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 #include <net/rawv6.h> 67 68 #include <linux/uaccess.h> 69 #include <linux/mroute6.h> 70 71 #include "ip6_offload.h" 72 73 MODULE_AUTHOR("Cast of dozens"); 74 MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); 75 MODULE_LICENSE("GPL"); 76 77 /* The inetsw6 table contains everything that inet6_create needs to 78 * build a new socket. 79 */ 80 static struct list_head inetsw6[SOCK_MAX]; 81 static DEFINE_SPINLOCK(inetsw6_lock); 82 83 struct ipv6_params ipv6_defaults = { 84 .disable_ipv6 = 0, 85 .autoconf = 1, 86 }; 87 88 static int disable_ipv6_mod; 89 90 module_param_named(disable, disable_ipv6_mod, int, 0444); 91 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional"); 92 93 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444); 94 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces"); 95 96 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444); 97 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces"); 98 99 bool ipv6_mod_enabled(void) 100 { 101 return disable_ipv6_mod == 0; 102 } 103 EXPORT_SYMBOL_GPL(ipv6_mod_enabled); 104 105 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) 106 { 107 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); 108 109 return (struct ipv6_pinfo *)(((u8 *)sk) + offset); 110 } 111 112 void inet6_sock_destruct(struct sock *sk) 113 { 114 inet6_cleanup_sock(sk); 115 inet_sock_destruct(sk); 116 } 117 EXPORT_SYMBOL_GPL(inet6_sock_destruct); 118 119 static int inet6_create(struct net *net, struct socket *sock, int protocol, 120 int kern) 121 { 122 struct inet_sock *inet; 123 struct ipv6_pinfo *np; 124 struct sock *sk; 125 struct inet_protosw *answer; 126 struct proto *answer_prot; 127 unsigned char answer_flags; 128 int try_loading_module = 0; 129 int err; 130 131 if (protocol < 0 || protocol >= IPPROTO_MAX) 132 return -EINVAL; 133 134 /* Look for the requested type/protocol pair. */ 135 lookup_protocol: 136 err = -ESOCKTNOSUPPORT; 137 rcu_read_lock(); 138 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) { 139 140 err = 0; 141 /* Check the non-wild match. */ 142 if (protocol == answer->protocol) { 143 if (protocol != IPPROTO_IP) 144 break; 145 } else { 146 /* Check for the two wild cases. */ 147 if (IPPROTO_IP == protocol) { 148 protocol = answer->protocol; 149 break; 150 } 151 if (IPPROTO_IP == answer->protocol) 152 break; 153 } 154 err = -EPROTONOSUPPORT; 155 } 156 157 if (err) { 158 if (try_loading_module < 2) { 159 rcu_read_unlock(); 160 /* 161 * Be more specific, e.g. net-pf-10-proto-132-type-1 162 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) 163 */ 164 if (++try_loading_module == 1) 165 request_module("net-pf-%d-proto-%d-type-%d", 166 PF_INET6, protocol, sock->type); 167 /* 168 * Fall back to generic, e.g. net-pf-10-proto-132 169 * (net-pf-PF_INET6-proto-IPPROTO_SCTP) 170 */ 171 else 172 request_module("net-pf-%d-proto-%d", 173 PF_INET6, protocol); 174 goto lookup_protocol; 175 } else 176 goto out_rcu_unlock; 177 } 178 179 err = -EPERM; 180 if (sock->type == SOCK_RAW && !kern && 181 !ns_capable(net->user_ns, CAP_NET_RAW)) 182 goto out_rcu_unlock; 183 184 sock->ops = answer->ops; 185 answer_prot = answer->prot; 186 answer_flags = answer->flags; 187 rcu_read_unlock(); 188 189 WARN_ON(!answer_prot->slab); 190 191 err = -ENOBUFS; 192 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern); 193 if (!sk) 194 goto out; 195 196 sock_init_data(sock, sk); 197 198 err = 0; 199 if (INET_PROTOSW_REUSE & answer_flags) 200 sk->sk_reuse = SK_CAN_REUSE; 201 202 inet = inet_sk(sk); 203 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; 204 205 if (SOCK_RAW == sock->type) { 206 inet->inet_num = protocol; 207 if (IPPROTO_RAW == protocol) 208 inet->hdrincl = 1; 209 } 210 211 sk->sk_destruct = inet6_sock_destruct; 212 sk->sk_family = PF_INET6; 213 sk->sk_protocol = protocol; 214 215 sk->sk_backlog_rcv = answer->prot->backlog_rcv; 216 217 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); 218 np->hop_limit = -1; 219 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS; 220 np->mc_loop = 1; 221 np->mc_all = 1; 222 np->pmtudisc = IPV6_PMTUDISC_WANT; 223 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED; 224 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only; 225 sk->sk_txrehash = READ_ONCE(net->core.sysctl_txrehash); 226 227 /* Init the ipv4 part of the socket since we can have sockets 228 * using v6 API for ipv4. 229 */ 230 inet->uc_ttl = -1; 231 232 inet->mc_loop = 1; 233 inet->mc_ttl = 1; 234 inet->mc_index = 0; 235 RCU_INIT_POINTER(inet->mc_list, NULL); 236 inet->rcv_tos = 0; 237 238 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc)) 239 inet->pmtudisc = IP_PMTUDISC_DONT; 240 else 241 inet->pmtudisc = IP_PMTUDISC_WANT; 242 243 if (inet->inet_num) { 244 /* It assumes that any protocol which allows 245 * the user to assign a number at socket 246 * creation time automatically shares. 247 */ 248 inet->inet_sport = htons(inet->inet_num); 249 err = sk->sk_prot->hash(sk); 250 if (err) { 251 sk_common_release(sk); 252 goto out; 253 } 254 } 255 if (sk->sk_prot->init) { 256 err = sk->sk_prot->init(sk); 257 if (err) { 258 sk_common_release(sk); 259 goto out; 260 } 261 } 262 263 if (!kern) { 264 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); 265 if (err) { 266 sk_common_release(sk); 267 goto out; 268 } 269 } 270 out: 271 return err; 272 out_rcu_unlock: 273 rcu_read_unlock(); 274 goto out; 275 } 276 277 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, 278 u32 flags) 279 { 280 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr; 281 struct inet_sock *inet = inet_sk(sk); 282 struct ipv6_pinfo *np = inet6_sk(sk); 283 struct net *net = sock_net(sk); 284 __be32 v4addr = 0; 285 unsigned short snum; 286 bool saved_ipv6only; 287 int addr_type = 0; 288 int err = 0; 289 290 if (addr->sin6_family != AF_INET6) 291 return -EAFNOSUPPORT; 292 293 addr_type = ipv6_addr_type(&addr->sin6_addr); 294 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM) 295 return -EINVAL; 296 297 snum = ntohs(addr->sin6_port); 298 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) && 299 snum && inet_port_requires_bind_service(net, snum) && 300 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 301 return -EACCES; 302 303 if (flags & BIND_WITH_LOCK) 304 lock_sock(sk); 305 306 /* Check these errors (active socket, double bind). */ 307 if (sk->sk_state != TCP_CLOSE || inet->inet_num) { 308 err = -EINVAL; 309 goto out; 310 } 311 312 /* Check if the address belongs to the host. */ 313 if (addr_type == IPV6_ADDR_MAPPED) { 314 struct net_device *dev = NULL; 315 int chk_addr_ret; 316 317 /* Binding to v4-mapped address on a v6-only socket 318 * makes no sense 319 */ 320 if (ipv6_only_sock(sk)) { 321 err = -EINVAL; 322 goto out; 323 } 324 325 rcu_read_lock(); 326 if (sk->sk_bound_dev_if) { 327 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 328 if (!dev) { 329 err = -ENODEV; 330 goto out_unlock; 331 } 332 } 333 334 /* Reproduce AF_INET checks to make the bindings consistent */ 335 v4addr = addr->sin6_addr.s6_addr32[3]; 336 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr); 337 rcu_read_unlock(); 338 339 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr, 340 chk_addr_ret)) { 341 err = -EADDRNOTAVAIL; 342 goto out; 343 } 344 } else { 345 if (addr_type != IPV6_ADDR_ANY) { 346 struct net_device *dev = NULL; 347 348 rcu_read_lock(); 349 if (__ipv6_addr_needs_scope_id(addr_type)) { 350 if (addr_len >= sizeof(struct sockaddr_in6) && 351 addr->sin6_scope_id) { 352 /* Override any existing binding, if another one 353 * is supplied by user. 354 */ 355 sk->sk_bound_dev_if = addr->sin6_scope_id; 356 } 357 358 /* Binding to link-local address requires an interface */ 359 if (!sk->sk_bound_dev_if) { 360 err = -EINVAL; 361 goto out_unlock; 362 } 363 } 364 365 if (sk->sk_bound_dev_if) { 366 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 367 if (!dev) { 368 err = -ENODEV; 369 goto out_unlock; 370 } 371 } 372 373 /* ipv4 addr of the socket is invalid. Only the 374 * unspecified and mapped address have a v4 equivalent. 375 */ 376 v4addr = LOOPBACK4_IPV6; 377 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 378 if (!ipv6_can_nonlocal_bind(net, inet) && 379 !ipv6_chk_addr(net, &addr->sin6_addr, 380 dev, 0)) { 381 err = -EADDRNOTAVAIL; 382 goto out_unlock; 383 } 384 } 385 rcu_read_unlock(); 386 } 387 } 388 389 inet->inet_rcv_saddr = v4addr; 390 inet->inet_saddr = v4addr; 391 392 sk->sk_v6_rcv_saddr = addr->sin6_addr; 393 394 if (!(addr_type & IPV6_ADDR_MULTICAST)) 395 np->saddr = addr->sin6_addr; 396 397 saved_ipv6only = sk->sk_ipv6only; 398 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED) 399 sk->sk_ipv6only = 1; 400 401 /* Make sure we are allowed to bind here. */ 402 if (snum || !(inet->bind_address_no_port || 403 (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 404 err = sk->sk_prot->get_port(sk, snum); 405 if (err) { 406 sk->sk_ipv6only = saved_ipv6only; 407 inet_reset_saddr(sk); 408 goto out; 409 } 410 if (!(flags & BIND_FROM_BPF)) { 411 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk); 412 if (err) { 413 sk->sk_ipv6only = saved_ipv6only; 414 inet_reset_saddr(sk); 415 if (sk->sk_prot->put_port) 416 sk->sk_prot->put_port(sk); 417 goto out; 418 } 419 } 420 } 421 422 if (addr_type != IPV6_ADDR_ANY) 423 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 424 if (snum) 425 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 426 inet->inet_sport = htons(inet->inet_num); 427 inet->inet_dport = 0; 428 inet->inet_daddr = 0; 429 out: 430 if (flags & BIND_WITH_LOCK) 431 release_sock(sk); 432 return err; 433 out_unlock: 434 rcu_read_unlock(); 435 goto out; 436 } 437 438 /* bind for INET6 API */ 439 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 440 { 441 struct sock *sk = sock->sk; 442 u32 flags = BIND_WITH_LOCK; 443 const struct proto *prot; 444 int err = 0; 445 446 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 447 prot = READ_ONCE(sk->sk_prot); 448 /* If the socket has its own bind function then use it. */ 449 if (prot->bind) 450 return prot->bind(sk, uaddr, addr_len); 451 452 if (addr_len < SIN6_LEN_RFC2133) 453 return -EINVAL; 454 455 /* BPF prog is run before any checks are done so that if the prog 456 * changes context in a wrong way it will be caught. 457 */ 458 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 459 CGROUP_INET6_BIND, &flags); 460 if (err) 461 return err; 462 463 return __inet6_bind(sk, uaddr, addr_len, flags); 464 } 465 EXPORT_SYMBOL(inet6_bind); 466 467 int inet6_release(struct socket *sock) 468 { 469 struct sock *sk = sock->sk; 470 471 if (!sk) 472 return -EINVAL; 473 474 /* Free mc lists */ 475 ipv6_sock_mc_close(sk); 476 477 /* Free ac lists */ 478 ipv6_sock_ac_close(sk); 479 480 return inet_release(sock); 481 } 482 EXPORT_SYMBOL(inet6_release); 483 484 void inet6_cleanup_sock(struct sock *sk) 485 { 486 struct ipv6_pinfo *np = inet6_sk(sk); 487 struct sk_buff *skb; 488 struct ipv6_txoptions *opt; 489 490 /* Release rx options */ 491 492 skb = xchg(&np->pktoptions, NULL); 493 kfree_skb(skb); 494 495 skb = xchg(&np->rxpmtu, NULL); 496 kfree_skb(skb); 497 498 /* Free flowlabels */ 499 fl6_free_socklist(sk); 500 501 /* Free tx options */ 502 503 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 504 if (opt) { 505 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 506 txopt_put(opt); 507 } 508 } 509 EXPORT_SYMBOL_GPL(inet6_cleanup_sock); 510 511 /* 512 * This does both peername and sockname. 513 */ 514 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 515 int peer) 516 { 517 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 518 struct sock *sk = sock->sk; 519 struct inet_sock *inet = inet_sk(sk); 520 struct ipv6_pinfo *np = inet6_sk(sk); 521 522 sin->sin6_family = AF_INET6; 523 sin->sin6_flowinfo = 0; 524 sin->sin6_scope_id = 0; 525 lock_sock(sk); 526 if (peer) { 527 if (!inet->inet_dport || 528 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 529 peer == 1)) { 530 release_sock(sk); 531 return -ENOTCONN; 532 } 533 sin->sin6_port = inet->inet_dport; 534 sin->sin6_addr = sk->sk_v6_daddr; 535 if (np->sndflow) 536 sin->sin6_flowinfo = np->flow_label; 537 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 538 CGROUP_INET6_GETPEERNAME); 539 } else { 540 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 541 sin->sin6_addr = np->saddr; 542 else 543 sin->sin6_addr = sk->sk_v6_rcv_saddr; 544 sin->sin6_port = inet->inet_sport; 545 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 546 CGROUP_INET6_GETSOCKNAME); 547 } 548 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 549 sk->sk_bound_dev_if); 550 release_sock(sk); 551 return sizeof(*sin); 552 } 553 EXPORT_SYMBOL(inet6_getname); 554 555 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 556 { 557 void __user *argp = (void __user *)arg; 558 struct sock *sk = sock->sk; 559 struct net *net = sock_net(sk); 560 const struct proto *prot; 561 562 switch (cmd) { 563 case SIOCADDRT: 564 case SIOCDELRT: { 565 struct in6_rtmsg rtmsg; 566 567 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg))) 568 return -EFAULT; 569 return ipv6_route_ioctl(net, cmd, &rtmsg); 570 } 571 case SIOCSIFADDR: 572 return addrconf_add_ifaddr(net, argp); 573 case SIOCDIFADDR: 574 return addrconf_del_ifaddr(net, argp); 575 case SIOCSIFDSTADDR: 576 return addrconf_set_dstaddr(net, argp); 577 default: 578 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 579 prot = READ_ONCE(sk->sk_prot); 580 if (!prot->ioctl) 581 return -ENOIOCTLCMD; 582 return sk_ioctl(sk, cmd, (void __user *)arg); 583 } 584 /*NOTREACHED*/ 585 return 0; 586 } 587 EXPORT_SYMBOL(inet6_ioctl); 588 589 #ifdef CONFIG_COMPAT 590 struct compat_in6_rtmsg { 591 struct in6_addr rtmsg_dst; 592 struct in6_addr rtmsg_src; 593 struct in6_addr rtmsg_gateway; 594 u32 rtmsg_type; 595 u16 rtmsg_dst_len; 596 u16 rtmsg_src_len; 597 u32 rtmsg_metric; 598 u32 rtmsg_info; 599 u32 rtmsg_flags; 600 s32 rtmsg_ifindex; 601 }; 602 603 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 604 struct compat_in6_rtmsg __user *ur) 605 { 606 struct in6_rtmsg rt; 607 608 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst, 609 3 * sizeof(struct in6_addr)) || 610 get_user(rt.rtmsg_type, &ur->rtmsg_type) || 611 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) || 612 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) || 613 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) || 614 get_user(rt.rtmsg_info, &ur->rtmsg_info) || 615 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) || 616 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex)) 617 return -EFAULT; 618 619 620 return ipv6_route_ioctl(sock_net(sk), cmd, &rt); 621 } 622 623 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 624 { 625 void __user *argp = compat_ptr(arg); 626 struct sock *sk = sock->sk; 627 628 switch (cmd) { 629 case SIOCADDRT: 630 case SIOCDELRT: 631 return inet6_compat_routing_ioctl(sk, cmd, argp); 632 default: 633 return -ENOIOCTLCMD; 634 } 635 } 636 EXPORT_SYMBOL_GPL(inet6_compat_ioctl); 637 #endif /* CONFIG_COMPAT */ 638 639 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 640 size_t)); 641 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 642 { 643 struct sock *sk = sock->sk; 644 const struct proto *prot; 645 646 if (unlikely(inet_send_prepare(sk))) 647 return -EAGAIN; 648 649 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 650 prot = READ_ONCE(sk->sk_prot); 651 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 652 sk, msg, size); 653 } 654 655 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 656 size_t, int, int *)); 657 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 658 int flags) 659 { 660 struct sock *sk = sock->sk; 661 const struct proto *prot; 662 int addr_len = 0; 663 int err; 664 665 if (likely(!(flags & MSG_ERRQUEUE))) 666 sock_rps_record_flow(sk); 667 668 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 669 prot = READ_ONCE(sk->sk_prot); 670 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 671 sk, msg, size, flags, &addr_len); 672 if (err >= 0) 673 msg->msg_namelen = addr_len; 674 return err; 675 } 676 677 const struct proto_ops inet6_stream_ops = { 678 .family = PF_INET6, 679 .owner = THIS_MODULE, 680 .release = inet6_release, 681 .bind = inet6_bind, 682 .connect = inet_stream_connect, /* ok */ 683 .socketpair = sock_no_socketpair, /* a do nothing */ 684 .accept = inet_accept, /* ok */ 685 .getname = inet6_getname, 686 .poll = tcp_poll, /* ok */ 687 .ioctl = inet6_ioctl, /* must change */ 688 .gettstamp = sock_gettstamp, 689 .listen = inet_listen, /* ok */ 690 .shutdown = inet_shutdown, /* ok */ 691 .setsockopt = sock_common_setsockopt, /* ok */ 692 .getsockopt = sock_common_getsockopt, /* ok */ 693 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 694 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 695 #ifdef CONFIG_MMU 696 .mmap = tcp_mmap, 697 #endif 698 .splice_eof = inet_splice_eof, 699 .sendmsg_locked = tcp_sendmsg_locked, 700 .splice_read = tcp_splice_read, 701 .read_sock = tcp_read_sock, 702 .read_skb = tcp_read_skb, 703 .peek_len = tcp_peek_len, 704 #ifdef CONFIG_COMPAT 705 .compat_ioctl = inet6_compat_ioctl, 706 #endif 707 .set_rcvlowat = tcp_set_rcvlowat, 708 }; 709 710 const struct proto_ops inet6_dgram_ops = { 711 .family = PF_INET6, 712 .owner = THIS_MODULE, 713 .release = inet6_release, 714 .bind = inet6_bind, 715 .connect = inet_dgram_connect, /* ok */ 716 .socketpair = sock_no_socketpair, /* a do nothing */ 717 .accept = sock_no_accept, /* a do nothing */ 718 .getname = inet6_getname, 719 .poll = udp_poll, /* ok */ 720 .ioctl = inet6_ioctl, /* must change */ 721 .gettstamp = sock_gettstamp, 722 .listen = sock_no_listen, /* ok */ 723 .shutdown = inet_shutdown, /* ok */ 724 .setsockopt = sock_common_setsockopt, /* ok */ 725 .getsockopt = sock_common_getsockopt, /* ok */ 726 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 727 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 728 .read_skb = udp_read_skb, 729 .mmap = sock_no_mmap, 730 .set_peek_off = sk_set_peek_off, 731 #ifdef CONFIG_COMPAT 732 .compat_ioctl = inet6_compat_ioctl, 733 #endif 734 }; 735 736 static const struct net_proto_family inet6_family_ops = { 737 .family = PF_INET6, 738 .create = inet6_create, 739 .owner = THIS_MODULE, 740 }; 741 742 int inet6_register_protosw(struct inet_protosw *p) 743 { 744 struct list_head *lh; 745 struct inet_protosw *answer; 746 struct list_head *last_perm; 747 int protocol = p->protocol; 748 int ret; 749 750 spin_lock_bh(&inetsw6_lock); 751 752 ret = -EINVAL; 753 if (p->type >= SOCK_MAX) 754 goto out_illegal; 755 756 /* If we are trying to override a permanent protocol, bail. */ 757 answer = NULL; 758 ret = -EPERM; 759 last_perm = &inetsw6[p->type]; 760 list_for_each(lh, &inetsw6[p->type]) { 761 answer = list_entry(lh, struct inet_protosw, list); 762 763 /* Check only the non-wild match. */ 764 if (INET_PROTOSW_PERMANENT & answer->flags) { 765 if (protocol == answer->protocol) 766 break; 767 last_perm = lh; 768 } 769 770 answer = NULL; 771 } 772 if (answer) 773 goto out_permanent; 774 775 /* Add the new entry after the last permanent entry if any, so that 776 * the new entry does not override a permanent entry when matched with 777 * a wild-card protocol. But it is allowed to override any existing 778 * non-permanent entry. This means that when we remove this entry, the 779 * system automatically returns to the old behavior. 780 */ 781 list_add_rcu(&p->list, last_perm); 782 ret = 0; 783 out: 784 spin_unlock_bh(&inetsw6_lock); 785 return ret; 786 787 out_permanent: 788 pr_err("Attempt to override permanent protocol %d\n", protocol); 789 goto out; 790 791 out_illegal: 792 pr_err("Ignoring attempt to register invalid socket type %d\n", 793 p->type); 794 goto out; 795 } 796 EXPORT_SYMBOL(inet6_register_protosw); 797 798 void 799 inet6_unregister_protosw(struct inet_protosw *p) 800 { 801 if (INET_PROTOSW_PERMANENT & p->flags) { 802 pr_err("Attempt to unregister permanent protocol %d\n", 803 p->protocol); 804 } else { 805 spin_lock_bh(&inetsw6_lock); 806 list_del_rcu(&p->list); 807 spin_unlock_bh(&inetsw6_lock); 808 809 synchronize_net(); 810 } 811 } 812 EXPORT_SYMBOL(inet6_unregister_protosw); 813 814 int inet6_sk_rebuild_header(struct sock *sk) 815 { 816 struct ipv6_pinfo *np = inet6_sk(sk); 817 struct dst_entry *dst; 818 819 dst = __sk_dst_check(sk, np->dst_cookie); 820 821 if (!dst) { 822 struct inet_sock *inet = inet_sk(sk); 823 struct in6_addr *final_p, final; 824 struct flowi6 fl6; 825 826 memset(&fl6, 0, sizeof(fl6)); 827 fl6.flowi6_proto = sk->sk_protocol; 828 fl6.daddr = sk->sk_v6_daddr; 829 fl6.saddr = np->saddr; 830 fl6.flowlabel = np->flow_label; 831 fl6.flowi6_oif = sk->sk_bound_dev_if; 832 fl6.flowi6_mark = sk->sk_mark; 833 fl6.fl6_dport = inet->inet_dport; 834 fl6.fl6_sport = inet->inet_sport; 835 fl6.flowi6_uid = sk->sk_uid; 836 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); 837 838 rcu_read_lock(); 839 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 840 &final); 841 rcu_read_unlock(); 842 843 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 844 if (IS_ERR(dst)) { 845 sk->sk_route_caps = 0; 846 WRITE_ONCE(sk->sk_err_soft, -PTR_ERR(dst)); 847 return PTR_ERR(dst); 848 } 849 850 ip6_dst_store(sk, dst, NULL, NULL); 851 } 852 853 return 0; 854 } 855 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 856 857 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 858 const struct inet6_skb_parm *opt) 859 { 860 const struct ipv6_pinfo *np = inet6_sk(sk); 861 862 if (np->rxopt.all) { 863 if (((opt->flags & IP6SKB_HOPBYHOP) && 864 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 865 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 866 np->rxopt.bits.rxflow) || 867 (opt->srcrt && (np->rxopt.bits.srcrt || 868 np->rxopt.bits.osrcrt)) || 869 ((opt->dst1 || opt->dst0) && 870 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 871 return true; 872 } 873 return false; 874 } 875 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 876 877 static struct packet_type ipv6_packet_type __read_mostly = { 878 .type = cpu_to_be16(ETH_P_IPV6), 879 .func = ipv6_rcv, 880 .list_func = ipv6_list_rcv, 881 }; 882 883 static int __init ipv6_packet_init(void) 884 { 885 dev_add_pack(&ipv6_packet_type); 886 return 0; 887 } 888 889 static void ipv6_packet_cleanup(void) 890 { 891 dev_remove_pack(&ipv6_packet_type); 892 } 893 894 static int __net_init ipv6_init_mibs(struct net *net) 895 { 896 int i; 897 898 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 899 if (!net->mib.udp_stats_in6) 900 return -ENOMEM; 901 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 902 if (!net->mib.udplite_stats_in6) 903 goto err_udplite_mib; 904 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 905 if (!net->mib.ipv6_statistics) 906 goto err_ip_mib; 907 908 for_each_possible_cpu(i) { 909 struct ipstats_mib *af_inet6_stats; 910 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 911 u64_stats_init(&af_inet6_stats->syncp); 912 } 913 914 915 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 916 if (!net->mib.icmpv6_statistics) 917 goto err_icmp_mib; 918 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 919 GFP_KERNEL); 920 if (!net->mib.icmpv6msg_statistics) 921 goto err_icmpmsg_mib; 922 return 0; 923 924 err_icmpmsg_mib: 925 free_percpu(net->mib.icmpv6_statistics); 926 err_icmp_mib: 927 free_percpu(net->mib.ipv6_statistics); 928 err_ip_mib: 929 free_percpu(net->mib.udplite_stats_in6); 930 err_udplite_mib: 931 free_percpu(net->mib.udp_stats_in6); 932 return -ENOMEM; 933 } 934 935 static void ipv6_cleanup_mibs(struct net *net) 936 { 937 free_percpu(net->mib.udp_stats_in6); 938 free_percpu(net->mib.udplite_stats_in6); 939 free_percpu(net->mib.ipv6_statistics); 940 free_percpu(net->mib.icmpv6_statistics); 941 kfree(net->mib.icmpv6msg_statistics); 942 } 943 944 static int __net_init inet6_net_init(struct net *net) 945 { 946 int err = 0; 947 948 net->ipv6.sysctl.bindv6only = 0; 949 net->ipv6.sysctl.icmpv6_time = 1*HZ; 950 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 951 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0; 952 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0; 953 net->ipv6.sysctl.icmpv6_error_anycast_as_unicast = 0; 954 955 /* By default, rate limit error messages. 956 * Except for pmtu discovery, it would break it. 957 * proc_do_large_bitmap needs pointer to the bitmap. 958 */ 959 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 960 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 961 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 962 963 net->ipv6.sysctl.flowlabel_consistency = 1; 964 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 965 net->ipv6.sysctl.idgen_retries = 3; 966 net->ipv6.sysctl.idgen_delay = 1 * HZ; 967 net->ipv6.sysctl.flowlabel_state_ranges = 0; 968 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 969 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 970 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 971 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 972 net->ipv6.sysctl.fib_notify_on_flag_change = 0; 973 atomic_set(&net->ipv6.fib6_sernum, 1); 974 975 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID; 976 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE; 977 978 err = ipv6_init_mibs(net); 979 if (err) 980 return err; 981 #ifdef CONFIG_PROC_FS 982 err = udp6_proc_init(net); 983 if (err) 984 goto out; 985 err = tcp6_proc_init(net); 986 if (err) 987 goto proc_tcp6_fail; 988 err = ac6_proc_init(net); 989 if (err) 990 goto proc_ac6_fail; 991 #endif 992 return err; 993 994 #ifdef CONFIG_PROC_FS 995 proc_ac6_fail: 996 tcp6_proc_exit(net); 997 proc_tcp6_fail: 998 udp6_proc_exit(net); 999 out: 1000 ipv6_cleanup_mibs(net); 1001 return err; 1002 #endif 1003 } 1004 1005 static void __net_exit inet6_net_exit(struct net *net) 1006 { 1007 #ifdef CONFIG_PROC_FS 1008 udp6_proc_exit(net); 1009 tcp6_proc_exit(net); 1010 ac6_proc_exit(net); 1011 #endif 1012 ipv6_cleanup_mibs(net); 1013 } 1014 1015 static struct pernet_operations inet6_net_ops = { 1016 .init = inet6_net_init, 1017 .exit = inet6_net_exit, 1018 }; 1019 1020 static int ipv6_route_input(struct sk_buff *skb) 1021 { 1022 ip6_route_input(skb); 1023 return skb_dst(skb)->error; 1024 } 1025 1026 static const struct ipv6_stub ipv6_stub_impl = { 1027 .ipv6_sock_mc_join = ipv6_sock_mc_join, 1028 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 1029 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 1030 .ipv6_route_input = ipv6_route_input, 1031 .fib6_get_table = fib6_get_table, 1032 .fib6_table_lookup = fib6_table_lookup, 1033 .fib6_lookup = fib6_lookup, 1034 .fib6_select_path = fib6_select_path, 1035 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 1036 .fib6_nh_init = fib6_nh_init, 1037 .fib6_nh_release = fib6_nh_release, 1038 .fib6_nh_release_dsts = fib6_nh_release_dsts, 1039 .fib6_update_sernum = fib6_update_sernum_stub, 1040 .fib6_rt_update = fib6_rt_update, 1041 .ip6_del_rt = ip6_del_rt, 1042 .udpv6_encap_enable = udpv6_encap_enable, 1043 .ndisc_send_na = ndisc_send_na, 1044 #if IS_ENABLED(CONFIG_XFRM) 1045 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu, 1046 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv, 1047 .xfrm6_rcv_encap = xfrm6_rcv_encap, 1048 #endif 1049 .nd_tbl = &nd_tbl, 1050 .ipv6_fragment = ip6_fragment, 1051 .ipv6_dev_find = ipv6_dev_find, 1052 }; 1053 1054 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 1055 .inet6_bind = __inet6_bind, 1056 .udp6_lib_lookup = __udp6_lib_lookup, 1057 .ipv6_setsockopt = do_ipv6_setsockopt, 1058 .ipv6_getsockopt = do_ipv6_getsockopt, 1059 }; 1060 1061 static int __init inet6_init(void) 1062 { 1063 struct list_head *r; 1064 int err = 0; 1065 1066 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 1067 1068 /* Register the socket-side information for inet6_create. */ 1069 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 1070 INIT_LIST_HEAD(r); 1071 1072 raw_hashinfo_init(&raw_v6_hashinfo); 1073 1074 if (disable_ipv6_mod) { 1075 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 1076 goto out; 1077 } 1078 1079 err = proto_register(&tcpv6_prot, 1); 1080 if (err) 1081 goto out; 1082 1083 err = proto_register(&udpv6_prot, 1); 1084 if (err) 1085 goto out_unregister_tcp_proto; 1086 1087 err = proto_register(&udplitev6_prot, 1); 1088 if (err) 1089 goto out_unregister_udp_proto; 1090 1091 err = proto_register(&rawv6_prot, 1); 1092 if (err) 1093 goto out_unregister_udplite_proto; 1094 1095 err = proto_register(&pingv6_prot, 1); 1096 if (err) 1097 goto out_unregister_raw_proto; 1098 1099 /* We MUST register RAW sockets before we create the ICMP6, 1100 * IGMP6, or NDISC control sockets. 1101 */ 1102 err = rawv6_init(); 1103 if (err) 1104 goto out_unregister_ping_proto; 1105 1106 /* Register the family here so that the init calls below will 1107 * be able to create sockets. (?? is this dangerous ??) 1108 */ 1109 err = sock_register(&inet6_family_ops); 1110 if (err) 1111 goto out_sock_register_fail; 1112 1113 /* 1114 * ipngwg API draft makes clear that the correct semantics 1115 * for TCP and UDP is to consider one TCP and UDP instance 1116 * in a host available by both INET and INET6 APIs and 1117 * able to communicate via both network protocols. 1118 */ 1119 1120 err = register_pernet_subsys(&inet6_net_ops); 1121 if (err) 1122 goto register_pernet_fail; 1123 err = ip6_mr_init(); 1124 if (err) 1125 goto ipmr_fail; 1126 err = icmpv6_init(); 1127 if (err) 1128 goto icmp_fail; 1129 err = ndisc_init(); 1130 if (err) 1131 goto ndisc_fail; 1132 err = igmp6_init(); 1133 if (err) 1134 goto igmp_fail; 1135 1136 err = ipv6_netfilter_init(); 1137 if (err) 1138 goto netfilter_fail; 1139 /* Create /proc/foo6 entries. */ 1140 #ifdef CONFIG_PROC_FS 1141 err = -ENOMEM; 1142 if (raw6_proc_init()) 1143 goto proc_raw6_fail; 1144 if (udplite6_proc_init()) 1145 goto proc_udplite6_fail; 1146 if (ipv6_misc_proc_init()) 1147 goto proc_misc6_fail; 1148 if (if6_proc_init()) 1149 goto proc_if6_fail; 1150 #endif 1151 err = ip6_route_init(); 1152 if (err) 1153 goto ip6_route_fail; 1154 err = ndisc_late_init(); 1155 if (err) 1156 goto ndisc_late_fail; 1157 err = ip6_flowlabel_init(); 1158 if (err) 1159 goto ip6_flowlabel_fail; 1160 err = ipv6_anycast_init(); 1161 if (err) 1162 goto ipv6_anycast_fail; 1163 err = addrconf_init(); 1164 if (err) 1165 goto addrconf_fail; 1166 1167 /* Init v6 extension headers. */ 1168 err = ipv6_exthdrs_init(); 1169 if (err) 1170 goto ipv6_exthdrs_fail; 1171 1172 err = ipv6_frag_init(); 1173 if (err) 1174 goto ipv6_frag_fail; 1175 1176 /* Init v6 transport protocols. */ 1177 err = udpv6_init(); 1178 if (err) 1179 goto udpv6_fail; 1180 1181 err = udplitev6_init(); 1182 if (err) 1183 goto udplitev6_fail; 1184 1185 err = udpv6_offload_init(); 1186 if (err) 1187 goto udpv6_offload_fail; 1188 1189 err = tcpv6_init(); 1190 if (err) 1191 goto tcpv6_fail; 1192 1193 err = ipv6_packet_init(); 1194 if (err) 1195 goto ipv6_packet_fail; 1196 1197 err = pingv6_init(); 1198 if (err) 1199 goto pingv6_fail; 1200 1201 err = calipso_init(); 1202 if (err) 1203 goto calipso_fail; 1204 1205 err = seg6_init(); 1206 if (err) 1207 goto seg6_fail; 1208 1209 err = rpl_init(); 1210 if (err) 1211 goto rpl_fail; 1212 1213 err = ioam6_init(); 1214 if (err) 1215 goto ioam6_fail; 1216 1217 err = igmp6_late_init(); 1218 if (err) 1219 goto igmp6_late_err; 1220 1221 #ifdef CONFIG_SYSCTL 1222 err = ipv6_sysctl_register(); 1223 if (err) 1224 goto sysctl_fail; 1225 #endif 1226 1227 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1228 wmb(); 1229 ipv6_stub = &ipv6_stub_impl; 1230 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1231 out: 1232 return err; 1233 1234 #ifdef CONFIG_SYSCTL 1235 sysctl_fail: 1236 igmp6_late_cleanup(); 1237 #endif 1238 igmp6_late_err: 1239 ioam6_exit(); 1240 ioam6_fail: 1241 rpl_exit(); 1242 rpl_fail: 1243 seg6_exit(); 1244 seg6_fail: 1245 calipso_exit(); 1246 calipso_fail: 1247 pingv6_exit(); 1248 pingv6_fail: 1249 ipv6_packet_cleanup(); 1250 ipv6_packet_fail: 1251 tcpv6_exit(); 1252 tcpv6_fail: 1253 udpv6_offload_exit(); 1254 udpv6_offload_fail: 1255 udplitev6_exit(); 1256 udplitev6_fail: 1257 udpv6_exit(); 1258 udpv6_fail: 1259 ipv6_frag_exit(); 1260 ipv6_frag_fail: 1261 ipv6_exthdrs_exit(); 1262 ipv6_exthdrs_fail: 1263 addrconf_cleanup(); 1264 addrconf_fail: 1265 ipv6_anycast_cleanup(); 1266 ipv6_anycast_fail: 1267 ip6_flowlabel_cleanup(); 1268 ip6_flowlabel_fail: 1269 ndisc_late_cleanup(); 1270 ndisc_late_fail: 1271 ip6_route_cleanup(); 1272 ip6_route_fail: 1273 #ifdef CONFIG_PROC_FS 1274 if6_proc_exit(); 1275 proc_if6_fail: 1276 ipv6_misc_proc_exit(); 1277 proc_misc6_fail: 1278 udplite6_proc_exit(); 1279 proc_udplite6_fail: 1280 raw6_proc_exit(); 1281 proc_raw6_fail: 1282 #endif 1283 ipv6_netfilter_fini(); 1284 netfilter_fail: 1285 igmp6_cleanup(); 1286 igmp_fail: 1287 ndisc_cleanup(); 1288 ndisc_fail: 1289 icmpv6_cleanup(); 1290 icmp_fail: 1291 ip6_mr_cleanup(); 1292 ipmr_fail: 1293 unregister_pernet_subsys(&inet6_net_ops); 1294 register_pernet_fail: 1295 sock_unregister(PF_INET6); 1296 rtnl_unregister_all(PF_INET6); 1297 out_sock_register_fail: 1298 rawv6_exit(); 1299 out_unregister_ping_proto: 1300 proto_unregister(&pingv6_prot); 1301 out_unregister_raw_proto: 1302 proto_unregister(&rawv6_prot); 1303 out_unregister_udplite_proto: 1304 proto_unregister(&udplitev6_prot); 1305 out_unregister_udp_proto: 1306 proto_unregister(&udpv6_prot); 1307 out_unregister_tcp_proto: 1308 proto_unregister(&tcpv6_prot); 1309 goto out; 1310 } 1311 module_init(inet6_init); 1312 1313 MODULE_ALIAS_NETPROTO(PF_INET6); 1314