1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * PF_INET6 socket protocol family 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * 9 * Adapted from linux/net/ipv4/af_inet.c 10 * 11 * Fixes: 12 * piggy, Karl Knutson : Socket protocol table 13 * Hideaki YOSHIFUJI : sin6_scope_id support 14 * Arnaldo Melo : check proc_net_create return, cleanups 15 */ 16 17 #define pr_fmt(fmt) "IPv6: " fmt 18 19 #include <linux/module.h> 20 #include <linux/capability.h> 21 #include <linux/errno.h> 22 #include <linux/types.h> 23 #include <linux/socket.h> 24 #include <linux/in.h> 25 #include <linux/kernel.h> 26 #include <linux/timer.h> 27 #include <linux/string.h> 28 #include <linux/sockios.h> 29 #include <linux/net.h> 30 #include <linux/fcntl.h> 31 #include <linux/mm.h> 32 #include <linux/interrupt.h> 33 #include <linux/proc_fs.h> 34 #include <linux/stat.h> 35 #include <linux/init.h> 36 #include <linux/slab.h> 37 38 #include <linux/inet.h> 39 #include <linux/netdevice.h> 40 #include <linux/icmpv6.h> 41 #include <linux/netfilter_ipv6.h> 42 43 #include <net/ip.h> 44 #include <net/ipv6.h> 45 #include <net/udp.h> 46 #include <net/udplite.h> 47 #include <net/tcp.h> 48 #include <net/ping.h> 49 #include <net/protocol.h> 50 #include <net/inet_common.h> 51 #include <net/route.h> 52 #include <net/transp_v6.h> 53 #include <net/ip6_route.h> 54 #include <net/addrconf.h> 55 #include <net/ipv6_stubs.h> 56 #include <net/ndisc.h> 57 #ifdef CONFIG_IPV6_TUNNEL 58 #include <net/ip6_tunnel.h> 59 #endif 60 #include <net/calipso.h> 61 #include <net/seg6.h> 62 #include <net/rpl.h> 63 64 #include <linux/uaccess.h> 65 #include <linux/mroute6.h> 66 67 #include "ip6_offload.h" 68 69 MODULE_AUTHOR("Cast of dozens"); 70 MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); 71 MODULE_LICENSE("GPL"); 72 73 /* The inetsw6 table contains everything that inet6_create needs to 74 * build a new socket. 75 */ 76 static struct list_head inetsw6[SOCK_MAX]; 77 static DEFINE_SPINLOCK(inetsw6_lock); 78 79 struct ipv6_params ipv6_defaults = { 80 .disable_ipv6 = 0, 81 .autoconf = 1, 82 }; 83 84 static int disable_ipv6_mod; 85 86 module_param_named(disable, disable_ipv6_mod, int, 0444); 87 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional"); 88 89 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444); 90 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces"); 91 92 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444); 93 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces"); 94 95 bool ipv6_mod_enabled(void) 96 { 97 return disable_ipv6_mod == 0; 98 } 99 EXPORT_SYMBOL_GPL(ipv6_mod_enabled); 100 101 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) 102 { 103 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); 104 105 return (struct ipv6_pinfo *)(((u8 *)sk) + offset); 106 } 107 108 static int inet6_create(struct net *net, struct socket *sock, int protocol, 109 int kern) 110 { 111 struct inet_sock *inet; 112 struct ipv6_pinfo *np; 113 struct sock *sk; 114 struct inet_protosw *answer; 115 struct proto *answer_prot; 116 unsigned char answer_flags; 117 int try_loading_module = 0; 118 int err; 119 120 if (protocol < 0 || protocol >= IPPROTO_MAX) 121 return -EINVAL; 122 123 /* Look for the requested type/protocol pair. */ 124 lookup_protocol: 125 err = -ESOCKTNOSUPPORT; 126 rcu_read_lock(); 127 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) { 128 129 err = 0; 130 /* Check the non-wild match. */ 131 if (protocol == answer->protocol) { 132 if (protocol != IPPROTO_IP) 133 break; 134 } else { 135 /* Check for the two wild cases. */ 136 if (IPPROTO_IP == protocol) { 137 protocol = answer->protocol; 138 break; 139 } 140 if (IPPROTO_IP == answer->protocol) 141 break; 142 } 143 err = -EPROTONOSUPPORT; 144 } 145 146 if (err) { 147 if (try_loading_module < 2) { 148 rcu_read_unlock(); 149 /* 150 * Be more specific, e.g. net-pf-10-proto-132-type-1 151 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) 152 */ 153 if (++try_loading_module == 1) 154 request_module("net-pf-%d-proto-%d-type-%d", 155 PF_INET6, protocol, sock->type); 156 /* 157 * Fall back to generic, e.g. net-pf-10-proto-132 158 * (net-pf-PF_INET6-proto-IPPROTO_SCTP) 159 */ 160 else 161 request_module("net-pf-%d-proto-%d", 162 PF_INET6, protocol); 163 goto lookup_protocol; 164 } else 165 goto out_rcu_unlock; 166 } 167 168 err = -EPERM; 169 if (sock->type == SOCK_RAW && !kern && 170 !ns_capable(net->user_ns, CAP_NET_RAW)) 171 goto out_rcu_unlock; 172 173 sock->ops = answer->ops; 174 answer_prot = answer->prot; 175 answer_flags = answer->flags; 176 rcu_read_unlock(); 177 178 WARN_ON(!answer_prot->slab); 179 180 err = -ENOBUFS; 181 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern); 182 if (!sk) 183 goto out; 184 185 sock_init_data(sock, sk); 186 187 err = 0; 188 if (INET_PROTOSW_REUSE & answer_flags) 189 sk->sk_reuse = SK_CAN_REUSE; 190 191 inet = inet_sk(sk); 192 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; 193 194 if (SOCK_RAW == sock->type) { 195 inet->inet_num = protocol; 196 if (IPPROTO_RAW == protocol) 197 inet->hdrincl = 1; 198 } 199 200 sk->sk_destruct = inet_sock_destruct; 201 sk->sk_family = PF_INET6; 202 sk->sk_protocol = protocol; 203 204 sk->sk_backlog_rcv = answer->prot->backlog_rcv; 205 206 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); 207 np->hop_limit = -1; 208 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS; 209 np->mc_loop = 1; 210 np->mc_all = 1; 211 np->pmtudisc = IPV6_PMTUDISC_WANT; 212 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED; 213 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only; 214 215 /* Init the ipv4 part of the socket since we can have sockets 216 * using v6 API for ipv4. 217 */ 218 inet->uc_ttl = -1; 219 220 inet->mc_loop = 1; 221 inet->mc_ttl = 1; 222 inet->mc_index = 0; 223 inet->mc_list = NULL; 224 inet->rcv_tos = 0; 225 226 if (net->ipv4.sysctl_ip_no_pmtu_disc) 227 inet->pmtudisc = IP_PMTUDISC_DONT; 228 else 229 inet->pmtudisc = IP_PMTUDISC_WANT; 230 /* 231 * Increment only the relevant sk_prot->socks debug field, this changes 232 * the previous behaviour of incrementing both the equivalent to 233 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr. 234 * 235 * This allows better debug granularity as we'll know exactly how many 236 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6 237 * transport protocol socks. -acme 238 */ 239 sk_refcnt_debug_inc(sk); 240 241 if (inet->inet_num) { 242 /* It assumes that any protocol which allows 243 * the user to assign a number at socket 244 * creation time automatically shares. 245 */ 246 inet->inet_sport = htons(inet->inet_num); 247 err = sk->sk_prot->hash(sk); 248 if (err) { 249 sk_common_release(sk); 250 goto out; 251 } 252 } 253 if (sk->sk_prot->init) { 254 err = sk->sk_prot->init(sk); 255 if (err) { 256 sk_common_release(sk); 257 goto out; 258 } 259 } 260 261 if (!kern) { 262 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); 263 if (err) { 264 sk_common_release(sk); 265 goto out; 266 } 267 } 268 out: 269 return err; 270 out_rcu_unlock: 271 rcu_read_unlock(); 272 goto out; 273 } 274 275 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, 276 u32 flags) 277 { 278 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr; 279 struct inet_sock *inet = inet_sk(sk); 280 struct ipv6_pinfo *np = inet6_sk(sk); 281 struct net *net = sock_net(sk); 282 __be32 v4addr = 0; 283 unsigned short snum; 284 bool saved_ipv6only; 285 int addr_type = 0; 286 int err = 0; 287 288 if (addr->sin6_family != AF_INET6) 289 return -EAFNOSUPPORT; 290 291 addr_type = ipv6_addr_type(&addr->sin6_addr); 292 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM) 293 return -EINVAL; 294 295 snum = ntohs(addr->sin6_port); 296 if (snum && inet_port_requires_bind_service(net, snum) && 297 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 298 return -EACCES; 299 300 if (flags & BIND_WITH_LOCK) 301 lock_sock(sk); 302 303 /* Check these errors (active socket, double bind). */ 304 if (sk->sk_state != TCP_CLOSE || inet->inet_num) { 305 err = -EINVAL; 306 goto out; 307 } 308 309 /* Check if the address belongs to the host. */ 310 if (addr_type == IPV6_ADDR_MAPPED) { 311 struct net_device *dev = NULL; 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 rcu_read_lock(); 323 if (sk->sk_bound_dev_if) { 324 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 325 if (!dev) { 326 err = -ENODEV; 327 goto out_unlock; 328 } 329 } 330 331 /* Reproduce AF_INET checks to make the bindings consistent */ 332 v4addr = addr->sin6_addr.s6_addr32[3]; 333 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr); 334 rcu_read_unlock(); 335 336 if (!inet_can_nonlocal_bind(net, inet) && 337 v4addr != htonl(INADDR_ANY) && 338 chk_addr_ret != RTN_LOCAL && 339 chk_addr_ret != RTN_MULTICAST && 340 chk_addr_ret != RTN_BROADCAST) { 341 err = -EADDRNOTAVAIL; 342 goto out; 343 } 344 } else { 345 if (addr_type != IPV6_ADDR_ANY) { 346 struct net_device *dev = NULL; 347 348 rcu_read_lock(); 349 if (__ipv6_addr_needs_scope_id(addr_type)) { 350 if (addr_len >= sizeof(struct sockaddr_in6) && 351 addr->sin6_scope_id) { 352 /* Override any existing binding, if another one 353 * is supplied by user. 354 */ 355 sk->sk_bound_dev_if = addr->sin6_scope_id; 356 } 357 358 /* Binding to link-local address requires an interface */ 359 if (!sk->sk_bound_dev_if) { 360 err = -EINVAL; 361 goto out_unlock; 362 } 363 } 364 365 if (sk->sk_bound_dev_if) { 366 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 367 if (!dev) { 368 err = -ENODEV; 369 goto out_unlock; 370 } 371 } 372 373 /* ipv4 addr of the socket is invalid. Only the 374 * unspecified and mapped address have a v4 equivalent. 375 */ 376 v4addr = LOOPBACK4_IPV6; 377 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 378 if (!ipv6_can_nonlocal_bind(net, inet) && 379 !ipv6_chk_addr(net, &addr->sin6_addr, 380 dev, 0)) { 381 err = -EADDRNOTAVAIL; 382 goto out_unlock; 383 } 384 } 385 rcu_read_unlock(); 386 } 387 } 388 389 inet->inet_rcv_saddr = v4addr; 390 inet->inet_saddr = v4addr; 391 392 sk->sk_v6_rcv_saddr = addr->sin6_addr; 393 394 if (!(addr_type & IPV6_ADDR_MULTICAST)) 395 np->saddr = addr->sin6_addr; 396 397 saved_ipv6only = sk->sk_ipv6only; 398 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED) 399 sk->sk_ipv6only = 1; 400 401 /* Make sure we are allowed to bind here. */ 402 if (snum || !(inet->bind_address_no_port || 403 (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 404 if (sk->sk_prot->get_port(sk, snum)) { 405 sk->sk_ipv6only = saved_ipv6only; 406 inet_reset_saddr(sk); 407 err = -EADDRINUSE; 408 goto out; 409 } 410 if (!(flags & BIND_FROM_BPF)) { 411 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk); 412 if (err) { 413 sk->sk_ipv6only = saved_ipv6only; 414 inet_reset_saddr(sk); 415 goto out; 416 } 417 } 418 } 419 420 if (addr_type != IPV6_ADDR_ANY) 421 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 422 if (snum) 423 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 424 inet->inet_sport = htons(inet->inet_num); 425 inet->inet_dport = 0; 426 inet->inet_daddr = 0; 427 out: 428 if (flags & BIND_WITH_LOCK) 429 release_sock(sk); 430 return err; 431 out_unlock: 432 rcu_read_unlock(); 433 goto out; 434 } 435 436 /* bind for INET6 API */ 437 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 438 { 439 struct sock *sk = sock->sk; 440 int err = 0; 441 442 /* If the socket has its own bind function then use it. */ 443 if (sk->sk_prot->bind) 444 return sk->sk_prot->bind(sk, uaddr, addr_len); 445 446 if (addr_len < SIN6_LEN_RFC2133) 447 return -EINVAL; 448 449 /* BPF prog is run before any checks are done so that if the prog 450 * changes context in a wrong way it will be caught. 451 */ 452 err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr); 453 if (err) 454 return err; 455 456 return __inet6_bind(sk, uaddr, addr_len, BIND_WITH_LOCK); 457 } 458 EXPORT_SYMBOL(inet6_bind); 459 460 int inet6_release(struct socket *sock) 461 { 462 struct sock *sk = sock->sk; 463 464 if (!sk) 465 return -EINVAL; 466 467 /* Free mc lists */ 468 ipv6_sock_mc_close(sk); 469 470 /* Free ac lists */ 471 ipv6_sock_ac_close(sk); 472 473 return inet_release(sock); 474 } 475 EXPORT_SYMBOL(inet6_release); 476 477 void inet6_destroy_sock(struct sock *sk) 478 { 479 struct ipv6_pinfo *np = inet6_sk(sk); 480 struct sk_buff *skb; 481 struct ipv6_txoptions *opt; 482 483 /* Release rx options */ 484 485 skb = xchg(&np->pktoptions, NULL); 486 kfree_skb(skb); 487 488 skb = xchg(&np->rxpmtu, NULL); 489 kfree_skb(skb); 490 491 /* Free flowlabels */ 492 fl6_free_socklist(sk); 493 494 /* Free tx options */ 495 496 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 497 if (opt) { 498 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 499 txopt_put(opt); 500 } 501 } 502 EXPORT_SYMBOL_GPL(inet6_destroy_sock); 503 504 /* 505 * This does both peername and sockname. 506 */ 507 508 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 509 int peer) 510 { 511 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 512 struct sock *sk = sock->sk; 513 struct inet_sock *inet = inet_sk(sk); 514 struct ipv6_pinfo *np = inet6_sk(sk); 515 516 sin->sin6_family = AF_INET6; 517 sin->sin6_flowinfo = 0; 518 sin->sin6_scope_id = 0; 519 if (peer) { 520 if (!inet->inet_dport) 521 return -ENOTCONN; 522 if (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 523 peer == 1) 524 return -ENOTCONN; 525 sin->sin6_port = inet->inet_dport; 526 sin->sin6_addr = sk->sk_v6_daddr; 527 if (np->sndflow) 528 sin->sin6_flowinfo = np->flow_label; 529 } else { 530 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 531 sin->sin6_addr = np->saddr; 532 else 533 sin->sin6_addr = sk->sk_v6_rcv_saddr; 534 535 sin->sin6_port = inet->inet_sport; 536 } 537 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 538 sk->sk_bound_dev_if); 539 return sizeof(*sin); 540 } 541 EXPORT_SYMBOL(inet6_getname); 542 543 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 544 { 545 struct sock *sk = sock->sk; 546 struct net *net = sock_net(sk); 547 548 switch (cmd) { 549 case SIOCADDRT: 550 case SIOCDELRT: 551 552 return ipv6_route_ioctl(net, cmd, (void __user *)arg); 553 554 case SIOCSIFADDR: 555 return addrconf_add_ifaddr(net, (void __user *) arg); 556 case SIOCDIFADDR: 557 return addrconf_del_ifaddr(net, (void __user *) arg); 558 case SIOCSIFDSTADDR: 559 return addrconf_set_dstaddr(net, (void __user *) arg); 560 default: 561 if (!sk->sk_prot->ioctl) 562 return -ENOIOCTLCMD; 563 return sk->sk_prot->ioctl(sk, cmd, arg); 564 } 565 /*NOTREACHED*/ 566 return 0; 567 } 568 EXPORT_SYMBOL(inet6_ioctl); 569 570 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 571 size_t)); 572 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 573 { 574 struct sock *sk = sock->sk; 575 576 if (unlikely(inet_send_prepare(sk))) 577 return -EAGAIN; 578 579 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 580 sk, msg, size); 581 } 582 583 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 584 size_t, int, int, int *)); 585 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 586 int flags) 587 { 588 struct sock *sk = sock->sk; 589 int addr_len = 0; 590 int err; 591 592 if (likely(!(flags & MSG_ERRQUEUE))) 593 sock_rps_record_flow(sk); 594 595 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 596 sk, msg, size, flags & MSG_DONTWAIT, 597 flags & ~MSG_DONTWAIT, &addr_len); 598 if (err >= 0) 599 msg->msg_namelen = addr_len; 600 return err; 601 } 602 603 const struct proto_ops inet6_stream_ops = { 604 .family = PF_INET6, 605 .owner = THIS_MODULE, 606 .release = inet6_release, 607 .bind = inet6_bind, 608 .connect = inet_stream_connect, /* ok */ 609 .socketpair = sock_no_socketpair, /* a do nothing */ 610 .accept = inet_accept, /* ok */ 611 .getname = inet6_getname, 612 .poll = tcp_poll, /* ok */ 613 .ioctl = inet6_ioctl, /* must change */ 614 .gettstamp = sock_gettstamp, 615 .listen = inet_listen, /* ok */ 616 .shutdown = inet_shutdown, /* ok */ 617 .setsockopt = sock_common_setsockopt, /* ok */ 618 .getsockopt = sock_common_getsockopt, /* ok */ 619 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 620 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 621 #ifdef CONFIG_MMU 622 .mmap = tcp_mmap, 623 #endif 624 .sendpage = inet_sendpage, 625 .sendmsg_locked = tcp_sendmsg_locked, 626 .sendpage_locked = tcp_sendpage_locked, 627 .splice_read = tcp_splice_read, 628 .read_sock = tcp_read_sock, 629 .peek_len = tcp_peek_len, 630 #ifdef CONFIG_COMPAT 631 .compat_setsockopt = compat_sock_common_setsockopt, 632 .compat_getsockopt = compat_sock_common_getsockopt, 633 #endif 634 .set_rcvlowat = tcp_set_rcvlowat, 635 }; 636 637 const struct proto_ops inet6_dgram_ops = { 638 .family = PF_INET6, 639 .owner = THIS_MODULE, 640 .release = inet6_release, 641 .bind = inet6_bind, 642 .connect = inet_dgram_connect, /* ok */ 643 .socketpair = sock_no_socketpair, /* a do nothing */ 644 .accept = sock_no_accept, /* a do nothing */ 645 .getname = inet6_getname, 646 .poll = udp_poll, /* ok */ 647 .ioctl = inet6_ioctl, /* must change */ 648 .gettstamp = sock_gettstamp, 649 .listen = sock_no_listen, /* ok */ 650 .shutdown = inet_shutdown, /* ok */ 651 .setsockopt = sock_common_setsockopt, /* ok */ 652 .getsockopt = sock_common_getsockopt, /* ok */ 653 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 654 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 655 .mmap = sock_no_mmap, 656 .sendpage = sock_no_sendpage, 657 .set_peek_off = sk_set_peek_off, 658 #ifdef CONFIG_COMPAT 659 .compat_setsockopt = compat_sock_common_setsockopt, 660 .compat_getsockopt = compat_sock_common_getsockopt, 661 #endif 662 }; 663 664 static const struct net_proto_family inet6_family_ops = { 665 .family = PF_INET6, 666 .create = inet6_create, 667 .owner = THIS_MODULE, 668 }; 669 670 int inet6_register_protosw(struct inet_protosw *p) 671 { 672 struct list_head *lh; 673 struct inet_protosw *answer; 674 struct list_head *last_perm; 675 int protocol = p->protocol; 676 int ret; 677 678 spin_lock_bh(&inetsw6_lock); 679 680 ret = -EINVAL; 681 if (p->type >= SOCK_MAX) 682 goto out_illegal; 683 684 /* If we are trying to override a permanent protocol, bail. */ 685 answer = NULL; 686 ret = -EPERM; 687 last_perm = &inetsw6[p->type]; 688 list_for_each(lh, &inetsw6[p->type]) { 689 answer = list_entry(lh, struct inet_protosw, list); 690 691 /* Check only the non-wild match. */ 692 if (INET_PROTOSW_PERMANENT & answer->flags) { 693 if (protocol == answer->protocol) 694 break; 695 last_perm = lh; 696 } 697 698 answer = NULL; 699 } 700 if (answer) 701 goto out_permanent; 702 703 /* Add the new entry after the last permanent entry if any, so that 704 * the new entry does not override a permanent entry when matched with 705 * a wild-card protocol. But it is allowed to override any existing 706 * non-permanent entry. This means that when we remove this entry, the 707 * system automatically returns to the old behavior. 708 */ 709 list_add_rcu(&p->list, last_perm); 710 ret = 0; 711 out: 712 spin_unlock_bh(&inetsw6_lock); 713 return ret; 714 715 out_permanent: 716 pr_err("Attempt to override permanent protocol %d\n", protocol); 717 goto out; 718 719 out_illegal: 720 pr_err("Ignoring attempt to register invalid socket type %d\n", 721 p->type); 722 goto out; 723 } 724 EXPORT_SYMBOL(inet6_register_protosw); 725 726 void 727 inet6_unregister_protosw(struct inet_protosw *p) 728 { 729 if (INET_PROTOSW_PERMANENT & p->flags) { 730 pr_err("Attempt to unregister permanent protocol %d\n", 731 p->protocol); 732 } else { 733 spin_lock_bh(&inetsw6_lock); 734 list_del_rcu(&p->list); 735 spin_unlock_bh(&inetsw6_lock); 736 737 synchronize_net(); 738 } 739 } 740 EXPORT_SYMBOL(inet6_unregister_protosw); 741 742 int inet6_sk_rebuild_header(struct sock *sk) 743 { 744 struct ipv6_pinfo *np = inet6_sk(sk); 745 struct dst_entry *dst; 746 747 dst = __sk_dst_check(sk, np->dst_cookie); 748 749 if (!dst) { 750 struct inet_sock *inet = inet_sk(sk); 751 struct in6_addr *final_p, final; 752 struct flowi6 fl6; 753 754 memset(&fl6, 0, sizeof(fl6)); 755 fl6.flowi6_proto = sk->sk_protocol; 756 fl6.daddr = sk->sk_v6_daddr; 757 fl6.saddr = np->saddr; 758 fl6.flowlabel = np->flow_label; 759 fl6.flowi6_oif = sk->sk_bound_dev_if; 760 fl6.flowi6_mark = sk->sk_mark; 761 fl6.fl6_dport = inet->inet_dport; 762 fl6.fl6_sport = inet->inet_sport; 763 fl6.flowi6_uid = sk->sk_uid; 764 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 765 766 rcu_read_lock(); 767 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 768 &final); 769 rcu_read_unlock(); 770 771 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 772 if (IS_ERR(dst)) { 773 sk->sk_route_caps = 0; 774 sk->sk_err_soft = -PTR_ERR(dst); 775 return PTR_ERR(dst); 776 } 777 778 ip6_dst_store(sk, dst, NULL, NULL); 779 } 780 781 return 0; 782 } 783 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 784 785 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 786 const struct inet6_skb_parm *opt) 787 { 788 const struct ipv6_pinfo *np = inet6_sk(sk); 789 790 if (np->rxopt.all) { 791 if (((opt->flags & IP6SKB_HOPBYHOP) && 792 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 793 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 794 np->rxopt.bits.rxflow) || 795 (opt->srcrt && (np->rxopt.bits.srcrt || 796 np->rxopt.bits.osrcrt)) || 797 ((opt->dst1 || opt->dst0) && 798 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 799 return true; 800 } 801 return false; 802 } 803 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 804 805 static struct packet_type ipv6_packet_type __read_mostly = { 806 .type = cpu_to_be16(ETH_P_IPV6), 807 .func = ipv6_rcv, 808 .list_func = ipv6_list_rcv, 809 }; 810 811 static int __init ipv6_packet_init(void) 812 { 813 dev_add_pack(&ipv6_packet_type); 814 return 0; 815 } 816 817 static void ipv6_packet_cleanup(void) 818 { 819 dev_remove_pack(&ipv6_packet_type); 820 } 821 822 static int __net_init ipv6_init_mibs(struct net *net) 823 { 824 int i; 825 826 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 827 if (!net->mib.udp_stats_in6) 828 return -ENOMEM; 829 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 830 if (!net->mib.udplite_stats_in6) 831 goto err_udplite_mib; 832 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 833 if (!net->mib.ipv6_statistics) 834 goto err_ip_mib; 835 836 for_each_possible_cpu(i) { 837 struct ipstats_mib *af_inet6_stats; 838 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 839 u64_stats_init(&af_inet6_stats->syncp); 840 } 841 842 843 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 844 if (!net->mib.icmpv6_statistics) 845 goto err_icmp_mib; 846 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 847 GFP_KERNEL); 848 if (!net->mib.icmpv6msg_statistics) 849 goto err_icmpmsg_mib; 850 return 0; 851 852 err_icmpmsg_mib: 853 free_percpu(net->mib.icmpv6_statistics); 854 err_icmp_mib: 855 free_percpu(net->mib.ipv6_statistics); 856 err_ip_mib: 857 free_percpu(net->mib.udplite_stats_in6); 858 err_udplite_mib: 859 free_percpu(net->mib.udp_stats_in6); 860 return -ENOMEM; 861 } 862 863 static void ipv6_cleanup_mibs(struct net *net) 864 { 865 free_percpu(net->mib.udp_stats_in6); 866 free_percpu(net->mib.udplite_stats_in6); 867 free_percpu(net->mib.ipv6_statistics); 868 free_percpu(net->mib.icmpv6_statistics); 869 kfree(net->mib.icmpv6msg_statistics); 870 } 871 872 static int __net_init inet6_net_init(struct net *net) 873 { 874 int err = 0; 875 876 net->ipv6.sysctl.bindv6only = 0; 877 net->ipv6.sysctl.icmpv6_time = 1*HZ; 878 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 879 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0; 880 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0; 881 882 /* By default, rate limit error messages. 883 * Except for pmtu discovery, it would break it. 884 * proc_do_large_bitmap needs pointer to the bitmap. 885 */ 886 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 887 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 888 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 889 890 net->ipv6.sysctl.flowlabel_consistency = 1; 891 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 892 net->ipv6.sysctl.idgen_retries = 3; 893 net->ipv6.sysctl.idgen_delay = 1 * HZ; 894 net->ipv6.sysctl.flowlabel_state_ranges = 0; 895 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 896 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 897 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 898 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 899 atomic_set(&net->ipv6.fib6_sernum, 1); 900 901 err = ipv6_init_mibs(net); 902 if (err) 903 return err; 904 #ifdef CONFIG_PROC_FS 905 err = udp6_proc_init(net); 906 if (err) 907 goto out; 908 err = tcp6_proc_init(net); 909 if (err) 910 goto proc_tcp6_fail; 911 err = ac6_proc_init(net); 912 if (err) 913 goto proc_ac6_fail; 914 #endif 915 return err; 916 917 #ifdef CONFIG_PROC_FS 918 proc_ac6_fail: 919 tcp6_proc_exit(net); 920 proc_tcp6_fail: 921 udp6_proc_exit(net); 922 out: 923 ipv6_cleanup_mibs(net); 924 return err; 925 #endif 926 } 927 928 static void __net_exit inet6_net_exit(struct net *net) 929 { 930 #ifdef CONFIG_PROC_FS 931 udp6_proc_exit(net); 932 tcp6_proc_exit(net); 933 ac6_proc_exit(net); 934 #endif 935 ipv6_cleanup_mibs(net); 936 } 937 938 static struct pernet_operations inet6_net_ops = { 939 .init = inet6_net_init, 940 .exit = inet6_net_exit, 941 }; 942 943 static int ipv6_route_input(struct sk_buff *skb) 944 { 945 ip6_route_input(skb); 946 return skb_dst(skb)->error; 947 } 948 949 static const struct ipv6_stub ipv6_stub_impl = { 950 .ipv6_sock_mc_join = ipv6_sock_mc_join, 951 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 952 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 953 .ipv6_route_input = ipv6_route_input, 954 .fib6_get_table = fib6_get_table, 955 .fib6_table_lookup = fib6_table_lookup, 956 .fib6_lookup = fib6_lookup, 957 .fib6_select_path = fib6_select_path, 958 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 959 .fib6_nh_init = fib6_nh_init, 960 .fib6_nh_release = fib6_nh_release, 961 .fib6_update_sernum = fib6_update_sernum_stub, 962 .fib6_rt_update = fib6_rt_update, 963 .ip6_del_rt = ip6_del_rt, 964 .udpv6_encap_enable = udpv6_encap_enable, 965 .ndisc_send_na = ndisc_send_na, 966 .nd_tbl = &nd_tbl, 967 }; 968 969 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 970 .inet6_bind = __inet6_bind, 971 .udp6_lib_lookup = __udp6_lib_lookup, 972 }; 973 974 static int __init inet6_init(void) 975 { 976 struct list_head *r; 977 int err = 0; 978 979 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 980 981 /* Register the socket-side information for inet6_create. */ 982 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 983 INIT_LIST_HEAD(r); 984 985 if (disable_ipv6_mod) { 986 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 987 goto out; 988 } 989 990 err = proto_register(&tcpv6_prot, 1); 991 if (err) 992 goto out; 993 994 err = proto_register(&udpv6_prot, 1); 995 if (err) 996 goto out_unregister_tcp_proto; 997 998 err = proto_register(&udplitev6_prot, 1); 999 if (err) 1000 goto out_unregister_udp_proto; 1001 1002 err = proto_register(&rawv6_prot, 1); 1003 if (err) 1004 goto out_unregister_udplite_proto; 1005 1006 err = proto_register(&pingv6_prot, 1); 1007 if (err) 1008 goto out_unregister_raw_proto; 1009 1010 /* We MUST register RAW sockets before we create the ICMP6, 1011 * IGMP6, or NDISC control sockets. 1012 */ 1013 err = rawv6_init(); 1014 if (err) 1015 goto out_unregister_ping_proto; 1016 1017 /* Register the family here so that the init calls below will 1018 * be able to create sockets. (?? is this dangerous ??) 1019 */ 1020 err = sock_register(&inet6_family_ops); 1021 if (err) 1022 goto out_sock_register_fail; 1023 1024 /* 1025 * ipngwg API draft makes clear that the correct semantics 1026 * for TCP and UDP is to consider one TCP and UDP instance 1027 * in a host available by both INET and INET6 APIs and 1028 * able to communicate via both network protocols. 1029 */ 1030 1031 err = register_pernet_subsys(&inet6_net_ops); 1032 if (err) 1033 goto register_pernet_fail; 1034 err = ip6_mr_init(); 1035 if (err) 1036 goto ipmr_fail; 1037 err = icmpv6_init(); 1038 if (err) 1039 goto icmp_fail; 1040 err = ndisc_init(); 1041 if (err) 1042 goto ndisc_fail; 1043 err = igmp6_init(); 1044 if (err) 1045 goto igmp_fail; 1046 1047 err = ipv6_netfilter_init(); 1048 if (err) 1049 goto netfilter_fail; 1050 /* Create /proc/foo6 entries. */ 1051 #ifdef CONFIG_PROC_FS 1052 err = -ENOMEM; 1053 if (raw6_proc_init()) 1054 goto proc_raw6_fail; 1055 if (udplite6_proc_init()) 1056 goto proc_udplite6_fail; 1057 if (ipv6_misc_proc_init()) 1058 goto proc_misc6_fail; 1059 if (if6_proc_init()) 1060 goto proc_if6_fail; 1061 #endif 1062 err = ip6_route_init(); 1063 if (err) 1064 goto ip6_route_fail; 1065 err = ndisc_late_init(); 1066 if (err) 1067 goto ndisc_late_fail; 1068 err = ip6_flowlabel_init(); 1069 if (err) 1070 goto ip6_flowlabel_fail; 1071 err = ipv6_anycast_init(); 1072 if (err) 1073 goto ipv6_anycast_fail; 1074 err = addrconf_init(); 1075 if (err) 1076 goto addrconf_fail; 1077 1078 /* Init v6 extension headers. */ 1079 err = ipv6_exthdrs_init(); 1080 if (err) 1081 goto ipv6_exthdrs_fail; 1082 1083 err = ipv6_frag_init(); 1084 if (err) 1085 goto ipv6_frag_fail; 1086 1087 /* Init v6 transport protocols. */ 1088 err = udpv6_init(); 1089 if (err) 1090 goto udpv6_fail; 1091 1092 err = udplitev6_init(); 1093 if (err) 1094 goto udplitev6_fail; 1095 1096 err = udpv6_offload_init(); 1097 if (err) 1098 goto udpv6_offload_fail; 1099 1100 err = tcpv6_init(); 1101 if (err) 1102 goto tcpv6_fail; 1103 1104 err = ipv6_packet_init(); 1105 if (err) 1106 goto ipv6_packet_fail; 1107 1108 err = pingv6_init(); 1109 if (err) 1110 goto pingv6_fail; 1111 1112 err = calipso_init(); 1113 if (err) 1114 goto calipso_fail; 1115 1116 err = seg6_init(); 1117 if (err) 1118 goto seg6_fail; 1119 1120 err = rpl_init(); 1121 if (err) 1122 goto rpl_fail; 1123 1124 err = igmp6_late_init(); 1125 if (err) 1126 goto igmp6_late_err; 1127 1128 #ifdef CONFIG_SYSCTL 1129 err = ipv6_sysctl_register(); 1130 if (err) 1131 goto sysctl_fail; 1132 #endif 1133 1134 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1135 wmb(); 1136 ipv6_stub = &ipv6_stub_impl; 1137 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1138 out: 1139 return err; 1140 1141 #ifdef CONFIG_SYSCTL 1142 sysctl_fail: 1143 igmp6_late_cleanup(); 1144 #endif 1145 igmp6_late_err: 1146 rpl_exit(); 1147 rpl_fail: 1148 seg6_exit(); 1149 seg6_fail: 1150 calipso_exit(); 1151 calipso_fail: 1152 pingv6_exit(); 1153 pingv6_fail: 1154 ipv6_packet_cleanup(); 1155 ipv6_packet_fail: 1156 tcpv6_exit(); 1157 tcpv6_fail: 1158 udpv6_offload_exit(); 1159 udpv6_offload_fail: 1160 udplitev6_exit(); 1161 udplitev6_fail: 1162 udpv6_exit(); 1163 udpv6_fail: 1164 ipv6_frag_exit(); 1165 ipv6_frag_fail: 1166 ipv6_exthdrs_exit(); 1167 ipv6_exthdrs_fail: 1168 addrconf_cleanup(); 1169 addrconf_fail: 1170 ipv6_anycast_cleanup(); 1171 ipv6_anycast_fail: 1172 ip6_flowlabel_cleanup(); 1173 ip6_flowlabel_fail: 1174 ndisc_late_cleanup(); 1175 ndisc_late_fail: 1176 ip6_route_cleanup(); 1177 ip6_route_fail: 1178 #ifdef CONFIG_PROC_FS 1179 if6_proc_exit(); 1180 proc_if6_fail: 1181 ipv6_misc_proc_exit(); 1182 proc_misc6_fail: 1183 udplite6_proc_exit(); 1184 proc_udplite6_fail: 1185 raw6_proc_exit(); 1186 proc_raw6_fail: 1187 #endif 1188 ipv6_netfilter_fini(); 1189 netfilter_fail: 1190 igmp6_cleanup(); 1191 igmp_fail: 1192 ndisc_cleanup(); 1193 ndisc_fail: 1194 icmpv6_cleanup(); 1195 icmp_fail: 1196 ip6_mr_cleanup(); 1197 ipmr_fail: 1198 unregister_pernet_subsys(&inet6_net_ops); 1199 register_pernet_fail: 1200 sock_unregister(PF_INET6); 1201 rtnl_unregister_all(PF_INET6); 1202 out_sock_register_fail: 1203 rawv6_exit(); 1204 out_unregister_ping_proto: 1205 proto_unregister(&pingv6_prot); 1206 out_unregister_raw_proto: 1207 proto_unregister(&rawv6_prot); 1208 out_unregister_udplite_proto: 1209 proto_unregister(&udplitev6_prot); 1210 out_unregister_udp_proto: 1211 proto_unregister(&udpv6_prot); 1212 out_unregister_tcp_proto: 1213 proto_unregister(&tcpv6_prot); 1214 goto out; 1215 } 1216 module_init(inet6_init); 1217 1218 MODULE_ALIAS_NETPROTO(PF_INET6); 1219