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 #include <net/xfrm.h> 65 #include <net/ioam6.h> 66 67 #include <linux/uaccess.h> 68 #include <linux/mroute6.h> 69 70 #include "ip6_offload.h" 71 72 MODULE_AUTHOR("Cast of dozens"); 73 MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); 74 MODULE_LICENSE("GPL"); 75 76 /* The inetsw6 table contains everything that inet6_create needs to 77 * build a new socket. 78 */ 79 static struct list_head inetsw6[SOCK_MAX]; 80 static DEFINE_SPINLOCK(inetsw6_lock); 81 82 struct ipv6_params ipv6_defaults = { 83 .disable_ipv6 = 0, 84 .autoconf = 1, 85 }; 86 87 static int disable_ipv6_mod; 88 89 module_param_named(disable, disable_ipv6_mod, int, 0444); 90 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional"); 91 92 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444); 93 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces"); 94 95 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444); 96 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces"); 97 98 bool ipv6_mod_enabled(void) 99 { 100 return disable_ipv6_mod == 0; 101 } 102 EXPORT_SYMBOL_GPL(ipv6_mod_enabled); 103 104 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) 105 { 106 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); 107 108 return (struct ipv6_pinfo *)(((u8 *)sk) + offset); 109 } 110 111 static int inet6_create(struct net *net, struct socket *sock, int protocol, 112 int kern) 113 { 114 struct inet_sock *inet; 115 struct ipv6_pinfo *np; 116 struct sock *sk; 117 struct inet_protosw *answer; 118 struct proto *answer_prot; 119 unsigned char answer_flags; 120 int try_loading_module = 0; 121 int err; 122 123 if (protocol < 0 || protocol >= IPPROTO_MAX) 124 return -EINVAL; 125 126 /* Look for the requested type/protocol pair. */ 127 lookup_protocol: 128 err = -ESOCKTNOSUPPORT; 129 rcu_read_lock(); 130 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) { 131 132 err = 0; 133 /* Check the non-wild match. */ 134 if (protocol == answer->protocol) { 135 if (protocol != IPPROTO_IP) 136 break; 137 } else { 138 /* Check for the two wild cases. */ 139 if (IPPROTO_IP == protocol) { 140 protocol = answer->protocol; 141 break; 142 } 143 if (IPPROTO_IP == answer->protocol) 144 break; 145 } 146 err = -EPROTONOSUPPORT; 147 } 148 149 if (err) { 150 if (try_loading_module < 2) { 151 rcu_read_unlock(); 152 /* 153 * Be more specific, e.g. net-pf-10-proto-132-type-1 154 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) 155 */ 156 if (++try_loading_module == 1) 157 request_module("net-pf-%d-proto-%d-type-%d", 158 PF_INET6, protocol, sock->type); 159 /* 160 * Fall back to generic, e.g. net-pf-10-proto-132 161 * (net-pf-PF_INET6-proto-IPPROTO_SCTP) 162 */ 163 else 164 request_module("net-pf-%d-proto-%d", 165 PF_INET6, protocol); 166 goto lookup_protocol; 167 } else 168 goto out_rcu_unlock; 169 } 170 171 err = -EPERM; 172 if (sock->type == SOCK_RAW && !kern && 173 !ns_capable(net->user_ns, CAP_NET_RAW)) 174 goto out_rcu_unlock; 175 176 sock->ops = answer->ops; 177 answer_prot = answer->prot; 178 answer_flags = answer->flags; 179 rcu_read_unlock(); 180 181 WARN_ON(!answer_prot->slab); 182 183 err = -ENOBUFS; 184 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern); 185 if (!sk) 186 goto out; 187 188 sock_init_data(sock, sk); 189 190 err = 0; 191 if (INET_PROTOSW_REUSE & answer_flags) 192 sk->sk_reuse = SK_CAN_REUSE; 193 194 inet = inet_sk(sk); 195 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; 196 197 if (SOCK_RAW == sock->type) { 198 inet->inet_num = protocol; 199 if (IPPROTO_RAW == protocol) 200 inet->hdrincl = 1; 201 } 202 203 sk->sk_destruct = inet_sock_destruct; 204 sk->sk_family = PF_INET6; 205 sk->sk_protocol = protocol; 206 207 sk->sk_backlog_rcv = answer->prot->backlog_rcv; 208 209 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); 210 np->hop_limit = -1; 211 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS; 212 np->mc_loop = 1; 213 np->mc_all = 1; 214 np->pmtudisc = IPV6_PMTUDISC_WANT; 215 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED; 216 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only; 217 218 /* Init the ipv4 part of the socket since we can have sockets 219 * using v6 API for ipv4. 220 */ 221 inet->uc_ttl = -1; 222 223 inet->mc_loop = 1; 224 inet->mc_ttl = 1; 225 inet->mc_index = 0; 226 RCU_INIT_POINTER(inet->mc_list, NULL); 227 inet->rcv_tos = 0; 228 229 if (net->ipv4.sysctl_ip_no_pmtu_disc) 230 inet->pmtudisc = IP_PMTUDISC_DONT; 231 else 232 inet->pmtudisc = IP_PMTUDISC_WANT; 233 /* 234 * Increment only the relevant sk_prot->socks debug field, this changes 235 * the previous behaviour of incrementing both the equivalent to 236 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr. 237 * 238 * This allows better debug granularity as we'll know exactly how many 239 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6 240 * transport protocol socks. -acme 241 */ 242 sk_refcnt_debug_inc(sk); 243 244 if (inet->inet_num) { 245 /* It assumes that any protocol which allows 246 * the user to assign a number at socket 247 * creation time automatically shares. 248 */ 249 inet->inet_sport = htons(inet->inet_num); 250 err = sk->sk_prot->hash(sk); 251 if (err) { 252 sk_common_release(sk); 253 goto out; 254 } 255 } 256 if (sk->sk_prot->init) { 257 err = sk->sk_prot->init(sk); 258 if (err) { 259 sk_common_release(sk); 260 goto out; 261 } 262 } 263 264 if (!kern) { 265 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); 266 if (err) { 267 sk_common_release(sk); 268 goto out; 269 } 270 } 271 out: 272 return err; 273 out_rcu_unlock: 274 rcu_read_unlock(); 275 goto out; 276 } 277 278 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, 279 u32 flags) 280 { 281 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr; 282 struct inet_sock *inet = inet_sk(sk); 283 struct ipv6_pinfo *np = inet6_sk(sk); 284 struct net *net = sock_net(sk); 285 __be32 v4addr = 0; 286 unsigned short snum; 287 bool saved_ipv6only; 288 int addr_type = 0; 289 int err = 0; 290 291 if (addr->sin6_family != AF_INET6) 292 return -EAFNOSUPPORT; 293 294 addr_type = ipv6_addr_type(&addr->sin6_addr); 295 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM) 296 return -EINVAL; 297 298 snum = ntohs(addr->sin6_port); 299 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) && 300 snum && inet_port_requires_bind_service(net, snum) && 301 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 302 return -EACCES; 303 304 if (flags & BIND_WITH_LOCK) 305 lock_sock(sk); 306 307 /* Check these errors (active socket, double bind). */ 308 if (sk->sk_state != TCP_CLOSE || inet->inet_num) { 309 err = -EINVAL; 310 goto out; 311 } 312 313 /* Check if the address belongs to the host. */ 314 if (addr_type == IPV6_ADDR_MAPPED) { 315 struct net_device *dev = NULL; 316 int chk_addr_ret; 317 318 /* Binding to v4-mapped address on a v6-only socket 319 * makes no sense 320 */ 321 if (sk->sk_ipv6only) { 322 err = -EINVAL; 323 goto out; 324 } 325 326 rcu_read_lock(); 327 if (sk->sk_bound_dev_if) { 328 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 329 if (!dev) { 330 err = -ENODEV; 331 goto out_unlock; 332 } 333 } 334 335 /* Reproduce AF_INET checks to make the bindings consistent */ 336 v4addr = addr->sin6_addr.s6_addr32[3]; 337 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr); 338 rcu_read_unlock(); 339 340 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr, 341 chk_addr_ret)) { 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 u32 flags = BIND_WITH_LOCK; 442 int err = 0; 443 444 /* If the socket has its own bind function then use it. */ 445 if (sk->sk_prot->bind) 446 return sk->sk_prot->bind(sk, uaddr, addr_len); 447 448 if (addr_len < SIN6_LEN_RFC2133) 449 return -EINVAL; 450 451 /* BPF prog is run before any checks are done so that if the prog 452 * changes context in a wrong way it will be caught. 453 */ 454 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 455 CGROUP_INET6_BIND, &flags); 456 if (err) 457 return err; 458 459 return __inet6_bind(sk, uaddr, addr_len, flags); 460 } 461 EXPORT_SYMBOL(inet6_bind); 462 463 int inet6_release(struct socket *sock) 464 { 465 struct sock *sk = sock->sk; 466 467 if (!sk) 468 return -EINVAL; 469 470 /* Free mc lists */ 471 ipv6_sock_mc_close(sk); 472 473 /* Free ac lists */ 474 ipv6_sock_ac_close(sk); 475 476 return inet_release(sock); 477 } 478 EXPORT_SYMBOL(inet6_release); 479 480 void inet6_destroy_sock(struct sock *sk) 481 { 482 struct ipv6_pinfo *np = inet6_sk(sk); 483 struct sk_buff *skb; 484 struct ipv6_txoptions *opt; 485 486 /* Release rx options */ 487 488 skb = xchg(&np->pktoptions, NULL); 489 kfree_skb(skb); 490 491 skb = xchg(&np->rxpmtu, NULL); 492 kfree_skb(skb); 493 494 /* Free flowlabels */ 495 fl6_free_socklist(sk); 496 497 /* Free tx options */ 498 499 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 500 if (opt) { 501 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 502 txopt_put(opt); 503 } 504 } 505 EXPORT_SYMBOL_GPL(inet6_destroy_sock); 506 507 /* 508 * This does both peername and sockname. 509 */ 510 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 511 int peer) 512 { 513 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 514 struct sock *sk = sock->sk; 515 struct inet_sock *inet = inet_sk(sk); 516 struct ipv6_pinfo *np = inet6_sk(sk); 517 518 sin->sin6_family = AF_INET6; 519 sin->sin6_flowinfo = 0; 520 sin->sin6_scope_id = 0; 521 lock_sock(sk); 522 if (peer) { 523 if (!inet->inet_dport || 524 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 525 peer == 1)) { 526 release_sock(sk); 527 return -ENOTCONN; 528 } 529 sin->sin6_port = inet->inet_dport; 530 sin->sin6_addr = sk->sk_v6_daddr; 531 if (np->sndflow) 532 sin->sin6_flowinfo = np->flow_label; 533 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 534 CGROUP_INET6_GETPEERNAME); 535 } else { 536 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 537 sin->sin6_addr = np->saddr; 538 else 539 sin->sin6_addr = sk->sk_v6_rcv_saddr; 540 sin->sin6_port = inet->inet_sport; 541 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 542 CGROUP_INET6_GETSOCKNAME); 543 } 544 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 545 sk->sk_bound_dev_if); 546 release_sock(sk); 547 return sizeof(*sin); 548 } 549 EXPORT_SYMBOL(inet6_getname); 550 551 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 552 { 553 void __user *argp = (void __user *)arg; 554 struct sock *sk = sock->sk; 555 struct net *net = sock_net(sk); 556 557 switch (cmd) { 558 case SIOCADDRT: 559 case SIOCDELRT: { 560 struct in6_rtmsg rtmsg; 561 562 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg))) 563 return -EFAULT; 564 return ipv6_route_ioctl(net, cmd, &rtmsg); 565 } 566 case SIOCSIFADDR: 567 return addrconf_add_ifaddr(net, argp); 568 case SIOCDIFADDR: 569 return addrconf_del_ifaddr(net, argp); 570 case SIOCSIFDSTADDR: 571 return addrconf_set_dstaddr(net, argp); 572 default: 573 if (!sk->sk_prot->ioctl) 574 return -ENOIOCTLCMD; 575 return sk->sk_prot->ioctl(sk, cmd, arg); 576 } 577 /*NOTREACHED*/ 578 return 0; 579 } 580 EXPORT_SYMBOL(inet6_ioctl); 581 582 #ifdef CONFIG_COMPAT 583 struct compat_in6_rtmsg { 584 struct in6_addr rtmsg_dst; 585 struct in6_addr rtmsg_src; 586 struct in6_addr rtmsg_gateway; 587 u32 rtmsg_type; 588 u16 rtmsg_dst_len; 589 u16 rtmsg_src_len; 590 u32 rtmsg_metric; 591 u32 rtmsg_info; 592 u32 rtmsg_flags; 593 s32 rtmsg_ifindex; 594 }; 595 596 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 597 struct compat_in6_rtmsg __user *ur) 598 { 599 struct in6_rtmsg rt; 600 601 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst, 602 3 * sizeof(struct in6_addr)) || 603 get_user(rt.rtmsg_type, &ur->rtmsg_type) || 604 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) || 605 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) || 606 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) || 607 get_user(rt.rtmsg_info, &ur->rtmsg_info) || 608 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) || 609 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex)) 610 return -EFAULT; 611 612 613 return ipv6_route_ioctl(sock_net(sk), cmd, &rt); 614 } 615 616 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 617 { 618 void __user *argp = compat_ptr(arg); 619 struct sock *sk = sock->sk; 620 621 switch (cmd) { 622 case SIOCADDRT: 623 case SIOCDELRT: 624 return inet6_compat_routing_ioctl(sk, cmd, argp); 625 default: 626 return -ENOIOCTLCMD; 627 } 628 } 629 EXPORT_SYMBOL_GPL(inet6_compat_ioctl); 630 #endif /* CONFIG_COMPAT */ 631 632 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 633 size_t)); 634 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 635 { 636 struct sock *sk = sock->sk; 637 638 if (unlikely(inet_send_prepare(sk))) 639 return -EAGAIN; 640 641 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 642 sk, msg, size); 643 } 644 645 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 646 size_t, int, int, int *)); 647 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 648 int flags) 649 { 650 struct sock *sk = sock->sk; 651 int addr_len = 0; 652 int err; 653 654 if (likely(!(flags & MSG_ERRQUEUE))) 655 sock_rps_record_flow(sk); 656 657 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 658 sk, msg, size, flags & MSG_DONTWAIT, 659 flags & ~MSG_DONTWAIT, &addr_len); 660 if (err >= 0) 661 msg->msg_namelen = addr_len; 662 return err; 663 } 664 665 const struct proto_ops inet6_stream_ops = { 666 .family = PF_INET6, 667 .owner = THIS_MODULE, 668 .release = inet6_release, 669 .bind = inet6_bind, 670 .connect = inet_stream_connect, /* ok */ 671 .socketpair = sock_no_socketpair, /* a do nothing */ 672 .accept = inet_accept, /* ok */ 673 .getname = inet6_getname, 674 .poll = tcp_poll, /* ok */ 675 .ioctl = inet6_ioctl, /* must change */ 676 .gettstamp = sock_gettstamp, 677 .listen = inet_listen, /* ok */ 678 .shutdown = inet_shutdown, /* ok */ 679 .setsockopt = sock_common_setsockopt, /* ok */ 680 .getsockopt = sock_common_getsockopt, /* ok */ 681 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 682 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 683 #ifdef CONFIG_MMU 684 .mmap = tcp_mmap, 685 #endif 686 .sendpage = inet_sendpage, 687 .sendmsg_locked = tcp_sendmsg_locked, 688 .sendpage_locked = tcp_sendpage_locked, 689 .splice_read = tcp_splice_read, 690 .read_sock = tcp_read_sock, 691 .peek_len = tcp_peek_len, 692 #ifdef CONFIG_COMPAT 693 .compat_ioctl = inet6_compat_ioctl, 694 #endif 695 .set_rcvlowat = tcp_set_rcvlowat, 696 }; 697 698 const struct proto_ops inet6_dgram_ops = { 699 .family = PF_INET6, 700 .owner = THIS_MODULE, 701 .release = inet6_release, 702 .bind = inet6_bind, 703 .connect = inet_dgram_connect, /* ok */ 704 .socketpair = sock_no_socketpair, /* a do nothing */ 705 .accept = sock_no_accept, /* a do nothing */ 706 .getname = inet6_getname, 707 .poll = udp_poll, /* ok */ 708 .ioctl = inet6_ioctl, /* must change */ 709 .gettstamp = sock_gettstamp, 710 .listen = sock_no_listen, /* ok */ 711 .shutdown = inet_shutdown, /* ok */ 712 .setsockopt = sock_common_setsockopt, /* ok */ 713 .getsockopt = sock_common_getsockopt, /* ok */ 714 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 715 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 716 .read_sock = udp_read_sock, 717 .mmap = sock_no_mmap, 718 .sendpage = sock_no_sendpage, 719 .set_peek_off = sk_set_peek_off, 720 #ifdef CONFIG_COMPAT 721 .compat_ioctl = inet6_compat_ioctl, 722 #endif 723 }; 724 725 static const struct net_proto_family inet6_family_ops = { 726 .family = PF_INET6, 727 .create = inet6_create, 728 .owner = THIS_MODULE, 729 }; 730 731 int inet6_register_protosw(struct inet_protosw *p) 732 { 733 struct list_head *lh; 734 struct inet_protosw *answer; 735 struct list_head *last_perm; 736 int protocol = p->protocol; 737 int ret; 738 739 spin_lock_bh(&inetsw6_lock); 740 741 ret = -EINVAL; 742 if (p->type >= SOCK_MAX) 743 goto out_illegal; 744 745 /* If we are trying to override a permanent protocol, bail. */ 746 answer = NULL; 747 ret = -EPERM; 748 last_perm = &inetsw6[p->type]; 749 list_for_each(lh, &inetsw6[p->type]) { 750 answer = list_entry(lh, struct inet_protosw, list); 751 752 /* Check only the non-wild match. */ 753 if (INET_PROTOSW_PERMANENT & answer->flags) { 754 if (protocol == answer->protocol) 755 break; 756 last_perm = lh; 757 } 758 759 answer = NULL; 760 } 761 if (answer) 762 goto out_permanent; 763 764 /* Add the new entry after the last permanent entry if any, so that 765 * the new entry does not override a permanent entry when matched with 766 * a wild-card protocol. But it is allowed to override any existing 767 * non-permanent entry. This means that when we remove this entry, the 768 * system automatically returns to the old behavior. 769 */ 770 list_add_rcu(&p->list, last_perm); 771 ret = 0; 772 out: 773 spin_unlock_bh(&inetsw6_lock); 774 return ret; 775 776 out_permanent: 777 pr_err("Attempt to override permanent protocol %d\n", protocol); 778 goto out; 779 780 out_illegal: 781 pr_err("Ignoring attempt to register invalid socket type %d\n", 782 p->type); 783 goto out; 784 } 785 EXPORT_SYMBOL(inet6_register_protosw); 786 787 void 788 inet6_unregister_protosw(struct inet_protosw *p) 789 { 790 if (INET_PROTOSW_PERMANENT & p->flags) { 791 pr_err("Attempt to unregister permanent protocol %d\n", 792 p->protocol); 793 } else { 794 spin_lock_bh(&inetsw6_lock); 795 list_del_rcu(&p->list); 796 spin_unlock_bh(&inetsw6_lock); 797 798 synchronize_net(); 799 } 800 } 801 EXPORT_SYMBOL(inet6_unregister_protosw); 802 803 int inet6_sk_rebuild_header(struct sock *sk) 804 { 805 struct ipv6_pinfo *np = inet6_sk(sk); 806 struct dst_entry *dst; 807 808 dst = __sk_dst_check(sk, np->dst_cookie); 809 810 if (!dst) { 811 struct inet_sock *inet = inet_sk(sk); 812 struct in6_addr *final_p, final; 813 struct flowi6 fl6; 814 815 memset(&fl6, 0, sizeof(fl6)); 816 fl6.flowi6_proto = sk->sk_protocol; 817 fl6.daddr = sk->sk_v6_daddr; 818 fl6.saddr = np->saddr; 819 fl6.flowlabel = np->flow_label; 820 fl6.flowi6_oif = sk->sk_bound_dev_if; 821 fl6.flowi6_mark = sk->sk_mark; 822 fl6.fl6_dport = inet->inet_dport; 823 fl6.fl6_sport = inet->inet_sport; 824 fl6.flowi6_uid = sk->sk_uid; 825 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); 826 827 rcu_read_lock(); 828 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 829 &final); 830 rcu_read_unlock(); 831 832 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 833 if (IS_ERR(dst)) { 834 sk->sk_route_caps = 0; 835 sk->sk_err_soft = -PTR_ERR(dst); 836 return PTR_ERR(dst); 837 } 838 839 ip6_dst_store(sk, dst, NULL, NULL); 840 } 841 842 return 0; 843 } 844 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 845 846 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 847 const struct inet6_skb_parm *opt) 848 { 849 const struct ipv6_pinfo *np = inet6_sk(sk); 850 851 if (np->rxopt.all) { 852 if (((opt->flags & IP6SKB_HOPBYHOP) && 853 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 854 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 855 np->rxopt.bits.rxflow) || 856 (opt->srcrt && (np->rxopt.bits.srcrt || 857 np->rxopt.bits.osrcrt)) || 858 ((opt->dst1 || opt->dst0) && 859 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 860 return true; 861 } 862 return false; 863 } 864 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 865 866 static struct packet_type ipv6_packet_type __read_mostly = { 867 .type = cpu_to_be16(ETH_P_IPV6), 868 .func = ipv6_rcv, 869 .list_func = ipv6_list_rcv, 870 }; 871 872 static int __init ipv6_packet_init(void) 873 { 874 dev_add_pack(&ipv6_packet_type); 875 return 0; 876 } 877 878 static void ipv6_packet_cleanup(void) 879 { 880 dev_remove_pack(&ipv6_packet_type); 881 } 882 883 static int __net_init ipv6_init_mibs(struct net *net) 884 { 885 int i; 886 887 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 888 if (!net->mib.udp_stats_in6) 889 return -ENOMEM; 890 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 891 if (!net->mib.udplite_stats_in6) 892 goto err_udplite_mib; 893 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 894 if (!net->mib.ipv6_statistics) 895 goto err_ip_mib; 896 897 for_each_possible_cpu(i) { 898 struct ipstats_mib *af_inet6_stats; 899 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 900 u64_stats_init(&af_inet6_stats->syncp); 901 } 902 903 904 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 905 if (!net->mib.icmpv6_statistics) 906 goto err_icmp_mib; 907 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 908 GFP_KERNEL); 909 if (!net->mib.icmpv6msg_statistics) 910 goto err_icmpmsg_mib; 911 return 0; 912 913 err_icmpmsg_mib: 914 free_percpu(net->mib.icmpv6_statistics); 915 err_icmp_mib: 916 free_percpu(net->mib.ipv6_statistics); 917 err_ip_mib: 918 free_percpu(net->mib.udplite_stats_in6); 919 err_udplite_mib: 920 free_percpu(net->mib.udp_stats_in6); 921 return -ENOMEM; 922 } 923 924 static void ipv6_cleanup_mibs(struct net *net) 925 { 926 free_percpu(net->mib.udp_stats_in6); 927 free_percpu(net->mib.udplite_stats_in6); 928 free_percpu(net->mib.ipv6_statistics); 929 free_percpu(net->mib.icmpv6_statistics); 930 kfree(net->mib.icmpv6msg_statistics); 931 } 932 933 static int __net_init inet6_net_init(struct net *net) 934 { 935 int err = 0; 936 937 net->ipv6.sysctl.bindv6only = 0; 938 net->ipv6.sysctl.icmpv6_time = 1*HZ; 939 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 940 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0; 941 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0; 942 943 /* By default, rate limit error messages. 944 * Except for pmtu discovery, it would break it. 945 * proc_do_large_bitmap needs pointer to the bitmap. 946 */ 947 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 948 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 949 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 950 951 net->ipv6.sysctl.flowlabel_consistency = 1; 952 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 953 net->ipv6.sysctl.idgen_retries = 3; 954 net->ipv6.sysctl.idgen_delay = 1 * HZ; 955 net->ipv6.sysctl.flowlabel_state_ranges = 0; 956 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 957 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 958 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 959 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 960 net->ipv6.sysctl.fib_notify_on_flag_change = 0; 961 atomic_set(&net->ipv6.fib6_sernum, 1); 962 963 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID; 964 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE; 965 966 err = ipv6_init_mibs(net); 967 if (err) 968 return err; 969 #ifdef CONFIG_PROC_FS 970 err = udp6_proc_init(net); 971 if (err) 972 goto out; 973 err = tcp6_proc_init(net); 974 if (err) 975 goto proc_tcp6_fail; 976 err = ac6_proc_init(net); 977 if (err) 978 goto proc_ac6_fail; 979 #endif 980 return err; 981 982 #ifdef CONFIG_PROC_FS 983 proc_ac6_fail: 984 tcp6_proc_exit(net); 985 proc_tcp6_fail: 986 udp6_proc_exit(net); 987 out: 988 ipv6_cleanup_mibs(net); 989 return err; 990 #endif 991 } 992 993 static void __net_exit inet6_net_exit(struct net *net) 994 { 995 #ifdef CONFIG_PROC_FS 996 udp6_proc_exit(net); 997 tcp6_proc_exit(net); 998 ac6_proc_exit(net); 999 #endif 1000 ipv6_cleanup_mibs(net); 1001 } 1002 1003 static struct pernet_operations inet6_net_ops = { 1004 .init = inet6_net_init, 1005 .exit = inet6_net_exit, 1006 }; 1007 1008 static int ipv6_route_input(struct sk_buff *skb) 1009 { 1010 ip6_route_input(skb); 1011 return skb_dst(skb)->error; 1012 } 1013 1014 static const struct ipv6_stub ipv6_stub_impl = { 1015 .ipv6_sock_mc_join = ipv6_sock_mc_join, 1016 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 1017 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 1018 .ipv6_route_input = ipv6_route_input, 1019 .fib6_get_table = fib6_get_table, 1020 .fib6_table_lookup = fib6_table_lookup, 1021 .fib6_lookup = fib6_lookup, 1022 .fib6_select_path = fib6_select_path, 1023 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 1024 .fib6_nh_init = fib6_nh_init, 1025 .fib6_nh_release = fib6_nh_release, 1026 .fib6_update_sernum = fib6_update_sernum_stub, 1027 .fib6_rt_update = fib6_rt_update, 1028 .ip6_del_rt = ip6_del_rt, 1029 .udpv6_encap_enable = udpv6_encap_enable, 1030 .ndisc_send_na = ndisc_send_na, 1031 #if IS_ENABLED(CONFIG_XFRM) 1032 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu, 1033 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv, 1034 .xfrm6_rcv_encap = xfrm6_rcv_encap, 1035 #endif 1036 .nd_tbl = &nd_tbl, 1037 .ipv6_fragment = ip6_fragment, 1038 .ipv6_dev_find = ipv6_dev_find, 1039 }; 1040 1041 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 1042 .inet6_bind = __inet6_bind, 1043 .udp6_lib_lookup = __udp6_lib_lookup, 1044 }; 1045 1046 static int __init inet6_init(void) 1047 { 1048 struct list_head *r; 1049 int err = 0; 1050 1051 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 1052 1053 /* Register the socket-side information for inet6_create. */ 1054 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 1055 INIT_LIST_HEAD(r); 1056 1057 if (disable_ipv6_mod) { 1058 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 1059 goto out; 1060 } 1061 1062 err = proto_register(&tcpv6_prot, 1); 1063 if (err) 1064 goto out; 1065 1066 err = proto_register(&udpv6_prot, 1); 1067 if (err) 1068 goto out_unregister_tcp_proto; 1069 1070 err = proto_register(&udplitev6_prot, 1); 1071 if (err) 1072 goto out_unregister_udp_proto; 1073 1074 err = proto_register(&rawv6_prot, 1); 1075 if (err) 1076 goto out_unregister_udplite_proto; 1077 1078 err = proto_register(&pingv6_prot, 1); 1079 if (err) 1080 goto out_unregister_raw_proto; 1081 1082 /* We MUST register RAW sockets before we create the ICMP6, 1083 * IGMP6, or NDISC control sockets. 1084 */ 1085 err = rawv6_init(); 1086 if (err) 1087 goto out_unregister_ping_proto; 1088 1089 /* Register the family here so that the init calls below will 1090 * be able to create sockets. (?? is this dangerous ??) 1091 */ 1092 err = sock_register(&inet6_family_ops); 1093 if (err) 1094 goto out_sock_register_fail; 1095 1096 /* 1097 * ipngwg API draft makes clear that the correct semantics 1098 * for TCP and UDP is to consider one TCP and UDP instance 1099 * in a host available by both INET and INET6 APIs and 1100 * able to communicate via both network protocols. 1101 */ 1102 1103 err = register_pernet_subsys(&inet6_net_ops); 1104 if (err) 1105 goto register_pernet_fail; 1106 err = ip6_mr_init(); 1107 if (err) 1108 goto ipmr_fail; 1109 err = icmpv6_init(); 1110 if (err) 1111 goto icmp_fail; 1112 err = ndisc_init(); 1113 if (err) 1114 goto ndisc_fail; 1115 err = igmp6_init(); 1116 if (err) 1117 goto igmp_fail; 1118 1119 err = ipv6_netfilter_init(); 1120 if (err) 1121 goto netfilter_fail; 1122 /* Create /proc/foo6 entries. */ 1123 #ifdef CONFIG_PROC_FS 1124 err = -ENOMEM; 1125 if (raw6_proc_init()) 1126 goto proc_raw6_fail; 1127 if (udplite6_proc_init()) 1128 goto proc_udplite6_fail; 1129 if (ipv6_misc_proc_init()) 1130 goto proc_misc6_fail; 1131 if (if6_proc_init()) 1132 goto proc_if6_fail; 1133 #endif 1134 err = ip6_route_init(); 1135 if (err) 1136 goto ip6_route_fail; 1137 err = ndisc_late_init(); 1138 if (err) 1139 goto ndisc_late_fail; 1140 err = ip6_flowlabel_init(); 1141 if (err) 1142 goto ip6_flowlabel_fail; 1143 err = ipv6_anycast_init(); 1144 if (err) 1145 goto ipv6_anycast_fail; 1146 err = addrconf_init(); 1147 if (err) 1148 goto addrconf_fail; 1149 1150 /* Init v6 extension headers. */ 1151 err = ipv6_exthdrs_init(); 1152 if (err) 1153 goto ipv6_exthdrs_fail; 1154 1155 err = ipv6_frag_init(); 1156 if (err) 1157 goto ipv6_frag_fail; 1158 1159 /* Init v6 transport protocols. */ 1160 err = udpv6_init(); 1161 if (err) 1162 goto udpv6_fail; 1163 1164 err = udplitev6_init(); 1165 if (err) 1166 goto udplitev6_fail; 1167 1168 err = udpv6_offload_init(); 1169 if (err) 1170 goto udpv6_offload_fail; 1171 1172 err = tcpv6_init(); 1173 if (err) 1174 goto tcpv6_fail; 1175 1176 err = ipv6_packet_init(); 1177 if (err) 1178 goto ipv6_packet_fail; 1179 1180 err = pingv6_init(); 1181 if (err) 1182 goto pingv6_fail; 1183 1184 err = calipso_init(); 1185 if (err) 1186 goto calipso_fail; 1187 1188 err = seg6_init(); 1189 if (err) 1190 goto seg6_fail; 1191 1192 err = rpl_init(); 1193 if (err) 1194 goto rpl_fail; 1195 1196 err = ioam6_init(); 1197 if (err) 1198 goto ioam6_fail; 1199 1200 err = igmp6_late_init(); 1201 if (err) 1202 goto igmp6_late_err; 1203 1204 #ifdef CONFIG_SYSCTL 1205 err = ipv6_sysctl_register(); 1206 if (err) 1207 goto sysctl_fail; 1208 #endif 1209 1210 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1211 wmb(); 1212 ipv6_stub = &ipv6_stub_impl; 1213 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1214 out: 1215 return err; 1216 1217 #ifdef CONFIG_SYSCTL 1218 sysctl_fail: 1219 igmp6_late_cleanup(); 1220 #endif 1221 igmp6_late_err: 1222 ioam6_exit(); 1223 ioam6_fail: 1224 rpl_exit(); 1225 rpl_fail: 1226 seg6_exit(); 1227 seg6_fail: 1228 calipso_exit(); 1229 calipso_fail: 1230 pingv6_exit(); 1231 pingv6_fail: 1232 ipv6_packet_cleanup(); 1233 ipv6_packet_fail: 1234 tcpv6_exit(); 1235 tcpv6_fail: 1236 udpv6_offload_exit(); 1237 udpv6_offload_fail: 1238 udplitev6_exit(); 1239 udplitev6_fail: 1240 udpv6_exit(); 1241 udpv6_fail: 1242 ipv6_frag_exit(); 1243 ipv6_frag_fail: 1244 ipv6_exthdrs_exit(); 1245 ipv6_exthdrs_fail: 1246 addrconf_cleanup(); 1247 addrconf_fail: 1248 ipv6_anycast_cleanup(); 1249 ipv6_anycast_fail: 1250 ip6_flowlabel_cleanup(); 1251 ip6_flowlabel_fail: 1252 ndisc_late_cleanup(); 1253 ndisc_late_fail: 1254 ip6_route_cleanup(); 1255 ip6_route_fail: 1256 #ifdef CONFIG_PROC_FS 1257 if6_proc_exit(); 1258 proc_if6_fail: 1259 ipv6_misc_proc_exit(); 1260 proc_misc6_fail: 1261 udplite6_proc_exit(); 1262 proc_udplite6_fail: 1263 raw6_proc_exit(); 1264 proc_raw6_fail: 1265 #endif 1266 ipv6_netfilter_fini(); 1267 netfilter_fail: 1268 igmp6_cleanup(); 1269 igmp_fail: 1270 ndisc_cleanup(); 1271 ndisc_fail: 1272 icmpv6_cleanup(); 1273 icmp_fail: 1274 ip6_mr_cleanup(); 1275 ipmr_fail: 1276 unregister_pernet_subsys(&inet6_net_ops); 1277 register_pernet_fail: 1278 sock_unregister(PF_INET6); 1279 rtnl_unregister_all(PF_INET6); 1280 out_sock_register_fail: 1281 rawv6_exit(); 1282 out_unregister_ping_proto: 1283 proto_unregister(&pingv6_prot); 1284 out_unregister_raw_proto: 1285 proto_unregister(&rawv6_prot); 1286 out_unregister_udplite_proto: 1287 proto_unregister(&udplitev6_prot); 1288 out_unregister_udp_proto: 1289 proto_unregister(&udpv6_prot); 1290 out_unregister_tcp_proto: 1291 proto_unregister(&tcpv6_prot); 1292 goto out; 1293 } 1294 module_init(inet6_init); 1295 1296 MODULE_ALIAS_NETPROTO(PF_INET6); 1297