net/ipv6/af_inet6.c

   1 /*
   2  *      PF_INET6 socket protocol family
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Adapted from linux/net/ipv4/af_inet.c
   9  *
  10  *      $Id: af_inet6.c,v 1.55 2000/02/27 19:51:47 davem Exp $
  11  *
  12  *      Fixes:
  13  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  14  *
  15  *      This program is free software; you can redistribute it and/or
  16  *      modify it under the terms of the GNU General Public License
  17  *      as published by the Free Software Foundation; either version
  18  *      2 of the License, or (at your option) any later version.
  19  */
  20
  21
  22 #include <linux/module.h>
  23 #include <linux/config.h>
  24 #include <linux/errno.h>
  25 #include <linux/types.h>
  26 #include <linux/socket.h>
  27 #include <linux/in.h>
  28 #include <linux/kernel.h>
  29 #include <linux/major.h>
  30 #include <linux/sched.h>
  31 #include <linux/timer.h>
  32 #include <linux/string.h>
  33 #include <linux/sockios.h>
  34 #include <linux/net.h>
  35 #include <linux/fcntl.h>
  36 #include <linux/mm.h>
  37 #include <linux/interrupt.h>
  38 #include <linux/proc_fs.h>
  39 #include <linux/stat.h>
  40 #include <linux/init.h>
  41 #include <linux/version.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <linux/icmpv6.h>
  46 #include <linux/smp_lock.h>
  47
  48 #include <net/ip.h>
  49 #include <net/ipv6.h>
  50 #include <net/udp.h>
  51 #include <net/tcp.h>
  52 #include <net/ipip.h>
  53 #include <net/protocol.h>
  54 #include <net/inet_common.h>
  55 #include <net/transp_v6.h>
  56 #include <net/ip6_route.h>
  57 #include <net/addrconf.h>
  58
  59 #include <asm/uaccess.h>
  60 #include <asm/system.h>
  61
  62 #ifdef MODULE
  63 static int unloadable = 0; /* XX: Turn to one when all is ok within the
  64                               module for allowing unload */
  65 #endif
  66
  67 #if defined(MODULE) && LINUX_VERSION_CODE > 0x20115
  68 MODULE_AUTHOR("Cast of dozens");
  69 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  70 MODULE_PARM(unloadable, "i");
  71 #endif
  72
  73 extern struct proto_ops inet6_stream_ops;
  74 extern struct proto_ops inet6_dgram_ops;
  75
  76 /* IPv6 procfs goodies... */
  77
  78 #ifdef CONFIG_PROC_FS
  79 extern int raw6_get_info(char *, char **, off_t, int);
  80 extern int tcp6_get_info(char *, char **, off_t, int);
  81 extern int udp6_get_info(char *, char **, off_t, int);
  82 extern int afinet6_get_info(char *, char **, off_t, int);
  83 extern int afinet6_get_snmp(char *, char **, off_t, int);
  84 #endif
  85
  86 #ifdef CONFIG_SYSCTL
  87 extern void ipv6_sysctl_register(void);
  88 extern void ipv6_sysctl_unregister(void);
  89 #endif
  90
  91 #ifdef INET_REFCNT_DEBUG
  92 atomic_t inet6_sock_nr;
  93 #endif
  94
  95 static void inet6_sock_destruct(struct sock *sk)
  96 {
  97         inet_sock_destruct(sk);
  98
  99 #ifdef INET_REFCNT_DEBUG
 100         atomic_dec(&inet6_sock_nr);
 101 #endif
 102         MOD_DEC_USE_COUNT;
 103 }
 104
 105 static int inet6_create(struct socket *sock, int protocol)
 106 {
 107         struct sock *sk;
 108         struct proto *prot;
 109
 110         sk = sk_alloc(PF_INET6, GFP_KERNEL, 1);
 111         if (sk == NULL)
 112                 goto do_oom;
 113
 114         if(sock->type == SOCK_STREAM || sock->type == SOCK_SEQPACKET) {
 115                 if (protocol && protocol != IPPROTO_TCP)
 116                         goto free_and_noproto;
 117                 protocol = IPPROTO_TCP;
 118                 prot = &tcpv6_prot;
 119                 sock->ops = &inet6_stream_ops;
 120         } else if(sock->type == SOCK_DGRAM) {
 121                 if (protocol && protocol != IPPROTO_UDP)
 122                         goto free_and_noproto;
 123                 protocol = IPPROTO_UDP;
 124                 sk->no_check = UDP_CSUM_DEFAULT;
 125                 prot=&udpv6_prot;
 126                 sock->ops = &inet6_dgram_ops;
 127         } else if(sock->type == SOCK_RAW) {
 128                 if (!capable(CAP_NET_RAW))
 129                         goto free_and_badperm;
 130                 if (!protocol)
 131                         goto free_and_noproto;
 132                 prot = &rawv6_prot;
 133                 sock->ops = &inet6_dgram_ops;
 134                 sk->reuse = 1;
 135                 sk->num = protocol;
 136         } else {
 137                 goto free_and_badtype;
 138         }
 139
 140         sock_init_data(sock, sk);
 141
 142         sk->destruct            = inet6_sock_destruct;
 143         sk->zapped              = 0;
 144         sk->family              = PF_INET6;
 145         sk->protocol            = protocol;
 146
 147         sk->prot                = prot;
 148         sk->backlog_rcv         = prot->backlog_rcv;
 149
 150         sk->net_pinfo.af_inet6.hop_limit  = -1;
 151         sk->net_pinfo.af_inet6.mcast_hops = -1;
 152         sk->net_pinfo.af_inet6.mc_loop    = 1;
 153         sk->net_pinfo.af_inet6.pmtudisc   = IPV6_PMTUDISC_WANT;
 154
 155         /* Init the ipv4 part of the socket since we can have sockets
 156          * using v6 API for ipv4.
 157          */
 158         sk->protinfo.af_inet.ttl        = 64;
 159
 160         sk->protinfo.af_inet.mc_loop    = 1;
 161         sk->protinfo.af_inet.mc_ttl     = 1;
 162         sk->protinfo.af_inet.mc_index   = 0;
 163         sk->protinfo.af_inet.mc_list    = NULL;
 164
 165         if (ipv4_config.no_pmtu_disc)
 166                 sk->protinfo.af_inet.pmtudisc = IP_PMTUDISC_DONT;
 167         else
 168                 sk->protinfo.af_inet.pmtudisc = IP_PMTUDISC_WANT;
 169
 170
 171 #ifdef INET_REFCNT_DEBUG
 172         atomic_inc(&inet6_sock_nr);
 173         atomic_inc(&inet_sock_nr);
 174 #endif
 175         MOD_INC_USE_COUNT;
 176
 177         if (sk->type==SOCK_RAW && protocol==IPPROTO_RAW)
 178                 sk->protinfo.af_inet.hdrincl=1;
 179
 180         if (sk->num) {
 181                 /* It assumes that any protocol which allows
 182                  * the user to assign a number at socket
 183                  * creation time automatically shares.
 184                  */
 185                 sk->sport = ntohs(sk->num);
 186                 sk->prot->hash(sk);
 187         }
 188
 189         if (sk->prot->init) {
 190                 int err = sk->prot->init(sk);
 191                 if (err != 0) {
 192                         sk->dead = 1;
 193                         inet_sock_release(sk);
 194                         return(err);
 195                 }
 196         }
 197         return(0);
 198
 199 free_and_badtype:
 200         sk_free(sk);
 201         return -ESOCKTNOSUPPORT;
 202 free_and_badperm:
 203         sk_free(sk);
 204         return -EPERM;
 205 free_and_noproto:
 206         sk_free(sk);
 207         return -EPROTONOSUPPORT;
 208 do_oom:
 209         return -ENOBUFS;
 210 }
 211
 212
 213 /* bind for INET6 API */
 214 static int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 215 {
 216         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 217         struct sock *sk = sock->sk;
 218         __u32 v4addr = 0;
 219         unsigned short snum;
 220         int addr_type = 0;
 221
 222         /* If the socket has its own bind function then use it. */
 223         if(sk->prot->bind)
 224                 return sk->prot->bind(sk, uaddr, addr_len);
 225
 226         if (addr_len < SIN6_LEN_RFC2133)
 227                 return -EINVAL;
 228         addr_type = ipv6_addr_type(&addr->sin6_addr);
 229         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 230                 return -EINVAL;
 231
 232         /* Check if the address belongs to the host. */
 233         if (addr_type == IPV6_ADDR_MAPPED) {
 234                 v4addr = addr->sin6_addr.s6_addr32[3];
 235                 if (inet_addr_type(v4addr) != RTN_LOCAL)
 236                         return -EADDRNOTAVAIL;
 237         } else {
 238                 if (addr_type != IPV6_ADDR_ANY) {
 239                         /* ipv4 addr of the socket is invalid.  Only the
 240                          * unpecified and mapped address have a v4 equivalent.
 241                          */
 242                         v4addr = LOOPBACK4_IPV6;
 243                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 244                                 if (!ipv6_chk_addr(&addr->sin6_addr, NULL))
 245                                         return -EADDRNOTAVAIL;
 246                         }
 247                 }
 248         }
 249
 250         snum = ntohs(addr->sin6_port);
 251         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 252                 return -EACCES;
 253
 254         lock_sock(sk);
 255
 256         /* Check these errors (active socket, double bind). */
 257         if ((sk->state != TCP_CLOSE)                    ||
 258             (sk->num != 0)) {
 259                 release_sock(sk);
 260                 return -EINVAL;
 261         }
 262
 263         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 264                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 265                     addr->sin6_scope_id) {
 266                         /* Override any existing binding, if another one
 267                          * is supplied by user.
 268                          */
 269                         sk->bound_dev_if = addr->sin6_scope_id;
 270                 }
 271
 272                 /* Binding to link-local address requires an interface */
 273                 if (sk->bound_dev_if == 0) {
 274                         release_sock(sk);
 275                         return -EINVAL;
 276                 }
 277         }
 278
 279         sk->rcv_saddr = v4addr;
 280         sk->saddr = v4addr;
 281
 282         ipv6_addr_copy(&sk->net_pinfo.af_inet6.rcv_saddr, &addr->sin6_addr);
 283
 284         if (!(addr_type & IPV6_ADDR_MULTICAST))
 285                 ipv6_addr_copy(&sk->net_pinfo.af_inet6.saddr, &addr->sin6_addr);
 286
 287         /* Make sure we are allowed to bind here. */
 288         if (sk->prot->get_port(sk, snum) != 0) {
 289                 sk->rcv_saddr = 0;
 290                 sk->saddr = 0;
 291                 memset(&sk->net_pinfo.af_inet6.rcv_saddr, 0, sizeof(struct in6_addr));
 292                 memset(&sk->net_pinfo.af_inet6.saddr, 0, sizeof(struct in6_addr));
 293
 294                 release_sock(sk);
 295                 return -EADDRINUSE;
 296         }
 297
 298         sk->sport = ntohs(sk->num);
 299         sk->dport = 0;
 300         sk->daddr = 0;
 301         sk->prot->hash(sk);
 302         release_sock(sk);
 303
 304         return 0;
 305 }
 306
 307 static int inet6_release(struct socket *sock)
 308 {
 309         struct sock *sk = sock->sk;
 310
 311         if (sk == NULL)
 312                 return -EINVAL;
 313
 314         /* Free mc lists */
 315         ipv6_sock_mc_close(sk);
 316
 317         return inet_release(sock);
 318 }
 319
 320 int inet6_destroy_sock(struct sock *sk)
 321 {
 322         struct sk_buff *skb;
 323         struct ipv6_txoptions *opt;
 324
 325         /*
 326          *      Release destination entry
 327          */
 328
 329         sk_dst_reset(sk);
 330
 331         /* Release rx options */
 332
 333         if ((skb = xchg(&sk->net_pinfo.af_inet6.pktoptions, NULL)) != NULL)
 334                 kfree_skb(skb);
 335
 336         /* Free flowlabels */
 337         fl6_free_socklist(sk);
 338
 339         /* Free tx options */
 340
 341         if ((opt = xchg(&sk->net_pinfo.af_inet6.opt, NULL)) != NULL)
 342                 sock_kfree_s(sk, opt, opt->tot_len);
 343
 344         return 0;
 345 }
 346
 347 /*
 348  *      This does both peername and sockname.
 349  */
 350
 351 static int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 352                  int *uaddr_len, int peer)
 353 {
 354         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 355         struct sock *sk = sock->sk;
 356
 357         sin->sin6_family = AF_INET6;
 358         sin->sin6_flowinfo = 0;
 359         sin->sin6_scope_id = 0;
 360         if (peer) {
 361                 if (!sk->dport)
 362                         return -ENOTCONN;
 363                 if (((1<<sk->state)&(TCPF_CLOSE|TCPF_SYN_SENT)) && peer == 1)
 364                         return -ENOTCONN;
 365                 sin->sin6_port = sk->dport;
 366                 memcpy(&sin->sin6_addr, &sk->net_pinfo.af_inet6.daddr,
 367                        sizeof(struct in6_addr));
 368                 if (sk->net_pinfo.af_inet6.sndflow)
 369                         sin->sin6_flowinfo = sk->net_pinfo.af_inet6.flow_label;
 370         } else {
 371                 if (ipv6_addr_type(&sk->net_pinfo.af_inet6.rcv_saddr) == IPV6_ADDR_ANY)
 372                         memcpy(&sin->sin6_addr,
 373                                &sk->net_pinfo.af_inet6.saddr,
 374                                sizeof(struct in6_addr));
 375                 else
 376                         memcpy(&sin->sin6_addr,
 377                                &sk->net_pinfo.af_inet6.rcv_saddr,
 378                                sizeof(struct in6_addr));
 379
 380                 sin->sin6_port = sk->sport;
 381         }
 382         if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 383                 sin->sin6_scope_id = sk->bound_dev_if;
 384         *uaddr_len = sizeof(*sin);
 385         return(0);
 386 }
 387
 388 static int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 389 {
 390         struct sock *sk = sock->sk;
 391         int err = -EINVAL;
 392         int pid;
 393
 394         switch(cmd)
 395         {
 396         case FIOSETOWN:
 397         case SIOCSPGRP:
 398                 err = get_user(pid, (int *) arg);
 399                 if(err)
 400                         return err;
 401
 402                 /* see sock_no_fcntl */
 403                 if (current->pid != pid && current->pgrp != -pid &&
 404                     !capable(CAP_NET_ADMIN))
 405                         return -EPERM;
 406                 sk->proc = pid;
 407                 return(0);
 408         case FIOGETOWN:
 409         case SIOCGPGRP:
 410                 err = put_user(sk->proc,(int *)arg);
 411                 if(err)
 412                         return err;
 413                 return(0);
 414         case SIOCGSTAMP:
 415                 if(sk->stamp.tv_sec==0)
 416                         return -ENOENT;
 417                 err = copy_to_user((void *)arg, &sk->stamp,
 418                                    sizeof(struct timeval));
 419                 if (err)
 420                         return -EFAULT;
 421                 return 0;
 422
 423         case SIOCADDRT:
 424         case SIOCDELRT:
 425
 426                 return(ipv6_route_ioctl(cmd,(void *)arg));
 427
 428         case SIOCSIFADDR:
 429                 return addrconf_add_ifaddr((void *) arg);
 430         case SIOCDIFADDR:
 431                 return addrconf_del_ifaddr((void *) arg);
 432         case SIOCSIFDSTADDR:
 433                 return addrconf_set_dstaddr((void *) arg);
 434         default:
 435                 if ((cmd >= SIOCDEVPRIVATE) &&
 436                     (cmd <= (SIOCDEVPRIVATE + 15)))
 437                         return(dev_ioctl(cmd,(void *) arg));
 438
 439                 if(sk->prot->ioctl==0 || (err=sk->prot->ioctl(sk, cmd, arg))==-ENOIOCTLCMD)
 440                         return(dev_ioctl(cmd,(void *) arg));
 441                 return err;
 442         }
 443         /*NOTREACHED*/
 444         return(0);
 445 }
 446
 447 struct proto_ops inet6_stream_ops = {
 448         PF_INET6,
 449
 450         inet6_release,
 451         inet6_bind,
 452         inet_stream_connect,            /* ok           */
 453         sock_no_socketpair,             /* a do nothing */
 454         inet_accept,                    /* ok           */
 455         inet6_getname,
 456         tcp_poll,                       /* ok           */
 457         inet6_ioctl,                    /* must change  */
 458         inet_listen,                    /* ok           */
 459         inet_shutdown,                  /* ok           */
 460         inet_setsockopt,                /* ok           */
 461         inet_getsockopt,                /* ok           */
 462         sock_no_fcntl,                  /* ok           */
 463         inet_sendmsg,                   /* ok           */
 464         inet_recvmsg,                   /* ok           */
 465         sock_no_mmap
 466 };
 467
 468 struct proto_ops inet6_dgram_ops = {
 469         PF_INET6,
 470
 471         inet6_release,
 472         inet6_bind,
 473         inet_dgram_connect,             /* ok           */
 474         sock_no_socketpair,             /* a do nothing */
 475         sock_no_accept,                 /* a do nothing */
 476         inet6_getname,
 477         datagram_poll,                  /* ok           */
 478         inet6_ioctl,                    /* must change  */
 479         sock_no_listen,                 /* ok           */
 480         inet_shutdown,                  /* ok           */
 481         inet_setsockopt,                /* ok           */
 482         inet_getsockopt,                /* ok           */
 483         sock_no_fcntl,                  /* ok           */
 484         inet_sendmsg,                   /* ok           */
 485         inet_recvmsg,                   /* ok           */
 486         sock_no_mmap,
 487 };
 488
 489 struct net_proto_family inet6_family_ops = {
 490         PF_INET6,
 491         inet6_create
 492 };
 493
 494 #ifdef MODULE
 495 int ipv6_unload(void)
 496 {
 497         if (!unloadable) return 1;
 498         /* We keep internally 3 raw sockets */
 499         return atomic_read(&(__this_module.uc.usecount)) - 3;
 500 }
 501 #endif
 502
 503 #if defined(MODULE) && defined(CONFIG_SYSCTL)
 504 extern void ipv6_sysctl_register(void);
 505 extern void ipv6_sysctl_unregister(void);
 506 #endif
 507
 508 #ifdef MODULE
 509 int init_module(void)
 510 #else
 511 void __init inet6_proto_init(struct net_proto *pro)
 512 #endif
 513 {
 514         struct sk_buff *dummy_skb;
 515         int err;
 516
 517 #ifdef MODULE
 518         if (!mod_member_present(&__this_module, can_unload))
 519           return -EINVAL;
 520
 521         __this_module.can_unload = &ipv6_unload;
 522 #endif
 523
 524         printk(KERN_INFO "IPv6 v0.8 for NET4.0\n");
 525
 526         if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb))
 527         {
 528                 printk(KERN_CRIT "inet6_proto_init: size fault\n");
 529 #ifdef MODULE
 530                 return -EINVAL;
 531 #else
 532                 return;
 533 #endif
 534         }
 535
 536         /*
 537          *      ipngwg API draft makes clear that the correct semantics
 538          *      for TCP and UDP is to consider one TCP and UDP instance
 539          *      in a host availiable by both INET and INET6 APIs and
 540          *      able to communicate via both network protocols.
 541          */
 542
 543 #if defined(MODULE) && defined(CONFIG_SYSCTL)
 544         ipv6_sysctl_register();
 545 #endif
 546         err = icmpv6_init(&inet6_family_ops);
 547         if (err)
 548                 goto icmp_fail;
 549         err = ndisc_init(&inet6_family_ops);
 550         if (err)
 551                 goto ndisc_fail;
 552         err = igmp6_init(&inet6_family_ops);
 553         if (err)
 554                 goto igmp_fail;
 555         ipv6_netdev_notif_init();
 556         ipv6_packet_init();
 557         ip6_route_init();
 558         ip6_flowlabel_init();
 559         addrconf_init();
 560         sit_init();
 561
 562         /* Init v6 transport protocols. */
 563         udpv6_init();
 564         tcpv6_init();
 565
 566         /* Create /proc/foo6 entries. */
 567 #ifdef CONFIG_PROC_FS
 568         proc_net_create("raw6", 0, raw6_get_info);
 569         proc_net_create("tcp6", 0, tcp6_get_info);
 570         proc_net_create("udp6", 0, udp6_get_info);
 571         proc_net_create("sockstat6", 0, afinet6_get_info);
 572         proc_net_create("snmp6", 0, afinet6_get_snmp);
 573 #endif
 574
 575         /* Now the userspace is allowed to create INET6 sockets. */
 576         (void) sock_register(&inet6_family_ops);
 577
 578 #ifdef MODULE
 579         return 0;
 580 #else
 581         return;
 582 #endif
 583
 584 igmp_fail:
 585         ndisc_cleanup();
 586 ndisc_fail:
 587         icmpv6_cleanup();
 588 icmp_fail:
 589 #if defined(MODULE) && defined(CONFIG_SYSCTL)
 590         ipv6_sysctl_unregister();
 591 #endif
 592 #ifdef MODULE
 593         return err;
 594 #else
 595         return;
 596 #endif
 597 }
 598
 599 #ifdef MODULE
 600 void cleanup_module(void)
 601 {
 602         /* First of all disallow new sockets creation. */
 603         sock_unregister(PF_INET6);
 604 #ifdef CONFIG_PROC_FS
 605         proc_net_remove("raw6");
 606         proc_net_remove("tcp6");
 607         proc_net_remove("udp6");
 608         proc_net_remove("sockstat6");
 609         proc_net_remove("snmp6");
 610 #endif
 611         /* Cleanup code parts. */
 612         sit_cleanup();
 613         ipv6_netdev_notif_cleanup();
 614         ip6_flowlabel_cleanup();
 615         addrconf_cleanup();
 616         ip6_route_cleanup();
 617         ipv6_packet_cleanup();
 618         igmp6_cleanup();
 619         ndisc_cleanup();
 620         icmpv6_cleanup();
 621 #ifdef CONFIG_SYSCTL
 622         ipv6_sysctl_unregister();
 623 #endif
 624 }
 625 #endif  /* MODULE */