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 #include <net/calipso.h> 64 #include <net/seg6.h> 65 66 #include <linux/uaccess.h> 67 #include <linux/mroute6.h> 68 69 #include "ip6_offload.h" 70 71 MODULE_AUTHOR("Cast of dozens"); 72 MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); 73 MODULE_LICENSE("GPL"); 74 75 /* The inetsw6 table contains everything that inet6_create needs to 76 * build a new socket. 77 */ 78 static struct list_head inetsw6[SOCK_MAX]; 79 static DEFINE_SPINLOCK(inetsw6_lock); 80 81 struct ipv6_params ipv6_defaults = { 82 .disable_ipv6 = 0, 83 .autoconf = 1, 84 }; 85 86 static int disable_ipv6_mod; 87 88 module_param_named(disable, disable_ipv6_mod, int, 0444); 89 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional"); 90 91 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444); 92 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces"); 93 94 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444); 95 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces"); 96 97 bool ipv6_mod_enabled(void) 98 { 99 return disable_ipv6_mod == 0; 100 } 101 EXPORT_SYMBOL_GPL(ipv6_mod_enabled); 102 103 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) 104 { 105 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); 106 107 return (struct ipv6_pinfo *)(((u8 *)sk) + offset); 108 } 109 110 static int inet6_create(struct net *net, struct socket *sock, int protocol, 111 int kern) 112 { 113 struct inet_sock *inet; 114 struct ipv6_pinfo *np; 115 struct sock *sk; 116 struct inet_protosw *answer; 117 struct proto *answer_prot; 118 unsigned char answer_flags; 119 int try_loading_module = 0; 120 int err; 121 122 if (protocol < 0 || protocol >= IPPROTO_MAX) 123 return -EINVAL; 124 125 /* Look for the requested type/protocol pair. */ 126 lookup_protocol: 127 err = -ESOCKTNOSUPPORT; 128 rcu_read_lock(); 129 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) { 130 131 err = 0; 132 /* Check the non-wild match. */ 133 if (protocol == answer->protocol) { 134 if (protocol != IPPROTO_IP) 135 break; 136 } else { 137 /* Check for the two wild cases. */ 138 if (IPPROTO_IP == protocol) { 139 protocol = answer->protocol; 140 break; 141 } 142 if (IPPROTO_IP == answer->protocol) 143 break; 144 } 145 err = -EPROTONOSUPPORT; 146 } 147 148 if (err) { 149 if (try_loading_module < 2) { 150 rcu_read_unlock(); 151 /* 152 * Be more specific, e.g. net-pf-10-proto-132-type-1 153 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) 154 */ 155 if (++try_loading_module == 1) 156 request_module("net-pf-%d-proto-%d-type-%d", 157 PF_INET6, protocol, sock->type); 158 /* 159 * Fall back to generic, e.g. net-pf-10-proto-132 160 * (net-pf-PF_INET6-proto-IPPROTO_SCTP) 161 */ 162 else 163 request_module("net-pf-%d-proto-%d", 164 PF_INET6, protocol); 165 goto lookup_protocol; 166 } else 167 goto out_rcu_unlock; 168 } 169 170 err = -EPERM; 171 if (sock->type == SOCK_RAW && !kern && 172 !ns_capable(net->user_ns, CAP_NET_RAW)) 173 goto out_rcu_unlock; 174 175 sock->ops = answer->ops; 176 answer_prot = answer->prot; 177 answer_flags = answer->flags; 178 rcu_read_unlock(); 179 180 WARN_ON(!answer_prot->slab); 181 182 err = -ENOBUFS; 183 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern); 184 if (!sk) 185 goto out; 186 187 sock_init_data(sock, sk); 188 189 err = 0; 190 if (INET_PROTOSW_REUSE & answer_flags) 191 sk->sk_reuse = SK_CAN_REUSE; 192 193 inet = inet_sk(sk); 194 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; 195 196 if (SOCK_RAW == sock->type) { 197 inet->inet_num = protocol; 198 if (IPPROTO_RAW == protocol) 199 inet->hdrincl = 1; 200 } 201 202 sk->sk_destruct = inet_sock_destruct; 203 sk->sk_family = PF_INET6; 204 sk->sk_protocol = protocol; 205 206 sk->sk_backlog_rcv = answer->prot->backlog_rcv; 207 208 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); 209 np->hop_limit = -1; 210 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS; 211 np->mc_loop = 1; 212 np->mc_all = 1; 213 np->pmtudisc = IPV6_PMTUDISC_WANT; 214 np->repflow = net->ipv6.sysctl.flowlabel_reflect; 215 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only; 216 217 /* Init the ipv4 part of the socket since we can have sockets 218 * using v6 API for ipv4. 219 */ 220 inet->uc_ttl = -1; 221 222 inet->mc_loop = 1; 223 inet->mc_ttl = 1; 224 inet->mc_index = 0; 225 inet->mc_list = NULL; 226 inet->rcv_tos = 0; 227 228 if (net->ipv4.sysctl_ip_no_pmtu_disc) 229 inet->pmtudisc = IP_PMTUDISC_DONT; 230 else 231 inet->pmtudisc = IP_PMTUDISC_WANT; 232 /* 233 * Increment only the relevant sk_prot->socks debug field, this changes 234 * the previous behaviour of incrementing both the equivalent to 235 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr. 236 * 237 * This allows better debug granularity as we'll know exactly how many 238 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6 239 * transport protocol socks. -acme 240 */ 241 sk_refcnt_debug_inc(sk); 242 243 if (inet->inet_num) { 244 /* It assumes that any protocol which allows 245 * the user to assign a number at socket 246 * creation time automatically shares. 247 */ 248 inet->inet_sport = htons(inet->inet_num); 249 err = sk->sk_prot->hash(sk); 250 if (err) { 251 sk_common_release(sk); 252 goto out; 253 } 254 } 255 if (sk->sk_prot->init) { 256 err = sk->sk_prot->init(sk); 257 if (err) { 258 sk_common_release(sk); 259 goto out; 260 } 261 } 262 263 if (!kern) { 264 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); 265 if (err) { 266 sk_common_release(sk); 267 goto out; 268 } 269 } 270 out: 271 return err; 272 out_rcu_unlock: 273 rcu_read_unlock(); 274 goto out; 275 } 276 277 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, 278 bool force_bind_address_no_port, bool with_lock) 279 { 280 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr; 281 struct inet_sock *inet = inet_sk(sk); 282 struct ipv6_pinfo *np = inet6_sk(sk); 283 struct net *net = sock_net(sk); 284 __be32 v4addr = 0; 285 unsigned short snum; 286 bool saved_ipv6only; 287 int addr_type = 0; 288 int err = 0; 289 290 if (addr->sin6_family != AF_INET6) 291 return -EAFNOSUPPORT; 292 293 addr_type = ipv6_addr_type(&addr->sin6_addr); 294 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM) 295 return -EINVAL; 296 297 snum = ntohs(addr->sin6_port); 298 if (snum && snum < inet_prot_sock(net) && 299 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 300 return -EACCES; 301 302 if (with_lock) 303 lock_sock(sk); 304 305 /* Check these errors (active socket, double bind). */ 306 if (sk->sk_state != TCP_CLOSE || inet->inet_num) { 307 err = -EINVAL; 308 goto out; 309 } 310 311 /* Check if the address belongs to the host. */ 312 if (addr_type == IPV6_ADDR_MAPPED) { 313 int chk_addr_ret; 314 315 /* Binding to v4-mapped address on a v6-only socket 316 * makes no sense 317 */ 318 if (sk->sk_ipv6only) { 319 err = -EINVAL; 320 goto out; 321 } 322 323 /* Reproduce AF_INET checks to make the bindings consistent */ 324 v4addr = addr->sin6_addr.s6_addr32[3]; 325 chk_addr_ret = inet_addr_type(net, v4addr); 326 if (!inet_can_nonlocal_bind(net, inet) && 327 v4addr != htonl(INADDR_ANY) && 328 chk_addr_ret != RTN_LOCAL && 329 chk_addr_ret != RTN_MULTICAST && 330 chk_addr_ret != RTN_BROADCAST) { 331 err = -EADDRNOTAVAIL; 332 goto out; 333 } 334 } else { 335 if (addr_type != IPV6_ADDR_ANY) { 336 struct net_device *dev = NULL; 337 338 rcu_read_lock(); 339 if (__ipv6_addr_needs_scope_id(addr_type)) { 340 if (addr_len >= sizeof(struct sockaddr_in6) && 341 addr->sin6_scope_id) { 342 /* Override any existing binding, if another one 343 * is supplied by user. 344 */ 345 sk->sk_bound_dev_if = addr->sin6_scope_id; 346 } 347 348 /* Binding to link-local address requires an interface */ 349 if (!sk->sk_bound_dev_if) { 350 err = -EINVAL; 351 goto out_unlock; 352 } 353 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 354 if (!dev) { 355 err = -ENODEV; 356 goto out_unlock; 357 } 358 } 359 360 /* ipv4 addr of the socket is invalid. Only the 361 * unspecified and mapped address have a v4 equivalent. 362 */ 363 v4addr = LOOPBACK4_IPV6; 364 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 365 if (!ipv6_can_nonlocal_bind(net, inet) && 366 !ipv6_chk_addr(net, &addr->sin6_addr, 367 dev, 0)) { 368 err = -EADDRNOTAVAIL; 369 goto out_unlock; 370 } 371 } 372 rcu_read_unlock(); 373 } 374 } 375 376 inet->inet_rcv_saddr = v4addr; 377 inet->inet_saddr = v4addr; 378 379 sk->sk_v6_rcv_saddr = addr->sin6_addr; 380 381 if (!(addr_type & IPV6_ADDR_MULTICAST)) 382 np->saddr = addr->sin6_addr; 383 384 saved_ipv6only = sk->sk_ipv6only; 385 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED) 386 sk->sk_ipv6only = 1; 387 388 /* Make sure we are allowed to bind here. */ 389 if (snum || !(inet->bind_address_no_port || 390 force_bind_address_no_port)) { 391 if (sk->sk_prot->get_port(sk, snum)) { 392 sk->sk_ipv6only = saved_ipv6only; 393 inet_reset_saddr(sk); 394 err = -EADDRINUSE; 395 goto out; 396 } 397 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk); 398 if (err) { 399 sk->sk_ipv6only = saved_ipv6only; 400 inet_reset_saddr(sk); 401 goto out; 402 } 403 } 404 405 if (addr_type != IPV6_ADDR_ANY) 406 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 407 if (snum) 408 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 409 inet->inet_sport = htons(inet->inet_num); 410 inet->inet_dport = 0; 411 inet->inet_daddr = 0; 412 out: 413 if (with_lock) 414 release_sock(sk); 415 return err; 416 out_unlock: 417 rcu_read_unlock(); 418 goto out; 419 } 420 421 /* bind for INET6 API */ 422 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 423 { 424 struct sock *sk = sock->sk; 425 int err = 0; 426 427 /* If the socket has its own bind function then use it. */ 428 if (sk->sk_prot->bind) 429 return sk->sk_prot->bind(sk, uaddr, addr_len); 430 431 if (addr_len < SIN6_LEN_RFC2133) 432 return -EINVAL; 433 434 /* BPF prog is run before any checks are done so that if the prog 435 * changes context in a wrong way it will be caught. 436 */ 437 err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr); 438 if (err) 439 return err; 440 441 return __inet6_bind(sk, uaddr, addr_len, false, true); 442 } 443 EXPORT_SYMBOL(inet6_bind); 444 445 int inet6_release(struct socket *sock) 446 { 447 struct sock *sk = sock->sk; 448 449 if (!sk) 450 return -EINVAL; 451 452 /* Free mc lists */ 453 ipv6_sock_mc_close(sk); 454 455 /* Free ac lists */ 456 ipv6_sock_ac_close(sk); 457 458 return inet_release(sock); 459 } 460 EXPORT_SYMBOL(inet6_release); 461 462 void inet6_destroy_sock(struct sock *sk) 463 { 464 struct ipv6_pinfo *np = inet6_sk(sk); 465 struct sk_buff *skb; 466 struct ipv6_txoptions *opt; 467 468 /* Release rx options */ 469 470 skb = xchg(&np->pktoptions, NULL); 471 if (skb) 472 kfree_skb(skb); 473 474 skb = xchg(&np->rxpmtu, NULL); 475 if (skb) 476 kfree_skb(skb); 477 478 /* Free flowlabels */ 479 fl6_free_socklist(sk); 480 481 /* Free tx options */ 482 483 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 484 if (opt) { 485 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 486 txopt_put(opt); 487 } 488 } 489 EXPORT_SYMBOL_GPL(inet6_destroy_sock); 490 491 /* 492 * This does both peername and sockname. 493 */ 494 495 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 496 int peer) 497 { 498 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 499 struct sock *sk = sock->sk; 500 struct inet_sock *inet = inet_sk(sk); 501 struct ipv6_pinfo *np = inet6_sk(sk); 502 503 sin->sin6_family = AF_INET6; 504 sin->sin6_flowinfo = 0; 505 sin->sin6_scope_id = 0; 506 if (peer) { 507 if (!inet->inet_dport) 508 return -ENOTCONN; 509 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 510 peer == 1) 511 return -ENOTCONN; 512 sin->sin6_port = inet->inet_dport; 513 sin->sin6_addr = sk->sk_v6_daddr; 514 if (np->sndflow) 515 sin->sin6_flowinfo = np->flow_label; 516 } else { 517 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 518 sin->sin6_addr = np->saddr; 519 else 520 sin->sin6_addr = sk->sk_v6_rcv_saddr; 521 522 sin->sin6_port = inet->inet_sport; 523 } 524 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 525 sk->sk_bound_dev_if); 526 return sizeof(*sin); 527 } 528 EXPORT_SYMBOL(inet6_getname); 529 530 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 531 { 532 struct sock *sk = sock->sk; 533 struct net *net = sock_net(sk); 534 535 switch (cmd) { 536 case SIOCGSTAMP: 537 return sock_get_timestamp(sk, (struct timeval __user *)arg); 538 539 case SIOCGSTAMPNS: 540 return sock_get_timestampns(sk, (struct timespec __user *)arg); 541 542 case SIOCADDRT: 543 case SIOCDELRT: 544 545 return ipv6_route_ioctl(net, cmd, (void __user *)arg); 546 547 case SIOCSIFADDR: 548 return addrconf_add_ifaddr(net, (void __user *) arg); 549 case SIOCDIFADDR: 550 return addrconf_del_ifaddr(net, (void __user *) arg); 551 case SIOCSIFDSTADDR: 552 return addrconf_set_dstaddr(net, (void __user *) arg); 553 default: 554 if (!sk->sk_prot->ioctl) 555 return -ENOIOCTLCMD; 556 return sk->sk_prot->ioctl(sk, cmd, arg); 557 } 558 /*NOTREACHED*/ 559 return 0; 560 } 561 EXPORT_SYMBOL(inet6_ioctl); 562 563 const struct proto_ops inet6_stream_ops = { 564 .family = PF_INET6, 565 .owner = THIS_MODULE, 566 .release = inet6_release, 567 .bind = inet6_bind, 568 .connect = inet_stream_connect, /* ok */ 569 .socketpair = sock_no_socketpair, /* a do nothing */ 570 .accept = inet_accept, /* ok */ 571 .getname = inet6_getname, 572 .poll = tcp_poll, /* ok */ 573 .ioctl = inet6_ioctl, /* must change */ 574 .listen = inet_listen, /* ok */ 575 .shutdown = inet_shutdown, /* ok */ 576 .setsockopt = sock_common_setsockopt, /* ok */ 577 .getsockopt = sock_common_getsockopt, /* ok */ 578 .sendmsg = inet_sendmsg, /* ok */ 579 .recvmsg = inet_recvmsg, /* ok */ 580 #ifdef CONFIG_MMU 581 .mmap = tcp_mmap, 582 #endif 583 .sendpage = inet_sendpage, 584 .sendmsg_locked = tcp_sendmsg_locked, 585 .sendpage_locked = tcp_sendpage_locked, 586 .splice_read = tcp_splice_read, 587 .read_sock = tcp_read_sock, 588 .peek_len = tcp_peek_len, 589 #ifdef CONFIG_COMPAT 590 .compat_setsockopt = compat_sock_common_setsockopt, 591 .compat_getsockopt = compat_sock_common_getsockopt, 592 #endif 593 .set_rcvlowat = tcp_set_rcvlowat, 594 }; 595 596 const struct proto_ops inet6_dgram_ops = { 597 .family = PF_INET6, 598 .owner = THIS_MODULE, 599 .release = inet6_release, 600 .bind = inet6_bind, 601 .connect = inet_dgram_connect, /* ok */ 602 .socketpair = sock_no_socketpair, /* a do nothing */ 603 .accept = sock_no_accept, /* a do nothing */ 604 .getname = inet6_getname, 605 .poll = udp_poll, /* ok */ 606 .ioctl = inet6_ioctl, /* must change */ 607 .listen = sock_no_listen, /* ok */ 608 .shutdown = inet_shutdown, /* ok */ 609 .setsockopt = sock_common_setsockopt, /* ok */ 610 .getsockopt = sock_common_getsockopt, /* ok */ 611 .sendmsg = inet_sendmsg, /* ok */ 612 .recvmsg = inet_recvmsg, /* ok */ 613 .mmap = sock_no_mmap, 614 .sendpage = sock_no_sendpage, 615 .set_peek_off = sk_set_peek_off, 616 #ifdef CONFIG_COMPAT 617 .compat_setsockopt = compat_sock_common_setsockopt, 618 .compat_getsockopt = compat_sock_common_getsockopt, 619 #endif 620 }; 621 622 static const struct net_proto_family inet6_family_ops = { 623 .family = PF_INET6, 624 .create = inet6_create, 625 .owner = THIS_MODULE, 626 }; 627 628 int inet6_register_protosw(struct inet_protosw *p) 629 { 630 struct list_head *lh; 631 struct inet_protosw *answer; 632 struct list_head *last_perm; 633 int protocol = p->protocol; 634 int ret; 635 636 spin_lock_bh(&inetsw6_lock); 637 638 ret = -EINVAL; 639 if (p->type >= SOCK_MAX) 640 goto out_illegal; 641 642 /* If we are trying to override a permanent protocol, bail. */ 643 answer = NULL; 644 ret = -EPERM; 645 last_perm = &inetsw6[p->type]; 646 list_for_each(lh, &inetsw6[p->type]) { 647 answer = list_entry(lh, struct inet_protosw, list); 648 649 /* Check only the non-wild match. */ 650 if (INET_PROTOSW_PERMANENT & answer->flags) { 651 if (protocol == answer->protocol) 652 break; 653 last_perm = lh; 654 } 655 656 answer = NULL; 657 } 658 if (answer) 659 goto out_permanent; 660 661 /* Add the new entry after the last permanent entry if any, so that 662 * the new entry does not override a permanent entry when matched with 663 * a wild-card protocol. But it is allowed to override any existing 664 * non-permanent entry. This means that when we remove this entry, the 665 * system automatically returns to the old behavior. 666 */ 667 list_add_rcu(&p->list, last_perm); 668 ret = 0; 669 out: 670 spin_unlock_bh(&inetsw6_lock); 671 return ret; 672 673 out_permanent: 674 pr_err("Attempt to override permanent protocol %d\n", protocol); 675 goto out; 676 677 out_illegal: 678 pr_err("Ignoring attempt to register invalid socket type %d\n", 679 p->type); 680 goto out; 681 } 682 EXPORT_SYMBOL(inet6_register_protosw); 683 684 void 685 inet6_unregister_protosw(struct inet_protosw *p) 686 { 687 if (INET_PROTOSW_PERMANENT & p->flags) { 688 pr_err("Attempt to unregister permanent protocol %d\n", 689 p->protocol); 690 } else { 691 spin_lock_bh(&inetsw6_lock); 692 list_del_rcu(&p->list); 693 spin_unlock_bh(&inetsw6_lock); 694 695 synchronize_net(); 696 } 697 } 698 EXPORT_SYMBOL(inet6_unregister_protosw); 699 700 int inet6_sk_rebuild_header(struct sock *sk) 701 { 702 struct ipv6_pinfo *np = inet6_sk(sk); 703 struct dst_entry *dst; 704 705 dst = __sk_dst_check(sk, np->dst_cookie); 706 707 if (!dst) { 708 struct inet_sock *inet = inet_sk(sk); 709 struct in6_addr *final_p, final; 710 struct flowi6 fl6; 711 712 memset(&fl6, 0, sizeof(fl6)); 713 fl6.flowi6_proto = sk->sk_protocol; 714 fl6.daddr = sk->sk_v6_daddr; 715 fl6.saddr = np->saddr; 716 fl6.flowlabel = np->flow_label; 717 fl6.flowi6_oif = sk->sk_bound_dev_if; 718 fl6.flowi6_mark = sk->sk_mark; 719 fl6.fl6_dport = inet->inet_dport; 720 fl6.fl6_sport = inet->inet_sport; 721 fl6.flowi6_uid = sk->sk_uid; 722 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 723 724 rcu_read_lock(); 725 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 726 &final); 727 rcu_read_unlock(); 728 729 dst = ip6_dst_lookup_flow(sk, &fl6, final_p); 730 if (IS_ERR(dst)) { 731 sk->sk_route_caps = 0; 732 sk->sk_err_soft = -PTR_ERR(dst); 733 return PTR_ERR(dst); 734 } 735 736 ip6_dst_store(sk, dst, NULL, NULL); 737 } 738 739 return 0; 740 } 741 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 742 743 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 744 const struct inet6_skb_parm *opt) 745 { 746 const struct ipv6_pinfo *np = inet6_sk(sk); 747 748 if (np->rxopt.all) { 749 if (((opt->flags & IP6SKB_HOPBYHOP) && 750 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 751 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 752 np->rxopt.bits.rxflow) || 753 (opt->srcrt && (np->rxopt.bits.srcrt || 754 np->rxopt.bits.osrcrt)) || 755 ((opt->dst1 || opt->dst0) && 756 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 757 return true; 758 } 759 return false; 760 } 761 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 762 763 static struct packet_type ipv6_packet_type __read_mostly = { 764 .type = cpu_to_be16(ETH_P_IPV6), 765 .func = ipv6_rcv, 766 .list_func = ipv6_list_rcv, 767 }; 768 769 static int __init ipv6_packet_init(void) 770 { 771 dev_add_pack(&ipv6_packet_type); 772 return 0; 773 } 774 775 static void ipv6_packet_cleanup(void) 776 { 777 dev_remove_pack(&ipv6_packet_type); 778 } 779 780 static int __net_init ipv6_init_mibs(struct net *net) 781 { 782 int i; 783 784 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 785 if (!net->mib.udp_stats_in6) 786 return -ENOMEM; 787 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 788 if (!net->mib.udplite_stats_in6) 789 goto err_udplite_mib; 790 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 791 if (!net->mib.ipv6_statistics) 792 goto err_ip_mib; 793 794 for_each_possible_cpu(i) { 795 struct ipstats_mib *af_inet6_stats; 796 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 797 u64_stats_init(&af_inet6_stats->syncp); 798 } 799 800 801 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 802 if (!net->mib.icmpv6_statistics) 803 goto err_icmp_mib; 804 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 805 GFP_KERNEL); 806 if (!net->mib.icmpv6msg_statistics) 807 goto err_icmpmsg_mib; 808 return 0; 809 810 err_icmpmsg_mib: 811 free_percpu(net->mib.icmpv6_statistics); 812 err_icmp_mib: 813 free_percpu(net->mib.ipv6_statistics); 814 err_ip_mib: 815 free_percpu(net->mib.udplite_stats_in6); 816 err_udplite_mib: 817 free_percpu(net->mib.udp_stats_in6); 818 return -ENOMEM; 819 } 820 821 static void ipv6_cleanup_mibs(struct net *net) 822 { 823 free_percpu(net->mib.udp_stats_in6); 824 free_percpu(net->mib.udplite_stats_in6); 825 free_percpu(net->mib.ipv6_statistics); 826 free_percpu(net->mib.icmpv6_statistics); 827 kfree(net->mib.icmpv6msg_statistics); 828 } 829 830 static int __net_init inet6_net_init(struct net *net) 831 { 832 int err = 0; 833 834 net->ipv6.sysctl.bindv6only = 0; 835 net->ipv6.sysctl.icmpv6_time = 1*HZ; 836 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 837 net->ipv6.sysctl.flowlabel_consistency = 1; 838 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 839 net->ipv6.sysctl.idgen_retries = 3; 840 net->ipv6.sysctl.idgen_delay = 1 * HZ; 841 net->ipv6.sysctl.flowlabel_state_ranges = 0; 842 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 843 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 844 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 845 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 846 atomic_set(&net->ipv6.fib6_sernum, 1); 847 848 err = ipv6_init_mibs(net); 849 if (err) 850 return err; 851 #ifdef CONFIG_PROC_FS 852 err = udp6_proc_init(net); 853 if (err) 854 goto out; 855 err = tcp6_proc_init(net); 856 if (err) 857 goto proc_tcp6_fail; 858 err = ac6_proc_init(net); 859 if (err) 860 goto proc_ac6_fail; 861 #endif 862 return err; 863 864 #ifdef CONFIG_PROC_FS 865 proc_ac6_fail: 866 tcp6_proc_exit(net); 867 proc_tcp6_fail: 868 udp6_proc_exit(net); 869 out: 870 ipv6_cleanup_mibs(net); 871 return err; 872 #endif 873 } 874 875 static void __net_exit inet6_net_exit(struct net *net) 876 { 877 #ifdef CONFIG_PROC_FS 878 udp6_proc_exit(net); 879 tcp6_proc_exit(net); 880 ac6_proc_exit(net); 881 #endif 882 ipv6_cleanup_mibs(net); 883 } 884 885 static struct pernet_operations inet6_net_ops = { 886 .init = inet6_net_init, 887 .exit = inet6_net_exit, 888 }; 889 890 static const struct ipv6_stub ipv6_stub_impl = { 891 .ipv6_sock_mc_join = ipv6_sock_mc_join, 892 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 893 .ipv6_dst_lookup = ip6_dst_lookup, 894 .fib6_get_table = fib6_get_table, 895 .fib6_table_lookup = fib6_table_lookup, 896 .fib6_lookup = fib6_lookup, 897 .fib6_multipath_select = fib6_multipath_select, 898 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 899 .udpv6_encap_enable = udpv6_encap_enable, 900 .ndisc_send_na = ndisc_send_na, 901 .nd_tbl = &nd_tbl, 902 }; 903 904 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 905 .inet6_bind = __inet6_bind, 906 }; 907 908 static int __init inet6_init(void) 909 { 910 struct list_head *r; 911 int err = 0; 912 913 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 914 915 /* Register the socket-side information for inet6_create. */ 916 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 917 INIT_LIST_HEAD(r); 918 919 if (disable_ipv6_mod) { 920 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 921 goto out; 922 } 923 924 err = proto_register(&tcpv6_prot, 1); 925 if (err) 926 goto out; 927 928 err = proto_register(&udpv6_prot, 1); 929 if (err) 930 goto out_unregister_tcp_proto; 931 932 err = proto_register(&udplitev6_prot, 1); 933 if (err) 934 goto out_unregister_udp_proto; 935 936 err = proto_register(&rawv6_prot, 1); 937 if (err) 938 goto out_unregister_udplite_proto; 939 940 err = proto_register(&pingv6_prot, 1); 941 if (err) 942 goto out_unregister_raw_proto; 943 944 /* We MUST register RAW sockets before we create the ICMP6, 945 * IGMP6, or NDISC control sockets. 946 */ 947 err = rawv6_init(); 948 if (err) 949 goto out_unregister_ping_proto; 950 951 /* Register the family here so that the init calls below will 952 * be able to create sockets. (?? is this dangerous ??) 953 */ 954 err = sock_register(&inet6_family_ops); 955 if (err) 956 goto out_sock_register_fail; 957 958 /* 959 * ipngwg API draft makes clear that the correct semantics 960 * for TCP and UDP is to consider one TCP and UDP instance 961 * in a host available by both INET and INET6 APIs and 962 * able to communicate via both network protocols. 963 */ 964 965 err = register_pernet_subsys(&inet6_net_ops); 966 if (err) 967 goto register_pernet_fail; 968 err = ip6_mr_init(); 969 if (err) 970 goto ipmr_fail; 971 err = icmpv6_init(); 972 if (err) 973 goto icmp_fail; 974 err = ndisc_init(); 975 if (err) 976 goto ndisc_fail; 977 err = igmp6_init(); 978 if (err) 979 goto igmp_fail; 980 981 err = ipv6_netfilter_init(); 982 if (err) 983 goto netfilter_fail; 984 /* Create /proc/foo6 entries. */ 985 #ifdef CONFIG_PROC_FS 986 err = -ENOMEM; 987 if (raw6_proc_init()) 988 goto proc_raw6_fail; 989 if (udplite6_proc_init()) 990 goto proc_udplite6_fail; 991 if (ipv6_misc_proc_init()) 992 goto proc_misc6_fail; 993 if (if6_proc_init()) 994 goto proc_if6_fail; 995 #endif 996 err = ip6_route_init(); 997 if (err) 998 goto ip6_route_fail; 999 err = ndisc_late_init(); 1000 if (err) 1001 goto ndisc_late_fail; 1002 err = ip6_flowlabel_init(); 1003 if (err) 1004 goto ip6_flowlabel_fail; 1005 err = addrconf_init(); 1006 if (err) 1007 goto addrconf_fail; 1008 1009 /* Init v6 extension headers. */ 1010 err = ipv6_exthdrs_init(); 1011 if (err) 1012 goto ipv6_exthdrs_fail; 1013 1014 err = ipv6_frag_init(); 1015 if (err) 1016 goto ipv6_frag_fail; 1017 1018 /* Init v6 transport protocols. */ 1019 err = udpv6_init(); 1020 if (err) 1021 goto udpv6_fail; 1022 1023 err = udplitev6_init(); 1024 if (err) 1025 goto udplitev6_fail; 1026 1027 err = udpv6_offload_init(); 1028 if (err) 1029 goto udpv6_offload_fail; 1030 1031 err = tcpv6_init(); 1032 if (err) 1033 goto tcpv6_fail; 1034 1035 err = ipv6_packet_init(); 1036 if (err) 1037 goto ipv6_packet_fail; 1038 1039 err = pingv6_init(); 1040 if (err) 1041 goto pingv6_fail; 1042 1043 err = calipso_init(); 1044 if (err) 1045 goto calipso_fail; 1046 1047 err = seg6_init(); 1048 if (err) 1049 goto seg6_fail; 1050 1051 err = igmp6_late_init(); 1052 if (err) 1053 goto igmp6_late_err; 1054 1055 #ifdef CONFIG_SYSCTL 1056 err = ipv6_sysctl_register(); 1057 if (err) 1058 goto sysctl_fail; 1059 #endif 1060 1061 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1062 wmb(); 1063 ipv6_stub = &ipv6_stub_impl; 1064 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1065 out: 1066 return err; 1067 1068 #ifdef CONFIG_SYSCTL 1069 sysctl_fail: 1070 igmp6_late_cleanup(); 1071 #endif 1072 igmp6_late_err: 1073 seg6_exit(); 1074 seg6_fail: 1075 calipso_exit(); 1076 calipso_fail: 1077 pingv6_exit(); 1078 pingv6_fail: 1079 ipv6_packet_cleanup(); 1080 ipv6_packet_fail: 1081 tcpv6_exit(); 1082 tcpv6_fail: 1083 udpv6_offload_exit(); 1084 udpv6_offload_fail: 1085 udplitev6_exit(); 1086 udplitev6_fail: 1087 udpv6_exit(); 1088 udpv6_fail: 1089 ipv6_frag_exit(); 1090 ipv6_frag_fail: 1091 ipv6_exthdrs_exit(); 1092 ipv6_exthdrs_fail: 1093 addrconf_cleanup(); 1094 addrconf_fail: 1095 ip6_flowlabel_cleanup(); 1096 ip6_flowlabel_fail: 1097 ndisc_late_cleanup(); 1098 ndisc_late_fail: 1099 ip6_route_cleanup(); 1100 ip6_route_fail: 1101 #ifdef CONFIG_PROC_FS 1102 if6_proc_exit(); 1103 proc_if6_fail: 1104 ipv6_misc_proc_exit(); 1105 proc_misc6_fail: 1106 udplite6_proc_exit(); 1107 proc_udplite6_fail: 1108 raw6_proc_exit(); 1109 proc_raw6_fail: 1110 #endif 1111 ipv6_netfilter_fini(); 1112 netfilter_fail: 1113 igmp6_cleanup(); 1114 igmp_fail: 1115 ndisc_cleanup(); 1116 ndisc_fail: 1117 icmpv6_cleanup(); 1118 icmp_fail: 1119 ip6_mr_cleanup(); 1120 ipmr_fail: 1121 unregister_pernet_subsys(&inet6_net_ops); 1122 register_pernet_fail: 1123 sock_unregister(PF_INET6); 1124 rtnl_unregister_all(PF_INET6); 1125 out_sock_register_fail: 1126 rawv6_exit(); 1127 out_unregister_ping_proto: 1128 proto_unregister(&pingv6_prot); 1129 out_unregister_raw_proto: 1130 proto_unregister(&rawv6_prot); 1131 out_unregister_udplite_proto: 1132 proto_unregister(&udplitev6_prot); 1133 out_unregister_udp_proto: 1134 proto_unregister(&udpv6_prot); 1135 out_unregister_tcp_proto: 1136 proto_unregister(&tcpv6_prot); 1137 goto out; 1138 } 1139 module_init(inet6_init); 1140 1141 MODULE_ALIAS_NETPROTO(PF_INET6); 1142