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_can_nonlocal_bind(net, inet) && 341 v4addr != htonl(INADDR_ANY) && 342 chk_addr_ret != RTN_LOCAL && 343 chk_addr_ret != RTN_MULTICAST && 344 chk_addr_ret != RTN_BROADCAST) { 345 err = -EADDRNOTAVAIL; 346 goto out; 347 } 348 } else { 349 if (addr_type != IPV6_ADDR_ANY) { 350 struct net_device *dev = NULL; 351 352 rcu_read_lock(); 353 if (__ipv6_addr_needs_scope_id(addr_type)) { 354 if (addr_len >= sizeof(struct sockaddr_in6) && 355 addr->sin6_scope_id) { 356 /* Override any existing binding, if another one 357 * is supplied by user. 358 */ 359 sk->sk_bound_dev_if = addr->sin6_scope_id; 360 } 361 362 /* Binding to link-local address requires an interface */ 363 if (!sk->sk_bound_dev_if) { 364 err = -EINVAL; 365 goto out_unlock; 366 } 367 } 368 369 if (sk->sk_bound_dev_if) { 370 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 371 if (!dev) { 372 err = -ENODEV; 373 goto out_unlock; 374 } 375 } 376 377 /* ipv4 addr of the socket is invalid. Only the 378 * unspecified and mapped address have a v4 equivalent. 379 */ 380 v4addr = LOOPBACK4_IPV6; 381 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 382 if (!ipv6_can_nonlocal_bind(net, inet) && 383 !ipv6_chk_addr(net, &addr->sin6_addr, 384 dev, 0)) { 385 err = -EADDRNOTAVAIL; 386 goto out_unlock; 387 } 388 } 389 rcu_read_unlock(); 390 } 391 } 392 393 inet->inet_rcv_saddr = v4addr; 394 inet->inet_saddr = v4addr; 395 396 sk->sk_v6_rcv_saddr = addr->sin6_addr; 397 398 if (!(addr_type & IPV6_ADDR_MULTICAST)) 399 np->saddr = addr->sin6_addr; 400 401 saved_ipv6only = sk->sk_ipv6only; 402 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED) 403 sk->sk_ipv6only = 1; 404 405 /* Make sure we are allowed to bind here. */ 406 if (snum || !(inet->bind_address_no_port || 407 (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 408 if (sk->sk_prot->get_port(sk, snum)) { 409 sk->sk_ipv6only = saved_ipv6only; 410 inet_reset_saddr(sk); 411 err = -EADDRINUSE; 412 goto out; 413 } 414 if (!(flags & BIND_FROM_BPF)) { 415 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk); 416 if (err) { 417 sk->sk_ipv6only = saved_ipv6only; 418 inet_reset_saddr(sk); 419 goto out; 420 } 421 } 422 } 423 424 if (addr_type != IPV6_ADDR_ANY) 425 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 426 if (snum) 427 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 428 inet->inet_sport = htons(inet->inet_num); 429 inet->inet_dport = 0; 430 inet->inet_daddr = 0; 431 out: 432 if (flags & BIND_WITH_LOCK) 433 release_sock(sk); 434 return err; 435 out_unlock: 436 rcu_read_unlock(); 437 goto out; 438 } 439 440 /* bind for INET6 API */ 441 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 442 { 443 struct sock *sk = sock->sk; 444 u32 flags = BIND_WITH_LOCK; 445 int err = 0; 446 447 /* If the socket has its own bind function then use it. */ 448 if (sk->sk_prot->bind) 449 return sk->sk_prot->bind(sk, uaddr, addr_len); 450 451 if (addr_len < SIN6_LEN_RFC2133) 452 return -EINVAL; 453 454 /* BPF prog is run before any checks are done so that if the prog 455 * changes context in a wrong way it will be caught. 456 */ 457 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 458 CGROUP_INET6_BIND, &flags); 459 if (err) 460 return err; 461 462 return __inet6_bind(sk, uaddr, addr_len, flags); 463 } 464 EXPORT_SYMBOL(inet6_bind); 465 466 int inet6_release(struct socket *sock) 467 { 468 struct sock *sk = sock->sk; 469 470 if (!sk) 471 return -EINVAL; 472 473 /* Free mc lists */ 474 ipv6_sock_mc_close(sk); 475 476 /* Free ac lists */ 477 ipv6_sock_ac_close(sk); 478 479 return inet_release(sock); 480 } 481 EXPORT_SYMBOL(inet6_release); 482 483 void inet6_destroy_sock(struct sock *sk) 484 { 485 struct ipv6_pinfo *np = inet6_sk(sk); 486 struct sk_buff *skb; 487 struct ipv6_txoptions *opt; 488 489 /* Release rx options */ 490 491 skb = xchg(&np->pktoptions, NULL); 492 kfree_skb(skb); 493 494 skb = xchg(&np->rxpmtu, NULL); 495 kfree_skb(skb); 496 497 /* Free flowlabels */ 498 fl6_free_socklist(sk); 499 500 /* Free tx options */ 501 502 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 503 if (opt) { 504 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 505 txopt_put(opt); 506 } 507 } 508 EXPORT_SYMBOL_GPL(inet6_destroy_sock); 509 510 /* 511 * This does both peername and sockname. 512 */ 513 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 514 int peer) 515 { 516 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 517 struct sock *sk = sock->sk; 518 struct inet_sock *inet = inet_sk(sk); 519 struct ipv6_pinfo *np = inet6_sk(sk); 520 521 sin->sin6_family = AF_INET6; 522 sin->sin6_flowinfo = 0; 523 sin->sin6_scope_id = 0; 524 lock_sock(sk); 525 if (peer) { 526 if (!inet->inet_dport || 527 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 528 peer == 1)) { 529 release_sock(sk); 530 return -ENOTCONN; 531 } 532 sin->sin6_port = inet->inet_dport; 533 sin->sin6_addr = sk->sk_v6_daddr; 534 if (np->sndflow) 535 sin->sin6_flowinfo = np->flow_label; 536 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 537 CGROUP_INET6_GETPEERNAME); 538 } else { 539 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 540 sin->sin6_addr = np->saddr; 541 else 542 sin->sin6_addr = sk->sk_v6_rcv_saddr; 543 sin->sin6_port = inet->inet_sport; 544 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 545 CGROUP_INET6_GETSOCKNAME); 546 } 547 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 548 sk->sk_bound_dev_if); 549 release_sock(sk); 550 return sizeof(*sin); 551 } 552 EXPORT_SYMBOL(inet6_getname); 553 554 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 555 { 556 void __user *argp = (void __user *)arg; 557 struct sock *sk = sock->sk; 558 struct net *net = sock_net(sk); 559 560 switch (cmd) { 561 case SIOCADDRT: 562 case SIOCDELRT: { 563 struct in6_rtmsg rtmsg; 564 565 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg))) 566 return -EFAULT; 567 return ipv6_route_ioctl(net, cmd, &rtmsg); 568 } 569 case SIOCSIFADDR: 570 return addrconf_add_ifaddr(net, argp); 571 case SIOCDIFADDR: 572 return addrconf_del_ifaddr(net, argp); 573 case SIOCSIFDSTADDR: 574 return addrconf_set_dstaddr(net, argp); 575 default: 576 if (!sk->sk_prot->ioctl) 577 return -ENOIOCTLCMD; 578 return sk->sk_prot->ioctl(sk, cmd, arg); 579 } 580 /*NOTREACHED*/ 581 return 0; 582 } 583 EXPORT_SYMBOL(inet6_ioctl); 584 585 #ifdef CONFIG_COMPAT 586 struct compat_in6_rtmsg { 587 struct in6_addr rtmsg_dst; 588 struct in6_addr rtmsg_src; 589 struct in6_addr rtmsg_gateway; 590 u32 rtmsg_type; 591 u16 rtmsg_dst_len; 592 u16 rtmsg_src_len; 593 u32 rtmsg_metric; 594 u32 rtmsg_info; 595 u32 rtmsg_flags; 596 s32 rtmsg_ifindex; 597 }; 598 599 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 600 struct compat_in6_rtmsg __user *ur) 601 { 602 struct in6_rtmsg rt; 603 604 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst, 605 3 * sizeof(struct in6_addr)) || 606 get_user(rt.rtmsg_type, &ur->rtmsg_type) || 607 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) || 608 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) || 609 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) || 610 get_user(rt.rtmsg_info, &ur->rtmsg_info) || 611 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) || 612 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex)) 613 return -EFAULT; 614 615 616 return ipv6_route_ioctl(sock_net(sk), cmd, &rt); 617 } 618 619 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 620 { 621 void __user *argp = compat_ptr(arg); 622 struct sock *sk = sock->sk; 623 624 switch (cmd) { 625 case SIOCADDRT: 626 case SIOCDELRT: 627 return inet6_compat_routing_ioctl(sk, cmd, argp); 628 default: 629 return -ENOIOCTLCMD; 630 } 631 } 632 EXPORT_SYMBOL_GPL(inet6_compat_ioctl); 633 #endif /* CONFIG_COMPAT */ 634 635 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 636 size_t)); 637 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 638 { 639 struct sock *sk = sock->sk; 640 641 if (unlikely(inet_send_prepare(sk))) 642 return -EAGAIN; 643 644 return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 645 sk, msg, size); 646 } 647 648 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 649 size_t, int, int, int *)); 650 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 651 int flags) 652 { 653 struct sock *sk = sock->sk; 654 int addr_len = 0; 655 int err; 656 657 if (likely(!(flags & MSG_ERRQUEUE))) 658 sock_rps_record_flow(sk); 659 660 err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 661 sk, msg, size, flags & MSG_DONTWAIT, 662 flags & ~MSG_DONTWAIT, &addr_len); 663 if (err >= 0) 664 msg->msg_namelen = addr_len; 665 return err; 666 } 667 668 const struct proto_ops inet6_stream_ops = { 669 .family = PF_INET6, 670 .owner = THIS_MODULE, 671 .release = inet6_release, 672 .bind = inet6_bind, 673 .connect = inet_stream_connect, /* ok */ 674 .socketpair = sock_no_socketpair, /* a do nothing */ 675 .accept = inet_accept, /* ok */ 676 .getname = inet6_getname, 677 .poll = tcp_poll, /* ok */ 678 .ioctl = inet6_ioctl, /* must change */ 679 .gettstamp = sock_gettstamp, 680 .listen = inet_listen, /* ok */ 681 .shutdown = inet_shutdown, /* ok */ 682 .setsockopt = sock_common_setsockopt, /* ok */ 683 .getsockopt = sock_common_getsockopt, /* ok */ 684 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 685 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 686 #ifdef CONFIG_MMU 687 .mmap = tcp_mmap, 688 #endif 689 .sendpage = inet_sendpage, 690 .sendmsg_locked = tcp_sendmsg_locked, 691 .sendpage_locked = tcp_sendpage_locked, 692 .splice_read = tcp_splice_read, 693 .read_sock = tcp_read_sock, 694 .peek_len = tcp_peek_len, 695 #ifdef CONFIG_COMPAT 696 .compat_ioctl = inet6_compat_ioctl, 697 #endif 698 .set_rcvlowat = tcp_set_rcvlowat, 699 }; 700 701 const struct proto_ops inet6_dgram_ops = { 702 .family = PF_INET6, 703 .owner = THIS_MODULE, 704 .release = inet6_release, 705 .bind = inet6_bind, 706 .connect = inet_dgram_connect, /* ok */ 707 .socketpair = sock_no_socketpair, /* a do nothing */ 708 .accept = sock_no_accept, /* a do nothing */ 709 .getname = inet6_getname, 710 .poll = udp_poll, /* ok */ 711 .ioctl = inet6_ioctl, /* must change */ 712 .gettstamp = sock_gettstamp, 713 .listen = sock_no_listen, /* ok */ 714 .shutdown = inet_shutdown, /* ok */ 715 .setsockopt = sock_common_setsockopt, /* ok */ 716 .getsockopt = sock_common_getsockopt, /* ok */ 717 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 718 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 719 .read_sock = udp_read_sock, 720 .mmap = sock_no_mmap, 721 .sendpage = sock_no_sendpage, 722 .set_peek_off = sk_set_peek_off, 723 #ifdef CONFIG_COMPAT 724 .compat_ioctl = inet6_compat_ioctl, 725 #endif 726 }; 727 728 static const struct net_proto_family inet6_family_ops = { 729 .family = PF_INET6, 730 .create = inet6_create, 731 .owner = THIS_MODULE, 732 }; 733 734 int inet6_register_protosw(struct inet_protosw *p) 735 { 736 struct list_head *lh; 737 struct inet_protosw *answer; 738 struct list_head *last_perm; 739 int protocol = p->protocol; 740 int ret; 741 742 spin_lock_bh(&inetsw6_lock); 743 744 ret = -EINVAL; 745 if (p->type >= SOCK_MAX) 746 goto out_illegal; 747 748 /* If we are trying to override a permanent protocol, bail. */ 749 answer = NULL; 750 ret = -EPERM; 751 last_perm = &inetsw6[p->type]; 752 list_for_each(lh, &inetsw6[p->type]) { 753 answer = list_entry(lh, struct inet_protosw, list); 754 755 /* Check only the non-wild match. */ 756 if (INET_PROTOSW_PERMANENT & answer->flags) { 757 if (protocol == answer->protocol) 758 break; 759 last_perm = lh; 760 } 761 762 answer = NULL; 763 } 764 if (answer) 765 goto out_permanent; 766 767 /* Add the new entry after the last permanent entry if any, so that 768 * the new entry does not override a permanent entry when matched with 769 * a wild-card protocol. But it is allowed to override any existing 770 * non-permanent entry. This means that when we remove this entry, the 771 * system automatically returns to the old behavior. 772 */ 773 list_add_rcu(&p->list, last_perm); 774 ret = 0; 775 out: 776 spin_unlock_bh(&inetsw6_lock); 777 return ret; 778 779 out_permanent: 780 pr_err("Attempt to override permanent protocol %d\n", protocol); 781 goto out; 782 783 out_illegal: 784 pr_err("Ignoring attempt to register invalid socket type %d\n", 785 p->type); 786 goto out; 787 } 788 EXPORT_SYMBOL(inet6_register_protosw); 789 790 void 791 inet6_unregister_protosw(struct inet_protosw *p) 792 { 793 if (INET_PROTOSW_PERMANENT & p->flags) { 794 pr_err("Attempt to unregister permanent protocol %d\n", 795 p->protocol); 796 } else { 797 spin_lock_bh(&inetsw6_lock); 798 list_del_rcu(&p->list); 799 spin_unlock_bh(&inetsw6_lock); 800 801 synchronize_net(); 802 } 803 } 804 EXPORT_SYMBOL(inet6_unregister_protosw); 805 806 int inet6_sk_rebuild_header(struct sock *sk) 807 { 808 struct ipv6_pinfo *np = inet6_sk(sk); 809 struct dst_entry *dst; 810 811 dst = __sk_dst_check(sk, np->dst_cookie); 812 813 if (!dst) { 814 struct inet_sock *inet = inet_sk(sk); 815 struct in6_addr *final_p, final; 816 struct flowi6 fl6; 817 818 memset(&fl6, 0, sizeof(fl6)); 819 fl6.flowi6_proto = sk->sk_protocol; 820 fl6.daddr = sk->sk_v6_daddr; 821 fl6.saddr = np->saddr; 822 fl6.flowlabel = np->flow_label; 823 fl6.flowi6_oif = sk->sk_bound_dev_if; 824 fl6.flowi6_mark = sk->sk_mark; 825 fl6.fl6_dport = inet->inet_dport; 826 fl6.fl6_sport = inet->inet_sport; 827 fl6.flowi6_uid = sk->sk_uid; 828 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); 829 830 rcu_read_lock(); 831 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 832 &final); 833 rcu_read_unlock(); 834 835 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 836 if (IS_ERR(dst)) { 837 sk->sk_route_caps = 0; 838 sk->sk_err_soft = -PTR_ERR(dst); 839 return PTR_ERR(dst); 840 } 841 842 ip6_dst_store(sk, dst, NULL, NULL); 843 } 844 845 return 0; 846 } 847 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 848 849 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 850 const struct inet6_skb_parm *opt) 851 { 852 const struct ipv6_pinfo *np = inet6_sk(sk); 853 854 if (np->rxopt.all) { 855 if (((opt->flags & IP6SKB_HOPBYHOP) && 856 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 857 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 858 np->rxopt.bits.rxflow) || 859 (opt->srcrt && (np->rxopt.bits.srcrt || 860 np->rxopt.bits.osrcrt)) || 861 ((opt->dst1 || opt->dst0) && 862 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 863 return true; 864 } 865 return false; 866 } 867 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 868 869 static struct packet_type ipv6_packet_type __read_mostly = { 870 .type = cpu_to_be16(ETH_P_IPV6), 871 .func = ipv6_rcv, 872 .list_func = ipv6_list_rcv, 873 }; 874 875 static int __init ipv6_packet_init(void) 876 { 877 dev_add_pack(&ipv6_packet_type); 878 return 0; 879 } 880 881 static void ipv6_packet_cleanup(void) 882 { 883 dev_remove_pack(&ipv6_packet_type); 884 } 885 886 static int __net_init ipv6_init_mibs(struct net *net) 887 { 888 int i; 889 890 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 891 if (!net->mib.udp_stats_in6) 892 return -ENOMEM; 893 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 894 if (!net->mib.udplite_stats_in6) 895 goto err_udplite_mib; 896 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 897 if (!net->mib.ipv6_statistics) 898 goto err_ip_mib; 899 900 for_each_possible_cpu(i) { 901 struct ipstats_mib *af_inet6_stats; 902 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 903 u64_stats_init(&af_inet6_stats->syncp); 904 } 905 906 907 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 908 if (!net->mib.icmpv6_statistics) 909 goto err_icmp_mib; 910 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 911 GFP_KERNEL); 912 if (!net->mib.icmpv6msg_statistics) 913 goto err_icmpmsg_mib; 914 return 0; 915 916 err_icmpmsg_mib: 917 free_percpu(net->mib.icmpv6_statistics); 918 err_icmp_mib: 919 free_percpu(net->mib.ipv6_statistics); 920 err_ip_mib: 921 free_percpu(net->mib.udplite_stats_in6); 922 err_udplite_mib: 923 free_percpu(net->mib.udp_stats_in6); 924 return -ENOMEM; 925 } 926 927 static void ipv6_cleanup_mibs(struct net *net) 928 { 929 free_percpu(net->mib.udp_stats_in6); 930 free_percpu(net->mib.udplite_stats_in6); 931 free_percpu(net->mib.ipv6_statistics); 932 free_percpu(net->mib.icmpv6_statistics); 933 kfree(net->mib.icmpv6msg_statistics); 934 } 935 936 static int __net_init inet6_net_init(struct net *net) 937 { 938 int err = 0; 939 940 net->ipv6.sysctl.bindv6only = 0; 941 net->ipv6.sysctl.icmpv6_time = 1*HZ; 942 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 943 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0; 944 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0; 945 946 /* By default, rate limit error messages. 947 * Except for pmtu discovery, it would break it. 948 * proc_do_large_bitmap needs pointer to the bitmap. 949 */ 950 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 951 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 952 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 953 954 net->ipv6.sysctl.flowlabel_consistency = 1; 955 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 956 net->ipv6.sysctl.idgen_retries = 3; 957 net->ipv6.sysctl.idgen_delay = 1 * HZ; 958 net->ipv6.sysctl.flowlabel_state_ranges = 0; 959 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 960 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 961 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 962 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 963 net->ipv6.sysctl.fib_notify_on_flag_change = 0; 964 atomic_set(&net->ipv6.fib6_sernum, 1); 965 966 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID; 967 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE; 968 969 err = ipv6_init_mibs(net); 970 if (err) 971 return err; 972 #ifdef CONFIG_PROC_FS 973 err = udp6_proc_init(net); 974 if (err) 975 goto out; 976 err = tcp6_proc_init(net); 977 if (err) 978 goto proc_tcp6_fail; 979 err = ac6_proc_init(net); 980 if (err) 981 goto proc_ac6_fail; 982 #endif 983 return err; 984 985 #ifdef CONFIG_PROC_FS 986 proc_ac6_fail: 987 tcp6_proc_exit(net); 988 proc_tcp6_fail: 989 udp6_proc_exit(net); 990 out: 991 ipv6_cleanup_mibs(net); 992 return err; 993 #endif 994 } 995 996 static void __net_exit inet6_net_exit(struct net *net) 997 { 998 #ifdef CONFIG_PROC_FS 999 udp6_proc_exit(net); 1000 tcp6_proc_exit(net); 1001 ac6_proc_exit(net); 1002 #endif 1003 ipv6_cleanup_mibs(net); 1004 } 1005 1006 static struct pernet_operations inet6_net_ops = { 1007 .init = inet6_net_init, 1008 .exit = inet6_net_exit, 1009 }; 1010 1011 static int ipv6_route_input(struct sk_buff *skb) 1012 { 1013 ip6_route_input(skb); 1014 return skb_dst(skb)->error; 1015 } 1016 1017 static const struct ipv6_stub ipv6_stub_impl = { 1018 .ipv6_sock_mc_join = ipv6_sock_mc_join, 1019 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 1020 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 1021 .ipv6_route_input = ipv6_route_input, 1022 .fib6_get_table = fib6_get_table, 1023 .fib6_table_lookup = fib6_table_lookup, 1024 .fib6_lookup = fib6_lookup, 1025 .fib6_select_path = fib6_select_path, 1026 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 1027 .fib6_nh_init = fib6_nh_init, 1028 .fib6_nh_release = fib6_nh_release, 1029 .fib6_nh_release_dsts = fib6_nh_release_dsts, 1030 .fib6_update_sernum = fib6_update_sernum_stub, 1031 .fib6_rt_update = fib6_rt_update, 1032 .ip6_del_rt = ip6_del_rt, 1033 .udpv6_encap_enable = udpv6_encap_enable, 1034 .ndisc_send_na = ndisc_send_na, 1035 #if IS_ENABLED(CONFIG_XFRM) 1036 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu, 1037 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv, 1038 .xfrm6_rcv_encap = xfrm6_rcv_encap, 1039 #endif 1040 .nd_tbl = &nd_tbl, 1041 .ipv6_fragment = ip6_fragment, 1042 .ipv6_dev_find = ipv6_dev_find, 1043 }; 1044 1045 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 1046 .inet6_bind = __inet6_bind, 1047 .udp6_lib_lookup = __udp6_lib_lookup, 1048 }; 1049 1050 static int __init inet6_init(void) 1051 { 1052 struct list_head *r; 1053 int err = 0; 1054 1055 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 1056 1057 /* Register the socket-side information for inet6_create. */ 1058 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 1059 INIT_LIST_HEAD(r); 1060 1061 if (disable_ipv6_mod) { 1062 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 1063 goto out; 1064 } 1065 1066 err = proto_register(&tcpv6_prot, 1); 1067 if (err) 1068 goto out; 1069 1070 err = proto_register(&udpv6_prot, 1); 1071 if (err) 1072 goto out_unregister_tcp_proto; 1073 1074 err = proto_register(&udplitev6_prot, 1); 1075 if (err) 1076 goto out_unregister_udp_proto; 1077 1078 err = proto_register(&rawv6_prot, 1); 1079 if (err) 1080 goto out_unregister_udplite_proto; 1081 1082 err = proto_register(&pingv6_prot, 1); 1083 if (err) 1084 goto out_unregister_raw_proto; 1085 1086 /* We MUST register RAW sockets before we create the ICMP6, 1087 * IGMP6, or NDISC control sockets. 1088 */ 1089 err = rawv6_init(); 1090 if (err) 1091 goto out_unregister_ping_proto; 1092 1093 /* Register the family here so that the init calls below will 1094 * be able to create sockets. (?? is this dangerous ??) 1095 */ 1096 err = sock_register(&inet6_family_ops); 1097 if (err) 1098 goto out_sock_register_fail; 1099 1100 /* 1101 * ipngwg API draft makes clear that the correct semantics 1102 * for TCP and UDP is to consider one TCP and UDP instance 1103 * in a host available by both INET and INET6 APIs and 1104 * able to communicate via both network protocols. 1105 */ 1106 1107 err = register_pernet_subsys(&inet6_net_ops); 1108 if (err) 1109 goto register_pernet_fail; 1110 err = ip6_mr_init(); 1111 if (err) 1112 goto ipmr_fail; 1113 err = icmpv6_init(); 1114 if (err) 1115 goto icmp_fail; 1116 err = ndisc_init(); 1117 if (err) 1118 goto ndisc_fail; 1119 err = igmp6_init(); 1120 if (err) 1121 goto igmp_fail; 1122 1123 err = ipv6_netfilter_init(); 1124 if (err) 1125 goto netfilter_fail; 1126 /* Create /proc/foo6 entries. */ 1127 #ifdef CONFIG_PROC_FS 1128 err = -ENOMEM; 1129 if (raw6_proc_init()) 1130 goto proc_raw6_fail; 1131 if (udplite6_proc_init()) 1132 goto proc_udplite6_fail; 1133 if (ipv6_misc_proc_init()) 1134 goto proc_misc6_fail; 1135 if (if6_proc_init()) 1136 goto proc_if6_fail; 1137 #endif 1138 err = ip6_route_init(); 1139 if (err) 1140 goto ip6_route_fail; 1141 err = ndisc_late_init(); 1142 if (err) 1143 goto ndisc_late_fail; 1144 err = ip6_flowlabel_init(); 1145 if (err) 1146 goto ip6_flowlabel_fail; 1147 err = ipv6_anycast_init(); 1148 if (err) 1149 goto ipv6_anycast_fail; 1150 err = addrconf_init(); 1151 if (err) 1152 goto addrconf_fail; 1153 1154 /* Init v6 extension headers. */ 1155 err = ipv6_exthdrs_init(); 1156 if (err) 1157 goto ipv6_exthdrs_fail; 1158 1159 err = ipv6_frag_init(); 1160 if (err) 1161 goto ipv6_frag_fail; 1162 1163 /* Init v6 transport protocols. */ 1164 err = udpv6_init(); 1165 if (err) 1166 goto udpv6_fail; 1167 1168 err = udplitev6_init(); 1169 if (err) 1170 goto udplitev6_fail; 1171 1172 err = udpv6_offload_init(); 1173 if (err) 1174 goto udpv6_offload_fail; 1175 1176 err = tcpv6_init(); 1177 if (err) 1178 goto tcpv6_fail; 1179 1180 err = ipv6_packet_init(); 1181 if (err) 1182 goto ipv6_packet_fail; 1183 1184 err = pingv6_init(); 1185 if (err) 1186 goto pingv6_fail; 1187 1188 err = calipso_init(); 1189 if (err) 1190 goto calipso_fail; 1191 1192 err = seg6_init(); 1193 if (err) 1194 goto seg6_fail; 1195 1196 err = rpl_init(); 1197 if (err) 1198 goto rpl_fail; 1199 1200 err = ioam6_init(); 1201 if (err) 1202 goto ioam6_fail; 1203 1204 err = igmp6_late_init(); 1205 if (err) 1206 goto igmp6_late_err; 1207 1208 #ifdef CONFIG_SYSCTL 1209 err = ipv6_sysctl_register(); 1210 if (err) 1211 goto sysctl_fail; 1212 #endif 1213 1214 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1215 wmb(); 1216 ipv6_stub = &ipv6_stub_impl; 1217 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1218 out: 1219 return err; 1220 1221 #ifdef CONFIG_SYSCTL 1222 sysctl_fail: 1223 igmp6_late_cleanup(); 1224 #endif 1225 igmp6_late_err: 1226 ioam6_exit(); 1227 ioam6_fail: 1228 rpl_exit(); 1229 rpl_fail: 1230 seg6_exit(); 1231 seg6_fail: 1232 calipso_exit(); 1233 calipso_fail: 1234 pingv6_exit(); 1235 pingv6_fail: 1236 ipv6_packet_cleanup(); 1237 ipv6_packet_fail: 1238 tcpv6_exit(); 1239 tcpv6_fail: 1240 udpv6_offload_exit(); 1241 udpv6_offload_fail: 1242 udplitev6_exit(); 1243 udplitev6_fail: 1244 udpv6_exit(); 1245 udpv6_fail: 1246 ipv6_frag_exit(); 1247 ipv6_frag_fail: 1248 ipv6_exthdrs_exit(); 1249 ipv6_exthdrs_fail: 1250 addrconf_cleanup(); 1251 addrconf_fail: 1252 ipv6_anycast_cleanup(); 1253 ipv6_anycast_fail: 1254 ip6_flowlabel_cleanup(); 1255 ip6_flowlabel_fail: 1256 ndisc_late_cleanup(); 1257 ndisc_late_fail: 1258 ip6_route_cleanup(); 1259 ip6_route_fail: 1260 #ifdef CONFIG_PROC_FS 1261 if6_proc_exit(); 1262 proc_if6_fail: 1263 ipv6_misc_proc_exit(); 1264 proc_misc6_fail: 1265 udplite6_proc_exit(); 1266 proc_udplite6_fail: 1267 raw6_proc_exit(); 1268 proc_raw6_fail: 1269 #endif 1270 ipv6_netfilter_fini(); 1271 netfilter_fail: 1272 igmp6_cleanup(); 1273 igmp_fail: 1274 ndisc_cleanup(); 1275 ndisc_fail: 1276 icmpv6_cleanup(); 1277 icmp_fail: 1278 ip6_mr_cleanup(); 1279 ipmr_fail: 1280 unregister_pernet_subsys(&inet6_net_ops); 1281 register_pernet_fail: 1282 sock_unregister(PF_INET6); 1283 rtnl_unregister_all(PF_INET6); 1284 out_sock_register_fail: 1285 rawv6_exit(); 1286 out_unregister_ping_proto: 1287 proto_unregister(&pingv6_prot); 1288 out_unregister_raw_proto: 1289 proto_unregister(&rawv6_prot); 1290 out_unregister_udplite_proto: 1291 proto_unregister(&udplitev6_prot); 1292 out_unregister_udp_proto: 1293 proto_unregister(&udpv6_prot); 1294 out_unregister_tcp_proto: 1295 proto_unregister(&tcpv6_prot); 1296 goto out; 1297 } 1298 module_init(inet6_init); 1299 1300 MODULE_ALIAS_NETPROTO(PF_INET6); 1301