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 1.52 2000/01/18 08:24:21 davem Exp $
  11  *
  12  *      This program is free software; you can redistribute it and/or
  13  *      modify it under the terms of the GNU General Public License
  14  *      as published by the Free Software Foundation; either version
  15  *      2 of the License, or (at your option) any later version.
  16  */
  17
  18
  19 #include <linux/module.h>
  20 #include <linux/config.h>
  21 #include <linux/errno.h>
  22 #include <linux/types.h>
  23 #include <linux/socket.h>
  24 #include <linux/in.h>
  25 #include <linux/kernel.h>
  26 #include <linux/major.h>
  27 #include <linux/sched.h>
  28 #include <linux/timer.h>
  29 #include <linux/string.h>
  30 #include <linux/sockios.h>
  31 #include <linux/net.h>
  32 #include <linux/fcntl.h>
  33 #include <linux/mm.h>
  34 #include <linux/interrupt.h>
  35 #include <linux/proc_fs.h>
  36 #include <linux/stat.h>
  37 #include <linux/init.h>
  38 #include <linux/version.h>
  39
  40 #include <linux/inet.h>
  41 #include <linux/netdevice.h>
  42 #include <linux/icmpv6.h>
  43 #include <linux/smp_lock.h>
  44
  45 #include <net/ip.h>
  46 #include <net/ipv6.h>
  47 #include <net/udp.h>
  48 #include <net/tcp.h>
  49 #include <net/ipip.h>
  50 #include <net/protocol.h>
  51 #include <net/inet_common.h>
  52 #include <net/transp_v6.h>
  53 #include <net/ip6_route.h>
  54 #include <net/addrconf.h>
  55
  56 #include <asm/uaccess.h>
  57 #include <asm/system.h>
  58
  59 #ifdef MODULE
  60 static int unloadable = 0; /* XX: Turn to one when all is ok within the
  61                               module for allowing unload */
  62 #endif
  63
  64 #if defined(MODULE) && LINUX_VERSION_CODE > 0x20115
  65 MODULE_AUTHOR("Cast of dozens");
  66 MODULE_DESCRIPTION("IPv6 protocol stack for Linux");
  67 MODULE_PARM(unloadable, "i");
  68 #endif
  69
  70 extern struct proto_ops inet6_stream_ops;
  71 extern struct proto_ops inet6_dgram_ops;
  72
  73 /* IPv6 procfs goodies... */
  74
  75 #ifdef CONFIG_PROC_FS
  76 extern int raw6_get_info(char *, char **, off_t, int);
  77 extern int tcp6_get_info(char *, char **, off_t, int);
  78 extern int udp6_get_info(char *, char **, off_t, int);
  79 extern int afinet6_get_info(char *, char **, off_t, int);
  80 extern int afinet6_get_snmp(char *, char **, off_t, int);
  81 #endif
  82
  83 #ifdef CONFIG_SYSCTL
  84 extern void ipv6_sysctl_register(void);
  85 extern void ipv6_sysctl_unregister(void);
  86 #endif
  87
  88 #ifdef INET_REFCNT_DEBUG
  89 atomic_t inet6_sock_nr;
  90 #endif
  91
  92 static void inet6_sock_destruct(struct sock *sk)
  93 {
  94         inet_sock_destruct(sk);
  95
  96 #ifdef INET_REFCNT_DEBUG
  97         atomic_dec(&inet6_sock_nr);
  98 #endif
  99         MOD_DEC_USE_COUNT;
 100 }
 101
 102 static int inet6_create(struct socket *sock, int protocol)
 103 {
 104         struct sock *sk;
 105         struct proto *prot;
 106
 107         sk = sk_alloc(PF_INET6, GFP_KERNEL, 1);
 108         if (sk == NULL)
 109                 goto do_oom;
 110
 111         if(sock->type == SOCK_STREAM || sock->type == SOCK_SEQPACKET) {
 112                 if (protocol && protocol != IPPROTO_TCP)
 113                         goto free_and_noproto;
 114                 protocol = IPPROTO_TCP;
 115                 prot = &tcpv6_prot;
 116                 sock->ops = &inet6_stream_ops;
 117         } else if(sock->type == SOCK_DGRAM) {
 118                 if (protocol && protocol != IPPROTO_UDP)
 119                         goto free_and_noproto;
 120                 protocol = IPPROTO_UDP;
 121                 sk->no_check = UDP_CSUM_DEFAULT;
 122                 prot=&udpv6_prot;
 123                 sock->ops = &inet6_dgram_ops;
 124         } else if(sock->type == SOCK_RAW) {
 125                 if (!capable(CAP_NET_RAW))
 126                         goto free_and_badperm;
 127                 if (!protocol)
 128                         goto free_and_noproto;
 129                 prot = &rawv6_prot;
 130                 sock->ops = &inet6_dgram_ops;
 131                 sk->reuse = 1;
 132                 sk->num = protocol;
 133         } else {
 134                 goto free_and_badtype;
 135         }
 136
 137         sock_init_data(sock, sk);
 138
 139         sk->destruct            = inet6_sock_destruct;
 140         sk->zapped              = 0;
 141         sk->family              = PF_INET6;
 142         sk->protocol            = protocol;
 143
 144         sk->prot                = prot;
 145         sk->backlog_rcv         = prot->backlog_rcv;
 146
 147         sk->net_pinfo.af_inet6.hop_limit  = -1;
 148         sk->net_pinfo.af_inet6.mcast_hops = -1;
 149         sk->net_pinfo.af_inet6.mc_loop    = 1;
 150         sk->net_pinfo.af_inet6.pmtudisc   = IPV6_PMTUDISC_WANT;
 151
 152         /* Init the ipv4 part of the socket since we can have sockets
 153          * using v6 API for ipv4.
 154          */
 155         sk->protinfo.af_inet.ttl        = 64;
 156
 157         sk->protinfo.af_inet.mc_loop    = 1;
 158         sk->protinfo.af_inet.mc_ttl     = 1;
 159         sk->protinfo.af_inet.mc_index   = 0;
 160         sk->protinfo.af_inet.mc_list    = NULL;
 161
 162         if (ipv4_config.no_pmtu_disc)
 163                 sk->protinfo.af_inet.pmtudisc = IP_PMTUDISC_DONT;
 164         else
 165                 sk->protinfo.af_inet.pmtudisc = IP_PMTUDISC_WANT;
 166
 167
 168 #ifdef INET_REFCNT_DEBUG
 169         atomic_inc(&inet6_sock_nr);
 170         atomic_inc(&inet_sock_nr);
 171 #endif
 172         MOD_INC_USE_COUNT;
 173
 174         if (sk->type==SOCK_RAW && protocol==IPPROTO_RAW)
 175                 sk->protinfo.af_inet.hdrincl=1;
 176
 177         if (sk->num) {
 178                 /* It assumes that any protocol which allows
 179                  * the user to assign a number at socket
 180                  * creation time automatically shares.
 181                  */
 182                 sk->sport = ntohs(sk->num);
 183                 sk->prot->hash(sk);
 184         }
 185
 186         if (sk->prot->init) {
 187                 int err = sk->prot->init(sk);
 188                 if (err != 0) {
 189                         sk->dead = 1;
 190                         inet_sock_release(sk);
 191                         return(err);
 192                 }
 193         }
 194         return(0);
 195
 196 free_and_badtype:
 197         sk_free(sk);
 198         return -ESOCKTNOSUPPORT;
 199 free_and_badperm:
 200         sk_free(sk);
 201         return -EPERM;
 202 free_and_noproto:
 203         sk_free(sk);
 204         return -EPROTONOSUPPORT;
 205 do_oom:
 206         return -ENOBUFS;
 207 }
 208
 209
 210 /* bind for INET6 API */
 211 static int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 212 {
 213         struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr;
 214         struct sock *sk = sock->sk;
 215         __u32 v4addr = 0;
 216         unsigned short snum;
 217         int addr_type = 0;
 218
 219         /* If the socket has its own bind function then use it. */
 220         if(sk->prot->bind)
 221                 return sk->prot->bind(sk, uaddr, addr_len);
 222
 223         if (addr_len < sizeof(struct sockaddr_in6))
 224                 return -EINVAL;
 225
 226         addr_type = ipv6_addr_type(&addr->sin6_addr);
 227         if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM)
 228                 return -EINVAL;
 229
 230         /* Check if the address belongs to the host. */
 231         if (addr_type == IPV6_ADDR_MAPPED) {
 232                 v4addr = addr->sin6_addr.s6_addr32[3];
 233                 if (inet_addr_type(v4addr) != RTN_LOCAL)
 234                         return -EADDRNOTAVAIL;
 235         } else {
 236                 if (addr_type != IPV6_ADDR_ANY) {
 237                         /* ipv4 addr of the socket is invalid.  Only the
 238                          * unpecified and mapped address have a v4 equivalent.
 239                          */
 240                         v4addr = LOOPBACK4_IPV6;
 241                         if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 242                                 if (!ipv6_chk_addr(&addr->sin6_addr, NULL))
 243                                         return -EADDRNOTAVAIL;
 244                         }
 245                 }
 246         }
 247
 248         snum = ntohs(addr->sin6_port);
 249         if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
 250                 return -EACCES;
 251
 252         lock_sock(sk);
 253
 254         /* Check these errors (active socket, double bind). */
 255         if ((sk->state != TCP_CLOSE)                    ||
 256             (sk->num != 0)) {
 257                 release_sock(sk);
 258                 return -EINVAL;
 259         }
 260
 261         sk->rcv_saddr = v4addr;
 262         sk->saddr = v4addr;
 263
 264         ipv6_addr_copy(&sk->net_pinfo.af_inet6.rcv_saddr, &addr->sin6_addr);
 265
 266         if (!(addr_type & IPV6_ADDR_MULTICAST))
 267                 ipv6_addr_copy(&sk->net_pinfo.af_inet6.saddr, &addr->sin6_addr);
 268
 269         /* Make sure we are allowed to bind here. */
 270         if (sk->prot->get_port(sk, snum) != 0) {
 271                 sk->rcv_saddr = 0;
 272                 sk->saddr = 0;
 273                 memset(&sk->net_pinfo.af_inet6.rcv_saddr, 0, sizeof(struct in6_addr));
 274                 memset(&sk->net_pinfo.af_inet6.saddr, 0, sizeof(struct in6_addr));
 275
 276                 release_sock(sk);
 277                 return -EADDRINUSE;
 278         }
 279
 280         sk->sport = ntohs(sk->num);
 281         sk->dport = 0;
 282         sk->daddr = 0;
 283         sk->prot->hash(sk);
 284         release_sock(sk);
 285
 286         return 0;
 287 }
 288
 289 static int inet6_release(struct socket *sock)
 290 {
 291         struct sock *sk = sock->sk;
 292
 293         if (sk == NULL)
 294                 return -EINVAL;
 295
 296         /* Free mc lists */
 297         ipv6_sock_mc_close(sk);
 298
 299         return inet_release(sock);
 300 }
 301
 302 int inet6_destroy_sock(struct sock *sk)
 303 {
 304         struct sk_buff *skb;
 305         struct ipv6_txoptions *opt;
 306
 307         /*
 308          *      Release destination entry
 309          */
 310
 311         sk_dst_reset(sk);
 312
 313         /* Release rx options */
 314
 315         if ((skb = xchg(&sk->net_pinfo.af_inet6.pktoptions, NULL)) != NULL)
 316                 kfree_skb(skb);
 317
 318         /* Free flowlabels */
 319         fl6_free_socklist(sk);
 320
 321         /* Free tx options */
 322
 323         if ((opt = xchg(&sk->net_pinfo.af_inet6.opt, NULL)) != NULL)
 324                 sock_kfree_s(sk, opt, opt->tot_len);
 325
 326         return 0;
 327 }
 328
 329 /*
 330  *      This does both peername and sockname.
 331  */
 332
 333 static int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 334                  int *uaddr_len, int peer)
 335 {
 336         struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr;
 337         struct sock *sk = sock->sk;
 338
 339         sin->sin6_family = AF_INET6;
 340         sin->sin6_flowinfo = 0;
 341         if (peer) {
 342                 if (!sk->dport)
 343                         return -ENOTCONN;
 344                 sin->sin6_port = sk->dport;
 345                 memcpy(&sin->sin6_addr, &sk->net_pinfo.af_inet6.daddr,
 346                        sizeof(struct in6_addr));
 347                 if (sk->net_pinfo.af_inet6.sndflow)
 348                         sin->sin6_flowinfo = sk->net_pinfo.af_inet6.flow_label;
 349         } else {
 350                 if (ipv6_addr_type(&sk->net_pinfo.af_inet6.rcv_saddr) == IPV6_ADDR_ANY)
 351                         memcpy(&sin->sin6_addr,
 352                                &sk->net_pinfo.af_inet6.saddr,
 353                                sizeof(struct in6_addr));
 354                 else
 355                         memcpy(&sin->sin6_addr,
 356                                &sk->net_pinfo.af_inet6.rcv_saddr,
 357                                sizeof(struct in6_addr));
 358
 359                 sin->sin6_port = sk->sport;
 360         }
 361         *uaddr_len = sizeof(*sin);
 362         return(0);
 363 }
 364
 365 static int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
 366 {
 367         struct sock *sk = sock->sk;
 368         int err = -EINVAL;
 369         int pid;
 370
 371         switch(cmd)
 372         {
 373         case FIOSETOWN:
 374         case SIOCSPGRP:
 375                 err = get_user(pid, (int *) arg);
 376                 if(err)
 377                         return err;
 378
 379                 /* see sock_no_fcntl */
 380                 if (current->pid != pid && current->pgrp != -pid &&
 381                     !capable(CAP_NET_ADMIN))
 382                         return -EPERM;
 383                 sk->proc = pid;
 384                 return(0);
 385         case FIOGETOWN:
 386         case SIOCGPGRP:
 387                 err = put_user(sk->proc,(int *)arg);
 388                 if(err)
 389                         return err;
 390                 return(0);
 391         case SIOCGSTAMP:
 392                 if(sk->stamp.tv_sec==0)
 393                         return -ENOENT;
 394                 err = copy_to_user((void *)arg, &sk->stamp,
 395                                    sizeof(struct timeval));
 396                 if (err)
 397                         return -EFAULT;
 398                 return 0;
 399
 400         case SIOCADDRT:
 401         case SIOCDELRT:
 402
 403                 return(ipv6_route_ioctl(cmd,(void *)arg));
 404
 405         case SIOCSIFADDR:
 406                 return addrconf_add_ifaddr((void *) arg);
 407         case SIOCDIFADDR:
 408                 return addrconf_del_ifaddr((void *) arg);
 409         case SIOCSIFDSTADDR:
 410                 return addrconf_set_dstaddr((void *) arg);
 411         default:
 412                 if ((cmd >= SIOCDEVPRIVATE) &&
 413                     (cmd <= (SIOCDEVPRIVATE + 15)))
 414                         return(dev_ioctl(cmd,(void *) arg));
 415
 416                 if(sk->prot->ioctl==0 || (err=sk->prot->ioctl(sk, cmd, arg))==-ENOIOCTLCMD)
 417                         return(dev_ioctl(cmd,(void *) arg));
 418                 return err;
 419         }
 420         /*NOTREACHED*/
 421         return(0);
 422 }
 423
 424 struct proto_ops inet6_stream_ops = {
 425         PF_INET6,
 426
 427         inet6_release,
 428         inet6_bind,
 429         inet_stream_connect,            /* ok           */
 430         sock_no_socketpair,             /* a do nothing */
 431         inet_accept,                    /* ok           */
 432         inet6_getname,
 433         tcp_poll,                       /* ok           */
 434         inet6_ioctl,                    /* must change  */
 435         inet_listen,                    /* ok           */
 436         inet_shutdown,                  /* ok           */
 437         inet_setsockopt,                /* ok           */
 438         inet_getsockopt,                /* ok           */
 439         sock_no_fcntl,                  /* ok           */
 440         inet_sendmsg,                   /* ok           */
 441         inet_recvmsg,                   /* ok           */
 442         sock_no_mmap
 443 };
 444
 445 struct proto_ops inet6_dgram_ops = {
 446         PF_INET6,
 447
 448         inet6_release,
 449         inet6_bind,
 450         inet_dgram_connect,             /* ok           */
 451         sock_no_socketpair,             /* a do nothing */
 452         inet_accept,                    /* ok           */
 453         inet6_getname,
 454         datagram_poll,                  /* ok           */
 455         inet6_ioctl,                    /* must change  */
 456         sock_no_listen,                 /* ok           */
 457         inet_shutdown,                  /* ok           */
 458         inet_setsockopt,                /* ok           */
 459         inet_getsockopt,                /* ok           */
 460         sock_no_fcntl,                  /* ok           */
 461         inet_sendmsg,                   /* ok           */
 462         inet_recvmsg,                   /* ok           */
 463         sock_no_mmap,
 464 };
 465
 466 struct net_proto_family inet6_family_ops = {
 467         PF_INET6,
 468         inet6_create
 469 };
 470
 471 #ifdef MODULE
 472 int ipv6_unload(void)
 473 {
 474         if (!unloadable) return 1;
 475         /* We keep internally 3 raw sockets */
 476         return atomic_read(&(__this_module.uc.usecount)) - 3;
 477 }
 478 #endif
 479
 480 #if defined(MODULE) && defined(CONFIG_SYSCTL)
 481 extern void ipv6_sysctl_register(void);
 482 extern void ipv6_sysctl_unregister(void);
 483 #endif
 484
 485 #ifdef MODULE
 486 int init_module(void)
 487 #else
 488 void __init inet6_proto_init(struct net_proto *pro)
 489 #endif
 490 {
 491         struct sk_buff *dummy_skb;
 492         int err;
 493
 494 #ifdef MODULE
 495         if (!mod_member_present(&__this_module, can_unload))
 496           return -EINVAL;
 497
 498         __this_module.can_unload = &ipv6_unload;
 499 #endif
 500
 501         printk(KERN_INFO "IPv6 v0.8 for NET4.0\n");
 502
 503         if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb))
 504         {
 505                 printk(KERN_CRIT "inet6_proto_init: size fault\n");
 506 #ifdef MODULE
 507                 return -EINVAL;
 508 #else
 509                 return;
 510 #endif
 511         }
 512
 513         /*
 514          *      ipngwg API draft makes clear that the correct semantics
 515          *      for TCP and UDP is to consider one TCP and UDP instance
 516          *      in a host availiable by both INET and INET6 APIs and
 517          *      able to communicate via both network protocols.
 518          */
 519
 520 #if defined(MODULE) && defined(CONFIG_SYSCTL)
 521         ipv6_sysctl_register();
 522 #endif
 523         err = icmpv6_init(&inet6_family_ops);
 524         if (err)
 525                 goto icmp_fail;
 526         err = ndisc_init(&inet6_family_ops);
 527         if (err)
 528                 goto ndisc_fail;
 529         err = igmp6_init(&inet6_family_ops);
 530         if (err)
 531                 goto igmp_fail;
 532         ipv6_netdev_notif_init();
 533         ipv6_packet_init();
 534         ip6_route_init();
 535         ip6_flowlabel_init();
 536         addrconf_init();
 537         sit_init();
 538
 539         /* Init v6 transport protocols. */
 540         udpv6_init();
 541         tcpv6_init();
 542
 543         /* Create /proc/foo6 entries. */
 544 #ifdef CONFIG_PROC_FS
 545         proc_net_create("raw6", 0, raw6_get_info);
 546         proc_net_create("tcp6", 0, tcp6_get_info);
 547         proc_net_create("udp6", 0, udp6_get_info);
 548         proc_net_create("sockstat6", 0, afinet6_get_info);
 549         proc_net_create("snmp6", 0, afinet6_get_snmp);
 550 #endif
 551
 552         /* Now the userspace is allowed to create INET6 sockets. */
 553         (void) sock_register(&inet6_family_ops);
 554
 555 #ifdef MODULE
 556         return 0;
 557 #else
 558         return;
 559 #endif
 560
 561 igmp_fail:
 562         ndisc_cleanup();
 563 ndisc_fail:
 564         icmpv6_cleanup();
 565 icmp_fail:
 566 #if defined(MODULE) && defined(CONFIG_SYSCTL)
 567         ipv6_sysctl_unregister();
 568 #endif
 569 #ifdef MODULE
 570         return err;
 571 #else
 572         return;
 573 #endif
 574 }
 575
 576 #ifdef MODULE
 577 void cleanup_module(void)
 578 {
 579         /* First of all disallow new sockets creation. */
 580         sock_unregister(PF_INET6);
 581 #ifdef CONFIG_PROC_FS
 582         proc_net_remove("raw6");
 583         proc_net_remove("tcp6");
 584         proc_net_remove("udp6");
 585         proc_net_remove("sockstat6");
 586         proc_net_remove("snmp6");
 587 #endif
 588         /* Cleanup code parts. */
 589         sit_cleanup();
 590         ipv6_netdev_notif_cleanup();
 591         ip6_flowlabel_cleanup();
 592         addrconf_cleanup();
 593         ip6_route_cleanup();
 594         ipv6_packet_cleanup();
 595         igmp6_cleanup();
 596         ndisc_cleanup();
 597         icmpv6_cleanup();
 598 #ifdef CONFIG_SYSCTL
 599         ipv6_sysctl_unregister();
 600 #endif
 601 }
 602 #endif  /* MODULE */