Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 1.66 2002/02/01 22:01:04 davem Exp $ |
| 11 | * |
| 12 | * Fixes: |
| 13 | * piggy, Karl Knutson : Socket protocol table |
| 14 | * Hideaki YOSHIFUJI : sin6_scope_id support |
| 15 | * Arnaldo Melo : check proc_net_create return, cleanups |
| 16 | * |
| 17 | * This program is free software; you can redistribute it and/or |
| 18 | * modify it under the terms of the GNU General Public License |
| 19 | * as published by the Free Software Foundation; either version |
| 20 | * 2 of the License, or (at your option) any later version. |
| 21 | */ |
| 22 | |
| 23 | |
| 24 | #include <linux/module.h> |
| 25 | #include <linux/config.h> |
| 26 | #include <linux/errno.h> |
| 27 | #include <linux/types.h> |
| 28 | #include <linux/socket.h> |
| 29 | #include <linux/in.h> |
| 30 | #include <linux/kernel.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | #include <linux/sched.h> |
| 32 | #include <linux/timer.h> |
| 33 | #include <linux/string.h> |
| 34 | #include <linux/sockios.h> |
| 35 | #include <linux/net.h> |
| 36 | #include <linux/fcntl.h> |
| 37 | #include <linux/mm.h> |
| 38 | #include <linux/interrupt.h> |
| 39 | #include <linux/proc_fs.h> |
| 40 | #include <linux/stat.h> |
| 41 | #include <linux/init.h> |
| 42 | |
| 43 | #include <linux/inet.h> |
| 44 | #include <linux/netdevice.h> |
| 45 | #include <linux/icmpv6.h> |
| 46 | #include <linux/smp_lock.h> |
Harald Welte | 2cc7d57 | 2005-08-09 19:42:34 -0700 | [diff] [blame] | 47 | #include <linux/netfilter_ipv6.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 48 | |
| 49 | #include <net/ip.h> |
| 50 | #include <net/ipv6.h> |
| 51 | #include <net/udp.h> |
| 52 | #include <net/tcp.h> |
| 53 | #include <net/ipip.h> |
| 54 | #include <net/protocol.h> |
| 55 | #include <net/inet_common.h> |
| 56 | #include <net/transp_v6.h> |
| 57 | #include <net/ip6_route.h> |
| 58 | #include <net/addrconf.h> |
| 59 | #ifdef CONFIG_IPV6_TUNNEL |
| 60 | #include <net/ip6_tunnel.h> |
| 61 | #endif |
| 62 | |
| 63 | #include <asm/uaccess.h> |
| 64 | #include <asm/system.h> |
| 65 | |
| 66 | MODULE_AUTHOR("Cast of dozens"); |
| 67 | MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); |
| 68 | MODULE_LICENSE("GPL"); |
| 69 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 70 | int sysctl_ipv6_bindv6only; |
| 71 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 72 | /* The inetsw table contains everything that inet_create needs to |
| 73 | * build a new socket. |
| 74 | */ |
| 75 | static struct list_head inetsw6[SOCK_MAX]; |
| 76 | static DEFINE_SPINLOCK(inetsw6_lock); |
| 77 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 78 | static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) |
| 79 | { |
| 80 | const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); |
| 81 | |
| 82 | return (struct ipv6_pinfo *)(((u8 *)sk) + offset); |
| 83 | } |
| 84 | |
| 85 | static int inet6_create(struct socket *sock, int protocol) |
| 86 | { |
| 87 | struct inet_sock *inet; |
| 88 | struct ipv6_pinfo *np; |
| 89 | struct sock *sk; |
| 90 | struct list_head *p; |
| 91 | struct inet_protosw *answer; |
| 92 | struct proto *answer_prot; |
| 93 | unsigned char answer_flags; |
| 94 | char answer_no_check; |
| 95 | int rc; |
| 96 | |
| 97 | /* Look for the requested type/protocol pair. */ |
| 98 | answer = NULL; |
| 99 | rcu_read_lock(); |
| 100 | list_for_each_rcu(p, &inetsw6[sock->type]) { |
| 101 | answer = list_entry(p, struct inet_protosw, list); |
| 102 | |
| 103 | /* Check the non-wild match. */ |
| 104 | if (protocol == answer->protocol) { |
| 105 | if (protocol != IPPROTO_IP) |
| 106 | break; |
| 107 | } else { |
| 108 | /* Check for the two wild cases. */ |
| 109 | if (IPPROTO_IP == protocol) { |
| 110 | protocol = answer->protocol; |
| 111 | break; |
| 112 | } |
| 113 | if (IPPROTO_IP == answer->protocol) |
| 114 | break; |
| 115 | } |
| 116 | answer = NULL; |
| 117 | } |
| 118 | |
| 119 | rc = -ESOCKTNOSUPPORT; |
| 120 | if (!answer) |
| 121 | goto out_rcu_unlock; |
| 122 | rc = -EPERM; |
| 123 | if (answer->capability > 0 && !capable(answer->capability)) |
| 124 | goto out_rcu_unlock; |
| 125 | rc = -EPROTONOSUPPORT; |
| 126 | if (!protocol) |
| 127 | goto out_rcu_unlock; |
| 128 | |
| 129 | sock->ops = answer->ops; |
| 130 | |
| 131 | answer_prot = answer->prot; |
| 132 | answer_no_check = answer->no_check; |
| 133 | answer_flags = answer->flags; |
| 134 | rcu_read_unlock(); |
| 135 | |
| 136 | BUG_TRAP(answer_prot->slab != NULL); |
| 137 | |
| 138 | rc = -ENOBUFS; |
| 139 | sk = sk_alloc(PF_INET6, GFP_KERNEL, answer_prot, 1); |
| 140 | if (sk == NULL) |
| 141 | goto out; |
| 142 | |
| 143 | sock_init_data(sock, sk); |
| 144 | |
| 145 | rc = 0; |
| 146 | sk->sk_no_check = answer_no_check; |
| 147 | if (INET_PROTOSW_REUSE & answer_flags) |
| 148 | sk->sk_reuse = 1; |
| 149 | |
| 150 | inet = inet_sk(sk); |
| 151 | |
| 152 | if (SOCK_RAW == sock->type) { |
| 153 | inet->num = protocol; |
| 154 | if (IPPROTO_RAW == protocol) |
| 155 | inet->hdrincl = 1; |
| 156 | } |
| 157 | |
Arnaldo Carvalho de Melo | e684897 | 2005-08-09 19:45:38 -0700 | [diff] [blame] | 158 | sk->sk_destruct = inet_sock_destruct; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 159 | sk->sk_family = PF_INET6; |
| 160 | sk->sk_protocol = protocol; |
| 161 | |
| 162 | sk->sk_backlog_rcv = answer->prot->backlog_rcv; |
| 163 | |
| 164 | inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); |
| 165 | np->hop_limit = -1; |
| 166 | np->mcast_hops = -1; |
| 167 | np->mc_loop = 1; |
| 168 | np->pmtudisc = IPV6_PMTUDISC_WANT; |
| 169 | np->ipv6only = sysctl_ipv6_bindv6only; |
| 170 | |
| 171 | /* Init the ipv4 part of the socket since we can have sockets |
| 172 | * using v6 API for ipv4. |
| 173 | */ |
| 174 | inet->uc_ttl = -1; |
| 175 | |
| 176 | inet->mc_loop = 1; |
| 177 | inet->mc_ttl = 1; |
| 178 | inet->mc_index = 0; |
| 179 | inet->mc_list = NULL; |
| 180 | |
| 181 | if (ipv4_config.no_pmtu_disc) |
| 182 | inet->pmtudisc = IP_PMTUDISC_DONT; |
| 183 | else |
| 184 | inet->pmtudisc = IP_PMTUDISC_WANT; |
Arnaldo Carvalho de Melo | e684897 | 2005-08-09 19:45:38 -0700 | [diff] [blame] | 185 | /* |
| 186 | * Increment only the relevant sk_prot->socks debug field, this changes |
| 187 | * the previous behaviour of incrementing both the equivalent to |
| 188 | * answer->prot->socks (inet6_sock_nr) and inet_sock_nr. |
| 189 | * |
| 190 | * This allows better debug granularity as we'll know exactly how many |
| 191 | * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6 |
| 192 | * transport protocol socks. -acme |
| 193 | */ |
| 194 | sk_refcnt_debug_inc(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 195 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 196 | if (inet->num) { |
| 197 | /* It assumes that any protocol which allows |
| 198 | * the user to assign a number at socket |
| 199 | * creation time automatically shares. |
| 200 | */ |
| 201 | inet->sport = ntohs(inet->num); |
| 202 | sk->sk_prot->hash(sk); |
| 203 | } |
| 204 | if (sk->sk_prot->init) { |
| 205 | rc = sk->sk_prot->init(sk); |
| 206 | if (rc) { |
| 207 | sk_common_release(sk); |
| 208 | goto out; |
| 209 | } |
| 210 | } |
| 211 | out: |
| 212 | return rc; |
| 213 | out_rcu_unlock: |
| 214 | rcu_read_unlock(); |
| 215 | goto out; |
| 216 | } |
| 217 | |
| 218 | |
| 219 | /* bind for INET6 API */ |
| 220 | int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) |
| 221 | { |
| 222 | struct sockaddr_in6 *addr=(struct sockaddr_in6 *)uaddr; |
| 223 | struct sock *sk = sock->sk; |
| 224 | struct inet_sock *inet = inet_sk(sk); |
| 225 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 226 | __u32 v4addr = 0; |
| 227 | unsigned short snum; |
| 228 | int addr_type = 0; |
| 229 | int err = 0; |
| 230 | |
| 231 | /* If the socket has its own bind function then use it. */ |
| 232 | if (sk->sk_prot->bind) |
| 233 | return sk->sk_prot->bind(sk, uaddr, addr_len); |
| 234 | |
| 235 | if (addr_len < SIN6_LEN_RFC2133) |
| 236 | return -EINVAL; |
| 237 | addr_type = ipv6_addr_type(&addr->sin6_addr); |
| 238 | if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM) |
| 239 | return -EINVAL; |
| 240 | |
| 241 | snum = ntohs(addr->sin6_port); |
| 242 | if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE)) |
| 243 | return -EACCES; |
| 244 | |
| 245 | lock_sock(sk); |
| 246 | |
| 247 | /* Check these errors (active socket, double bind). */ |
| 248 | if (sk->sk_state != TCP_CLOSE || inet->num) { |
| 249 | err = -EINVAL; |
| 250 | goto out; |
| 251 | } |
| 252 | |
| 253 | /* Check if the address belongs to the host. */ |
| 254 | if (addr_type == IPV6_ADDR_MAPPED) { |
| 255 | v4addr = addr->sin6_addr.s6_addr32[3]; |
| 256 | if (inet_addr_type(v4addr) != RTN_LOCAL) { |
| 257 | err = -EADDRNOTAVAIL; |
| 258 | goto out; |
| 259 | } |
| 260 | } else { |
| 261 | if (addr_type != IPV6_ADDR_ANY) { |
| 262 | struct net_device *dev = NULL; |
| 263 | |
| 264 | if (addr_type & IPV6_ADDR_LINKLOCAL) { |
| 265 | if (addr_len >= sizeof(struct sockaddr_in6) && |
| 266 | addr->sin6_scope_id) { |
| 267 | /* Override any existing binding, if another one |
| 268 | * is supplied by user. |
| 269 | */ |
| 270 | sk->sk_bound_dev_if = addr->sin6_scope_id; |
| 271 | } |
| 272 | |
| 273 | /* Binding to link-local address requires an interface */ |
| 274 | if (!sk->sk_bound_dev_if) { |
| 275 | err = -EINVAL; |
| 276 | goto out; |
| 277 | } |
| 278 | dev = dev_get_by_index(sk->sk_bound_dev_if); |
| 279 | if (!dev) { |
| 280 | err = -ENODEV; |
| 281 | goto out; |
| 282 | } |
| 283 | } |
| 284 | |
| 285 | /* ipv4 addr of the socket is invalid. Only the |
| 286 | * unspecified and mapped address have a v4 equivalent. |
| 287 | */ |
| 288 | v4addr = LOOPBACK4_IPV6; |
| 289 | if (!(addr_type & IPV6_ADDR_MULTICAST)) { |
| 290 | if (!ipv6_chk_addr(&addr->sin6_addr, dev, 0)) { |
| 291 | if (dev) |
| 292 | dev_put(dev); |
| 293 | err = -EADDRNOTAVAIL; |
| 294 | goto out; |
| 295 | } |
| 296 | } |
| 297 | if (dev) |
| 298 | dev_put(dev); |
| 299 | } |
| 300 | } |
| 301 | |
| 302 | inet->rcv_saddr = v4addr; |
| 303 | inet->saddr = v4addr; |
| 304 | |
| 305 | ipv6_addr_copy(&np->rcv_saddr, &addr->sin6_addr); |
| 306 | |
| 307 | if (!(addr_type & IPV6_ADDR_MULTICAST)) |
| 308 | ipv6_addr_copy(&np->saddr, &addr->sin6_addr); |
| 309 | |
| 310 | /* Make sure we are allowed to bind here. */ |
| 311 | if (sk->sk_prot->get_port(sk, snum)) { |
| 312 | inet_reset_saddr(sk); |
| 313 | err = -EADDRINUSE; |
| 314 | goto out; |
| 315 | } |
| 316 | |
| 317 | if (addr_type != IPV6_ADDR_ANY) |
| 318 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; |
| 319 | if (snum) |
| 320 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; |
| 321 | inet->sport = ntohs(inet->num); |
| 322 | inet->dport = 0; |
| 323 | inet->daddr = 0; |
| 324 | out: |
| 325 | release_sock(sk); |
| 326 | return err; |
| 327 | } |
| 328 | |
| 329 | int inet6_release(struct socket *sock) |
| 330 | { |
| 331 | struct sock *sk = sock->sk; |
| 332 | |
| 333 | if (sk == NULL) |
| 334 | return -EINVAL; |
| 335 | |
| 336 | /* Free mc lists */ |
| 337 | ipv6_sock_mc_close(sk); |
| 338 | |
| 339 | /* Free ac lists */ |
| 340 | ipv6_sock_ac_close(sk); |
| 341 | |
| 342 | return inet_release(sock); |
| 343 | } |
| 344 | |
| 345 | int inet6_destroy_sock(struct sock *sk) |
| 346 | { |
| 347 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 348 | struct sk_buff *skb; |
| 349 | struct ipv6_txoptions *opt; |
| 350 | |
| 351 | /* |
| 352 | * Release destination entry |
| 353 | */ |
| 354 | |
| 355 | sk_dst_reset(sk); |
| 356 | |
| 357 | /* Release rx options */ |
| 358 | |
| 359 | if ((skb = xchg(&np->pktoptions, NULL)) != NULL) |
| 360 | kfree_skb(skb); |
| 361 | |
| 362 | /* Free flowlabels */ |
| 363 | fl6_free_socklist(sk); |
| 364 | |
| 365 | /* Free tx options */ |
| 366 | |
| 367 | if ((opt = xchg(&np->opt, NULL)) != NULL) |
| 368 | sock_kfree_s(sk, opt, opt->tot_len); |
| 369 | |
| 370 | return 0; |
| 371 | } |
| 372 | |
| 373 | /* |
| 374 | * This does both peername and sockname. |
| 375 | */ |
| 376 | |
| 377 | int inet6_getname(struct socket *sock, struct sockaddr *uaddr, |
| 378 | int *uaddr_len, int peer) |
| 379 | { |
| 380 | struct sockaddr_in6 *sin=(struct sockaddr_in6 *)uaddr; |
| 381 | struct sock *sk = sock->sk; |
| 382 | struct inet_sock *inet = inet_sk(sk); |
| 383 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 384 | |
| 385 | sin->sin6_family = AF_INET6; |
| 386 | sin->sin6_flowinfo = 0; |
| 387 | sin->sin6_scope_id = 0; |
| 388 | if (peer) { |
| 389 | if (!inet->dport) |
| 390 | return -ENOTCONN; |
| 391 | if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && |
| 392 | peer == 1) |
| 393 | return -ENOTCONN; |
| 394 | sin->sin6_port = inet->dport; |
| 395 | ipv6_addr_copy(&sin->sin6_addr, &np->daddr); |
| 396 | if (np->sndflow) |
| 397 | sin->sin6_flowinfo = np->flow_label; |
| 398 | } else { |
| 399 | if (ipv6_addr_any(&np->rcv_saddr)) |
| 400 | ipv6_addr_copy(&sin->sin6_addr, &np->saddr); |
| 401 | else |
| 402 | ipv6_addr_copy(&sin->sin6_addr, &np->rcv_saddr); |
| 403 | |
| 404 | sin->sin6_port = inet->sport; |
| 405 | } |
| 406 | if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL) |
| 407 | sin->sin6_scope_id = sk->sk_bound_dev_if; |
| 408 | *uaddr_len = sizeof(*sin); |
| 409 | return(0); |
| 410 | } |
| 411 | |
| 412 | int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
| 413 | { |
| 414 | struct sock *sk = sock->sk; |
| 415 | int err = -EINVAL; |
| 416 | |
| 417 | switch(cmd) |
| 418 | { |
| 419 | case SIOCGSTAMP: |
| 420 | return sock_get_timestamp(sk, (struct timeval __user *)arg); |
| 421 | |
| 422 | case SIOCADDRT: |
| 423 | case SIOCDELRT: |
| 424 | |
| 425 | return(ipv6_route_ioctl(cmd,(void __user *)arg)); |
| 426 | |
| 427 | case SIOCSIFADDR: |
| 428 | return addrconf_add_ifaddr((void __user *) arg); |
| 429 | case SIOCDIFADDR: |
| 430 | return addrconf_del_ifaddr((void __user *) arg); |
| 431 | case SIOCSIFDSTADDR: |
| 432 | return addrconf_set_dstaddr((void __user *) arg); |
| 433 | default: |
| 434 | if (!sk->sk_prot->ioctl || |
| 435 | (err = sk->sk_prot->ioctl(sk, cmd, arg)) == -ENOIOCTLCMD) |
| 436 | return(dev_ioctl(cmd,(void __user *) arg)); |
| 437 | return err; |
| 438 | } |
| 439 | /*NOTREACHED*/ |
| 440 | return(0); |
| 441 | } |
| 442 | |
| 443 | struct proto_ops inet6_stream_ops = { |
| 444 | .family = PF_INET6, |
| 445 | .owner = THIS_MODULE, |
| 446 | .release = inet6_release, |
| 447 | .bind = inet6_bind, |
| 448 | .connect = inet_stream_connect, /* ok */ |
| 449 | .socketpair = sock_no_socketpair, /* a do nothing */ |
| 450 | .accept = inet_accept, /* ok */ |
| 451 | .getname = inet6_getname, |
| 452 | .poll = tcp_poll, /* ok */ |
| 453 | .ioctl = inet6_ioctl, /* must change */ |
| 454 | .listen = inet_listen, /* ok */ |
| 455 | .shutdown = inet_shutdown, /* ok */ |
| 456 | .setsockopt = sock_common_setsockopt, /* ok */ |
| 457 | .getsockopt = sock_common_getsockopt, /* ok */ |
| 458 | .sendmsg = inet_sendmsg, /* ok */ |
| 459 | .recvmsg = sock_common_recvmsg, /* ok */ |
| 460 | .mmap = sock_no_mmap, |
| 461 | .sendpage = tcp_sendpage |
| 462 | }; |
| 463 | |
| 464 | struct proto_ops inet6_dgram_ops = { |
| 465 | .family = PF_INET6, |
| 466 | .owner = THIS_MODULE, |
| 467 | .release = inet6_release, |
| 468 | .bind = inet6_bind, |
| 469 | .connect = inet_dgram_connect, /* ok */ |
| 470 | .socketpair = sock_no_socketpair, /* a do nothing */ |
| 471 | .accept = sock_no_accept, /* a do nothing */ |
| 472 | .getname = inet6_getname, |
| 473 | .poll = udp_poll, /* ok */ |
| 474 | .ioctl = inet6_ioctl, /* must change */ |
| 475 | .listen = sock_no_listen, /* ok */ |
| 476 | .shutdown = inet_shutdown, /* ok */ |
| 477 | .setsockopt = sock_common_setsockopt, /* ok */ |
| 478 | .getsockopt = sock_common_getsockopt, /* ok */ |
| 479 | .sendmsg = inet_sendmsg, /* ok */ |
| 480 | .recvmsg = sock_common_recvmsg, /* ok */ |
| 481 | .mmap = sock_no_mmap, |
| 482 | .sendpage = sock_no_sendpage, |
| 483 | }; |
| 484 | |
| 485 | static struct net_proto_family inet6_family_ops = { |
| 486 | .family = PF_INET6, |
| 487 | .create = inet6_create, |
| 488 | .owner = THIS_MODULE, |
| 489 | }; |
| 490 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 491 | /* Same as inet6_dgram_ops, sans udp_poll. */ |
| 492 | static struct proto_ops inet6_sockraw_ops = { |
| 493 | .family = PF_INET6, |
| 494 | .owner = THIS_MODULE, |
| 495 | .release = inet6_release, |
| 496 | .bind = inet6_bind, |
| 497 | .connect = inet_dgram_connect, /* ok */ |
| 498 | .socketpair = sock_no_socketpair, /* a do nothing */ |
| 499 | .accept = sock_no_accept, /* a do nothing */ |
| 500 | .getname = inet6_getname, |
| 501 | .poll = datagram_poll, /* ok */ |
| 502 | .ioctl = inet6_ioctl, /* must change */ |
| 503 | .listen = sock_no_listen, /* ok */ |
| 504 | .shutdown = inet_shutdown, /* ok */ |
| 505 | .setsockopt = sock_common_setsockopt, /* ok */ |
| 506 | .getsockopt = sock_common_getsockopt, /* ok */ |
| 507 | .sendmsg = inet_sendmsg, /* ok */ |
| 508 | .recvmsg = sock_common_recvmsg, /* ok */ |
| 509 | .mmap = sock_no_mmap, |
| 510 | .sendpage = sock_no_sendpage, |
| 511 | }; |
| 512 | |
| 513 | static struct inet_protosw rawv6_protosw = { |
| 514 | .type = SOCK_RAW, |
| 515 | .protocol = IPPROTO_IP, /* wild card */ |
| 516 | .prot = &rawv6_prot, |
| 517 | .ops = &inet6_sockraw_ops, |
| 518 | .capability = CAP_NET_RAW, |
| 519 | .no_check = UDP_CSUM_DEFAULT, |
| 520 | .flags = INET_PROTOSW_REUSE, |
| 521 | }; |
| 522 | |
| 523 | void |
| 524 | inet6_register_protosw(struct inet_protosw *p) |
| 525 | { |
| 526 | struct list_head *lh; |
| 527 | struct inet_protosw *answer; |
| 528 | int protocol = p->protocol; |
| 529 | struct list_head *last_perm; |
| 530 | |
| 531 | spin_lock_bh(&inetsw6_lock); |
| 532 | |
| 533 | if (p->type >= SOCK_MAX) |
| 534 | goto out_illegal; |
| 535 | |
| 536 | /* If we are trying to override a permanent protocol, bail. */ |
| 537 | answer = NULL; |
| 538 | last_perm = &inetsw6[p->type]; |
| 539 | list_for_each(lh, &inetsw6[p->type]) { |
| 540 | answer = list_entry(lh, struct inet_protosw, list); |
| 541 | |
| 542 | /* Check only the non-wild match. */ |
| 543 | if (INET_PROTOSW_PERMANENT & answer->flags) { |
| 544 | if (protocol == answer->protocol) |
| 545 | break; |
| 546 | last_perm = lh; |
| 547 | } |
| 548 | |
| 549 | answer = NULL; |
| 550 | } |
| 551 | if (answer) |
| 552 | goto out_permanent; |
| 553 | |
| 554 | /* Add the new entry after the last permanent entry if any, so that |
| 555 | * the new entry does not override a permanent entry when matched with |
| 556 | * a wild-card protocol. But it is allowed to override any existing |
| 557 | * non-permanent entry. This means that when we remove this entry, the |
| 558 | * system automatically returns to the old behavior. |
| 559 | */ |
| 560 | list_add_rcu(&p->list, last_perm); |
| 561 | out: |
| 562 | spin_unlock_bh(&inetsw6_lock); |
| 563 | return; |
| 564 | |
| 565 | out_permanent: |
| 566 | printk(KERN_ERR "Attempt to override permanent protocol %d.\n", |
| 567 | protocol); |
| 568 | goto out; |
| 569 | |
| 570 | out_illegal: |
| 571 | printk(KERN_ERR |
| 572 | "Ignoring attempt to register invalid socket type %d.\n", |
| 573 | p->type); |
| 574 | goto out; |
| 575 | } |
| 576 | |
| 577 | void |
| 578 | inet6_unregister_protosw(struct inet_protosw *p) |
| 579 | { |
| 580 | if (INET_PROTOSW_PERMANENT & p->flags) { |
| 581 | printk(KERN_ERR |
| 582 | "Attempt to unregister permanent protocol %d.\n", |
| 583 | p->protocol); |
| 584 | } else { |
| 585 | spin_lock_bh(&inetsw6_lock); |
| 586 | list_del_rcu(&p->list); |
| 587 | spin_unlock_bh(&inetsw6_lock); |
| 588 | |
| 589 | synchronize_net(); |
| 590 | } |
| 591 | } |
| 592 | |
| 593 | int |
| 594 | snmp6_mib_init(void *ptr[2], size_t mibsize, size_t mibalign) |
| 595 | { |
| 596 | if (ptr == NULL) |
| 597 | return -EINVAL; |
| 598 | |
| 599 | ptr[0] = __alloc_percpu(mibsize, mibalign); |
| 600 | if (!ptr[0]) |
| 601 | goto err0; |
| 602 | |
| 603 | ptr[1] = __alloc_percpu(mibsize, mibalign); |
| 604 | if (!ptr[1]) |
| 605 | goto err1; |
| 606 | |
| 607 | return 0; |
| 608 | |
| 609 | err1: |
| 610 | free_percpu(ptr[0]); |
| 611 | ptr[0] = NULL; |
| 612 | err0: |
| 613 | return -ENOMEM; |
| 614 | } |
| 615 | |
| 616 | void |
| 617 | snmp6_mib_free(void *ptr[2]) |
| 618 | { |
| 619 | if (ptr == NULL) |
| 620 | return; |
| 621 | if (ptr[0]) |
| 622 | free_percpu(ptr[0]); |
| 623 | if (ptr[1]) |
| 624 | free_percpu(ptr[1]); |
| 625 | ptr[0] = ptr[1] = NULL; |
| 626 | } |
| 627 | |
| 628 | static int __init init_ipv6_mibs(void) |
| 629 | { |
| 630 | if (snmp6_mib_init((void **)ipv6_statistics, sizeof (struct ipstats_mib), |
| 631 | __alignof__(struct ipstats_mib)) < 0) |
| 632 | goto err_ip_mib; |
| 633 | if (snmp6_mib_init((void **)icmpv6_statistics, sizeof (struct icmpv6_mib), |
| 634 | __alignof__(struct icmpv6_mib)) < 0) |
| 635 | goto err_icmp_mib; |
| 636 | if (snmp6_mib_init((void **)udp_stats_in6, sizeof (struct udp_mib), |
| 637 | __alignof__(struct udp_mib)) < 0) |
| 638 | goto err_udp_mib; |
| 639 | return 0; |
| 640 | |
| 641 | err_udp_mib: |
| 642 | snmp6_mib_free((void **)icmpv6_statistics); |
| 643 | err_icmp_mib: |
| 644 | snmp6_mib_free((void **)ipv6_statistics); |
| 645 | err_ip_mib: |
| 646 | return -ENOMEM; |
| 647 | |
| 648 | } |
| 649 | |
| 650 | static void cleanup_ipv6_mibs(void) |
| 651 | { |
| 652 | snmp6_mib_free((void **)ipv6_statistics); |
| 653 | snmp6_mib_free((void **)icmpv6_statistics); |
| 654 | snmp6_mib_free((void **)udp_stats_in6); |
| 655 | } |
| 656 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 657 | static int __init inet6_init(void) |
| 658 | { |
| 659 | struct sk_buff *dummy_skb; |
| 660 | struct list_head *r; |
| 661 | int err; |
| 662 | |
| 663 | #ifdef MODULE |
| 664 | #if 0 /* FIXME --RR */ |
| 665 | if (!mod_member_present(&__this_module, can_unload)) |
| 666 | return -EINVAL; |
| 667 | |
| 668 | __this_module.can_unload = &ipv6_unload; |
| 669 | #endif |
| 670 | #endif |
| 671 | |
| 672 | if (sizeof(struct inet6_skb_parm) > sizeof(dummy_skb->cb)) { |
| 673 | printk(KERN_CRIT "inet6_proto_init: size fault\n"); |
| 674 | return -EINVAL; |
| 675 | } |
| 676 | |
| 677 | err = proto_register(&tcpv6_prot, 1); |
| 678 | if (err) |
| 679 | goto out; |
| 680 | |
| 681 | err = proto_register(&udpv6_prot, 1); |
| 682 | if (err) |
| 683 | goto out_unregister_tcp_proto; |
| 684 | |
| 685 | err = proto_register(&rawv6_prot, 1); |
| 686 | if (err) |
| 687 | goto out_unregister_udp_proto; |
| 688 | |
| 689 | |
| 690 | /* Register the socket-side information for inet6_create. */ |
| 691 | for(r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) |
| 692 | INIT_LIST_HEAD(r); |
| 693 | |
| 694 | /* We MUST register RAW sockets before we create the ICMP6, |
| 695 | * IGMP6, or NDISC control sockets. |
| 696 | */ |
| 697 | inet6_register_protosw(&rawv6_protosw); |
| 698 | |
| 699 | /* Register the family here so that the init calls below will |
| 700 | * be able to create sockets. (?? is this dangerous ??) |
| 701 | */ |
| 702 | (void) sock_register(&inet6_family_ops); |
| 703 | |
| 704 | /* Initialise ipv6 mibs */ |
| 705 | err = init_ipv6_mibs(); |
| 706 | if (err) |
| 707 | goto out_unregister_raw_proto; |
| 708 | |
| 709 | /* |
| 710 | * ipngwg API draft makes clear that the correct semantics |
| 711 | * for TCP and UDP is to consider one TCP and UDP instance |
| 712 | * in a host availiable by both INET and INET6 APIs and |
| 713 | * able to communicate via both network protocols. |
| 714 | */ |
| 715 | |
| 716 | #ifdef CONFIG_SYSCTL |
| 717 | ipv6_sysctl_register(); |
| 718 | #endif |
| 719 | err = icmpv6_init(&inet6_family_ops); |
| 720 | if (err) |
| 721 | goto icmp_fail; |
| 722 | err = ndisc_init(&inet6_family_ops); |
| 723 | if (err) |
| 724 | goto ndisc_fail; |
| 725 | err = igmp6_init(&inet6_family_ops); |
| 726 | if (err) |
| 727 | goto igmp_fail; |
Harald Welte | 2cc7d57 | 2005-08-09 19:42:34 -0700 | [diff] [blame] | 728 | err = ipv6_netfilter_init(); |
| 729 | if (err) |
| 730 | goto netfilter_fail; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 731 | /* Create /proc/foo6 entries. */ |
| 732 | #ifdef CONFIG_PROC_FS |
| 733 | err = -ENOMEM; |
| 734 | if (raw6_proc_init()) |
| 735 | goto proc_raw6_fail; |
| 736 | if (tcp6_proc_init()) |
| 737 | goto proc_tcp6_fail; |
| 738 | if (udp6_proc_init()) |
| 739 | goto proc_udp6_fail; |
| 740 | if (ipv6_misc_proc_init()) |
| 741 | goto proc_misc6_fail; |
| 742 | |
| 743 | if (ac6_proc_init()) |
| 744 | goto proc_anycast6_fail; |
| 745 | if (if6_proc_init()) |
| 746 | goto proc_if6_fail; |
| 747 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 748 | ip6_route_init(); |
| 749 | ip6_flowlabel_init(); |
| 750 | err = addrconf_init(); |
| 751 | if (err) |
| 752 | goto addrconf_fail; |
| 753 | sit_init(); |
| 754 | |
| 755 | /* Init v6 extension headers. */ |
| 756 | ipv6_rthdr_init(); |
| 757 | ipv6_frag_init(); |
| 758 | ipv6_nodata_init(); |
| 759 | ipv6_destopt_init(); |
| 760 | |
| 761 | /* Init v6 transport protocols. */ |
| 762 | udpv6_init(); |
| 763 | tcpv6_init(); |
Herbert Xu | e2ed405 | 2005-07-05 14:41:20 -0700 | [diff] [blame] | 764 | |
| 765 | ipv6_packet_init(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 766 | err = 0; |
| 767 | out: |
| 768 | return err; |
| 769 | |
| 770 | addrconf_fail: |
| 771 | ip6_flowlabel_cleanup(); |
| 772 | ip6_route_cleanup(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 773 | #ifdef CONFIG_PROC_FS |
| 774 | if6_proc_exit(); |
| 775 | proc_if6_fail: |
| 776 | ac6_proc_exit(); |
| 777 | proc_anycast6_fail: |
| 778 | ipv6_misc_proc_exit(); |
| 779 | proc_misc6_fail: |
| 780 | udp6_proc_exit(); |
| 781 | proc_udp6_fail: |
| 782 | tcp6_proc_exit(); |
| 783 | proc_tcp6_fail: |
| 784 | raw6_proc_exit(); |
| 785 | proc_raw6_fail: |
| 786 | #endif |
Harald Welte | 2cc7d57 | 2005-08-09 19:42:34 -0700 | [diff] [blame] | 787 | ipv6_netfilter_fini(); |
| 788 | netfilter_fail: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 789 | igmp6_cleanup(); |
| 790 | igmp_fail: |
| 791 | ndisc_cleanup(); |
| 792 | ndisc_fail: |
| 793 | icmpv6_cleanup(); |
| 794 | icmp_fail: |
| 795 | #ifdef CONFIG_SYSCTL |
| 796 | ipv6_sysctl_unregister(); |
| 797 | #endif |
| 798 | cleanup_ipv6_mibs(); |
| 799 | out_unregister_raw_proto: |
| 800 | proto_unregister(&rawv6_prot); |
| 801 | out_unregister_udp_proto: |
| 802 | proto_unregister(&udpv6_prot); |
| 803 | out_unregister_tcp_proto: |
| 804 | proto_unregister(&tcpv6_prot); |
| 805 | goto out; |
| 806 | } |
| 807 | module_init(inet6_init); |
| 808 | |
| 809 | static void __exit inet6_exit(void) |
| 810 | { |
| 811 | /* First of all disallow new sockets creation. */ |
| 812 | sock_unregister(PF_INET6); |
| 813 | #ifdef CONFIG_PROC_FS |
| 814 | if6_proc_exit(); |
| 815 | ac6_proc_exit(); |
| 816 | ipv6_misc_proc_exit(); |
| 817 | udp6_proc_exit(); |
| 818 | tcp6_proc_exit(); |
| 819 | raw6_proc_exit(); |
| 820 | #endif |
| 821 | /* Cleanup code parts. */ |
| 822 | sit_cleanup(); |
| 823 | ip6_flowlabel_cleanup(); |
| 824 | addrconf_cleanup(); |
| 825 | ip6_route_cleanup(); |
| 826 | ipv6_packet_cleanup(); |
| 827 | igmp6_cleanup(); |
Harald Welte | 2cc7d57 | 2005-08-09 19:42:34 -0700 | [diff] [blame] | 828 | ipv6_netfilter_fini(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 829 | ndisc_cleanup(); |
| 830 | icmpv6_cleanup(); |
| 831 | #ifdef CONFIG_SYSCTL |
| 832 | ipv6_sysctl_unregister(); |
| 833 | #endif |
| 834 | cleanup_ipv6_mibs(); |
| 835 | proto_unregister(&rawv6_prot); |
| 836 | proto_unregister(&udpv6_prot); |
| 837 | proto_unregister(&tcpv6_prot); |
| 838 | } |
| 839 | module_exit(inet6_exit); |
| 840 | |
| 841 | MODULE_ALIAS_NETPROTO(PF_INET6); |