <netinet/ip_var.h>
  68 #include <netinet/ip_mroute.h>
  69
  70 #include <netinet/ip_fw.h>
  71 #include <netinet/ip_dummynet.h>
  72
  73 #ifdef IPSEC
  74 #include <netipsec/ipsec.h>
  75 #endif /*IPSEC*/
  76
  77 #include <security/mac/mac_framework.h>
  78
  79 struct  inpcbhead ripcb;
  80 struct  inpcbinfo ripcbinfo;
  81
  82 /* control hooks for ipfw and dummynet */
  83 ip_fw_ctl_t *ip_fw_ctl_ptr = NULL;
  84 ip_dn_ctl_t *ip_dn_ctl_ptr = NULL;
  85
  86 /*
  87  * Hooks for multicast routing. They all default to NULL, so leave them not
  88  * initialized and rely on BSS being set to 0.
  89  */
  90
  91 /*
  92  * The socket used to communicate with the multicast routing daemon.
  93  */
  94 struct socket  *ip_mrouter;
  95
  96 /*
  97  * The various mrouter and rsvp functions.
  98  */
  99 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
 100 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
 101 int (*ip_mrouter_done)(void);
 102 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
 103                    struct ip_moptions *);
 104 int (*mrt_ioctl)(int, caddr_t, int);
 105 int (*legal_vif_num)(int);
 106 u_long (*ip_mcast_src)(int);
 107
 108 void (*rsvp_input_p)(struct mbuf *m, int off);
 109 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
 110 void (*ip_rsvp_force_done)(struct socket *);
 111
 112 /*
 113  * Hash functions
 114  */
 115
 116 #define INP_PCBHASH_RAW_SIZE    256
 117 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
 118         (((proto) + (laddr) + (faddr)) % (mask) + 1)
 119
 120 static void
 121 rip_inshash(struct inpcb *inp)
 122 {
 123         struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
 124         struct inpcbhead *pcbhash;
 125         int hash;
 126
 127         INP_INFO_WLOCK_ASSERT(pcbinfo);
 128         INP_WLOCK_ASSERT(inp);
 129
 130         if (inp->inp_ip_p != 0 &&
 131             inp->inp_laddr.s_addr != INADDR_ANY &&
 132             inp->inp_faddr.s_addr != INADDR_ANY) {
 133                 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
 134                     inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
 135         } else
 136                 hash = 0;
 137         pcbhash = &pcbinfo->ipi_hashbase[hash];
 138         LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
 139 }
 140
 141 static void
 142 rip_delhash(struct inpcb *inp)
 143 {
 144
 145         INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
 146         INP_WLOCK_ASSERT(inp);
 147
 148         LIST_REMOVE(inp, inp_hash);
 149 }
 150
 151 /*
 152  * Raw interface to IP protocol.
 153  */
 154
 155 /*
 156  * Initialize raw connection block q.
 157  */
 158 static void
 159 rip_zone_change(void *tag)
 160 {
 161
 162         uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
 163 }
 164
 165 static int
 166 rip_inpcb_init(void *mem, int size, int flags)
 167 {
 168         struct inpcb *inp = mem;
 169
 170         INP_LOCK_INIT(inp, "inp", "rawinp");
 171         return (0);
 172 }
 173
 174 void
 175 rip_init(void)
 176 {
 177
 178         INP_INFO_LOCK_INIT(&V_ripcbinfo, "rip");
 179         LIST_INIT(&V_ripcb);
 180         V_ripcbinfo.ipi_listhead = &V_ripcb;
 181         V_ripcbinfo.ipi_hashbase =
 182             hashinit(INP_PCBHASH_RAW_SIZE, M_PCB, &V_ripcbinfo.ipi_hashmask);
 183         V_ripcbinfo.ipi_porthashbase =
 184             hashinit(1, M_PCB, &V_ripcbinfo.ipi_porthashmask);
 185         V_ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb),
 186             NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
 187         uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
 188         EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
 189             EVENTHANDLER_PRI_ANY);
 190 }
 191
 192 static int
 193 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
 194     struct sockaddr_in *ripsrc)
 195 {
 196         int policyfail = 0;
 197
 198         INP_RLOCK_ASSERT(last);
 199
 200 #ifdef IPSEC
 201         /* check AH/ESP integrity. */
 202         if (ipsec4_in_reject(n, last)) {
 203                 policyfail = 1;
 204         }
 205 #endif /* IPSEC */
 206 #ifdef MAC
 207         if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
 208                 policyfail = 1;
 209 #endif
 210         /* Check the minimum TTL for socket. */
 211         if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
 212                 policyfail = 1;
 213         if (!policyfail) {
 214                 struct mbuf *opts = NULL;
 215                 struct socket *so;
 216
 217                 so = last->inp_socket;
 218                 if ((last->inp_flags & INP_CONTROLOPTS) ||
 219                     (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
 220                         ip_savecontrol(last, &opts, ip, n);
 221                 SOCKBUF_LOCK(&so->so_rcv);
 222                 if (sbappendaddr_locked(&so->so_rcv,
 223                     (struct sockaddr *)ripsrc, n, opts) == 0) {
 224                         /* should notify about lost packet */
 225                         m_freem(n);
 226                         if (opts)
 227                                 m_freem(opts);
 228                         SOCKBUF_UNLOCK(&so->so_rcv);
 229                 } else
 230                         sorwakeup_locked(so);
 231         } else
 232                 m_freem(n);
 233         return (policyfail);
 234 }
 235
 236 /*
 237  * Setup generic address and protocol structures for raw_input routine, then
 238  * pass them along with mbuf chain.
 239  */
 240 void
 241 rip_input(struct mbuf *m, int off)
 242 {
 243         struct ip *ip = mtod(m, struct ip *);
 244         int proto = ip->ip_p;
 245         struct inpcb *inp, *last;
 246         struct sockaddr_in ripsrc;
 247         int hash;
 248
 249         bzero(&ripsrc, sizeof(ripsrc));
 250         ripsrc.sin_len = sizeof(ripsrc);
 251         ripsrc.sin_family = AF_INET;
 252         ripsrc.sin_addr = ip->ip_src;
 253         last = NULL;
 254         hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
 255             ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
 256         INP_INFO_RLOCK(&V_ripcbinfo);
 257         LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
 258                 if (inp->inp_ip_p != proto)
 259                         continue;
 260 #ifdef INET6
 261                 if ((inp->inp_vflag & INP_IPV4) == 0)
 262                         continue;
 263 #endif
 264                 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
 265                         continue;
 266                 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
 267                         continue;
 268                 INP_RLOCK(inp);
 269                 if (jailed(inp->inp_socket->so_cred) &&
 270                     (htonl(prison_getip(inp->inp_socket->so_cred)) !=
 271                     ip->ip_dst.s_addr)) {
 272                         INP_RUNLOCK(inp);
 273                         continue;
 274                 }
 275                 if (last) {
 276                         struct mbuf *n;
 277
 278                         n = m_copy(m, 0, (int)M_COPYALL);
 279                         if (n != NULL)
 280                             (void) rip_append(last, ip, n, &ripsrc);
 281                         /* XXX count dropped packet */
 282                         INP_RUNLOCK(last);
 283                 }
 284                 last = inp;
 285         }
 286         LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
 287                 if (inp->inp_ip_p && inp->inp_ip_p != proto)
 288                         continue;
 289 #ifdef INET6
 290                 if ((inp->inp_vflag & INP_IPV4) == 0)
 291                         continue;
 292 #endif
 293                 if (inp->inp_laddr.s_addr &&
 294                     inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
 295                         continue;
 296                 if (inp->inp_faddr.s_addr &&
 297                     inp->inp_faddr.s_addr != ip->ip_src.s_addr)
 298                         continue;
 299                 INP_RLOCK(inp);
 300                 if (jailed(inp->inp_socket->so_cred) &&
 301                     (htonl(prison_getip(inp->inp_socket->so_cred)) !=
 302                     ip->ip_dst.s_addr)) {
 303                         INP_RUNLOCK(inp);
 304                         continue;
 305                 }
 306                 if (last) {
 307                         struct mbuf *n;
 308
 309                         n = m_copy(m, 0, (int)M_COPYALL);
 310                         if (n != NULL)
 311                                 (void) rip_append(last, ip, n, &ripsrc);
 312                         /* XXX count dropped packet */
 313                         INP_RUNLOCK(last);
 314                 }
 315                 last = inp;
 316         }
 317         INP_INFO_RUNLOCK(&V_ripcbinfo);
 318         if (last != NULL) {
 319                 if (rip_append(last, ip, m, &ripsrc) != 0)
 320                         V_ipstat.ips_delivered--;
 321                 INP_RUNLOCK(last);
 322         } else {
 323                 m_freem(m);
 324                 V_ipstat.ips_noproto++;
 325                 V_ipstat.ips_delivered--;
 326         }
 327 }
 328
 329 /*
 330  * Generate IP header and pass packet to ip_output.  Tack on options user may
 331  * have setup with control call.
 332  */
 333 int
 334 rip_output(struct mbuf *m, struct socket *so, u_long dst)
 335 {
 336         struct ip *ip;
 337         int error;
 338         struct inpcb *inp = sotoinpcb(so);
 339         int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
 340             IP_ALLOWBROADCAST;
 341
 342         /*
 343          * If the user handed us a complete IP packet, use it.  Otherwise,
 344          * allocate an mbuf for a header and fill it in.
 345          */
 346         if ((inp->inp_flags & INP_HDRINCL) == 0) {
 347                 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
 348                         m_freem(m);
 349                         return(EMSGSIZE);
 350                 }
 351                 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
 352                 if (m == NULL)
 353                         return(ENOBUFS);
 354
 355                 INP_RLOCK(inp);
 356                 ip = mtod(m, struct ip *);
 357                 ip->ip_tos = inp->inp_ip_tos;
 358                 if (inp->inp_flags & INP_DONTFRAG)
 359                         ip->ip_off = IP_DF;
 360                 else
 361                         ip->ip_off = 0;
 362                 ip->ip_p = inp->inp_ip_p;
 363                 ip->ip_len = m->m_pkthdr.len;
 364                 if (jailed(inp->inp_socket->so_cred))
 365                         ip->ip_src.s_addr =
 366                             htonl(prison_getip(inp->inp_socket->so_cred));
 367                 else
 368                         ip->ip_src = inp->inp_laddr;
 369                 ip->ip_dst.s_addr = dst;
 370                 ip->ip_ttl = inp->inp_ip_ttl;
 371         } else {
 372                 if (m->m_pkthdr.len > IP_MAXPACKET) {
 373                         m_freem(m);
 374                         return(EMSGSIZE);
 375                 }
 376                 INP_RLOCK(inp);
 377                 ip = mtod(m, struct ip *);
 378                 if (jailed(inp->inp_socket->so_cred)) {
 379                         if (ip->ip_src.s_addr !=
 380                             htonl(prison_getip(inp->inp_socket->so_cred))) {
 381                                 INP_RUNLOCK(inp);
 382                                 m_freem(m);
 383                                 return (EPERM);
 384                         }
 385                 }
 386
 387                 /*
 388                  * Don't allow both user specified and setsockopt options,
 389                  * and don't allow packet length sizes that will crash.
 390                  */
 391                 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options)
 392                     || (ip->ip_len > m->m_pkthdr.len)
 393                     || (ip->ip_len < (ip->ip_hl << 2))) {
 394                         INP_RUNLOCK(inp);
 395                         m_freem(m);
 396                         return (EINVAL);
 397                 }
 398                 if (ip->ip_id == 0)
 399                         ip->ip_id = ip_newid();
 400
 401                 /*
 402                  * XXX prevent ip_output from overwriting header fields.
 403                  */
 404                 flags |= IP_RAWOUTPUT;
 405                 V_ipstat.ips_rawout++;
 406         }
 407
 408         if (inp->inp_flags & INP_ONESBCAST)
 409                 flags |= IP_SENDONES;
 410
 411 #ifdef MAC
 412         mac_inpcb_create_mbuf(inp, m);
 413 #endif
 414
 415         error = ip_output(m, inp->inp_options, NULL, flags,
 416             inp->inp_moptions, inp);
 417         INP_RUNLOCK(inp);
 418         return (error);
 419 }
 420
 421 /*
 422  * Raw IP socket option processing.
 423  *
 424  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
 425  * only be created by a privileged process, and as such, socket option
 426  * operations to manage system properties on any raw socket were allowed to
 427  * take place without explicit additional access control checks.  However,
 428  * raw sockets can now also be created in jail(), and therefore explicit
 429  * checks are now required.  Likewise, raw sockets can be used by a process
 430  * after it gives up privilege, so some caution is required.  For options
 431  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
 432  * performed in ip_ctloutput() and therefore no check occurs here.
 433  * Unilaterally checking priv_check() here breaks normal IP socket option
 434  * operations on raw sockets.
 435  *
 436  * When adding new socket options here, make sure to add access control
 437  * checks here as necessary.
 438  */
 439 int
 440 rip_ctloutput(struct socket *so, struct sockopt *sopt)
 441 {
 442         struct  inpcb *inp = sotoinpcb(so);
 443         int     error, optval;
 444
 445         if (sopt->sopt_level != IPPROTO_IP)
 446                 return (EINVAL);
 447
 448         error = 0;
 449         switch (sopt->sopt_dir) {
 450         case SOPT_GET:
 451                 switch (sopt->sopt_name) {
 452                 case IP_HDRINCL:
 453                         optval = inp->inp_flags & INP_HDRINCL;
 454                         error = sooptcopyout(sopt, &optval, sizeof optval);
 455                         break;
 456
 457                 case IP_FW_ADD: /* ADD actually returns the body... */
 458                 case IP_FW_GET:
 459                 case IP_FW_TABLE_GETSIZE:
 460                 case IP_FW_TABLE_LIST:
 461                 case IP_FW_NAT_GET_CONFIG:
 462                 case IP_FW_NAT_GET_LOG:
 463                         if (ip_fw_ctl_ptr != NULL)
 464                                 error = ip_fw_ctl_ptr(sopt);
 465                         else
 466                                 error = ENOPROTOOPT;
 467                         break;
 468
 469                 case IP_DUMMYNET_GET:
 470                         if (ip_dn_ctl_ptr != NULL)
 471                                 error = ip_dn_ctl_ptr(sopt);
 472                         else
 473                                 error = ENOPROTOOPT;
 474                         break ;
 475
 476                 case MRT_INIT:
 477                 case MRT_DONE:
 478                 case MRT_ADD_VIF:
 479                 case MRT_DEL_VIF:
 480                 case MRT_ADD_MFC:
 481                 case MRT_DEL_MFC:
 482                 case MRT_VERSION:
 483                 case MRT_ASSERT:
 484                 case MRT_API_SUPPORT:
 485                 case MRT_API_CONFIG:
 486                 case MRT_ADD_BW_UPCALL:
 487                 case MRT_DEL_BW_UPCALL:
 488                         error = priv_check(curthread, PRIV_NETINET_MROUTE);
 489                         if (error != 0)
 490                                 return (error);
 491                         error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
 492                                 EOPNOTSUPP;
 493                         break;
 494
 495                 default:
 496                         error = ip_ctloutput(so, sopt);
 497                         break;
 498                 }
 499                 break;
 500
 501         case SOPT_SET:
 502                 switch (sopt->sopt_name) {
 503                 case IP_HDRINCL:
 504                         error = sooptcopyin(sopt, &optval, sizeof optval,
 505                                             sizeof optval);
 506                         if (error)
 507                                 break;
 508                         if (optval)
 509                                 inp->inp_flags |= INP_HDRINCL;
 510                         else
 511                                 inp->inp_flags &= ~INP_HDRINCL;
 512                         break;
 513
 514                 case IP_FW_ADD:
 515                 case IP_FW_DEL:
 516                 case IP_FW_FLUSH:
 517                 case IP_FW_ZERO:
 518                 case IP_FW_RESETLOG:
 519                 case IP_FW_TABLE_ADD:
 520                 case IP_FW_TABLE_DEL:
 521                 case IP_FW_TABLE_FLUSH:
 522                 case IP_FW_NAT_CFG:
 523                 case IP_FW_NAT_DEL:
 524                         if (ip_fw_ctl_ptr != NULL)
 525                                 error = ip_fw_ctl_ptr(sopt);
 526                         else
 527                                 error = ENOPROTOOPT;
 528                         break;
 529
 530                 case IP_DUMMYNET_CONFIGURE:
 531                 case IP_DUMMYNET_DEL:
 532                 case IP_DUMMYNET_FLUSH:
 533                         if (ip_dn_ctl_ptr != NULL)
 534                                 error = ip_dn_ctl_ptr(sopt);
 535                         else
 536                                 error = ENOPROTOOPT ;
 537                         break ;
 538
 539                 case IP_RSVP_ON:
 540                         error = priv_check(curthread, PRIV_NETINET_MROUTE);
 541                         if (error != 0)
 542                                 return (error);
 543                         error = ip_rsvp_init(so);
 544                         break;
 545
 546                 case IP_RSVP_OFF:
 547                         error = priv_check(curthread, PRIV_NETINET_MROUTE);
 548                         if (error != 0)
 549                                 return (error);
 550                         error = ip_rsvp_done();
 551                         break;
 552
 553                 case IP_RSVP_VIF_ON:
 554                 case IP_RSVP_VIF_OFF:
 555                         error = priv_check(curthread, PRIV_NETINET_MROUTE);
 556                         if (error != 0)
 557                                 return (error);
 558                         error = ip_rsvp_vif ?
 559                                 ip_rsvp_vif(so, sopt) : EINVAL;
 560                         break;
 561
 562                 case MRT_INIT:
 563                 case MRT_DONE:
 564                 case MRT_ADD_VIF:
 565                 case MRT_DEL_VIF:
 566                 case MRT_ADD_MFC:
 567                 case MRT_DEL_MFC:
 568                 case MRT_VERSION:
 569                 case MRT_ASSERT:
 570                 case MRT_API_SUPPORT:
 571                 case MRT_API_CONFIG:
 572                 case MRT_ADD_BW_UPCALL:
 573                 case MRT_DEL_BW_UPCALL:
 574                         error = priv_check(curthread, PRIV_NETINET_MROUTE);
 575                         if (error != 0)
 576                                 return (error);
 577                         error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
 578                                         EOPNOTSUPP;
 579                         break;
 580
 581                 default:
 582                         error = ip_ctloutput(so, sopt);
 583                         break;
 584                 }
 585                 break;
 586         }
 587
 588         return (error);
 589 }
 590
 591 /*
 592  * This function exists solely to receive the PRC_IFDOWN messages which are
 593  * sent by if_down().  It looks for an ifaddr whose ifa_addr is sa, and calls
 594  * in_ifadown() to remove all routes corresponding to that address.  It also
 595  * receives the PRC_IFUP messages from if_up() and reinstalls the interface
 596  * routes.
 597  */
 598 void
 599 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
 600 {
 601         struct in_ifaddr *ia;
 602         struct ifnet *ifp;
 603         int err;
 604         int flags;
 605
 606         switch (cmd) {
 607         case PRC_IFDOWN:
 608                 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
 609                         if (ia->ia_ifa.ifa_addr == sa
 610                             && (ia->ia_flags & IFA_ROUTE)) {
 611                                 /*
 612                                  * in_ifscrub kills the interface route.
 613                                  */
 614                                 in_ifscrub(ia->ia_ifp, ia);
 615                                 /*
 616                                  * in_ifadown gets rid of all the rest of the
 617                                  * routes.  This is not quite the right thing
 618                                  * to do, but at least if we are running a
 619                                  * routing process they will come back.
 620                                  */
 621                                 in_ifadown(&ia->ia_ifa, 0);
 622                                 break;
 623                         }
 624                 }
 625                 break;
 626
 627         case PRC_IFUP:
 628                 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
 629                         if (ia->ia_ifa.ifa_addr == sa)
 630                                 break;
 631                 }
 632                 if (ia == 0 || (ia->ia_flags & IFA_ROUTE))
 633                         return;
 634                 flags = RTF_UP;
 635                 ifp = ia->ia_ifa.ifa_ifp;
 636
 637                 if ((ifp->if_flags & IFF_LOOPBACK)
 638                     || (ifp->if_flags & IFF_POINTOPOINT))
 639                         flags |= RTF_HOST;
 640
 641                 err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
 642                 if (err == 0)
 643                         ia->ia_flags |= IFA_ROUTE;
 644                 break;
 645         }
 646 }
 647
 648 u_long  rip_sendspace = 9216;
 649 u_long  rip_recvspace = 9216;
 650
 651 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
 652     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
 653 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
 654     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
 655
 656 static int
 657 rip_attach(struct socket *so, int proto, struct thread *td)
 658 {
 659         struct inpcb *inp;
 660         int error;
 661
 662         inp = sotoinpcb(so);
 663         KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
 664
 665         error = priv_check(td, PRIV_NETINET_RAW);
 666         if (error)
 667                 return (error);
 668         if (proto >= IPPROTO_MAX || proto < 0)
 669                 return EPROTONOSUPPORT;
 670         error = soreserve(so, rip_sendspace, rip_recvspace);
 671         if (error)
 672                 return (error);
 673         INP_INFO_WLOCK(&V_ripcbinfo);
 674         error = in_pcballoc(so, &V_ripcbinfo);
 675         if (error) {
 676                 INP_INFO_WUNLOCK(&V_ripcbinfo);
 677                 return (error);
 678         }
 679         inp = (struct inpcb *)so->so_pcb;
 680         inp->inp_vflag |= INP_IPV4;
 681         inp->inp_ip_p = proto;
 682         inp->inp_ip_ttl = V_ip_defttl;
 683         rip_inshash(inp);
 684         INP_INFO_WUNLOCK(&V_ripcbinfo);
 685         INP_WUNLOCK(inp);
 686         return (0);
 687 }
 688
 689 static void
 690 rip_detach(struct socket *so)
 691 {
 692         struct inpcb *inp;
 693
 694         inp = sotoinpcb(so);
 695         KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
 696         KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
 697             ("rip_detach: not closed"));
 698
 699         INP_INFO_WLOCK(&V_ripcbinfo);
 700         INP_WLOCK(inp);
 701         rip_delhash(inp);
 702         if (so == V_ip_mrouter && ip_mrouter_done)
 703                 ip_mrouter_done();
 704         if (ip_rsvp_force_done)
 705                 ip_rsvp_force_done(so);
 706         if (so == V_ip_rsvpd)
 707                 ip_rsvp_done();
 708         in_pcbdetach(inp);
 709         in_pcbfree(inp);
 710         INP_INFO_WUNLOCK(&V_ripcbinfo);
 711 }
 712
 713 static void
 714 rip_dodisconnect(struct socket *so, struct inpcb *inp)
 715 {
 716
 717         INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
 718         INP_WLOCK_ASSERT(inp);
 719
 720         rip_delhash(inp);
 721         inp->inp_faddr.s_addr = INADDR_ANY;
 722         rip_inshash(inp);
 723         SOCK_LOCK(so);
 724         so->so_state &= ~SS_ISCONNECTED;
 725         SOCK_UNLOCK(so);
 726 }
 727
 728 static void
 729 rip_abort(struct socket *so)
 730 {
 731         struct inpcb *inp;
 732
 733         inp = sotoinpcb(so);
 734         KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
 735
 736         INP_INFO_WLOCK(&V_ripcbinfo);
 737         INP_WLOCK(inp);
 738         rip_dodisconnect(so, inp);
 739         INP_WUNLOCK(inp);
 740         INP_INFO_WUNLOCK(&V_ripcbinfo);
 741 }
 742
 743 static void
 744 rip_close(struct socket *so)
 745 {
 746         struct inpcb *inp;
 747
 748         inp = sotoinpcb(so);
 749         KASSERT(inp != NULL, ("rip_close: inp == NULL"));
 750
 751         INP_INFO_WLOCK(&V_ripcbinfo);
 752         INP_WLOCK(inp);
 753         rip_dodisconnect(so, inp);
 754         INP_WUNLOCK(inp);
 755         INP_INFO_WUNLOCK(&V_ripcbinfo);
 756 }
 757
 758 static int
 759 rip_disconnect(struct socket *so)
 760 {
 761         struct inpcb *inp;
 762
 763         if ((so->so_state & SS_ISCONNECTED) == 0)
 764                 return (ENOTCONN);
 765
 766         inp = sotoinpcb(so);
 767         KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
 768
 769         INP_INFO_WLOCK(&V_ripcbinfo);
 770         INP_WLOCK(inp);
 771         rip_dodisconnect(so, inp);
 772         INP_WUNLOCK(inp);
 773         INP_INFO_WUNLOCK(&V_ripcbinfo);
 774         return (0);
 775 }
 776
 777 static int
 778 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
 779 {
 780         struct sockaddr_in *addr = (struct sockaddr_in *)nam;
 781         struct inpcb *inp;
 782
 783         if (nam->sa_len != sizeof(*addr))
 784                 return (EINVAL);
 785
 786         if (jailed(td->td_ucred)) {
 787                 if (addr->sin_addr.s_addr == INADDR_ANY)
 788                         addr->sin_addr.s_addr =
 789                             htonl(prison_getip(td->td_ucred));
 790                 if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr)
 791                         return (EADDRNOTAVAIL);
 792         }
 793
 794         if (TAILQ_EMPTY(&V_ifnet) ||
 795             (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
 796             (addr->sin_addr.s_addr &&
 797              ifa_ifwithaddr((struct sockaddr *)addr) == 0))
 798                 return (EADDRNOTAVAIL);
 799
 800         inp = sotoinpcb(so);
 801         KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
 802
 803         INP_INFO_WLOCK(&V_ripcbinfo);
 804         INP_WLOCK(inp);
 805         rip_delhash(inp);
 806         inp->inp_laddr = addr->sin_addr;
 807         rip_inshash(inp);
 808         INP_WUNLOCK(inp);
 809         INP_INFO_WUNLOCK(&V_ripcbinfo);
 810         return (0);
 811 }
 812
 813 static int
 814 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
 815 {
 816         struct sockaddr_in *addr = (struct sockaddr_in *)nam;
 817         struct inpcb *inp;
 818
 819         if (nam->sa_len != sizeof(*addr))
 820                 return (EINVAL);
 821         if (TAILQ_EMPTY(&V_ifnet))
 822                 return (EADDRNOTAVAIL);
 823         if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
 824                 return (EAFNOSUPPORT);
 825
 826         inp = sotoinpcb(so);
 827         KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
 828
 829         INP_INFO_WLOCK(&V_ripcbinfo);
 830         INP_WLOCK(inp);
 831         rip_delhash(inp);
 832         inp->inp_faddr = addr->sin_addr;
 833         rip_inshash(inp);
 834         soisconnected(so);
 835         INP_WUNLOCK(inp);
 836         INP_INFO_WUNLOCK(&V_ripcbinfo);
 837         return (0);
 838 }
 839
 840 static int
 841 rip_shutdown(struct socket *so)
 842 {
 843         struct inpcb *inp;
 844
 845         inp = sotoinpcb(so);
 846         KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
 847
 848         INP_WLOCK(inp);
 849         socantsendmore(so);
 850         INP_WUNLOCK(inp);
 851         return (0);
 852 }
 853
 854 static int
 855 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
 856     struct mbuf *control, struct thread *td)
 857 {
 858         struct inpcb *inp;
 859         u_long dst;
 860
 861         inp = sotoinpcb(so);
 862         KASSERT(inp != NULL, ("rip_send: inp == NULL"));
 863
 864         /*
 865          * Note: 'dst' reads below are unlocked.
 866          */
 867         if (so->so_state & SS_ISCONNECTED) {
 868                 if (nam) {
 869                         m_freem(m);
 870                         return (EISCONN);
 871                 }
 872                 dst = inp->inp_faddr.s_addr;    /* Unlocked read. */
 873         } else {
 874                 if (nam == NULL) {
 875                         m_freem(m);
 876                         return (ENOTCONN);
 877                 }
 878                 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
 879         }
 880         return (rip_output(m, so, dst));
 881 }
 882
 883 static int
 884 rip_pcblist(SYSCTL_HANDLER_ARGS)
 885 {
 886         int error, i, n;
 887         struct inpcb *inp, **inp_list;
 888         inp_gen_t gencnt;
 889         struct xinpgen xig;
 890
 891         /*
 892          * The process of preparing the TCB list is too time-consuming and
 893          * resource-intensive to repeat twice on every request.
 894          */
 895         if (req->oldptr == 0) {
 896                 n = V_ripcbinfo.ipi_count;
 897                 req->oldidx = 2 * (sizeof xig)
 898                     + (n + n/8) * sizeof(struct xinpcb);
 899                 return (0);
 900         }
 901
 902         if (req->newptr != 0)
 903                 return (EPERM);
 904
 905         /*
 906          * OK, now we're committed to doing something.
 907          */
 908         INP_INFO_RLOCK(&V_ripcbinfo);
 909         gencnt = V_ripcbinfo.ipi_gencnt;
 910         n = V_ripcbinfo.ipi_count;
 911         INP_INFO_RUNLOCK(&V_ripcbinfo);
 912
 913         xig.xig_len = sizeof xig;
 914         xig.xig_count = n;
 915         xig.xig_gen = gencnt;
 916         xig.xig_sogen = so_gencnt;
 917         error = SYSCTL_OUT(req, &xig, sizeof xig);
 918         if (error)
 919                 return (error);
 920
 921         inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
 922         if (inp_list == 0)
 923                 return (ENOMEM);
 924
 925         INP_INFO_RLOCK(&V_ripcbinfo);
 926         for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n;
 927              inp = LIST_NEXT(inp, inp_list)) {
 928                 INP_RLOCK(inp);
 929                 if (inp->inp_gencnt <= gencnt &&
 930                     cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) {
 931                         /* XXX held references? */
 932                         inp_list[i++] = inp;
 933                 }
 934                 INP_RUNLOCK(inp);
 935         }
 936         INP_INFO_RUNLOCK(&V_ripcbinfo);
 937         n = i;
 938
 939         error = 0;
 940         for (i = 0; i < n; i++) {
 941                 inp = inp_list[i];
 942                 INP_RLOCK(inp);
 943                 if (inp->inp_gencnt <= gencnt) {
 944                         struct xinpcb xi;
 945                         bzero(&xi, sizeof(xi));
 946                         xi.xi_len = sizeof xi;
 947                         /* XXX should avoid extra copy */
 948                         bcopy(inp, &xi.xi_inp, sizeof *inp);
 949                         if (inp->inp_socket)
 950                                 sotoxsocket(inp->inp_socket, &xi.xi_socket);
 951                         INP_RUNLOCK(inp);
 952                         error = SYSCTL_OUT(req, &xi, sizeof xi);
 953                 } else
 954                         INP_RUNLOCK(inp);
 955         }
 956         if (!error) {
 957                 /*
 958                  * Give the user an updated idea of our state.  If the
 959                  * generation differs from what we told her before, she knows
 960                  * that something happened while we were processing this
 961                  * request, and it might be necessary to retry.
 962                  */
 963                 INP_INFO_RLOCK(&V_ripcbinfo);
 964                 xig.xig_gen = V_ripcbinfo.ipi_gencnt;
 965                 xig.xig_sogen = so_gencnt;
 966                 xig.xig_count = V_ripcbinfo.ipi_count;
 967                 INP_INFO_RUNLOCK(&V_ripcbinfo);
 968                 error = SYSCTL_OUT(req, &xig, sizeof xig);
 969         }
 970         free(inp_list, M_TEMP);
 971         return (error);
 972 }
 973
 974 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0,
 975     rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
 976
 977 struct pr_usrreqs rip_usrreqs = {
 978         .pru_abort =            rip_abort,
 979         .pru_attach =           rip_attach,
 980         .pru_bind =             rip_bind,
 981         .pru_connect =          rip_connect,
 982         .pru_control =          in_control,
 983         .pru_detach =           rip_detach,
 984         .pru_disconnect =       rip_disconnect,
 985         .pru_peeraddr =         in_getpeeraddr,
 986         .pru_send =             rip_send,
 987         .pru_shutdown =         rip_shutdown,
 988         .pru_sockaddr =         in_getsockaddr,
 989         .pru_sosetlabel =       in_pcbsosetlabel,
 990         .pru_close =            rip_close,
 991 };