| 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 | * Fixes: |
| 11 | * piggy, Karl Knutson : Socket protocol table |
| 12 | * Hideaki YOSHIFUJI : sin6_scope_id support |
| 13 | * Arnaldo Melo : check proc_net_create return, cleanups |
| 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 | #define pr_fmt(fmt) "IPv6: " fmt |
| 22 | |
| 23 | #include <linux/module.h> |
| 24 | #include <linux/capability.h> |
| 25 | #include <linux/errno.h> |
| 26 | #include <linux/types.h> |
| 27 | #include <linux/socket.h> |
| 28 | #include <linux/in.h> |
| 29 | #include <linux/kernel.h> |
| 30 | #include <linux/timer.h> |
| 31 | #include <linux/string.h> |
| 32 | #include <linux/sockios.h> |
| 33 | #include <linux/net.h> |
| 34 | #include <linux/fcntl.h> |
| 35 | #include <linux/mm.h> |
| 36 | #include <linux/interrupt.h> |
| 37 | #include <linux/proc_fs.h> |
| 38 | #include <linux/stat.h> |
| 39 | #include <linux/init.h> |
| 40 | #include <linux/slab.h> |
| 41 | |
| 42 | #include <linux/inet.h> |
| 43 | #include <linux/netdevice.h> |
| 44 | #include <linux/icmpv6.h> |
| 45 | #include <linux/netfilter_ipv6.h> |
| 46 | |
| 47 | #include <net/ip.h> |
| 48 | #include <net/ipv6.h> |
| 49 | #include <net/udp.h> |
| 50 | #include <net/udplite.h> |
| 51 | #include <net/tcp.h> |
| 52 | #include <net/ping.h> |
| 53 | #include <net/protocol.h> |
| 54 | #include <net/inet_common.h> |
| 55 | #include <net/route.h> |
| 56 | #include <net/transp_v6.h> |
| 57 | #include <net/ip6_route.h> |
| 58 | #include <net/addrconf.h> |
| 59 | #include <net/ndisc.h> |
| 60 | #ifdef CONFIG_IPV6_TUNNEL |
| 61 | #include <net/ip6_tunnel.h> |
| 62 | #endif |
| 63 | |
| 64 | #include <asm/uaccess.h> |
| 65 | #include <linux/mroute6.h> |
| 66 | |
| 67 | MODULE_AUTHOR("Cast of dozens"); |
| 68 | MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); |
| 69 | MODULE_LICENSE("GPL"); |
| 70 | |
| 71 | /* The inetsw6 table contains everything that inet6_create needs to |
| 72 | * build a new socket. |
| 73 | */ |
| 74 | static struct list_head inetsw6[SOCK_MAX]; |
| 75 | static DEFINE_SPINLOCK(inetsw6_lock); |
| 76 | |
| 77 | struct ipv6_params ipv6_defaults = { |
| 78 | .disable_ipv6 = 0, |
| 79 | .autoconf = 1, |
| 80 | }; |
| 81 | |
| 82 | static int disable_ipv6_mod; |
| 83 | |
| 84 | module_param_named(disable, disable_ipv6_mod, int, 0444); |
| 85 | MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional"); |
| 86 | |
| 87 | module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444); |
| 88 | MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces"); |
| 89 | |
| 90 | module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444); |
| 91 | MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces"); |
| 92 | |
| 93 | static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) |
| 94 | { |
| 95 | const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); |
| 96 | |
| 97 | return (struct ipv6_pinfo *)(((u8 *)sk) + offset); |
| 98 | } |
| 99 | |
| 100 | static int inet6_create(struct net *net, struct socket *sock, int protocol, |
| 101 | int kern) |
| 102 | { |
| 103 | struct inet_sock *inet; |
| 104 | struct ipv6_pinfo *np; |
| 105 | struct sock *sk; |
| 106 | struct inet_protosw *answer; |
| 107 | struct proto *answer_prot; |
| 108 | unsigned char answer_flags; |
| 109 | int try_loading_module = 0; |
| 110 | int err; |
| 111 | |
| 112 | if (protocol < 0 || protocol >= IPPROTO_MAX) |
| 113 | return -EINVAL; |
| 114 | |
| 115 | /* Look for the requested type/protocol pair. */ |
| 116 | lookup_protocol: |
| 117 | err = -ESOCKTNOSUPPORT; |
| 118 | rcu_read_lock(); |
| 119 | list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) { |
| 120 | |
| 121 | err = 0; |
| 122 | /* Check the non-wild match. */ |
| 123 | if (protocol == answer->protocol) { |
| 124 | if (protocol != IPPROTO_IP) |
| 125 | break; |
| 126 | } else { |
| 127 | /* Check for the two wild cases. */ |
| 128 | if (IPPROTO_IP == protocol) { |
| 129 | protocol = answer->protocol; |
| 130 | break; |
| 131 | } |
| 132 | if (IPPROTO_IP == answer->protocol) |
| 133 | break; |
| 134 | } |
| 135 | err = -EPROTONOSUPPORT; |
| 136 | } |
| 137 | |
| 138 | if (err) { |
| 139 | if (try_loading_module < 2) { |
| 140 | rcu_read_unlock(); |
| 141 | /* |
| 142 | * Be more specific, e.g. net-pf-10-proto-132-type-1 |
| 143 | * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) |
| 144 | */ |
| 145 | if (++try_loading_module == 1) |
| 146 | request_module("net-pf-%d-proto-%d-type-%d", |
| 147 | PF_INET6, protocol, sock->type); |
| 148 | /* |
| 149 | * Fall back to generic, e.g. net-pf-10-proto-132 |
| 150 | * (net-pf-PF_INET6-proto-IPPROTO_SCTP) |
| 151 | */ |
| 152 | else |
| 153 | request_module("net-pf-%d-proto-%d", |
| 154 | PF_INET6, protocol); |
| 155 | goto lookup_protocol; |
| 156 | } else |
| 157 | goto out_rcu_unlock; |
| 158 | } |
| 159 | |
| 160 | err = -EPERM; |
| 161 | if (sock->type == SOCK_RAW && !kern && |
| 162 | !ns_capable(net->user_ns, CAP_NET_RAW)) |
| 163 | goto out_rcu_unlock; |
| 164 | |
| 165 | sock->ops = answer->ops; |
| 166 | answer_prot = answer->prot; |
| 167 | answer_flags = answer->flags; |
| 168 | rcu_read_unlock(); |
| 169 | |
| 170 | WARN_ON(!answer_prot->slab); |
| 171 | |
| 172 | err = -ENOBUFS; |
| 173 | sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern); |
| 174 | if (!sk) |
| 175 | goto out; |
| 176 | |
| 177 | sock_init_data(sock, sk); |
| 178 | |
| 179 | err = 0; |
| 180 | if (INET_PROTOSW_REUSE & answer_flags) |
| 181 | sk->sk_reuse = SK_CAN_REUSE; |
| 182 | |
| 183 | inet = inet_sk(sk); |
| 184 | inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; |
| 185 | |
| 186 | if (SOCK_RAW == sock->type) { |
| 187 | inet->inet_num = protocol; |
| 188 | if (IPPROTO_RAW == protocol) |
| 189 | inet->hdrincl = 1; |
| 190 | } |
| 191 | |
| 192 | sk->sk_destruct = inet_sock_destruct; |
| 193 | sk->sk_family = PF_INET6; |
| 194 | sk->sk_protocol = protocol; |
| 195 | |
| 196 | sk->sk_backlog_rcv = answer->prot->backlog_rcv; |
| 197 | |
| 198 | inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); |
| 199 | np->hop_limit = -1; |
| 200 | np->mcast_hops = IPV6_DEFAULT_MCASTHOPS; |
| 201 | np->mc_loop = 1; |
| 202 | np->pmtudisc = IPV6_PMTUDISC_WANT; |
| 203 | np->autoflowlabel = ip6_default_np_autolabel(sock_net(sk)); |
| 204 | sk->sk_ipv6only = net->ipv6.sysctl.bindv6only; |
| 205 | |
| 206 | /* Init the ipv4 part of the socket since we can have sockets |
| 207 | * using v6 API for ipv4. |
| 208 | */ |
| 209 | inet->uc_ttl = -1; |
| 210 | |
| 211 | inet->mc_loop = 1; |
| 212 | inet->mc_ttl = 1; |
| 213 | inet->mc_index = 0; |
| 214 | inet->mc_list = NULL; |
| 215 | inet->rcv_tos = 0; |
| 216 | |
| 217 | if (net->ipv4.sysctl_ip_no_pmtu_disc) |
| 218 | inet->pmtudisc = IP_PMTUDISC_DONT; |
| 219 | else |
| 220 | inet->pmtudisc = IP_PMTUDISC_WANT; |
| 221 | /* |
| 222 | * Increment only the relevant sk_prot->socks debug field, this changes |
| 223 | * the previous behaviour of incrementing both the equivalent to |
| 224 | * answer->prot->socks (inet6_sock_nr) and inet_sock_nr. |
| 225 | * |
| 226 | * This allows better debug granularity as we'll know exactly how many |
| 227 | * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6 |
| 228 | * transport protocol socks. -acme |
| 229 | */ |
| 230 | sk_refcnt_debug_inc(sk); |
| 231 | |
| 232 | if (inet->inet_num) { |
| 233 | /* It assumes that any protocol which allows |
| 234 | * the user to assign a number at socket |
| 235 | * creation time automatically shares. |
| 236 | */ |
| 237 | inet->inet_sport = htons(inet->inet_num); |
| 238 | err = sk->sk_prot->hash(sk); |
| 239 | if (err) { |
| 240 | sk_common_release(sk); |
| 241 | goto out; |
| 242 | } |
| 243 | } |
| 244 | if (sk->sk_prot->init) { |
| 245 | err = sk->sk_prot->init(sk); |
| 246 | if (err) { |
| 247 | sk_common_release(sk); |
| 248 | goto out; |
| 249 | } |
| 250 | } |
| 251 | out: |
| 252 | return err; |
| 253 | out_rcu_unlock: |
| 254 | rcu_read_unlock(); |
| 255 | goto out; |
| 256 | } |
| 257 | |
| 258 | |
| 259 | /* bind for INET6 API */ |
| 260 | int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) |
| 261 | { |
| 262 | struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr; |
| 263 | struct sock *sk = sock->sk; |
| 264 | struct inet_sock *inet = inet_sk(sk); |
| 265 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 266 | struct net *net = sock_net(sk); |
| 267 | __be32 v4addr = 0; |
| 268 | unsigned short snum; |
| 269 | int addr_type = 0; |
| 270 | int err = 0; |
| 271 | |
| 272 | /* If the socket has its own bind function then use it. */ |
| 273 | if (sk->sk_prot->bind) |
| 274 | return sk->sk_prot->bind(sk, uaddr, addr_len); |
| 275 | |
| 276 | if (addr_len < SIN6_LEN_RFC2133) |
| 277 | return -EINVAL; |
| 278 | |
| 279 | if (addr->sin6_family != AF_INET6) |
| 280 | return -EAFNOSUPPORT; |
| 281 | |
| 282 | addr_type = ipv6_addr_type(&addr->sin6_addr); |
| 283 | if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM) |
| 284 | return -EINVAL; |
| 285 | |
| 286 | snum = ntohs(addr->sin6_port); |
| 287 | if (snum && snum < PROT_SOCK && !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) |
| 288 | return -EACCES; |
| 289 | |
| 290 | lock_sock(sk); |
| 291 | |
| 292 | /* Check these errors (active socket, double bind). */ |
| 293 | if (sk->sk_state != TCP_CLOSE || inet->inet_num) { |
| 294 | err = -EINVAL; |
| 295 | goto out; |
| 296 | } |
| 297 | |
| 298 | /* Check if the address belongs to the host. */ |
| 299 | if (addr_type == IPV6_ADDR_MAPPED) { |
| 300 | int chk_addr_ret; |
| 301 | |
| 302 | /* Binding to v4-mapped address on a v6-only socket |
| 303 | * makes no sense |
| 304 | */ |
| 305 | if (sk->sk_ipv6only) { |
| 306 | err = -EINVAL; |
| 307 | goto out; |
| 308 | } |
| 309 | |
| 310 | /* Reproduce AF_INET checks to make the bindings consistent */ |
| 311 | v4addr = addr->sin6_addr.s6_addr32[3]; |
| 312 | chk_addr_ret = inet_addr_type(net, v4addr); |
| 313 | if (!net->ipv4.sysctl_ip_nonlocal_bind && |
| 314 | !(inet->freebind || inet->transparent) && |
| 315 | v4addr != htonl(INADDR_ANY) && |
| 316 | chk_addr_ret != RTN_LOCAL && |
| 317 | chk_addr_ret != RTN_MULTICAST && |
| 318 | chk_addr_ret != RTN_BROADCAST) { |
| 319 | err = -EADDRNOTAVAIL; |
| 320 | goto out; |
| 321 | } |
| 322 | } else { |
| 323 | if (addr_type != IPV6_ADDR_ANY) { |
| 324 | struct net_device *dev = NULL; |
| 325 | |
| 326 | rcu_read_lock(); |
| 327 | if (__ipv6_addr_needs_scope_id(addr_type)) { |
| 328 | if (addr_len >= sizeof(struct sockaddr_in6) && |
| 329 | addr->sin6_scope_id) { |
| 330 | /* Override any existing binding, if another one |
| 331 | * is supplied by user. |
| 332 | */ |
| 333 | sk->sk_bound_dev_if = addr->sin6_scope_id; |
| 334 | } |
| 335 | |
| 336 | /* Binding to link-local address requires an interface */ |
| 337 | if (!sk->sk_bound_dev_if) { |
| 338 | err = -EINVAL; |
| 339 | goto out_unlock; |
| 340 | } |
| 341 | dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); |
| 342 | if (!dev) { |
| 343 | err = -ENODEV; |
| 344 | goto out_unlock; |
| 345 | } |
| 346 | } |
| 347 | |
| 348 | /* ipv4 addr of the socket is invalid. Only the |
| 349 | * unspecified and mapped address have a v4 equivalent. |
| 350 | */ |
| 351 | v4addr = LOOPBACK4_IPV6; |
| 352 | if (!(addr_type & IPV6_ADDR_MULTICAST)) { |
| 353 | if (!net->ipv6.sysctl.ip_nonlocal_bind && |
| 354 | !(inet->freebind || inet->transparent) && |
| 355 | !ipv6_chk_addr(net, &addr->sin6_addr, |
| 356 | dev, 0)) { |
| 357 | err = -EADDRNOTAVAIL; |
| 358 | goto out_unlock; |
| 359 | } |
| 360 | } |
| 361 | rcu_read_unlock(); |
| 362 | } |
| 363 | } |
| 364 | |
| 365 | inet->inet_rcv_saddr = v4addr; |
| 366 | inet->inet_saddr = v4addr; |
| 367 | |
| 368 | sk->sk_v6_rcv_saddr = addr->sin6_addr; |
| 369 | |
| 370 | if (!(addr_type & IPV6_ADDR_MULTICAST)) |
| 371 | np->saddr = addr->sin6_addr; |
| 372 | |
| 373 | /* Make sure we are allowed to bind here. */ |
| 374 | if ((snum || !inet->bind_address_no_port) && |
| 375 | sk->sk_prot->get_port(sk, snum)) { |
| 376 | inet_reset_saddr(sk); |
| 377 | err = -EADDRINUSE; |
| 378 | goto out; |
| 379 | } |
| 380 | |
| 381 | if (addr_type != IPV6_ADDR_ANY) { |
| 382 | sk->sk_userlocks |= SOCK_BINDADDR_LOCK; |
| 383 | if (addr_type != IPV6_ADDR_MAPPED) |
| 384 | sk->sk_ipv6only = 1; |
| 385 | } |
| 386 | if (snum) |
| 387 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; |
| 388 | inet->inet_sport = htons(inet->inet_num); |
| 389 | inet->inet_dport = 0; |
| 390 | inet->inet_daddr = 0; |
| 391 | out: |
| 392 | release_sock(sk); |
| 393 | return err; |
| 394 | out_unlock: |
| 395 | rcu_read_unlock(); |
| 396 | goto out; |
| 397 | } |
| 398 | EXPORT_SYMBOL(inet6_bind); |
| 399 | |
| 400 | int inet6_release(struct socket *sock) |
| 401 | { |
| 402 | struct sock *sk = sock->sk; |
| 403 | |
| 404 | if (!sk) |
| 405 | return -EINVAL; |
| 406 | |
| 407 | /* Free mc lists */ |
| 408 | ipv6_sock_mc_close(sk); |
| 409 | |
| 410 | /* Free ac lists */ |
| 411 | ipv6_sock_ac_close(sk); |
| 412 | |
| 413 | return inet_release(sock); |
| 414 | } |
| 415 | EXPORT_SYMBOL(inet6_release); |
| 416 | |
| 417 | void inet6_destroy_sock(struct sock *sk) |
| 418 | { |
| 419 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 420 | struct sk_buff *skb; |
| 421 | struct ipv6_txoptions *opt; |
| 422 | |
| 423 | /* Release rx options */ |
| 424 | |
| 425 | skb = xchg(&np->pktoptions, NULL); |
| 426 | if (skb) |
| 427 | kfree_skb(skb); |
| 428 | |
| 429 | skb = xchg(&np->rxpmtu, NULL); |
| 430 | if (skb) |
| 431 | kfree_skb(skb); |
| 432 | |
| 433 | /* Free flowlabels */ |
| 434 | fl6_free_socklist(sk); |
| 435 | |
| 436 | /* Free tx options */ |
| 437 | |
| 438 | opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); |
| 439 | if (opt) { |
| 440 | atomic_sub(opt->tot_len, &sk->sk_omem_alloc); |
| 441 | txopt_put(opt); |
| 442 | } |
| 443 | } |
| 444 | EXPORT_SYMBOL_GPL(inet6_destroy_sock); |
| 445 | |
| 446 | /* |
| 447 | * This does both peername and sockname. |
| 448 | */ |
| 449 | |
| 450 | int inet6_getname(struct socket *sock, struct sockaddr *uaddr, |
| 451 | int *uaddr_len, int peer) |
| 452 | { |
| 453 | struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; |
| 454 | struct sock *sk = sock->sk; |
| 455 | struct inet_sock *inet = inet_sk(sk); |
| 456 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 457 | |
| 458 | sin->sin6_family = AF_INET6; |
| 459 | sin->sin6_flowinfo = 0; |
| 460 | sin->sin6_scope_id = 0; |
| 461 | if (peer) { |
| 462 | if (!inet->inet_dport) |
| 463 | return -ENOTCONN; |
| 464 | if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && |
| 465 | peer == 1) |
| 466 | return -ENOTCONN; |
| 467 | sin->sin6_port = inet->inet_dport; |
| 468 | sin->sin6_addr = sk->sk_v6_daddr; |
| 469 | if (np->sndflow) |
| 470 | sin->sin6_flowinfo = np->flow_label; |
| 471 | } else { |
| 472 | if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) |
| 473 | sin->sin6_addr = np->saddr; |
| 474 | else |
| 475 | sin->sin6_addr = sk->sk_v6_rcv_saddr; |
| 476 | |
| 477 | sin->sin6_port = inet->inet_sport; |
| 478 | } |
| 479 | sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, |
| 480 | sk->sk_bound_dev_if); |
| 481 | *uaddr_len = sizeof(*sin); |
| 482 | return 0; |
| 483 | } |
| 484 | EXPORT_SYMBOL(inet6_getname); |
| 485 | |
| 486 | int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) |
| 487 | { |
| 488 | struct sock *sk = sock->sk; |
| 489 | struct net *net = sock_net(sk); |
| 490 | |
| 491 | switch (cmd) { |
| 492 | case SIOCGSTAMP: |
| 493 | return sock_get_timestamp(sk, (struct timeval __user *)arg); |
| 494 | |
| 495 | case SIOCGSTAMPNS: |
| 496 | return sock_get_timestampns(sk, (struct timespec __user *)arg); |
| 497 | |
| 498 | case SIOCADDRT: |
| 499 | case SIOCDELRT: |
| 500 | |
| 501 | return ipv6_route_ioctl(net, cmd, (void __user *)arg); |
| 502 | |
| 503 | case SIOCSIFADDR: |
| 504 | return addrconf_add_ifaddr(net, (void __user *) arg); |
| 505 | case SIOCDIFADDR: |
| 506 | return addrconf_del_ifaddr(net, (void __user *) arg); |
| 507 | case SIOCSIFDSTADDR: |
| 508 | return addrconf_set_dstaddr(net, (void __user *) arg); |
| 509 | default: |
| 510 | if (!sk->sk_prot->ioctl) |
| 511 | return -ENOIOCTLCMD; |
| 512 | return sk->sk_prot->ioctl(sk, cmd, arg); |
| 513 | } |
| 514 | /*NOTREACHED*/ |
| 515 | return 0; |
| 516 | } |
| 517 | EXPORT_SYMBOL(inet6_ioctl); |
| 518 | |
| 519 | const struct proto_ops inet6_stream_ops = { |
| 520 | .family = PF_INET6, |
| 521 | .owner = THIS_MODULE, |
| 522 | .release = inet6_release, |
| 523 | .bind = inet6_bind, |
| 524 | .connect = inet_stream_connect, /* ok */ |
| 525 | .socketpair = sock_no_socketpair, /* a do nothing */ |
| 526 | .accept = inet_accept, /* ok */ |
| 527 | .getname = inet6_getname, |
| 528 | .poll = tcp_poll, /* ok */ |
| 529 | .ioctl = inet6_ioctl, /* must change */ |
| 530 | .listen = inet_listen, /* ok */ |
| 531 | .shutdown = inet_shutdown, /* ok */ |
| 532 | .setsockopt = sock_common_setsockopt, /* ok */ |
| 533 | .getsockopt = sock_common_getsockopt, /* ok */ |
| 534 | .sendmsg = inet_sendmsg, /* ok */ |
| 535 | .recvmsg = inet_recvmsg, /* ok */ |
| 536 | .mmap = sock_no_mmap, |
| 537 | .sendpage = inet_sendpage, |
| 538 | .splice_read = tcp_splice_read, |
| 539 | #ifdef CONFIG_COMPAT |
| 540 | .compat_setsockopt = compat_sock_common_setsockopt, |
| 541 | .compat_getsockopt = compat_sock_common_getsockopt, |
| 542 | #endif |
| 543 | }; |
| 544 | |
| 545 | const struct proto_ops inet6_dgram_ops = { |
| 546 | .family = PF_INET6, |
| 547 | .owner = THIS_MODULE, |
| 548 | .release = inet6_release, |
| 549 | .bind = inet6_bind, |
| 550 | .connect = inet_dgram_connect, /* ok */ |
| 551 | .socketpair = sock_no_socketpair, /* a do nothing */ |
| 552 | .accept = sock_no_accept, /* a do nothing */ |
| 553 | .getname = inet6_getname, |
| 554 | .poll = udp_poll, /* ok */ |
| 555 | .ioctl = inet6_ioctl, /* must change */ |
| 556 | .listen = sock_no_listen, /* ok */ |
| 557 | .shutdown = inet_shutdown, /* ok */ |
| 558 | .setsockopt = sock_common_setsockopt, /* ok */ |
| 559 | .getsockopt = sock_common_getsockopt, /* ok */ |
| 560 | .sendmsg = inet_sendmsg, /* ok */ |
| 561 | .recvmsg = inet_recvmsg, /* ok */ |
| 562 | .mmap = sock_no_mmap, |
| 563 | .sendpage = sock_no_sendpage, |
| 564 | #ifdef CONFIG_COMPAT |
| 565 | .compat_setsockopt = compat_sock_common_setsockopt, |
| 566 | .compat_getsockopt = compat_sock_common_getsockopt, |
| 567 | #endif |
| 568 | }; |
| 569 | |
| 570 | static const struct net_proto_family inet6_family_ops = { |
| 571 | .family = PF_INET6, |
| 572 | .create = inet6_create, |
| 573 | .owner = THIS_MODULE, |
| 574 | }; |
| 575 | |
| 576 | int inet6_register_protosw(struct inet_protosw *p) |
| 577 | { |
| 578 | struct list_head *lh; |
| 579 | struct inet_protosw *answer; |
| 580 | struct list_head *last_perm; |
| 581 | int protocol = p->protocol; |
| 582 | int ret; |
| 583 | |
| 584 | spin_lock_bh(&inetsw6_lock); |
| 585 | |
| 586 | ret = -EINVAL; |
| 587 | if (p->type >= SOCK_MAX) |
| 588 | goto out_illegal; |
| 589 | |
| 590 | /* If we are trying to override a permanent protocol, bail. */ |
| 591 | answer = NULL; |
| 592 | ret = -EPERM; |
| 593 | last_perm = &inetsw6[p->type]; |
| 594 | list_for_each(lh, &inetsw6[p->type]) { |
| 595 | answer = list_entry(lh, struct inet_protosw, list); |
| 596 | |
| 597 | /* Check only the non-wild match. */ |
| 598 | if (INET_PROTOSW_PERMANENT & answer->flags) { |
| 599 | if (protocol == answer->protocol) |
| 600 | break; |
| 601 | last_perm = lh; |
| 602 | } |
| 603 | |
| 604 | answer = NULL; |
| 605 | } |
| 606 | if (answer) |
| 607 | goto out_permanent; |
| 608 | |
| 609 | /* Add the new entry after the last permanent entry if any, so that |
| 610 | * the new entry does not override a permanent entry when matched with |
| 611 | * a wild-card protocol. But it is allowed to override any existing |
| 612 | * non-permanent entry. This means that when we remove this entry, the |
| 613 | * system automatically returns to the old behavior. |
| 614 | */ |
| 615 | list_add_rcu(&p->list, last_perm); |
| 616 | ret = 0; |
| 617 | out: |
| 618 | spin_unlock_bh(&inetsw6_lock); |
| 619 | return ret; |
| 620 | |
| 621 | out_permanent: |
| 622 | pr_err("Attempt to override permanent protocol %d\n", protocol); |
| 623 | goto out; |
| 624 | |
| 625 | out_illegal: |
| 626 | pr_err("Ignoring attempt to register invalid socket type %d\n", |
| 627 | p->type); |
| 628 | goto out; |
| 629 | } |
| 630 | EXPORT_SYMBOL(inet6_register_protosw); |
| 631 | |
| 632 | void |
| 633 | inet6_unregister_protosw(struct inet_protosw *p) |
| 634 | { |
| 635 | if (INET_PROTOSW_PERMANENT & p->flags) { |
| 636 | pr_err("Attempt to unregister permanent protocol %d\n", |
| 637 | p->protocol); |
| 638 | } else { |
| 639 | spin_lock_bh(&inetsw6_lock); |
| 640 | list_del_rcu(&p->list); |
| 641 | spin_unlock_bh(&inetsw6_lock); |
| 642 | |
| 643 | synchronize_net(); |
| 644 | } |
| 645 | } |
| 646 | EXPORT_SYMBOL(inet6_unregister_protosw); |
| 647 | |
| 648 | int inet6_sk_rebuild_header(struct sock *sk) |
| 649 | { |
| 650 | struct ipv6_pinfo *np = inet6_sk(sk); |
| 651 | struct dst_entry *dst; |
| 652 | |
| 653 | dst = __sk_dst_check(sk, np->dst_cookie); |
| 654 | |
| 655 | if (!dst) { |
| 656 | struct inet_sock *inet = inet_sk(sk); |
| 657 | struct in6_addr *final_p, final; |
| 658 | struct flowi6 fl6; |
| 659 | |
| 660 | memset(&fl6, 0, sizeof(fl6)); |
| 661 | fl6.flowi6_proto = sk->sk_protocol; |
| 662 | fl6.daddr = sk->sk_v6_daddr; |
| 663 | fl6.saddr = np->saddr; |
| 664 | fl6.flowlabel = np->flow_label; |
| 665 | fl6.flowi6_oif = sk->sk_bound_dev_if; |
| 666 | fl6.flowi6_mark = sk->sk_mark; |
| 667 | fl6.fl6_dport = inet->inet_dport; |
| 668 | fl6.fl6_sport = inet->inet_sport; |
| 669 | security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); |
| 670 | |
| 671 | rcu_read_lock(); |
| 672 | final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), |
| 673 | &final); |
| 674 | rcu_read_unlock(); |
| 675 | |
| 676 | dst = ip6_dst_lookup_flow(sk, &fl6, final_p); |
| 677 | if (IS_ERR(dst)) { |
| 678 | sk->sk_route_caps = 0; |
| 679 | sk->sk_err_soft = -PTR_ERR(dst); |
| 680 | return PTR_ERR(dst); |
| 681 | } |
| 682 | |
| 683 | ip6_dst_store(sk, dst, NULL, NULL); |
| 684 | } |
| 685 | |
| 686 | return 0; |
| 687 | } |
| 688 | EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); |
| 689 | |
| 690 | bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, |
| 691 | const struct inet6_skb_parm *opt) |
| 692 | { |
| 693 | const struct ipv6_pinfo *np = inet6_sk(sk); |
| 694 | |
| 695 | if (np->rxopt.all) { |
| 696 | if (((opt->flags & IP6SKB_HOPBYHOP) && |
| 697 | (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || |
| 698 | (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && |
| 699 | np->rxopt.bits.rxflow) || |
| 700 | (opt->srcrt && (np->rxopt.bits.srcrt || |
| 701 | np->rxopt.bits.osrcrt)) || |
| 702 | ((opt->dst1 || opt->dst0) && |
| 703 | (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) |
| 704 | return true; |
| 705 | } |
| 706 | return false; |
| 707 | } |
| 708 | EXPORT_SYMBOL_GPL(ipv6_opt_accepted); |
| 709 | |
| 710 | static struct packet_type ipv6_packet_type __read_mostly = { |
| 711 | .type = cpu_to_be16(ETH_P_IPV6), |
| 712 | .func = ipv6_rcv, |
| 713 | }; |
| 714 | |
| 715 | static int __init ipv6_packet_init(void) |
| 716 | { |
| 717 | dev_add_pack(&ipv6_packet_type); |
| 718 | return 0; |
| 719 | } |
| 720 | |
| 721 | static void ipv6_packet_cleanup(void) |
| 722 | { |
| 723 | dev_remove_pack(&ipv6_packet_type); |
| 724 | } |
| 725 | |
| 726 | static int __net_init ipv6_init_mibs(struct net *net) |
| 727 | { |
| 728 | int i; |
| 729 | |
| 730 | net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); |
| 731 | if (!net->mib.udp_stats_in6) |
| 732 | return -ENOMEM; |
| 733 | net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); |
| 734 | if (!net->mib.udplite_stats_in6) |
| 735 | goto err_udplite_mib; |
| 736 | net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); |
| 737 | if (!net->mib.ipv6_statistics) |
| 738 | goto err_ip_mib; |
| 739 | |
| 740 | for_each_possible_cpu(i) { |
| 741 | struct ipstats_mib *af_inet6_stats; |
| 742 | af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); |
| 743 | u64_stats_init(&af_inet6_stats->syncp); |
| 744 | } |
| 745 | |
| 746 | |
| 747 | net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); |
| 748 | if (!net->mib.icmpv6_statistics) |
| 749 | goto err_icmp_mib; |
| 750 | net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), |
| 751 | GFP_KERNEL); |
| 752 | if (!net->mib.icmpv6msg_statistics) |
| 753 | goto err_icmpmsg_mib; |
| 754 | return 0; |
| 755 | |
| 756 | err_icmpmsg_mib: |
| 757 | free_percpu(net->mib.icmpv6_statistics); |
| 758 | err_icmp_mib: |
| 759 | free_percpu(net->mib.ipv6_statistics); |
| 760 | err_ip_mib: |
| 761 | free_percpu(net->mib.udplite_stats_in6); |
| 762 | err_udplite_mib: |
| 763 | free_percpu(net->mib.udp_stats_in6); |
| 764 | return -ENOMEM; |
| 765 | } |
| 766 | |
| 767 | static void ipv6_cleanup_mibs(struct net *net) |
| 768 | { |
| 769 | free_percpu(net->mib.udp_stats_in6); |
| 770 | free_percpu(net->mib.udplite_stats_in6); |
| 771 | free_percpu(net->mib.ipv6_statistics); |
| 772 | free_percpu(net->mib.icmpv6_statistics); |
| 773 | kfree(net->mib.icmpv6msg_statistics); |
| 774 | } |
| 775 | |
| 776 | static int __net_init inet6_net_init(struct net *net) |
| 777 | { |
| 778 | int err = 0; |
| 779 | |
| 780 | net->ipv6.sysctl.bindv6only = 0; |
| 781 | net->ipv6.sysctl.icmpv6_time = 1*HZ; |
| 782 | net->ipv6.sysctl.flowlabel_consistency = 1; |
| 783 | net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; |
| 784 | net->ipv6.sysctl.idgen_retries = 3; |
| 785 | net->ipv6.sysctl.idgen_delay = 1 * HZ; |
| 786 | net->ipv6.sysctl.flowlabel_state_ranges = 0; |
| 787 | atomic_set(&net->ipv6.fib6_sernum, 1); |
| 788 | |
| 789 | err = ipv6_init_mibs(net); |
| 790 | if (err) |
| 791 | return err; |
| 792 | #ifdef CONFIG_PROC_FS |
| 793 | err = udp6_proc_init(net); |
| 794 | if (err) |
| 795 | goto out; |
| 796 | err = tcp6_proc_init(net); |
| 797 | if (err) |
| 798 | goto proc_tcp6_fail; |
| 799 | err = ac6_proc_init(net); |
| 800 | if (err) |
| 801 | goto proc_ac6_fail; |
| 802 | #endif |
| 803 | return err; |
| 804 | |
| 805 | #ifdef CONFIG_PROC_FS |
| 806 | proc_ac6_fail: |
| 807 | tcp6_proc_exit(net); |
| 808 | proc_tcp6_fail: |
| 809 | udp6_proc_exit(net); |
| 810 | out: |
| 811 | ipv6_cleanup_mibs(net); |
| 812 | return err; |
| 813 | #endif |
| 814 | } |
| 815 | |
| 816 | static void __net_exit inet6_net_exit(struct net *net) |
| 817 | { |
| 818 | #ifdef CONFIG_PROC_FS |
| 819 | udp6_proc_exit(net); |
| 820 | tcp6_proc_exit(net); |
| 821 | ac6_proc_exit(net); |
| 822 | #endif |
| 823 | ipv6_cleanup_mibs(net); |
| 824 | } |
| 825 | |
| 826 | static struct pernet_operations inet6_net_ops = { |
| 827 | .init = inet6_net_init, |
| 828 | .exit = inet6_net_exit, |
| 829 | }; |
| 830 | |
| 831 | static const struct ipv6_stub ipv6_stub_impl = { |
| 832 | .ipv6_sock_mc_join = ipv6_sock_mc_join, |
| 833 | .ipv6_sock_mc_drop = ipv6_sock_mc_drop, |
| 834 | .ipv6_dst_lookup = ip6_dst_lookup, |
| 835 | .udpv6_encap_enable = udpv6_encap_enable, |
| 836 | .ndisc_send_na = ndisc_send_na, |
| 837 | .nd_tbl = &nd_tbl, |
| 838 | }; |
| 839 | |
| 840 | static int __init inet6_init(void) |
| 841 | { |
| 842 | struct list_head *r; |
| 843 | int err = 0; |
| 844 | |
| 845 | sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); |
| 846 | |
| 847 | /* Register the socket-side information for inet6_create. */ |
| 848 | for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) |
| 849 | INIT_LIST_HEAD(r); |
| 850 | |
| 851 | if (disable_ipv6_mod) { |
| 852 | pr_info("Loaded, but administratively disabled, reboot required to enable\n"); |
| 853 | goto out; |
| 854 | } |
| 855 | |
| 856 | err = proto_register(&tcpv6_prot, 1); |
| 857 | if (err) |
| 858 | goto out; |
| 859 | |
| 860 | err = proto_register(&udpv6_prot, 1); |
| 861 | if (err) |
| 862 | goto out_unregister_tcp_proto; |
| 863 | |
| 864 | err = proto_register(&udplitev6_prot, 1); |
| 865 | if (err) |
| 866 | goto out_unregister_udp_proto; |
| 867 | |
| 868 | err = proto_register(&rawv6_prot, 1); |
| 869 | if (err) |
| 870 | goto out_unregister_udplite_proto; |
| 871 | |
| 872 | err = proto_register(&pingv6_prot, 1); |
| 873 | if (err) |
| 874 | goto out_unregister_ping_proto; |
| 875 | |
| 876 | /* We MUST register RAW sockets before we create the ICMP6, |
| 877 | * IGMP6, or NDISC control sockets. |
| 878 | */ |
| 879 | err = rawv6_init(); |
| 880 | if (err) |
| 881 | goto out_unregister_raw_proto; |
| 882 | |
| 883 | /* Register the family here so that the init calls below will |
| 884 | * be able to create sockets. (?? is this dangerous ??) |
| 885 | */ |
| 886 | err = sock_register(&inet6_family_ops); |
| 887 | if (err) |
| 888 | goto out_sock_register_fail; |
| 889 | |
| 890 | /* |
| 891 | * ipngwg API draft makes clear that the correct semantics |
| 892 | * for TCP and UDP is to consider one TCP and UDP instance |
| 893 | * in a host available by both INET and INET6 APIs and |
| 894 | * able to communicate via both network protocols. |
| 895 | */ |
| 896 | |
| 897 | err = register_pernet_subsys(&inet6_net_ops); |
| 898 | if (err) |
| 899 | goto register_pernet_fail; |
| 900 | err = icmpv6_init(); |
| 901 | if (err) |
| 902 | goto icmp_fail; |
| 903 | err = ip6_mr_init(); |
| 904 | if (err) |
| 905 | goto ipmr_fail; |
| 906 | err = ndisc_init(); |
| 907 | if (err) |
| 908 | goto ndisc_fail; |
| 909 | err = igmp6_init(); |
| 910 | if (err) |
| 911 | goto igmp_fail; |
| 912 | |
| 913 | ipv6_stub = &ipv6_stub_impl; |
| 914 | |
| 915 | err = ipv6_netfilter_init(); |
| 916 | if (err) |
| 917 | goto netfilter_fail; |
| 918 | /* Create /proc/foo6 entries. */ |
| 919 | #ifdef CONFIG_PROC_FS |
| 920 | err = -ENOMEM; |
| 921 | if (raw6_proc_init()) |
| 922 | goto proc_raw6_fail; |
| 923 | if (udplite6_proc_init()) |
| 924 | goto proc_udplite6_fail; |
| 925 | if (ipv6_misc_proc_init()) |
| 926 | goto proc_misc6_fail; |
| 927 | if (if6_proc_init()) |
| 928 | goto proc_if6_fail; |
| 929 | #endif |
| 930 | err = ip6_route_init(); |
| 931 | if (err) |
| 932 | goto ip6_route_fail; |
| 933 | err = ndisc_late_init(); |
| 934 | if (err) |
| 935 | goto ndisc_late_fail; |
| 936 | err = ip6_flowlabel_init(); |
| 937 | if (err) |
| 938 | goto ip6_flowlabel_fail; |
| 939 | err = addrconf_init(); |
| 940 | if (err) |
| 941 | goto addrconf_fail; |
| 942 | |
| 943 | /* Init v6 extension headers. */ |
| 944 | err = ipv6_exthdrs_init(); |
| 945 | if (err) |
| 946 | goto ipv6_exthdrs_fail; |
| 947 | |
| 948 | err = ipv6_frag_init(); |
| 949 | if (err) |
| 950 | goto ipv6_frag_fail; |
| 951 | |
| 952 | /* Init v6 transport protocols. */ |
| 953 | err = udpv6_init(); |
| 954 | if (err) |
| 955 | goto udpv6_fail; |
| 956 | |
| 957 | err = udplitev6_init(); |
| 958 | if (err) |
| 959 | goto udplitev6_fail; |
| 960 | |
| 961 | err = tcpv6_init(); |
| 962 | if (err) |
| 963 | goto tcpv6_fail; |
| 964 | |
| 965 | err = ipv6_packet_init(); |
| 966 | if (err) |
| 967 | goto ipv6_packet_fail; |
| 968 | |
| 969 | err = pingv6_init(); |
| 970 | if (err) |
| 971 | goto pingv6_fail; |
| 972 | |
| 973 | #ifdef CONFIG_SYSCTL |
| 974 | err = ipv6_sysctl_register(); |
| 975 | if (err) |
| 976 | goto sysctl_fail; |
| 977 | #endif |
| 978 | out: |
| 979 | return err; |
| 980 | |
| 981 | #ifdef CONFIG_SYSCTL |
| 982 | sysctl_fail: |
| 983 | pingv6_exit(); |
| 984 | #endif |
| 985 | pingv6_fail: |
| 986 | ipv6_packet_cleanup(); |
| 987 | ipv6_packet_fail: |
| 988 | tcpv6_exit(); |
| 989 | tcpv6_fail: |
| 990 | udplitev6_exit(); |
| 991 | udplitev6_fail: |
| 992 | udpv6_exit(); |
| 993 | udpv6_fail: |
| 994 | ipv6_frag_exit(); |
| 995 | ipv6_frag_fail: |
| 996 | ipv6_exthdrs_exit(); |
| 997 | ipv6_exthdrs_fail: |
| 998 | addrconf_cleanup(); |
| 999 | addrconf_fail: |
| 1000 | ip6_flowlabel_cleanup(); |
| 1001 | ip6_flowlabel_fail: |
| 1002 | ndisc_late_cleanup(); |
| 1003 | ndisc_late_fail: |
| 1004 | ip6_route_cleanup(); |
| 1005 | ip6_route_fail: |
| 1006 | #ifdef CONFIG_PROC_FS |
| 1007 | if6_proc_exit(); |
| 1008 | proc_if6_fail: |
| 1009 | ipv6_misc_proc_exit(); |
| 1010 | proc_misc6_fail: |
| 1011 | udplite6_proc_exit(); |
| 1012 | proc_udplite6_fail: |
| 1013 | raw6_proc_exit(); |
| 1014 | proc_raw6_fail: |
| 1015 | #endif |
| 1016 | ipv6_netfilter_fini(); |
| 1017 | netfilter_fail: |
| 1018 | igmp6_cleanup(); |
| 1019 | igmp_fail: |
| 1020 | ndisc_cleanup(); |
| 1021 | ndisc_fail: |
| 1022 | ip6_mr_cleanup(); |
| 1023 | ipmr_fail: |
| 1024 | icmpv6_cleanup(); |
| 1025 | icmp_fail: |
| 1026 | unregister_pernet_subsys(&inet6_net_ops); |
| 1027 | register_pernet_fail: |
| 1028 | sock_unregister(PF_INET6); |
| 1029 | rtnl_unregister_all(PF_INET6); |
| 1030 | out_sock_register_fail: |
| 1031 | rawv6_exit(); |
| 1032 | out_unregister_ping_proto: |
| 1033 | proto_unregister(&pingv6_prot); |
| 1034 | out_unregister_raw_proto: |
| 1035 | proto_unregister(&rawv6_prot); |
| 1036 | out_unregister_udplite_proto: |
| 1037 | proto_unregister(&udplitev6_prot); |
| 1038 | out_unregister_udp_proto: |
| 1039 | proto_unregister(&udpv6_prot); |
| 1040 | out_unregister_tcp_proto: |
| 1041 | proto_unregister(&tcpv6_prot); |
| 1042 | goto out; |
| 1043 | } |
| 1044 | module_init(inet6_init); |
| 1045 | |
| 1046 | MODULE_ALIAS_NETPROTO(PF_INET6); |