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