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 static int inet6_create(struct net *net, struct socket *sock, int protocol, 113 int kern) 114 { 115 struct inet_sock *inet; 116 struct ipv6_pinfo *np; 117 struct sock *sk; 118 struct inet_protosw *answer; 119 struct proto *answer_prot; 120 unsigned char answer_flags; 121 int try_loading_module = 0; 122 int err; 123 124 if (protocol < 0 || protocol >= IPPROTO_MAX) 125 return -EINVAL; 126 127 /* Look for the requested type/protocol pair. */ 128 lookup_protocol: 129 err = -ESOCKTNOSUPPORT; 130 rcu_read_lock(); 131 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) { 132 133 err = 0; 134 /* Check the non-wild match. */ 135 if (protocol == answer->protocol) { 136 if (protocol != IPPROTO_IP) 137 break; 138 } else { 139 /* Check for the two wild cases. */ 140 if (IPPROTO_IP == protocol) { 141 protocol = answer->protocol; 142 break; 143 } 144 if (IPPROTO_IP == answer->protocol) 145 break; 146 } 147 err = -EPROTONOSUPPORT; 148 } 149 150 if (err) { 151 if (try_loading_module < 2) { 152 rcu_read_unlock(); 153 /* 154 * Be more specific, e.g. net-pf-10-proto-132-type-1 155 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) 156 */ 157 if (++try_loading_module == 1) 158 request_module("net-pf-%d-proto-%d-type-%d", 159 PF_INET6, protocol, sock->type); 160 /* 161 * Fall back to generic, e.g. net-pf-10-proto-132 162 * (net-pf-PF_INET6-proto-IPPROTO_SCTP) 163 */ 164 else 165 request_module("net-pf-%d-proto-%d", 166 PF_INET6, protocol); 167 goto lookup_protocol; 168 } else 169 goto out_rcu_unlock; 170 } 171 172 err = -EPERM; 173 if (sock->type == SOCK_RAW && !kern && 174 !ns_capable(net->user_ns, CAP_NET_RAW)) 175 goto out_rcu_unlock; 176 177 sock->ops = answer->ops; 178 answer_prot = answer->prot; 179 answer_flags = answer->flags; 180 rcu_read_unlock(); 181 182 WARN_ON(!answer_prot->slab); 183 184 err = -ENOBUFS; 185 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern); 186 if (!sk) 187 goto out; 188 189 sock_init_data(sock, sk); 190 191 err = 0; 192 if (INET_PROTOSW_REUSE & answer_flags) 193 sk->sk_reuse = SK_CAN_REUSE; 194 195 inet = inet_sk(sk); 196 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; 197 198 if (SOCK_RAW == sock->type) { 199 inet->inet_num = protocol; 200 if (IPPROTO_RAW == protocol) 201 inet->hdrincl = 1; 202 } 203 204 sk->sk_destruct = inet_sock_destruct; 205 sk->sk_family = PF_INET6; 206 sk->sk_protocol = protocol; 207 208 sk->sk_backlog_rcv = answer->prot->backlog_rcv; 209 210 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); 211 np->hop_limit = -1; 212 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS; 213 np->mc_loop = 1; 214 np->mc_all = 1; 215 np->pmtudisc = IPV6_PMTUDISC_WANT; 216 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED; 217 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only; 218 219 /* Init the ipv4 part of the socket since we can have sockets 220 * using v6 API for ipv4. 221 */ 222 inet->uc_ttl = -1; 223 224 inet->mc_loop = 1; 225 inet->mc_ttl = 1; 226 inet->mc_index = 0; 227 RCU_INIT_POINTER(inet->mc_list, NULL); 228 inet->rcv_tos = 0; 229 230 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc)) 231 inet->pmtudisc = IP_PMTUDISC_DONT; 232 else 233 inet->pmtudisc = IP_PMTUDISC_WANT; 234 /* 235 * Increment only the relevant sk_prot->socks debug field, this changes 236 * the previous behaviour of incrementing both the equivalent to 237 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr. 238 * 239 * This allows better debug granularity as we'll know exactly how many 240 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6 241 * transport protocol socks. -acme 242 */ 243 sk_refcnt_debug_inc(sk); 244 245 if (inet->inet_num) { 246 /* It assumes that any protocol which allows 247 * the user to assign a number at socket 248 * creation time automatically shares. 249 */ 250 inet->inet_sport = htons(inet->inet_num); 251 err = sk->sk_prot->hash(sk); 252 if (err) { 253 sk_common_release(sk); 254 goto out; 255 } 256 } 257 if (sk->sk_prot->init) { 258 err = sk->sk_prot->init(sk); 259 if (err) { 260 sk_common_release(sk); 261 goto out; 262 } 263 } 264 265 if (!kern) { 266 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); 267 if (err) { 268 sk_common_release(sk); 269 goto out; 270 } 271 } 272 out: 273 return err; 274 out_rcu_unlock: 275 rcu_read_unlock(); 276 goto out; 277 } 278 279 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, 280 u32 flags) 281 { 282 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr; 283 struct inet_sock *inet = inet_sk(sk); 284 struct ipv6_pinfo *np = inet6_sk(sk); 285 struct net *net = sock_net(sk); 286 __be32 v4addr = 0; 287 unsigned short snum; 288 bool saved_ipv6only; 289 int addr_type = 0; 290 int err = 0; 291 292 if (addr->sin6_family != AF_INET6) 293 return -EAFNOSUPPORT; 294 295 addr_type = ipv6_addr_type(&addr->sin6_addr); 296 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM) 297 return -EINVAL; 298 299 snum = ntohs(addr->sin6_port); 300 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) && 301 snum && inet_port_requires_bind_service(net, snum) && 302 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 303 return -EACCES; 304 305 if (flags & BIND_WITH_LOCK) 306 lock_sock(sk); 307 308 /* Check these errors (active socket, double bind). */ 309 if (sk->sk_state != TCP_CLOSE || inet->inet_num) { 310 err = -EINVAL; 311 goto out; 312 } 313 314 /* Check if the address belongs to the host. */ 315 if (addr_type == IPV6_ADDR_MAPPED) { 316 struct net_device *dev = NULL; 317 int chk_addr_ret; 318 319 /* Binding to v4-mapped address on a v6-only socket 320 * makes no sense 321 */ 322 if (ipv6_only_sock(sk)) { 323 err = -EINVAL; 324 goto out; 325 } 326 327 rcu_read_lock(); 328 if (sk->sk_bound_dev_if) { 329 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 330 if (!dev) { 331 err = -ENODEV; 332 goto out_unlock; 333 } 334 } 335 336 /* Reproduce AF_INET checks to make the bindings consistent */ 337 v4addr = addr->sin6_addr.s6_addr32[3]; 338 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr); 339 rcu_read_unlock(); 340 341 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr, 342 chk_addr_ret)) { 343 err = -EADDRNOTAVAIL; 344 goto out; 345 } 346 } else { 347 if (addr_type != IPV6_ADDR_ANY) { 348 struct net_device *dev = NULL; 349 350 rcu_read_lock(); 351 if (__ipv6_addr_needs_scope_id(addr_type)) { 352 if (addr_len >= sizeof(struct sockaddr_in6) && 353 addr->sin6_scope_id) { 354 /* Override any existing binding, if another one 355 * is supplied by user. 356 */ 357 sk->sk_bound_dev_if = addr->sin6_scope_id; 358 } 359 360 /* Binding to link-local address requires an interface */ 361 if (!sk->sk_bound_dev_if) { 362 err = -EINVAL; 363 goto out_unlock; 364 } 365 } 366 367 if (sk->sk_bound_dev_if) { 368 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 369 if (!dev) { 370 err = -ENODEV; 371 goto out_unlock; 372 } 373 } 374 375 /* ipv4 addr of the socket is invalid. Only the 376 * unspecified and mapped address have a v4 equivalent. 377 */ 378 v4addr = LOOPBACK4_IPV6; 379 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 380 if (!ipv6_can_nonlocal_bind(net, inet) && 381 !ipv6_chk_addr(net, &addr->sin6_addr, 382 dev, 0)) { 383 err = -EADDRNOTAVAIL; 384 goto out_unlock; 385 } 386 } 387 rcu_read_unlock(); 388 } 389 } 390 391 inet->inet_rcv_saddr = v4addr; 392 inet->inet_saddr = v4addr; 393 394 sk->sk_v6_rcv_saddr = addr->sin6_addr; 395 396 if (!(addr_type & IPV6_ADDR_MULTICAST)) 397 np->saddr = addr->sin6_addr; 398 399 saved_ipv6only = sk->sk_ipv6only; 400 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED) 401 sk->sk_ipv6only = 1; 402 403 /* Make sure we are allowed to bind here. */ 404 if (snum || !(inet->bind_address_no_port || 405 (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 406 if (sk->sk_prot->get_port(sk, snum)) { 407 sk->sk_ipv6only = saved_ipv6only; 408 inet_reset_saddr(sk); 409 err = -EADDRINUSE; 410 goto out; 411 } 412 if (!(flags & BIND_FROM_BPF)) { 413 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk); 414 if (err) { 415 sk->sk_ipv6only = saved_ipv6only; 416 inet_reset_saddr(sk); 417 if (sk->sk_prot->put_port) 418 sk->sk_prot->put_port(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 const struct proto *prot; 446 int err = 0; 447 448 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 449 prot = READ_ONCE(sk->sk_prot); 450 /* If the socket has its own bind function then use it. */ 451 if (prot->bind) 452 return prot->bind(sk, uaddr, addr_len); 453 454 if (addr_len < SIN6_LEN_RFC2133) 455 return -EINVAL; 456 457 /* BPF prog is run before any checks are done so that if the prog 458 * changes context in a wrong way it will be caught. 459 */ 460 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 461 CGROUP_INET6_BIND, &flags); 462 if (err) 463 return err; 464 465 return __inet6_bind(sk, uaddr, addr_len, flags); 466 } 467 EXPORT_SYMBOL(inet6_bind); 468 469 int inet6_release(struct socket *sock) 470 { 471 struct sock *sk = sock->sk; 472 473 if (!sk) 474 return -EINVAL; 475 476 /* Free mc lists */ 477 ipv6_sock_mc_close(sk); 478 479 /* Free ac lists */ 480 ipv6_sock_ac_close(sk); 481 482 return inet_release(sock); 483 } 484 EXPORT_SYMBOL(inet6_release); 485 486 void inet6_destroy_sock(struct sock *sk) 487 { 488 struct ipv6_pinfo *np = inet6_sk(sk); 489 struct sk_buff *skb; 490 struct ipv6_txoptions *opt; 491 492 /* Release rx options */ 493 494 skb = xchg(&np->pktoptions, NULL); 495 kfree_skb(skb); 496 497 skb = xchg(&np->rxpmtu, NULL); 498 kfree_skb(skb); 499 500 /* Free flowlabels */ 501 fl6_free_socklist(sk); 502 503 /* Free tx options */ 504 505 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 506 if (opt) { 507 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 508 txopt_put(opt); 509 } 510 } 511 EXPORT_SYMBOL_GPL(inet6_destroy_sock); 512 513 /* 514 * This does both peername and sockname. 515 */ 516 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 517 int peer) 518 { 519 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 520 struct sock *sk = sock->sk; 521 struct inet_sock *inet = inet_sk(sk); 522 struct ipv6_pinfo *np = inet6_sk(sk); 523 524 sin->sin6_family = AF_INET6; 525 sin->sin6_flowinfo = 0; 526 sin->sin6_scope_id = 0; 527 lock_sock(sk); 528 if (peer) { 529 if (!inet->inet_dport || 530 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 531 peer == 1)) { 532 release_sock(sk); 533 return -ENOTCONN; 534 } 535 sin->sin6_port = inet->inet_dport; 536 sin->sin6_addr = sk->sk_v6_daddr; 537 if (np->sndflow) 538 sin->sin6_flowinfo = np->flow_label; 539 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 540 CGROUP_INET6_GETPEERNAME); 541 } else { 542 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 543 sin->sin6_addr = np->saddr; 544 else 545 sin->sin6_addr = sk->sk_v6_rcv_saddr; 546 sin->sin6_port = inet->inet_sport; 547 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 548 CGROUP_INET6_GETSOCKNAME); 549 } 550 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 551 sk->sk_bound_dev_if); 552 release_sock(sk); 553 return sizeof(*sin); 554 } 555 EXPORT_SYMBOL(inet6_getname); 556 557 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 558 { 559 void __user *argp = (void __user *)arg; 560 struct sock *sk = sock->sk; 561 struct net *net = sock_net(sk); 562 const struct proto *prot; 563 564 switch (cmd) { 565 case SIOCADDRT: 566 case SIOCDELRT: { 567 struct in6_rtmsg rtmsg; 568 569 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg))) 570 return -EFAULT; 571 return ipv6_route_ioctl(net, cmd, &rtmsg); 572 } 573 case SIOCSIFADDR: 574 return addrconf_add_ifaddr(net, argp); 575 case SIOCDIFADDR: 576 return addrconf_del_ifaddr(net, argp); 577 case SIOCSIFDSTADDR: 578 return addrconf_set_dstaddr(net, argp); 579 default: 580 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 581 prot = READ_ONCE(sk->sk_prot); 582 if (!prot->ioctl) 583 return -ENOIOCTLCMD; 584 return prot->ioctl(sk, cmd, arg); 585 } 586 /*NOTREACHED*/ 587 return 0; 588 } 589 EXPORT_SYMBOL(inet6_ioctl); 590 591 #ifdef CONFIG_COMPAT 592 struct compat_in6_rtmsg { 593 struct in6_addr rtmsg_dst; 594 struct in6_addr rtmsg_src; 595 struct in6_addr rtmsg_gateway; 596 u32 rtmsg_type; 597 u16 rtmsg_dst_len; 598 u16 rtmsg_src_len; 599 u32 rtmsg_metric; 600 u32 rtmsg_info; 601 u32 rtmsg_flags; 602 s32 rtmsg_ifindex; 603 }; 604 605 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 606 struct compat_in6_rtmsg __user *ur) 607 { 608 struct in6_rtmsg rt; 609 610 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst, 611 3 * sizeof(struct in6_addr)) || 612 get_user(rt.rtmsg_type, &ur->rtmsg_type) || 613 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) || 614 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) || 615 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) || 616 get_user(rt.rtmsg_info, &ur->rtmsg_info) || 617 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) || 618 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex)) 619 return -EFAULT; 620 621 622 return ipv6_route_ioctl(sock_net(sk), cmd, &rt); 623 } 624 625 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 626 { 627 void __user *argp = compat_ptr(arg); 628 struct sock *sk = sock->sk; 629 630 switch (cmd) { 631 case SIOCADDRT: 632 case SIOCDELRT: 633 return inet6_compat_routing_ioctl(sk, cmd, argp); 634 default: 635 return -ENOIOCTLCMD; 636 } 637 } 638 EXPORT_SYMBOL_GPL(inet6_compat_ioctl); 639 #endif /* CONFIG_COMPAT */ 640 641 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 642 size_t)); 643 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 644 { 645 struct sock *sk = sock->sk; 646 const struct proto *prot; 647 648 if (unlikely(inet_send_prepare(sk))) 649 return -EAGAIN; 650 651 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 652 prot = READ_ONCE(sk->sk_prot); 653 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 654 sk, msg, size); 655 } 656 657 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 658 size_t, int, int *)); 659 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 660 int flags) 661 { 662 struct sock *sk = sock->sk; 663 const struct proto *prot; 664 int addr_len = 0; 665 int err; 666 667 if (likely(!(flags & MSG_ERRQUEUE))) 668 sock_rps_record_flow(sk); 669 670 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 671 prot = READ_ONCE(sk->sk_prot); 672 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 673 sk, msg, size, flags, &addr_len); 674 if (err >= 0) 675 msg->msg_namelen = addr_len; 676 return err; 677 } 678 679 const struct proto_ops inet6_stream_ops = { 680 .family = PF_INET6, 681 .owner = THIS_MODULE, 682 .release = inet6_release, 683 .bind = inet6_bind, 684 .connect = inet_stream_connect, /* ok */ 685 .socketpair = sock_no_socketpair, /* a do nothing */ 686 .accept = inet_accept, /* ok */ 687 .getname = inet6_getname, 688 .poll = tcp_poll, /* ok */ 689 .ioctl = inet6_ioctl, /* must change */ 690 .gettstamp = sock_gettstamp, 691 .listen = inet_listen, /* ok */ 692 .shutdown = inet_shutdown, /* ok */ 693 .setsockopt = sock_common_setsockopt, /* ok */ 694 .getsockopt = sock_common_getsockopt, /* ok */ 695 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 696 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 697 #ifdef CONFIG_MMU 698 .mmap = tcp_mmap, 699 #endif 700 .sendpage = inet_sendpage, 701 .sendmsg_locked = tcp_sendmsg_locked, 702 .sendpage_locked = tcp_sendpage_locked, 703 .splice_read = tcp_splice_read, 704 .read_sock = tcp_read_sock, 705 .read_skb = tcp_read_skb, 706 .peek_len = tcp_peek_len, 707 #ifdef CONFIG_COMPAT 708 .compat_ioctl = inet6_compat_ioctl, 709 #endif 710 .set_rcvlowat = tcp_set_rcvlowat, 711 }; 712 713 const struct proto_ops inet6_dgram_ops = { 714 .family = PF_INET6, 715 .owner = THIS_MODULE, 716 .release = inet6_release, 717 .bind = inet6_bind, 718 .connect = inet_dgram_connect, /* ok */ 719 .socketpair = sock_no_socketpair, /* a do nothing */ 720 .accept = sock_no_accept, /* a do nothing */ 721 .getname = inet6_getname, 722 .poll = udp_poll, /* ok */ 723 .ioctl = inet6_ioctl, /* must change */ 724 .gettstamp = sock_gettstamp, 725 .listen = sock_no_listen, /* ok */ 726 .shutdown = inet_shutdown, /* ok */ 727 .setsockopt = sock_common_setsockopt, /* ok */ 728 .getsockopt = sock_common_getsockopt, /* ok */ 729 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 730 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 731 .read_skb = udp_read_skb, 732 .mmap = sock_no_mmap, 733 .sendpage = sock_no_sendpage, 734 .set_peek_off = sk_set_peek_off, 735 #ifdef CONFIG_COMPAT 736 .compat_ioctl = inet6_compat_ioctl, 737 #endif 738 }; 739 740 static const struct net_proto_family inet6_family_ops = { 741 .family = PF_INET6, 742 .create = inet6_create, 743 .owner = THIS_MODULE, 744 }; 745 746 int inet6_register_protosw(struct inet_protosw *p) 747 { 748 struct list_head *lh; 749 struct inet_protosw *answer; 750 struct list_head *last_perm; 751 int protocol = p->protocol; 752 int ret; 753 754 spin_lock_bh(&inetsw6_lock); 755 756 ret = -EINVAL; 757 if (p->type >= SOCK_MAX) 758 goto out_illegal; 759 760 /* If we are trying to override a permanent protocol, bail. */ 761 answer = NULL; 762 ret = -EPERM; 763 last_perm = &inetsw6[p->type]; 764 list_for_each(lh, &inetsw6[p->type]) { 765 answer = list_entry(lh, struct inet_protosw, list); 766 767 /* Check only the non-wild match. */ 768 if (INET_PROTOSW_PERMANENT & answer->flags) { 769 if (protocol == answer->protocol) 770 break; 771 last_perm = lh; 772 } 773 774 answer = NULL; 775 } 776 if (answer) 777 goto out_permanent; 778 779 /* Add the new entry after the last permanent entry if any, so that 780 * the new entry does not override a permanent entry when matched with 781 * a wild-card protocol. But it is allowed to override any existing 782 * non-permanent entry. This means that when we remove this entry, the 783 * system automatically returns to the old behavior. 784 */ 785 list_add_rcu(&p->list, last_perm); 786 ret = 0; 787 out: 788 spin_unlock_bh(&inetsw6_lock); 789 return ret; 790 791 out_permanent: 792 pr_err("Attempt to override permanent protocol %d\n", protocol); 793 goto out; 794 795 out_illegal: 796 pr_err("Ignoring attempt to register invalid socket type %d\n", 797 p->type); 798 goto out; 799 } 800 EXPORT_SYMBOL(inet6_register_protosw); 801 802 void 803 inet6_unregister_protosw(struct inet_protosw *p) 804 { 805 if (INET_PROTOSW_PERMANENT & p->flags) { 806 pr_err("Attempt to unregister permanent protocol %d\n", 807 p->protocol); 808 } else { 809 spin_lock_bh(&inetsw6_lock); 810 list_del_rcu(&p->list); 811 spin_unlock_bh(&inetsw6_lock); 812 813 synchronize_net(); 814 } 815 } 816 EXPORT_SYMBOL(inet6_unregister_protosw); 817 818 int inet6_sk_rebuild_header(struct sock *sk) 819 { 820 struct ipv6_pinfo *np = inet6_sk(sk); 821 struct dst_entry *dst; 822 823 dst = __sk_dst_check(sk, np->dst_cookie); 824 825 if (!dst) { 826 struct inet_sock *inet = inet_sk(sk); 827 struct in6_addr *final_p, final; 828 struct flowi6 fl6; 829 830 memset(&fl6, 0, sizeof(fl6)); 831 fl6.flowi6_proto = sk->sk_protocol; 832 fl6.daddr = sk->sk_v6_daddr; 833 fl6.saddr = np->saddr; 834 fl6.flowlabel = np->flow_label; 835 fl6.flowi6_oif = sk->sk_bound_dev_if; 836 fl6.flowi6_mark = sk->sk_mark; 837 fl6.fl6_dport = inet->inet_dport; 838 fl6.fl6_sport = inet->inet_sport; 839 fl6.flowi6_uid = sk->sk_uid; 840 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); 841 842 rcu_read_lock(); 843 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 844 &final); 845 rcu_read_unlock(); 846 847 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 848 if (IS_ERR(dst)) { 849 sk->sk_route_caps = 0; 850 sk->sk_err_soft = -PTR_ERR(dst); 851 return PTR_ERR(dst); 852 } 853 854 ip6_dst_store(sk, dst, NULL, NULL); 855 } 856 857 return 0; 858 } 859 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 860 861 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 862 const struct inet6_skb_parm *opt) 863 { 864 const struct ipv6_pinfo *np = inet6_sk(sk); 865 866 if (np->rxopt.all) { 867 if (((opt->flags & IP6SKB_HOPBYHOP) && 868 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 869 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 870 np->rxopt.bits.rxflow) || 871 (opt->srcrt && (np->rxopt.bits.srcrt || 872 np->rxopt.bits.osrcrt)) || 873 ((opt->dst1 || opt->dst0) && 874 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 875 return true; 876 } 877 return false; 878 } 879 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 880 881 static struct packet_type ipv6_packet_type __read_mostly = { 882 .type = cpu_to_be16(ETH_P_IPV6), 883 .func = ipv6_rcv, 884 .list_func = ipv6_list_rcv, 885 }; 886 887 static int __init ipv6_packet_init(void) 888 { 889 dev_add_pack(&ipv6_packet_type); 890 return 0; 891 } 892 893 static void ipv6_packet_cleanup(void) 894 { 895 dev_remove_pack(&ipv6_packet_type); 896 } 897 898 static int __net_init ipv6_init_mibs(struct net *net) 899 { 900 int i; 901 902 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 903 if (!net->mib.udp_stats_in6) 904 return -ENOMEM; 905 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 906 if (!net->mib.udplite_stats_in6) 907 goto err_udplite_mib; 908 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 909 if (!net->mib.ipv6_statistics) 910 goto err_ip_mib; 911 912 for_each_possible_cpu(i) { 913 struct ipstats_mib *af_inet6_stats; 914 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 915 u64_stats_init(&af_inet6_stats->syncp); 916 } 917 918 919 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 920 if (!net->mib.icmpv6_statistics) 921 goto err_icmp_mib; 922 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 923 GFP_KERNEL); 924 if (!net->mib.icmpv6msg_statistics) 925 goto err_icmpmsg_mib; 926 return 0; 927 928 err_icmpmsg_mib: 929 free_percpu(net->mib.icmpv6_statistics); 930 err_icmp_mib: 931 free_percpu(net->mib.ipv6_statistics); 932 err_ip_mib: 933 free_percpu(net->mib.udplite_stats_in6); 934 err_udplite_mib: 935 free_percpu(net->mib.udp_stats_in6); 936 return -ENOMEM; 937 } 938 939 static void ipv6_cleanup_mibs(struct net *net) 940 { 941 free_percpu(net->mib.udp_stats_in6); 942 free_percpu(net->mib.udplite_stats_in6); 943 free_percpu(net->mib.ipv6_statistics); 944 free_percpu(net->mib.icmpv6_statistics); 945 kfree(net->mib.icmpv6msg_statistics); 946 } 947 948 static int __net_init inet6_net_init(struct net *net) 949 { 950 int err = 0; 951 952 net->ipv6.sysctl.bindv6only = 0; 953 net->ipv6.sysctl.icmpv6_time = 1*HZ; 954 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 955 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0; 956 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0; 957 958 /* By default, rate limit error messages. 959 * Except for pmtu discovery, it would break it. 960 * proc_do_large_bitmap needs pointer to the bitmap. 961 */ 962 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 963 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 964 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 965 966 net->ipv6.sysctl.flowlabel_consistency = 1; 967 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 968 net->ipv6.sysctl.idgen_retries = 3; 969 net->ipv6.sysctl.idgen_delay = 1 * HZ; 970 net->ipv6.sysctl.flowlabel_state_ranges = 0; 971 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 972 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 973 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 974 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 975 net->ipv6.sysctl.fib_notify_on_flag_change = 0; 976 atomic_set(&net->ipv6.fib6_sernum, 1); 977 978 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID; 979 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE; 980 981 err = ipv6_init_mibs(net); 982 if (err) 983 return err; 984 #ifdef CONFIG_PROC_FS 985 err = udp6_proc_init(net); 986 if (err) 987 goto out; 988 err = tcp6_proc_init(net); 989 if (err) 990 goto proc_tcp6_fail; 991 err = ac6_proc_init(net); 992 if (err) 993 goto proc_ac6_fail; 994 #endif 995 return err; 996 997 #ifdef CONFIG_PROC_FS 998 proc_ac6_fail: 999 tcp6_proc_exit(net); 1000 proc_tcp6_fail: 1001 udp6_proc_exit(net); 1002 out: 1003 ipv6_cleanup_mibs(net); 1004 return err; 1005 #endif 1006 } 1007 1008 static void __net_exit inet6_net_exit(struct net *net) 1009 { 1010 #ifdef CONFIG_PROC_FS 1011 udp6_proc_exit(net); 1012 tcp6_proc_exit(net); 1013 ac6_proc_exit(net); 1014 #endif 1015 ipv6_cleanup_mibs(net); 1016 } 1017 1018 static struct pernet_operations inet6_net_ops = { 1019 .init = inet6_net_init, 1020 .exit = inet6_net_exit, 1021 }; 1022 1023 static int ipv6_route_input(struct sk_buff *skb) 1024 { 1025 ip6_route_input(skb); 1026 return skb_dst(skb)->error; 1027 } 1028 1029 static const struct ipv6_stub ipv6_stub_impl = { 1030 .ipv6_sock_mc_join = ipv6_sock_mc_join, 1031 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 1032 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 1033 .ipv6_route_input = ipv6_route_input, 1034 .fib6_get_table = fib6_get_table, 1035 .fib6_table_lookup = fib6_table_lookup, 1036 .fib6_lookup = fib6_lookup, 1037 .fib6_select_path = fib6_select_path, 1038 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 1039 .fib6_nh_init = fib6_nh_init, 1040 .fib6_nh_release = fib6_nh_release, 1041 .fib6_nh_release_dsts = fib6_nh_release_dsts, 1042 .fib6_update_sernum = fib6_update_sernum_stub, 1043 .fib6_rt_update = fib6_rt_update, 1044 .ip6_del_rt = ip6_del_rt, 1045 .udpv6_encap_enable = udpv6_encap_enable, 1046 .ndisc_send_na = ndisc_send_na, 1047 #if IS_ENABLED(CONFIG_XFRM) 1048 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu, 1049 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv, 1050 .xfrm6_rcv_encap = xfrm6_rcv_encap, 1051 #endif 1052 .nd_tbl = &nd_tbl, 1053 .ipv6_fragment = ip6_fragment, 1054 .ipv6_dev_find = ipv6_dev_find, 1055 }; 1056 1057 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 1058 .inet6_bind = __inet6_bind, 1059 .udp6_lib_lookup = __udp6_lib_lookup, 1060 }; 1061 1062 static int __init inet6_init(void) 1063 { 1064 struct list_head *r; 1065 int err = 0; 1066 1067 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 1068 1069 /* Register the socket-side information for inet6_create. */ 1070 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 1071 INIT_LIST_HEAD(r); 1072 1073 raw_hashinfo_init(&raw_v6_hashinfo); 1074 1075 if (disable_ipv6_mod) { 1076 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 1077 goto out; 1078 } 1079 1080 err = proto_register(&tcpv6_prot, 1); 1081 if (err) 1082 goto out; 1083 1084 err = proto_register(&udpv6_prot, 1); 1085 if (err) 1086 goto out_unregister_tcp_proto; 1087 1088 err = proto_register(&udplitev6_prot, 1); 1089 if (err) 1090 goto out_unregister_udp_proto; 1091 1092 err = proto_register(&rawv6_prot, 1); 1093 if (err) 1094 goto out_unregister_udplite_proto; 1095 1096 err = proto_register(&pingv6_prot, 1); 1097 if (err) 1098 goto out_unregister_raw_proto; 1099 1100 /* We MUST register RAW sockets before we create the ICMP6, 1101 * IGMP6, or NDISC control sockets. 1102 */ 1103 err = rawv6_init(); 1104 if (err) 1105 goto out_unregister_ping_proto; 1106 1107 /* Register the family here so that the init calls below will 1108 * be able to create sockets. (?? is this dangerous ??) 1109 */ 1110 err = sock_register(&inet6_family_ops); 1111 if (err) 1112 goto out_sock_register_fail; 1113 1114 /* 1115 * ipngwg API draft makes clear that the correct semantics 1116 * for TCP and UDP is to consider one TCP and UDP instance 1117 * in a host available by both INET and INET6 APIs and 1118 * able to communicate via both network protocols. 1119 */ 1120 1121 err = register_pernet_subsys(&inet6_net_ops); 1122 if (err) 1123 goto register_pernet_fail; 1124 err = ip6_mr_init(); 1125 if (err) 1126 goto ipmr_fail; 1127 err = icmpv6_init(); 1128 if (err) 1129 goto icmp_fail; 1130 err = ndisc_init(); 1131 if (err) 1132 goto ndisc_fail; 1133 err = igmp6_init(); 1134 if (err) 1135 goto igmp_fail; 1136 1137 err = ipv6_netfilter_init(); 1138 if (err) 1139 goto netfilter_fail; 1140 /* Create /proc/foo6 entries. */ 1141 #ifdef CONFIG_PROC_FS 1142 err = -ENOMEM; 1143 if (raw6_proc_init()) 1144 goto proc_raw6_fail; 1145 if (udplite6_proc_init()) 1146 goto proc_udplite6_fail; 1147 if (ipv6_misc_proc_init()) 1148 goto proc_misc6_fail; 1149 if (if6_proc_init()) 1150 goto proc_if6_fail; 1151 #endif 1152 err = ip6_route_init(); 1153 if (err) 1154 goto ip6_route_fail; 1155 err = ndisc_late_init(); 1156 if (err) 1157 goto ndisc_late_fail; 1158 err = ip6_flowlabel_init(); 1159 if (err) 1160 goto ip6_flowlabel_fail; 1161 err = ipv6_anycast_init(); 1162 if (err) 1163 goto ipv6_anycast_fail; 1164 err = addrconf_init(); 1165 if (err) 1166 goto addrconf_fail; 1167 1168 /* Init v6 extension headers. */ 1169 err = ipv6_exthdrs_init(); 1170 if (err) 1171 goto ipv6_exthdrs_fail; 1172 1173 err = ipv6_frag_init(); 1174 if (err) 1175 goto ipv6_frag_fail; 1176 1177 /* Init v6 transport protocols. */ 1178 err = udpv6_init(); 1179 if (err) 1180 goto udpv6_fail; 1181 1182 err = udplitev6_init(); 1183 if (err) 1184 goto udplitev6_fail; 1185 1186 err = udpv6_offload_init(); 1187 if (err) 1188 goto udpv6_offload_fail; 1189 1190 err = tcpv6_init(); 1191 if (err) 1192 goto tcpv6_fail; 1193 1194 err = ipv6_packet_init(); 1195 if (err) 1196 goto ipv6_packet_fail; 1197 1198 err = pingv6_init(); 1199 if (err) 1200 goto pingv6_fail; 1201 1202 err = calipso_init(); 1203 if (err) 1204 goto calipso_fail; 1205 1206 err = seg6_init(); 1207 if (err) 1208 goto seg6_fail; 1209 1210 err = rpl_init(); 1211 if (err) 1212 goto rpl_fail; 1213 1214 err = ioam6_init(); 1215 if (err) 1216 goto ioam6_fail; 1217 1218 err = igmp6_late_init(); 1219 if (err) 1220 goto igmp6_late_err; 1221 1222 #ifdef CONFIG_SYSCTL 1223 err = ipv6_sysctl_register(); 1224 if (err) 1225 goto sysctl_fail; 1226 #endif 1227 1228 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1229 wmb(); 1230 ipv6_stub = &ipv6_stub_impl; 1231 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1232 out: 1233 return err; 1234 1235 #ifdef CONFIG_SYSCTL 1236 sysctl_fail: 1237 igmp6_late_cleanup(); 1238 #endif 1239 igmp6_late_err: 1240 ioam6_exit(); 1241 ioam6_fail: 1242 rpl_exit(); 1243 rpl_fail: 1244 seg6_exit(); 1245 seg6_fail: 1246 calipso_exit(); 1247 calipso_fail: 1248 pingv6_exit(); 1249 pingv6_fail: 1250 ipv6_packet_cleanup(); 1251 ipv6_packet_fail: 1252 tcpv6_exit(); 1253 tcpv6_fail: 1254 udpv6_offload_exit(); 1255 udpv6_offload_fail: 1256 udplitev6_exit(); 1257 udplitev6_fail: 1258 udpv6_exit(); 1259 udpv6_fail: 1260 ipv6_frag_exit(); 1261 ipv6_frag_fail: 1262 ipv6_exthdrs_exit(); 1263 ipv6_exthdrs_fail: 1264 addrconf_cleanup(); 1265 addrconf_fail: 1266 ipv6_anycast_cleanup(); 1267 ipv6_anycast_fail: 1268 ip6_flowlabel_cleanup(); 1269 ip6_flowlabel_fail: 1270 ndisc_late_cleanup(); 1271 ndisc_late_fail: 1272 ip6_route_cleanup(); 1273 ip6_route_fail: 1274 #ifdef CONFIG_PROC_FS 1275 if6_proc_exit(); 1276 proc_if6_fail: 1277 ipv6_misc_proc_exit(); 1278 proc_misc6_fail: 1279 udplite6_proc_exit(); 1280 proc_udplite6_fail: 1281 raw6_proc_exit(); 1282 proc_raw6_fail: 1283 #endif 1284 ipv6_netfilter_fini(); 1285 netfilter_fail: 1286 igmp6_cleanup(); 1287 igmp_fail: 1288 ndisc_cleanup(); 1289 ndisc_fail: 1290 icmpv6_cleanup(); 1291 icmp_fail: 1292 ip6_mr_cleanup(); 1293 ipmr_fail: 1294 unregister_pernet_subsys(&inet6_net_ops); 1295 register_pernet_fail: 1296 sock_unregister(PF_INET6); 1297 rtnl_unregister_all(PF_INET6); 1298 out_sock_register_fail: 1299 rawv6_exit(); 1300 out_unregister_ping_proto: 1301 proto_unregister(&pingv6_prot); 1302 out_unregister_raw_proto: 1303 proto_unregister(&rawv6_prot); 1304 out_unregister_udplite_proto: 1305 proto_unregister(&udplitev6_prot); 1306 out_unregister_udp_proto: 1307 proto_unregister(&udpv6_prot); 1308 out_unregister_tcp_proto: 1309 proto_unregister(&tcpv6_prot); 1310 goto out; 1311 } 1312 module_init(inet6_init); 1313 1314 MODULE_ALIAS_NETPROTO(PF_INET6); 1315