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 #include <net/compat.h> 64 65 #include <linux/uaccess.h> 66 #include <linux/mroute6.h> 67 68 #include "ip6_offload.h" 69 70 MODULE_AUTHOR("Cast of dozens"); 71 MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); 72 MODULE_LICENSE("GPL"); 73 74 /* The inetsw6 table contains everything that inet6_create needs to 75 * build a new socket. 76 */ 77 static struct list_head inetsw6[SOCK_MAX]; 78 static DEFINE_SPINLOCK(inetsw6_lock); 79 80 struct ipv6_params ipv6_defaults = { 81 .disable_ipv6 = 0, 82 .autoconf = 1, 83 }; 84 85 static int disable_ipv6_mod; 86 87 module_param_named(disable, disable_ipv6_mod, int, 0444); 88 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional"); 89 90 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444); 91 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces"); 92 93 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444); 94 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces"); 95 96 bool ipv6_mod_enabled(void) 97 { 98 return disable_ipv6_mod == 0; 99 } 100 EXPORT_SYMBOL_GPL(ipv6_mod_enabled); 101 102 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) 103 { 104 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); 105 106 return (struct ipv6_pinfo *)(((u8 *)sk) + offset); 107 } 108 109 static int inet6_create(struct net *net, struct socket *sock, int protocol, 110 int kern) 111 { 112 struct inet_sock *inet; 113 struct ipv6_pinfo *np; 114 struct sock *sk; 115 struct inet_protosw *answer; 116 struct proto *answer_prot; 117 unsigned char answer_flags; 118 int try_loading_module = 0; 119 int err; 120 121 if (protocol < 0 || protocol >= IPPROTO_MAX) 122 return -EINVAL; 123 124 /* Look for the requested type/protocol pair. */ 125 lookup_protocol: 126 err = -ESOCKTNOSUPPORT; 127 rcu_read_lock(); 128 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) { 129 130 err = 0; 131 /* Check the non-wild match. */ 132 if (protocol == answer->protocol) { 133 if (protocol != IPPROTO_IP) 134 break; 135 } else { 136 /* Check for the two wild cases. */ 137 if (IPPROTO_IP == protocol) { 138 protocol = answer->protocol; 139 break; 140 } 141 if (IPPROTO_IP == answer->protocol) 142 break; 143 } 144 err = -EPROTONOSUPPORT; 145 } 146 147 if (err) { 148 if (try_loading_module < 2) { 149 rcu_read_unlock(); 150 /* 151 * Be more specific, e.g. net-pf-10-proto-132-type-1 152 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) 153 */ 154 if (++try_loading_module == 1) 155 request_module("net-pf-%d-proto-%d-type-%d", 156 PF_INET6, protocol, sock->type); 157 /* 158 * Fall back to generic, e.g. net-pf-10-proto-132 159 * (net-pf-PF_INET6-proto-IPPROTO_SCTP) 160 */ 161 else 162 request_module("net-pf-%d-proto-%d", 163 PF_INET6, protocol); 164 goto lookup_protocol; 165 } else 166 goto out_rcu_unlock; 167 } 168 169 err = -EPERM; 170 if (sock->type == SOCK_RAW && !kern && 171 !ns_capable(net->user_ns, CAP_NET_RAW)) 172 goto out_rcu_unlock; 173 174 sock->ops = answer->ops; 175 answer_prot = answer->prot; 176 answer_flags = answer->flags; 177 rcu_read_unlock(); 178 179 WARN_ON(!answer_prot->slab); 180 181 err = -ENOBUFS; 182 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern); 183 if (!sk) 184 goto out; 185 186 sock_init_data(sock, sk); 187 188 err = 0; 189 if (INET_PROTOSW_REUSE & answer_flags) 190 sk->sk_reuse = SK_CAN_REUSE; 191 192 inet = inet_sk(sk); 193 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; 194 195 if (SOCK_RAW == sock->type) { 196 inet->inet_num = protocol; 197 if (IPPROTO_RAW == protocol) 198 inet->hdrincl = 1; 199 } 200 201 sk->sk_destruct = inet_sock_destruct; 202 sk->sk_family = PF_INET6; 203 sk->sk_protocol = protocol; 204 205 sk->sk_backlog_rcv = answer->prot->backlog_rcv; 206 207 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); 208 np->hop_limit = -1; 209 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS; 210 np->mc_loop = 1; 211 np->mc_all = 1; 212 np->pmtudisc = IPV6_PMTUDISC_WANT; 213 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED; 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 u32 flags) 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 && inet_port_requires_bind_service(net, snum) && 298 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 299 return -EACCES; 300 301 if (flags & BIND_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 struct net_device *dev = NULL; 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 rcu_read_lock(); 324 if (sk->sk_bound_dev_if) { 325 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 326 if (!dev) { 327 err = -ENODEV; 328 goto out_unlock; 329 } 330 } 331 332 /* Reproduce AF_INET checks to make the bindings consistent */ 333 v4addr = addr->sin6_addr.s6_addr32[3]; 334 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr); 335 rcu_read_unlock(); 336 337 if (!inet_can_nonlocal_bind(net, inet) && 338 v4addr != htonl(INADDR_ANY) && 339 chk_addr_ret != RTN_LOCAL && 340 chk_addr_ret != RTN_MULTICAST && 341 chk_addr_ret != RTN_BROADCAST) { 342 err = -EADDRNOTAVAIL; 343 goto out; 344 } 345 } else { 346 if (addr_type != IPV6_ADDR_ANY) { 347 struct net_device *dev = NULL; 348 349 rcu_read_lock(); 350 if (__ipv6_addr_needs_scope_id(addr_type)) { 351 if (addr_len >= sizeof(struct sockaddr_in6) && 352 addr->sin6_scope_id) { 353 /* Override any existing binding, if another one 354 * is supplied by user. 355 */ 356 sk->sk_bound_dev_if = addr->sin6_scope_id; 357 } 358 359 /* Binding to link-local address requires an interface */ 360 if (!sk->sk_bound_dev_if) { 361 err = -EINVAL; 362 goto out_unlock; 363 } 364 } 365 366 if (sk->sk_bound_dev_if) { 367 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 368 if (!dev) { 369 err = -ENODEV; 370 goto out_unlock; 371 } 372 } 373 374 /* ipv4 addr of the socket is invalid. Only the 375 * unspecified and mapped address have a v4 equivalent. 376 */ 377 v4addr = LOOPBACK4_IPV6; 378 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 379 if (!ipv6_can_nonlocal_bind(net, inet) && 380 !ipv6_chk_addr(net, &addr->sin6_addr, 381 dev, 0)) { 382 err = -EADDRNOTAVAIL; 383 goto out_unlock; 384 } 385 } 386 rcu_read_unlock(); 387 } 388 } 389 390 inet->inet_rcv_saddr = v4addr; 391 inet->inet_saddr = v4addr; 392 393 sk->sk_v6_rcv_saddr = addr->sin6_addr; 394 395 if (!(addr_type & IPV6_ADDR_MULTICAST)) 396 np->saddr = addr->sin6_addr; 397 398 saved_ipv6only = sk->sk_ipv6only; 399 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED) 400 sk->sk_ipv6only = 1; 401 402 /* Make sure we are allowed to bind here. */ 403 if (snum || !(inet->bind_address_no_port || 404 (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 405 if (sk->sk_prot->get_port(sk, snum)) { 406 sk->sk_ipv6only = saved_ipv6only; 407 inet_reset_saddr(sk); 408 err = -EADDRINUSE; 409 goto out; 410 } 411 if (!(flags & BIND_FROM_BPF)) { 412 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk); 413 if (err) { 414 sk->sk_ipv6only = saved_ipv6only; 415 inet_reset_saddr(sk); 416 goto out; 417 } 418 } 419 } 420 421 if (addr_type != IPV6_ADDR_ANY) 422 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 423 if (snum) 424 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 425 inet->inet_sport = htons(inet->inet_num); 426 inet->inet_dport = 0; 427 inet->inet_daddr = 0; 428 out: 429 if (flags & BIND_WITH_LOCK) 430 release_sock(sk); 431 return err; 432 out_unlock: 433 rcu_read_unlock(); 434 goto out; 435 } 436 437 /* bind for INET6 API */ 438 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 439 { 440 struct sock *sk = sock->sk; 441 int err = 0; 442 443 /* If the socket has its own bind function then use it. */ 444 if (sk->sk_prot->bind) 445 return sk->sk_prot->bind(sk, uaddr, addr_len); 446 447 if (addr_len < SIN6_LEN_RFC2133) 448 return -EINVAL; 449 450 /* BPF prog is run before any checks are done so that if the prog 451 * changes context in a wrong way it will be caught. 452 */ 453 err = BPF_CGROUP_RUN_PROG_INET6_BIND(sk, uaddr); 454 if (err) 455 return err; 456 457 return __inet6_bind(sk, uaddr, addr_len, BIND_WITH_LOCK); 458 } 459 EXPORT_SYMBOL(inet6_bind); 460 461 int inet6_release(struct socket *sock) 462 { 463 struct sock *sk = sock->sk; 464 465 if (!sk) 466 return -EINVAL; 467 468 /* Free mc lists */ 469 ipv6_sock_mc_close(sk); 470 471 /* Free ac lists */ 472 ipv6_sock_ac_close(sk); 473 474 return inet_release(sock); 475 } 476 EXPORT_SYMBOL(inet6_release); 477 478 void inet6_destroy_sock(struct sock *sk) 479 { 480 struct ipv6_pinfo *np = inet6_sk(sk); 481 struct sk_buff *skb; 482 struct ipv6_txoptions *opt; 483 484 /* Release rx options */ 485 486 skb = xchg(&np->pktoptions, NULL); 487 kfree_skb(skb); 488 489 skb = xchg(&np->rxpmtu, NULL); 490 kfree_skb(skb); 491 492 /* Free flowlabels */ 493 fl6_free_socklist(sk); 494 495 /* Free tx options */ 496 497 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 498 if (opt) { 499 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 500 txopt_put(opt); 501 } 502 } 503 EXPORT_SYMBOL_GPL(inet6_destroy_sock); 504 505 /* 506 * This does both peername and sockname. 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 sin->sin6_port = inet->inet_sport; 535 } 536 if (cgroup_bpf_enabled) 537 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, (struct sockaddr *)sin, 538 peer ? BPF_CGROUP_INET6_GETPEERNAME : 539 BPF_CGROUP_INET6_GETSOCKNAME, 540 NULL); 541 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 542 sk->sk_bound_dev_if); 543 return sizeof(*sin); 544 } 545 EXPORT_SYMBOL(inet6_getname); 546 547 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 548 { 549 void __user *argp = (void __user *)arg; 550 struct sock *sk = sock->sk; 551 struct net *net = sock_net(sk); 552 553 switch (cmd) { 554 case SIOCADDRT: 555 case SIOCDELRT: { 556 struct in6_rtmsg rtmsg; 557 558 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg))) 559 return -EFAULT; 560 return ipv6_route_ioctl(net, cmd, &rtmsg); 561 } 562 case SIOCSIFADDR: 563 return addrconf_add_ifaddr(net, argp); 564 case SIOCDIFADDR: 565 return addrconf_del_ifaddr(net, argp); 566 case SIOCSIFDSTADDR: 567 return addrconf_set_dstaddr(net, argp); 568 default: 569 if (!sk->sk_prot->ioctl) 570 return -ENOIOCTLCMD; 571 return sk->sk_prot->ioctl(sk, cmd, arg); 572 } 573 /*NOTREACHED*/ 574 return 0; 575 } 576 EXPORT_SYMBOL(inet6_ioctl); 577 578 #ifdef CONFIG_COMPAT 579 struct compat_in6_rtmsg { 580 struct in6_addr rtmsg_dst; 581 struct in6_addr rtmsg_src; 582 struct in6_addr rtmsg_gateway; 583 u32 rtmsg_type; 584 u16 rtmsg_dst_len; 585 u16 rtmsg_src_len; 586 u32 rtmsg_metric; 587 u32 rtmsg_info; 588 u32 rtmsg_flags; 589 s32 rtmsg_ifindex; 590 }; 591 592 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 593 struct compat_in6_rtmsg __user *ur) 594 { 595 struct in6_rtmsg rt; 596 597 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst, 598 3 * sizeof(struct in6_addr)) || 599 get_user(rt.rtmsg_type, &ur->rtmsg_type) || 600 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) || 601 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) || 602 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) || 603 get_user(rt.rtmsg_info, &ur->rtmsg_info) || 604 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) || 605 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex)) 606 return -EFAULT; 607 608 609 return ipv6_route_ioctl(sock_net(sk), cmd, &rt); 610 } 611 612 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 613 { 614 void __user *argp = compat_ptr(arg); 615 struct sock *sk = sock->sk; 616 617 switch (cmd) { 618 case SIOCADDRT: 619 case SIOCDELRT: 620 return inet6_compat_routing_ioctl(sk, cmd, argp); 621 default: 622 return -ENOIOCTLCMD; 623 } 624 } 625 EXPORT_SYMBOL_GPL(inet6_compat_ioctl); 626 #endif /* CONFIG_COMPAT */ 627 628 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 629 size_t)); 630 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 631 { 632 struct sock *sk = sock->sk; 633 634 if (unlikely(inet_send_prepare(sk))) 635 return -EAGAIN; 636 637 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 638 sk, msg, size); 639 } 640 641 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 642 size_t, int, int, int *)); 643 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 644 int flags) 645 { 646 struct sock *sk = sock->sk; 647 int addr_len = 0; 648 int err; 649 650 if (likely(!(flags & MSG_ERRQUEUE))) 651 sock_rps_record_flow(sk); 652 653 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 654 sk, msg, size, flags & MSG_DONTWAIT, 655 flags & ~MSG_DONTWAIT, &addr_len); 656 if (err >= 0) 657 msg->msg_namelen = addr_len; 658 return err; 659 } 660 661 const struct proto_ops inet6_stream_ops = { 662 .family = PF_INET6, 663 .owner = THIS_MODULE, 664 .release = inet6_release, 665 .bind = inet6_bind, 666 .connect = inet_stream_connect, /* ok */ 667 .socketpair = sock_no_socketpair, /* a do nothing */ 668 .accept = inet_accept, /* ok */ 669 .getname = inet6_getname, 670 .poll = tcp_poll, /* ok */ 671 .ioctl = inet6_ioctl, /* must change */ 672 .gettstamp = sock_gettstamp, 673 .listen = inet_listen, /* ok */ 674 .shutdown = inet_shutdown, /* ok */ 675 .setsockopt = sock_common_setsockopt, /* ok */ 676 .getsockopt = sock_common_getsockopt, /* ok */ 677 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 678 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 679 #ifdef CONFIG_MMU 680 .mmap = tcp_mmap, 681 #endif 682 .sendpage = inet_sendpage, 683 .sendmsg_locked = tcp_sendmsg_locked, 684 .sendpage_locked = tcp_sendpage_locked, 685 .splice_read = tcp_splice_read, 686 .read_sock = tcp_read_sock, 687 .peek_len = tcp_peek_len, 688 #ifdef CONFIG_COMPAT 689 .compat_ioctl = inet6_compat_ioctl, 690 .compat_setsockopt = compat_sock_common_setsockopt, 691 .compat_getsockopt = compat_sock_common_getsockopt, 692 #endif 693 .set_rcvlowat = tcp_set_rcvlowat, 694 }; 695 696 const struct proto_ops inet6_dgram_ops = { 697 .family = PF_INET6, 698 .owner = THIS_MODULE, 699 .release = inet6_release, 700 .bind = inet6_bind, 701 .connect = inet_dgram_connect, /* ok */ 702 .socketpair = sock_no_socketpair, /* a do nothing */ 703 .accept = sock_no_accept, /* a do nothing */ 704 .getname = inet6_getname, 705 .poll = udp_poll, /* ok */ 706 .ioctl = inet6_ioctl, /* must change */ 707 .gettstamp = sock_gettstamp, 708 .listen = sock_no_listen, /* ok */ 709 .shutdown = inet_shutdown, /* ok */ 710 .setsockopt = sock_common_setsockopt, /* ok */ 711 .getsockopt = sock_common_getsockopt, /* ok */ 712 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 713 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 714 .mmap = sock_no_mmap, 715 .sendpage = sock_no_sendpage, 716 .set_peek_off = sk_set_peek_off, 717 #ifdef CONFIG_COMPAT 718 .compat_ioctl = inet6_compat_ioctl, 719 .compat_setsockopt = compat_sock_common_setsockopt, 720 .compat_getsockopt = compat_sock_common_getsockopt, 721 #endif 722 }; 723 724 static const struct net_proto_family inet6_family_ops = { 725 .family = PF_INET6, 726 .create = inet6_create, 727 .owner = THIS_MODULE, 728 }; 729 730 int inet6_register_protosw(struct inet_protosw *p) 731 { 732 struct list_head *lh; 733 struct inet_protosw *answer; 734 struct list_head *last_perm; 735 int protocol = p->protocol; 736 int ret; 737 738 spin_lock_bh(&inetsw6_lock); 739 740 ret = -EINVAL; 741 if (p->type >= SOCK_MAX) 742 goto out_illegal; 743 744 /* If we are trying to override a permanent protocol, bail. */ 745 answer = NULL; 746 ret = -EPERM; 747 last_perm = &inetsw6[p->type]; 748 list_for_each(lh, &inetsw6[p->type]) { 749 answer = list_entry(lh, struct inet_protosw, list); 750 751 /* Check only the non-wild match. */ 752 if (INET_PROTOSW_PERMANENT & answer->flags) { 753 if (protocol == answer->protocol) 754 break; 755 last_perm = lh; 756 } 757 758 answer = NULL; 759 } 760 if (answer) 761 goto out_permanent; 762 763 /* Add the new entry after the last permanent entry if any, so that 764 * the new entry does not override a permanent entry when matched with 765 * a wild-card protocol. But it is allowed to override any existing 766 * non-permanent entry. This means that when we remove this entry, the 767 * system automatically returns to the old behavior. 768 */ 769 list_add_rcu(&p->list, last_perm); 770 ret = 0; 771 out: 772 spin_unlock_bh(&inetsw6_lock); 773 return ret; 774 775 out_permanent: 776 pr_err("Attempt to override permanent protocol %d\n", protocol); 777 goto out; 778 779 out_illegal: 780 pr_err("Ignoring attempt to register invalid socket type %d\n", 781 p->type); 782 goto out; 783 } 784 EXPORT_SYMBOL(inet6_register_protosw); 785 786 void 787 inet6_unregister_protosw(struct inet_protosw *p) 788 { 789 if (INET_PROTOSW_PERMANENT & p->flags) { 790 pr_err("Attempt to unregister permanent protocol %d\n", 791 p->protocol); 792 } else { 793 spin_lock_bh(&inetsw6_lock); 794 list_del_rcu(&p->list); 795 spin_unlock_bh(&inetsw6_lock); 796 797 synchronize_net(); 798 } 799 } 800 EXPORT_SYMBOL(inet6_unregister_protosw); 801 802 int inet6_sk_rebuild_header(struct sock *sk) 803 { 804 struct ipv6_pinfo *np = inet6_sk(sk); 805 struct dst_entry *dst; 806 807 dst = __sk_dst_check(sk, np->dst_cookie); 808 809 if (!dst) { 810 struct inet_sock *inet = inet_sk(sk); 811 struct in6_addr *final_p, final; 812 struct flowi6 fl6; 813 814 memset(&fl6, 0, sizeof(fl6)); 815 fl6.flowi6_proto = sk->sk_protocol; 816 fl6.daddr = sk->sk_v6_daddr; 817 fl6.saddr = np->saddr; 818 fl6.flowlabel = np->flow_label; 819 fl6.flowi6_oif = sk->sk_bound_dev_if; 820 fl6.flowi6_mark = sk->sk_mark; 821 fl6.fl6_dport = inet->inet_dport; 822 fl6.fl6_sport = inet->inet_sport; 823 fl6.flowi6_uid = sk->sk_uid; 824 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 825 826 rcu_read_lock(); 827 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 828 &final); 829 rcu_read_unlock(); 830 831 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 832 if (IS_ERR(dst)) { 833 sk->sk_route_caps = 0; 834 sk->sk_err_soft = -PTR_ERR(dst); 835 return PTR_ERR(dst); 836 } 837 838 ip6_dst_store(sk, dst, NULL, NULL); 839 } 840 841 return 0; 842 } 843 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 844 845 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 846 const struct inet6_skb_parm *opt) 847 { 848 const struct ipv6_pinfo *np = inet6_sk(sk); 849 850 if (np->rxopt.all) { 851 if (((opt->flags & IP6SKB_HOPBYHOP) && 852 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 853 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 854 np->rxopt.bits.rxflow) || 855 (opt->srcrt && (np->rxopt.bits.srcrt || 856 np->rxopt.bits.osrcrt)) || 857 ((opt->dst1 || opt->dst0) && 858 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 859 return true; 860 } 861 return false; 862 } 863 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 864 865 static struct packet_type ipv6_packet_type __read_mostly = { 866 .type = cpu_to_be16(ETH_P_IPV6), 867 .func = ipv6_rcv, 868 .list_func = ipv6_list_rcv, 869 }; 870 871 static int __init ipv6_packet_init(void) 872 { 873 dev_add_pack(&ipv6_packet_type); 874 return 0; 875 } 876 877 static void ipv6_packet_cleanup(void) 878 { 879 dev_remove_pack(&ipv6_packet_type); 880 } 881 882 static int __net_init ipv6_init_mibs(struct net *net) 883 { 884 int i; 885 886 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 887 if (!net->mib.udp_stats_in6) 888 return -ENOMEM; 889 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 890 if (!net->mib.udplite_stats_in6) 891 goto err_udplite_mib; 892 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 893 if (!net->mib.ipv6_statistics) 894 goto err_ip_mib; 895 896 for_each_possible_cpu(i) { 897 struct ipstats_mib *af_inet6_stats; 898 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 899 u64_stats_init(&af_inet6_stats->syncp); 900 } 901 902 903 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 904 if (!net->mib.icmpv6_statistics) 905 goto err_icmp_mib; 906 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 907 GFP_KERNEL); 908 if (!net->mib.icmpv6msg_statistics) 909 goto err_icmpmsg_mib; 910 return 0; 911 912 err_icmpmsg_mib: 913 free_percpu(net->mib.icmpv6_statistics); 914 err_icmp_mib: 915 free_percpu(net->mib.ipv6_statistics); 916 err_ip_mib: 917 free_percpu(net->mib.udplite_stats_in6); 918 err_udplite_mib: 919 free_percpu(net->mib.udp_stats_in6); 920 return -ENOMEM; 921 } 922 923 static void ipv6_cleanup_mibs(struct net *net) 924 { 925 free_percpu(net->mib.udp_stats_in6); 926 free_percpu(net->mib.udplite_stats_in6); 927 free_percpu(net->mib.ipv6_statistics); 928 free_percpu(net->mib.icmpv6_statistics); 929 kfree(net->mib.icmpv6msg_statistics); 930 } 931 932 static int __net_init inet6_net_init(struct net *net) 933 { 934 int err = 0; 935 936 net->ipv6.sysctl.bindv6only = 0; 937 net->ipv6.sysctl.icmpv6_time = 1*HZ; 938 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 939 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0; 940 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0; 941 942 /* By default, rate limit error messages. 943 * Except for pmtu discovery, it would break it. 944 * proc_do_large_bitmap needs pointer to the bitmap. 945 */ 946 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 947 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 948 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 949 950 net->ipv6.sysctl.flowlabel_consistency = 1; 951 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 952 net->ipv6.sysctl.idgen_retries = 3; 953 net->ipv6.sysctl.idgen_delay = 1 * HZ; 954 net->ipv6.sysctl.flowlabel_state_ranges = 0; 955 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 956 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 957 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 958 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 959 atomic_set(&net->ipv6.fib6_sernum, 1); 960 961 err = ipv6_init_mibs(net); 962 if (err) 963 return err; 964 #ifdef CONFIG_PROC_FS 965 err = udp6_proc_init(net); 966 if (err) 967 goto out; 968 err = tcp6_proc_init(net); 969 if (err) 970 goto proc_tcp6_fail; 971 err = ac6_proc_init(net); 972 if (err) 973 goto proc_ac6_fail; 974 #endif 975 return err; 976 977 #ifdef CONFIG_PROC_FS 978 proc_ac6_fail: 979 tcp6_proc_exit(net); 980 proc_tcp6_fail: 981 udp6_proc_exit(net); 982 out: 983 ipv6_cleanup_mibs(net); 984 return err; 985 #endif 986 } 987 988 static void __net_exit inet6_net_exit(struct net *net) 989 { 990 #ifdef CONFIG_PROC_FS 991 udp6_proc_exit(net); 992 tcp6_proc_exit(net); 993 ac6_proc_exit(net); 994 #endif 995 ipv6_cleanup_mibs(net); 996 } 997 998 static struct pernet_operations inet6_net_ops = { 999 .init = inet6_net_init, 1000 .exit = inet6_net_exit, 1001 }; 1002 1003 static int ipv6_route_input(struct sk_buff *skb) 1004 { 1005 ip6_route_input(skb); 1006 return skb_dst(skb)->error; 1007 } 1008 1009 static const struct ipv6_stub ipv6_stub_impl = { 1010 .ipv6_sock_mc_join = ipv6_sock_mc_join, 1011 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 1012 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 1013 .ipv6_route_input = ipv6_route_input, 1014 .fib6_get_table = fib6_get_table, 1015 .fib6_table_lookup = fib6_table_lookup, 1016 .fib6_lookup = fib6_lookup, 1017 .fib6_select_path = fib6_select_path, 1018 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 1019 .fib6_nh_init = fib6_nh_init, 1020 .fib6_nh_release = fib6_nh_release, 1021 .fib6_update_sernum = fib6_update_sernum_stub, 1022 .fib6_rt_update = fib6_rt_update, 1023 .ip6_del_rt = ip6_del_rt, 1024 .udpv6_encap_enable = udpv6_encap_enable, 1025 .ndisc_send_na = ndisc_send_na, 1026 .nd_tbl = &nd_tbl, 1027 }; 1028 1029 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 1030 .inet6_bind = __inet6_bind, 1031 .udp6_lib_lookup = __udp6_lib_lookup, 1032 }; 1033 1034 static int __init inet6_init(void) 1035 { 1036 struct list_head *r; 1037 int err = 0; 1038 1039 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 1040 1041 /* Register the socket-side information for inet6_create. */ 1042 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 1043 INIT_LIST_HEAD(r); 1044 1045 if (disable_ipv6_mod) { 1046 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 1047 goto out; 1048 } 1049 1050 err = proto_register(&tcpv6_prot, 1); 1051 if (err) 1052 goto out; 1053 1054 err = proto_register(&udpv6_prot, 1); 1055 if (err) 1056 goto out_unregister_tcp_proto; 1057 1058 err = proto_register(&udplitev6_prot, 1); 1059 if (err) 1060 goto out_unregister_udp_proto; 1061 1062 err = proto_register(&rawv6_prot, 1); 1063 if (err) 1064 goto out_unregister_udplite_proto; 1065 1066 err = proto_register(&pingv6_prot, 1); 1067 if (err) 1068 goto out_unregister_raw_proto; 1069 1070 /* We MUST register RAW sockets before we create the ICMP6, 1071 * IGMP6, or NDISC control sockets. 1072 */ 1073 err = rawv6_init(); 1074 if (err) 1075 goto out_unregister_ping_proto; 1076 1077 /* Register the family here so that the init calls below will 1078 * be able to create sockets. (?? is this dangerous ??) 1079 */ 1080 err = sock_register(&inet6_family_ops); 1081 if (err) 1082 goto out_sock_register_fail; 1083 1084 /* 1085 * ipngwg API draft makes clear that the correct semantics 1086 * for TCP and UDP is to consider one TCP and UDP instance 1087 * in a host available by both INET and INET6 APIs and 1088 * able to communicate via both network protocols. 1089 */ 1090 1091 err = register_pernet_subsys(&inet6_net_ops); 1092 if (err) 1093 goto register_pernet_fail; 1094 err = ip6_mr_init(); 1095 if (err) 1096 goto ipmr_fail; 1097 err = icmpv6_init(); 1098 if (err) 1099 goto icmp_fail; 1100 err = ndisc_init(); 1101 if (err) 1102 goto ndisc_fail; 1103 err = igmp6_init(); 1104 if (err) 1105 goto igmp_fail; 1106 1107 err = ipv6_netfilter_init(); 1108 if (err) 1109 goto netfilter_fail; 1110 /* Create /proc/foo6 entries. */ 1111 #ifdef CONFIG_PROC_FS 1112 err = -ENOMEM; 1113 if (raw6_proc_init()) 1114 goto proc_raw6_fail; 1115 if (udplite6_proc_init()) 1116 goto proc_udplite6_fail; 1117 if (ipv6_misc_proc_init()) 1118 goto proc_misc6_fail; 1119 if (if6_proc_init()) 1120 goto proc_if6_fail; 1121 #endif 1122 err = ip6_route_init(); 1123 if (err) 1124 goto ip6_route_fail; 1125 err = ndisc_late_init(); 1126 if (err) 1127 goto ndisc_late_fail; 1128 err = ip6_flowlabel_init(); 1129 if (err) 1130 goto ip6_flowlabel_fail; 1131 err = ipv6_anycast_init(); 1132 if (err) 1133 goto ipv6_anycast_fail; 1134 err = addrconf_init(); 1135 if (err) 1136 goto addrconf_fail; 1137 1138 /* Init v6 extension headers. */ 1139 err = ipv6_exthdrs_init(); 1140 if (err) 1141 goto ipv6_exthdrs_fail; 1142 1143 err = ipv6_frag_init(); 1144 if (err) 1145 goto ipv6_frag_fail; 1146 1147 /* Init v6 transport protocols. */ 1148 err = udpv6_init(); 1149 if (err) 1150 goto udpv6_fail; 1151 1152 err = udplitev6_init(); 1153 if (err) 1154 goto udplitev6_fail; 1155 1156 err = udpv6_offload_init(); 1157 if (err) 1158 goto udpv6_offload_fail; 1159 1160 err = tcpv6_init(); 1161 if (err) 1162 goto tcpv6_fail; 1163 1164 err = ipv6_packet_init(); 1165 if (err) 1166 goto ipv6_packet_fail; 1167 1168 err = pingv6_init(); 1169 if (err) 1170 goto pingv6_fail; 1171 1172 err = calipso_init(); 1173 if (err) 1174 goto calipso_fail; 1175 1176 err = seg6_init(); 1177 if (err) 1178 goto seg6_fail; 1179 1180 err = rpl_init(); 1181 if (err) 1182 goto rpl_fail; 1183 1184 err = igmp6_late_init(); 1185 if (err) 1186 goto igmp6_late_err; 1187 1188 #ifdef CONFIG_SYSCTL 1189 err = ipv6_sysctl_register(); 1190 if (err) 1191 goto sysctl_fail; 1192 #endif 1193 1194 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1195 wmb(); 1196 ipv6_stub = &ipv6_stub_impl; 1197 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1198 out: 1199 return err; 1200 1201 #ifdef CONFIG_SYSCTL 1202 sysctl_fail: 1203 igmp6_late_cleanup(); 1204 #endif 1205 igmp6_late_err: 1206 rpl_exit(); 1207 rpl_fail: 1208 seg6_exit(); 1209 seg6_fail: 1210 calipso_exit(); 1211 calipso_fail: 1212 pingv6_exit(); 1213 pingv6_fail: 1214 ipv6_packet_cleanup(); 1215 ipv6_packet_fail: 1216 tcpv6_exit(); 1217 tcpv6_fail: 1218 udpv6_offload_exit(); 1219 udpv6_offload_fail: 1220 udplitev6_exit(); 1221 udplitev6_fail: 1222 udpv6_exit(); 1223 udpv6_fail: 1224 ipv6_frag_exit(); 1225 ipv6_frag_fail: 1226 ipv6_exthdrs_exit(); 1227 ipv6_exthdrs_fail: 1228 addrconf_cleanup(); 1229 addrconf_fail: 1230 ipv6_anycast_cleanup(); 1231 ipv6_anycast_fail: 1232 ip6_flowlabel_cleanup(); 1233 ip6_flowlabel_fail: 1234 ndisc_late_cleanup(); 1235 ndisc_late_fail: 1236 ip6_route_cleanup(); 1237 ip6_route_fail: 1238 #ifdef CONFIG_PROC_FS 1239 if6_proc_exit(); 1240 proc_if6_fail: 1241 ipv6_misc_proc_exit(); 1242 proc_misc6_fail: 1243 udplite6_proc_exit(); 1244 proc_udplite6_fail: 1245 raw6_proc_exit(); 1246 proc_raw6_fail: 1247 #endif 1248 ipv6_netfilter_fini(); 1249 netfilter_fail: 1250 igmp6_cleanup(); 1251 igmp_fail: 1252 ndisc_cleanup(); 1253 ndisc_fail: 1254 icmpv6_cleanup(); 1255 icmp_fail: 1256 ip6_mr_cleanup(); 1257 ipmr_fail: 1258 unregister_pernet_subsys(&inet6_net_ops); 1259 register_pernet_fail: 1260 sock_unregister(PF_INET6); 1261 rtnl_unregister_all(PF_INET6); 1262 out_sock_register_fail: 1263 rawv6_exit(); 1264 out_unregister_ping_proto: 1265 proto_unregister(&pingv6_prot); 1266 out_unregister_raw_proto: 1267 proto_unregister(&rawv6_prot); 1268 out_unregister_udplite_proto: 1269 proto_unregister(&udplitev6_prot); 1270 out_unregister_udp_proto: 1271 proto_unregister(&udpv6_prot); 1272 out_unregister_tcp_proto: 1273 proto_unregister(&tcpv6_prot); 1274 goto out; 1275 } 1276 module_init(inet6_init); 1277 1278 MODULE_ALIAS_NETPROTO(PF_INET6); 1279