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->autoflowlabel = ip6_default_np_autolabel(net); 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 } 276 277 278 /* bind for INET6 API */ 279 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 280 { 281 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr; 282 struct sock *sk = sock->sk; 283 struct inet_sock *inet = inet_sk(sk); 284 struct ipv6_pinfo *np = inet6_sk(sk); 285 struct net *net = sock_net(sk); 286 __be32 v4addr = 0; 287 unsigned short snum; 288 int addr_type = 0; 289 int err = 0; 290 291 /* If the socket has its own bind function then use it. */ 292 if (sk->sk_prot->bind) 293 return sk->sk_prot->bind(sk, uaddr, addr_len); 294 295 if (addr_len < SIN6_LEN_RFC2133) 296 return -EINVAL; 297 298 if (addr->sin6_family != AF_INET6) 299 return -EAFNOSUPPORT; 300 301 addr_type = ipv6_addr_type(&addr->sin6_addr); 302 if ((addr_type & IPV6_ADDR_MULTICAST) && sock->type == SOCK_STREAM) 303 return -EINVAL; 304 305 snum = ntohs(addr->sin6_port); 306 if (snum && snum < inet_prot_sock(net) && 307 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 308 return -EACCES; 309 310 lock_sock(sk); 311 312 /* Check these errors (active socket, double bind). */ 313 if (sk->sk_state != TCP_CLOSE || inet->inet_num) { 314 err = -EINVAL; 315 goto out; 316 } 317 318 /* Check if the address belongs to the host. */ 319 if (addr_type == IPV6_ADDR_MAPPED) { 320 int chk_addr_ret; 321 322 /* Binding to v4-mapped address on a v6-only socket 323 * makes no sense 324 */ 325 if (sk->sk_ipv6only) { 326 err = -EINVAL; 327 goto out; 328 } 329 330 /* Reproduce AF_INET checks to make the bindings consistent */ 331 v4addr = addr->sin6_addr.s6_addr32[3]; 332 chk_addr_ret = inet_addr_type(net, v4addr); 333 if (!net->ipv4.sysctl_ip_nonlocal_bind && 334 !(inet->freebind || inet->transparent) && 335 v4addr != htonl(INADDR_ANY) && 336 chk_addr_ret != RTN_LOCAL && 337 chk_addr_ret != RTN_MULTICAST && 338 chk_addr_ret != RTN_BROADCAST) { 339 err = -EADDRNOTAVAIL; 340 goto out; 341 } 342 } else { 343 if (addr_type != IPV6_ADDR_ANY) { 344 struct net_device *dev = NULL; 345 346 rcu_read_lock(); 347 if (__ipv6_addr_needs_scope_id(addr_type)) { 348 if (addr_len >= sizeof(struct sockaddr_in6) && 349 addr->sin6_scope_id) { 350 /* Override any existing binding, if another one 351 * is supplied by user. 352 */ 353 sk->sk_bound_dev_if = addr->sin6_scope_id; 354 } 355 356 /* Binding to link-local address requires an interface */ 357 if (!sk->sk_bound_dev_if) { 358 err = -EINVAL; 359 goto out_unlock; 360 } 361 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 362 if (!dev) { 363 err = -ENODEV; 364 goto out_unlock; 365 } 366 } 367 368 /* ipv4 addr of the socket is invalid. Only the 369 * unspecified and mapped address have a v4 equivalent. 370 */ 371 v4addr = LOOPBACK4_IPV6; 372 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 373 if (!net->ipv6.sysctl.ip_nonlocal_bind && 374 !(inet->freebind || inet->transparent) && 375 !ipv6_chk_addr(net, &addr->sin6_addr, 376 dev, 0)) { 377 err = -EADDRNOTAVAIL; 378 goto out_unlock; 379 } 380 } 381 rcu_read_unlock(); 382 } 383 } 384 385 inet->inet_rcv_saddr = v4addr; 386 inet->inet_saddr = v4addr; 387 388 sk->sk_v6_rcv_saddr = addr->sin6_addr; 389 390 if (!(addr_type & IPV6_ADDR_MULTICAST)) 391 np->saddr = addr->sin6_addr; 392 393 /* Make sure we are allowed to bind here. */ 394 if ((snum || !inet->bind_address_no_port) && 395 sk->sk_prot->get_port(sk, snum)) { 396 inet_reset_saddr(sk); 397 err = -EADDRINUSE; 398 goto out; 399 } 400 401 if (addr_type != IPV6_ADDR_ANY) { 402 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 403 if (addr_type != IPV6_ADDR_MAPPED) 404 sk->sk_ipv6only = 1; 405 } 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 release_sock(sk); 413 return err; 414 out_unlock: 415 rcu_read_unlock(); 416 goto out; 417 } 418 EXPORT_SYMBOL(inet6_bind); 419 420 int inet6_release(struct socket *sock) 421 { 422 struct sock *sk = sock->sk; 423 424 if (!sk) 425 return -EINVAL; 426 427 /* Free mc lists */ 428 ipv6_sock_mc_close(sk); 429 430 /* Free ac lists */ 431 ipv6_sock_ac_close(sk); 432 433 return inet_release(sock); 434 } 435 EXPORT_SYMBOL(inet6_release); 436 437 void inet6_destroy_sock(struct sock *sk) 438 { 439 struct ipv6_pinfo *np = inet6_sk(sk); 440 struct sk_buff *skb; 441 struct ipv6_txoptions *opt; 442 443 /* Release rx options */ 444 445 skb = xchg(&np->pktoptions, NULL); 446 if (skb) 447 kfree_skb(skb); 448 449 skb = xchg(&np->rxpmtu, NULL); 450 if (skb) 451 kfree_skb(skb); 452 453 /* Free flowlabels */ 454 fl6_free_socklist(sk); 455 456 /* Free tx options */ 457 458 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 459 if (opt) { 460 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 461 txopt_put(opt); 462 } 463 } 464 EXPORT_SYMBOL_GPL(inet6_destroy_sock); 465 466 /* 467 * This does both peername and sockname. 468 */ 469 470 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 471 int *uaddr_len, int peer) 472 { 473 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 474 struct sock *sk = sock->sk; 475 struct inet_sock *inet = inet_sk(sk); 476 struct ipv6_pinfo *np = inet6_sk(sk); 477 478 sin->sin6_family = AF_INET6; 479 sin->sin6_flowinfo = 0; 480 sin->sin6_scope_id = 0; 481 if (peer) { 482 if (!inet->inet_dport) 483 return -ENOTCONN; 484 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 485 peer == 1) 486 return -ENOTCONN; 487 sin->sin6_port = inet->inet_dport; 488 sin->sin6_addr = sk->sk_v6_daddr; 489 if (np->sndflow) 490 sin->sin6_flowinfo = np->flow_label; 491 } else { 492 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 493 sin->sin6_addr = np->saddr; 494 else 495 sin->sin6_addr = sk->sk_v6_rcv_saddr; 496 497 sin->sin6_port = inet->inet_sport; 498 } 499 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 500 sk->sk_bound_dev_if); 501 *uaddr_len = sizeof(*sin); 502 return 0; 503 } 504 EXPORT_SYMBOL(inet6_getname); 505 506 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 507 { 508 struct sock *sk = sock->sk; 509 struct net *net = sock_net(sk); 510 511 switch (cmd) { 512 case SIOCGSTAMP: 513 return sock_get_timestamp(sk, (struct timeval __user *)arg); 514 515 case SIOCGSTAMPNS: 516 return sock_get_timestampns(sk, (struct timespec __user *)arg); 517 518 case SIOCADDRT: 519 case SIOCDELRT: 520 521 return ipv6_route_ioctl(net, cmd, (void __user *)arg); 522 523 case SIOCSIFADDR: 524 return addrconf_add_ifaddr(net, (void __user *) arg); 525 case SIOCDIFADDR: 526 return addrconf_del_ifaddr(net, (void __user *) arg); 527 case SIOCSIFDSTADDR: 528 return addrconf_set_dstaddr(net, (void __user *) arg); 529 default: 530 if (!sk->sk_prot->ioctl) 531 return -ENOIOCTLCMD; 532 return sk->sk_prot->ioctl(sk, cmd, arg); 533 } 534 /*NOTREACHED*/ 535 return 0; 536 } 537 EXPORT_SYMBOL(inet6_ioctl); 538 539 const struct proto_ops inet6_stream_ops = { 540 .family = PF_INET6, 541 .owner = THIS_MODULE, 542 .release = inet6_release, 543 .bind = inet6_bind, 544 .connect = inet_stream_connect, /* ok */ 545 .socketpair = sock_no_socketpair, /* a do nothing */ 546 .accept = inet_accept, /* ok */ 547 .getname = inet6_getname, 548 .poll = tcp_poll, /* ok */ 549 .ioctl = inet6_ioctl, /* must change */ 550 .listen = inet_listen, /* ok */ 551 .shutdown = inet_shutdown, /* ok */ 552 .setsockopt = sock_common_setsockopt, /* ok */ 553 .getsockopt = sock_common_getsockopt, /* ok */ 554 .sendmsg = inet_sendmsg, /* ok */ 555 .recvmsg = inet_recvmsg, /* ok */ 556 .mmap = sock_no_mmap, 557 .sendpage = inet_sendpage, 558 .sendmsg_locked = tcp_sendmsg_locked, 559 .sendpage_locked = tcp_sendpage_locked, 560 .splice_read = tcp_splice_read, 561 .read_sock = tcp_read_sock, 562 .peek_len = tcp_peek_len, 563 #ifdef CONFIG_COMPAT 564 .compat_setsockopt = compat_sock_common_setsockopt, 565 .compat_getsockopt = compat_sock_common_getsockopt, 566 #endif 567 }; 568 569 const struct proto_ops inet6_dgram_ops = { 570 .family = PF_INET6, 571 .owner = THIS_MODULE, 572 .release = inet6_release, 573 .bind = inet6_bind, 574 .connect = inet_dgram_connect, /* ok */ 575 .socketpair = sock_no_socketpair, /* a do nothing */ 576 .accept = sock_no_accept, /* a do nothing */ 577 .getname = inet6_getname, 578 .poll = udp_poll, /* ok */ 579 .ioctl = inet6_ioctl, /* must change */ 580 .listen = sock_no_listen, /* ok */ 581 .shutdown = inet_shutdown, /* ok */ 582 .setsockopt = sock_common_setsockopt, /* ok */ 583 .getsockopt = sock_common_getsockopt, /* ok */ 584 .sendmsg = inet_sendmsg, /* ok */ 585 .recvmsg = inet_recvmsg, /* ok */ 586 .mmap = sock_no_mmap, 587 .sendpage = sock_no_sendpage, 588 .set_peek_off = sk_set_peek_off, 589 #ifdef CONFIG_COMPAT 590 .compat_setsockopt = compat_sock_common_setsockopt, 591 .compat_getsockopt = compat_sock_common_getsockopt, 592 #endif 593 }; 594 595 static const struct net_proto_family inet6_family_ops = { 596 .family = PF_INET6, 597 .create = inet6_create, 598 .owner = THIS_MODULE, 599 }; 600 601 int inet6_register_protosw(struct inet_protosw *p) 602 { 603 struct list_head *lh; 604 struct inet_protosw *answer; 605 struct list_head *last_perm; 606 int protocol = p->protocol; 607 int ret; 608 609 spin_lock_bh(&inetsw6_lock); 610 611 ret = -EINVAL; 612 if (p->type >= SOCK_MAX) 613 goto out_illegal; 614 615 /* If we are trying to override a permanent protocol, bail. */ 616 answer = NULL; 617 ret = -EPERM; 618 last_perm = &inetsw6[p->type]; 619 list_for_each(lh, &inetsw6[p->type]) { 620 answer = list_entry(lh, struct inet_protosw, list); 621 622 /* Check only the non-wild match. */ 623 if (INET_PROTOSW_PERMANENT & answer->flags) { 624 if (protocol == answer->protocol) 625 break; 626 last_perm = lh; 627 } 628 629 answer = NULL; 630 } 631 if (answer) 632 goto out_permanent; 633 634 /* Add the new entry after the last permanent entry if any, so that 635 * the new entry does not override a permanent entry when matched with 636 * a wild-card protocol. But it is allowed to override any existing 637 * non-permanent entry. This means that when we remove this entry, the 638 * system automatically returns to the old behavior. 639 */ 640 list_add_rcu(&p->list, last_perm); 641 ret = 0; 642 out: 643 spin_unlock_bh(&inetsw6_lock); 644 return ret; 645 646 out_permanent: 647 pr_err("Attempt to override permanent protocol %d\n", protocol); 648 goto out; 649 650 out_illegal: 651 pr_err("Ignoring attempt to register invalid socket type %d\n", 652 p->type); 653 goto out; 654 } 655 EXPORT_SYMBOL(inet6_register_protosw); 656 657 void 658 inet6_unregister_protosw(struct inet_protosw *p) 659 { 660 if (INET_PROTOSW_PERMANENT & p->flags) { 661 pr_err("Attempt to unregister permanent protocol %d\n", 662 p->protocol); 663 } else { 664 spin_lock_bh(&inetsw6_lock); 665 list_del_rcu(&p->list); 666 spin_unlock_bh(&inetsw6_lock); 667 668 synchronize_net(); 669 } 670 } 671 EXPORT_SYMBOL(inet6_unregister_protosw); 672 673 int inet6_sk_rebuild_header(struct sock *sk) 674 { 675 struct ipv6_pinfo *np = inet6_sk(sk); 676 struct dst_entry *dst; 677 678 dst = __sk_dst_check(sk, np->dst_cookie); 679 680 if (!dst) { 681 struct inet_sock *inet = inet_sk(sk); 682 struct in6_addr *final_p, final; 683 struct flowi6 fl6; 684 685 memset(&fl6, 0, sizeof(fl6)); 686 fl6.flowi6_proto = sk->sk_protocol; 687 fl6.daddr = sk->sk_v6_daddr; 688 fl6.saddr = np->saddr; 689 fl6.flowlabel = np->flow_label; 690 fl6.flowi6_oif = sk->sk_bound_dev_if; 691 fl6.flowi6_mark = sk->sk_mark; 692 fl6.fl6_dport = inet->inet_dport; 693 fl6.fl6_sport = inet->inet_sport; 694 fl6.flowi6_uid = sk->sk_uid; 695 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 696 697 rcu_read_lock(); 698 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 699 &final); 700 rcu_read_unlock(); 701 702 dst = ip6_dst_lookup_flow(sk, &fl6, final_p); 703 if (IS_ERR(dst)) { 704 sk->sk_route_caps = 0; 705 sk->sk_err_soft = -PTR_ERR(dst); 706 return PTR_ERR(dst); 707 } 708 709 ip6_dst_store(sk, dst, NULL, NULL); 710 } 711 712 return 0; 713 } 714 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 715 716 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 717 const struct inet6_skb_parm *opt) 718 { 719 const struct ipv6_pinfo *np = inet6_sk(sk); 720 721 if (np->rxopt.all) { 722 if (((opt->flags & IP6SKB_HOPBYHOP) && 723 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 724 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 725 np->rxopt.bits.rxflow) || 726 (opt->srcrt && (np->rxopt.bits.srcrt || 727 np->rxopt.bits.osrcrt)) || 728 ((opt->dst1 || opt->dst0) && 729 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 730 return true; 731 } 732 return false; 733 } 734 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 735 736 static struct packet_type ipv6_packet_type __read_mostly = { 737 .type = cpu_to_be16(ETH_P_IPV6), 738 .func = ipv6_rcv, 739 }; 740 741 static int __init ipv6_packet_init(void) 742 { 743 dev_add_pack(&ipv6_packet_type); 744 return 0; 745 } 746 747 static void ipv6_packet_cleanup(void) 748 { 749 dev_remove_pack(&ipv6_packet_type); 750 } 751 752 static int __net_init ipv6_init_mibs(struct net *net) 753 { 754 int i; 755 756 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 757 if (!net->mib.udp_stats_in6) 758 return -ENOMEM; 759 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 760 if (!net->mib.udplite_stats_in6) 761 goto err_udplite_mib; 762 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 763 if (!net->mib.ipv6_statistics) 764 goto err_ip_mib; 765 766 for_each_possible_cpu(i) { 767 struct ipstats_mib *af_inet6_stats; 768 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 769 u64_stats_init(&af_inet6_stats->syncp); 770 } 771 772 773 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 774 if (!net->mib.icmpv6_statistics) 775 goto err_icmp_mib; 776 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 777 GFP_KERNEL); 778 if (!net->mib.icmpv6msg_statistics) 779 goto err_icmpmsg_mib; 780 return 0; 781 782 err_icmpmsg_mib: 783 free_percpu(net->mib.icmpv6_statistics); 784 err_icmp_mib: 785 free_percpu(net->mib.ipv6_statistics); 786 err_ip_mib: 787 free_percpu(net->mib.udplite_stats_in6); 788 err_udplite_mib: 789 free_percpu(net->mib.udp_stats_in6); 790 return -ENOMEM; 791 } 792 793 static void ipv6_cleanup_mibs(struct net *net) 794 { 795 free_percpu(net->mib.udp_stats_in6); 796 free_percpu(net->mib.udplite_stats_in6); 797 free_percpu(net->mib.ipv6_statistics); 798 free_percpu(net->mib.icmpv6_statistics); 799 kfree(net->mib.icmpv6msg_statistics); 800 } 801 802 static int __net_init inet6_net_init(struct net *net) 803 { 804 int err = 0; 805 806 net->ipv6.sysctl.bindv6only = 0; 807 net->ipv6.sysctl.icmpv6_time = 1*HZ; 808 net->ipv6.sysctl.flowlabel_consistency = 1; 809 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 810 net->ipv6.sysctl.idgen_retries = 3; 811 net->ipv6.sysctl.idgen_delay = 1 * HZ; 812 net->ipv6.sysctl.flowlabel_state_ranges = 0; 813 atomic_set(&net->ipv6.fib6_sernum, 1); 814 815 err = ipv6_init_mibs(net); 816 if (err) 817 return err; 818 #ifdef CONFIG_PROC_FS 819 err = udp6_proc_init(net); 820 if (err) 821 goto out; 822 err = tcp6_proc_init(net); 823 if (err) 824 goto proc_tcp6_fail; 825 err = ac6_proc_init(net); 826 if (err) 827 goto proc_ac6_fail; 828 #endif 829 return err; 830 831 #ifdef CONFIG_PROC_FS 832 proc_ac6_fail: 833 tcp6_proc_exit(net); 834 proc_tcp6_fail: 835 udp6_proc_exit(net); 836 out: 837 ipv6_cleanup_mibs(net); 838 return err; 839 #endif 840 } 841 842 static void __net_exit inet6_net_exit(struct net *net) 843 { 844 #ifdef CONFIG_PROC_FS 845 udp6_proc_exit(net); 846 tcp6_proc_exit(net); 847 ac6_proc_exit(net); 848 #endif 849 ipv6_cleanup_mibs(net); 850 } 851 852 static struct pernet_operations inet6_net_ops = { 853 .init = inet6_net_init, 854 .exit = inet6_net_exit, 855 }; 856 857 static const struct ipv6_stub ipv6_stub_impl = { 858 .ipv6_sock_mc_join = ipv6_sock_mc_join, 859 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 860 .ipv6_dst_lookup = ip6_dst_lookup, 861 .udpv6_encap_enable = udpv6_encap_enable, 862 .ndisc_send_na = ndisc_send_na, 863 .nd_tbl = &nd_tbl, 864 }; 865 866 static int __init inet6_init(void) 867 { 868 struct list_head *r; 869 int err = 0; 870 871 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 872 873 /* Register the socket-side information for inet6_create. */ 874 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 875 INIT_LIST_HEAD(r); 876 877 if (disable_ipv6_mod) { 878 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 879 goto out; 880 } 881 882 err = proto_register(&tcpv6_prot, 1); 883 if (err) 884 goto out; 885 886 err = proto_register(&udpv6_prot, 1); 887 if (err) 888 goto out_unregister_tcp_proto; 889 890 err = proto_register(&udplitev6_prot, 1); 891 if (err) 892 goto out_unregister_udp_proto; 893 894 err = proto_register(&rawv6_prot, 1); 895 if (err) 896 goto out_unregister_udplite_proto; 897 898 err = proto_register(&pingv6_prot, 1); 899 if (err) 900 goto out_unregister_ping_proto; 901 902 /* We MUST register RAW sockets before we create the ICMP6, 903 * IGMP6, or NDISC control sockets. 904 */ 905 err = rawv6_init(); 906 if (err) 907 goto out_unregister_raw_proto; 908 909 /* Register the family here so that the init calls below will 910 * be able to create sockets. (?? is this dangerous ??) 911 */ 912 err = sock_register(&inet6_family_ops); 913 if (err) 914 goto out_sock_register_fail; 915 916 /* 917 * ipngwg API draft makes clear that the correct semantics 918 * for TCP and UDP is to consider one TCP and UDP instance 919 * in a host available by both INET and INET6 APIs and 920 * able to communicate via both network protocols. 921 */ 922 923 err = register_pernet_subsys(&inet6_net_ops); 924 if (err) 925 goto register_pernet_fail; 926 err = ip6_mr_init(); 927 if (err) 928 goto ipmr_fail; 929 err = icmpv6_init(); 930 if (err) 931 goto icmp_fail; 932 err = ndisc_init(); 933 if (err) 934 goto ndisc_fail; 935 err = igmp6_init(); 936 if (err) 937 goto igmp_fail; 938 939 err = ipv6_netfilter_init(); 940 if (err) 941 goto netfilter_fail; 942 /* Create /proc/foo6 entries. */ 943 #ifdef CONFIG_PROC_FS 944 err = -ENOMEM; 945 if (raw6_proc_init()) 946 goto proc_raw6_fail; 947 if (udplite6_proc_init()) 948 goto proc_udplite6_fail; 949 if (ipv6_misc_proc_init()) 950 goto proc_misc6_fail; 951 if (if6_proc_init()) 952 goto proc_if6_fail; 953 #endif 954 err = ip6_route_init(); 955 if (err) 956 goto ip6_route_fail; 957 err = ndisc_late_init(); 958 if (err) 959 goto ndisc_late_fail; 960 err = ip6_flowlabel_init(); 961 if (err) 962 goto ip6_flowlabel_fail; 963 err = addrconf_init(); 964 if (err) 965 goto addrconf_fail; 966 967 /* Init v6 extension headers. */ 968 err = ipv6_exthdrs_init(); 969 if (err) 970 goto ipv6_exthdrs_fail; 971 972 err = ipv6_frag_init(); 973 if (err) 974 goto ipv6_frag_fail; 975 976 /* Init v6 transport protocols. */ 977 err = udpv6_init(); 978 if (err) 979 goto udpv6_fail; 980 981 err = udplitev6_init(); 982 if (err) 983 goto udplitev6_fail; 984 985 err = udpv6_offload_init(); 986 if (err) 987 goto udpv6_offload_fail; 988 989 err = tcpv6_init(); 990 if (err) 991 goto tcpv6_fail; 992 993 err = ipv6_packet_init(); 994 if (err) 995 goto ipv6_packet_fail; 996 997 err = pingv6_init(); 998 if (err) 999 goto pingv6_fail; 1000 1001 err = calipso_init(); 1002 if (err) 1003 goto calipso_fail; 1004 1005 err = seg6_init(); 1006 if (err) 1007 goto seg6_fail; 1008 1009 err = igmp6_late_init(); 1010 if (err) 1011 goto igmp6_late_err; 1012 1013 #ifdef CONFIG_SYSCTL 1014 err = ipv6_sysctl_register(); 1015 if (err) 1016 goto sysctl_fail; 1017 #endif 1018 1019 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1020 wmb(); 1021 ipv6_stub = &ipv6_stub_impl; 1022 out: 1023 return err; 1024 1025 #ifdef CONFIG_SYSCTL 1026 sysctl_fail: 1027 igmp6_late_cleanup(); 1028 #endif 1029 igmp6_late_err: 1030 seg6_exit(); 1031 seg6_fail: 1032 calipso_exit(); 1033 calipso_fail: 1034 pingv6_exit(); 1035 pingv6_fail: 1036 ipv6_packet_cleanup(); 1037 ipv6_packet_fail: 1038 tcpv6_exit(); 1039 tcpv6_fail: 1040 udpv6_offload_exit(); 1041 udpv6_offload_fail: 1042 udplitev6_exit(); 1043 udplitev6_fail: 1044 udpv6_exit(); 1045 udpv6_fail: 1046 ipv6_frag_exit(); 1047 ipv6_frag_fail: 1048 ipv6_exthdrs_exit(); 1049 ipv6_exthdrs_fail: 1050 addrconf_cleanup(); 1051 addrconf_fail: 1052 ip6_flowlabel_cleanup(); 1053 ip6_flowlabel_fail: 1054 ndisc_late_cleanup(); 1055 ndisc_late_fail: 1056 ip6_route_cleanup(); 1057 ip6_route_fail: 1058 #ifdef CONFIG_PROC_FS 1059 if6_proc_exit(); 1060 proc_if6_fail: 1061 ipv6_misc_proc_exit(); 1062 proc_misc6_fail: 1063 udplite6_proc_exit(); 1064 proc_udplite6_fail: 1065 raw6_proc_exit(); 1066 proc_raw6_fail: 1067 #endif 1068 ipv6_netfilter_fini(); 1069 netfilter_fail: 1070 igmp6_cleanup(); 1071 igmp_fail: 1072 ndisc_cleanup(); 1073 ndisc_fail: 1074 ip6_mr_cleanup(); 1075 icmp_fail: 1076 unregister_pernet_subsys(&inet6_net_ops); 1077 ipmr_fail: 1078 icmpv6_cleanup(); 1079 register_pernet_fail: 1080 sock_unregister(PF_INET6); 1081 rtnl_unregister_all(PF_INET6); 1082 out_sock_register_fail: 1083 rawv6_exit(); 1084 out_unregister_ping_proto: 1085 proto_unregister(&pingv6_prot); 1086 out_unregister_raw_proto: 1087 proto_unregister(&rawv6_prot); 1088 out_unregister_udplite_proto: 1089 proto_unregister(&udplitev6_prot); 1090 out_unregister_udp_proto: 1091 proto_unregister(&udpv6_prot); 1092 out_unregister_tcp_proto: 1093 proto_unregister(&tcpv6_prot); 1094 goto out; 1095 } 1096 module_init(inet6_init); 1097 1098 MODULE_ALIAS_NETPROTO(PF_INET6); 1099