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 (READ_ONCE(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 (ipv6_only_sock(sk)) { 322 err = -EINVAL; 323 goto out; 324 } 325 326 rcu_read_lock(); 327 if (sk->sk_bound_dev_if) { 328 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 329 if (!dev) { 330 err = -ENODEV; 331 goto out_unlock; 332 } 333 } 334 335 /* Reproduce AF_INET checks to make the bindings consistent */ 336 v4addr = addr->sin6_addr.s6_addr32[3]; 337 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr); 338 rcu_read_unlock(); 339 340 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr, 341 chk_addr_ret)) { 342 err = -EADDRNOTAVAIL; 343 goto out; 344 } 345 } else { 346 if (addr_type != IPV6_ADDR_ANY) { 347 struct net_device *dev = NULL; 348 349 rcu_read_lock(); 350 if (__ipv6_addr_needs_scope_id(addr_type)) { 351 if (addr_len >= sizeof(struct sockaddr_in6) && 352 addr->sin6_scope_id) { 353 /* Override any existing binding, if another one 354 * is supplied by user. 355 */ 356 sk->sk_bound_dev_if = addr->sin6_scope_id; 357 } 358 359 /* Binding to link-local address requires an interface */ 360 if (!sk->sk_bound_dev_if) { 361 err = -EINVAL; 362 goto out_unlock; 363 } 364 } 365 366 if (sk->sk_bound_dev_if) { 367 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 368 if (!dev) { 369 err = -ENODEV; 370 goto out_unlock; 371 } 372 } 373 374 /* ipv4 addr of the socket is invalid. Only the 375 * unspecified and mapped address have a v4 equivalent. 376 */ 377 v4addr = LOOPBACK4_IPV6; 378 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 379 if (!ipv6_can_nonlocal_bind(net, inet) && 380 !ipv6_chk_addr(net, &addr->sin6_addr, 381 dev, 0)) { 382 err = -EADDRNOTAVAIL; 383 goto out_unlock; 384 } 385 } 386 rcu_read_unlock(); 387 } 388 } 389 390 inet->inet_rcv_saddr = v4addr; 391 inet->inet_saddr = v4addr; 392 393 sk->sk_v6_rcv_saddr = addr->sin6_addr; 394 395 if (!(addr_type & IPV6_ADDR_MULTICAST)) 396 np->saddr = addr->sin6_addr; 397 398 saved_ipv6only = sk->sk_ipv6only; 399 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED) 400 sk->sk_ipv6only = 1; 401 402 /* Make sure we are allowed to bind here. */ 403 if (snum || !(inet->bind_address_no_port || 404 (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 405 if (sk->sk_prot->get_port(sk, snum)) { 406 sk->sk_ipv6only = saved_ipv6only; 407 inet_reset_saddr(sk); 408 err = -EADDRINUSE; 409 goto out; 410 } 411 if (!(flags & BIND_FROM_BPF)) { 412 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk); 413 if (err) { 414 sk->sk_ipv6only = saved_ipv6only; 415 inet_reset_saddr(sk); 416 if (sk->sk_prot->put_port) 417 sk->sk_prot->put_port(sk); 418 goto out; 419 } 420 } 421 } 422 423 if (addr_type != IPV6_ADDR_ANY) 424 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 425 if (snum) 426 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 427 inet->inet_sport = htons(inet->inet_num); 428 inet->inet_dport = 0; 429 inet->inet_daddr = 0; 430 out: 431 if (flags & BIND_WITH_LOCK) 432 release_sock(sk); 433 return err; 434 out_unlock: 435 rcu_read_unlock(); 436 goto out; 437 } 438 439 /* bind for INET6 API */ 440 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 441 { 442 struct sock *sk = sock->sk; 443 u32 flags = BIND_WITH_LOCK; 444 const struct proto *prot; 445 int err = 0; 446 447 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 448 prot = READ_ONCE(sk->sk_prot); 449 /* If the socket has its own bind function then use it. */ 450 if (prot->bind) 451 return prot->bind(sk, uaddr, addr_len); 452 453 if (addr_len < SIN6_LEN_RFC2133) 454 return -EINVAL; 455 456 /* BPF prog is run before any checks are done so that if the prog 457 * changes context in a wrong way it will be caught. 458 */ 459 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 460 CGROUP_INET6_BIND, &flags); 461 if (err) 462 return err; 463 464 return __inet6_bind(sk, uaddr, addr_len, flags); 465 } 466 EXPORT_SYMBOL(inet6_bind); 467 468 int inet6_release(struct socket *sock) 469 { 470 struct sock *sk = sock->sk; 471 472 if (!sk) 473 return -EINVAL; 474 475 /* Free mc lists */ 476 ipv6_sock_mc_close(sk); 477 478 /* Free ac lists */ 479 ipv6_sock_ac_close(sk); 480 481 return inet_release(sock); 482 } 483 EXPORT_SYMBOL(inet6_release); 484 485 void inet6_destroy_sock(struct sock *sk) 486 { 487 struct ipv6_pinfo *np = inet6_sk(sk); 488 struct sk_buff *skb; 489 struct ipv6_txoptions *opt; 490 491 /* Release rx options */ 492 493 skb = xchg(&np->pktoptions, NULL); 494 kfree_skb(skb); 495 496 skb = xchg(&np->rxpmtu, NULL); 497 kfree_skb(skb); 498 499 /* Free flowlabels */ 500 fl6_free_socklist(sk); 501 502 /* Free tx options */ 503 504 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 505 if (opt) { 506 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 507 txopt_put(opt); 508 } 509 } 510 EXPORT_SYMBOL_GPL(inet6_destroy_sock); 511 512 /* 513 * This does both peername and sockname. 514 */ 515 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 516 int peer) 517 { 518 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 519 struct sock *sk = sock->sk; 520 struct inet_sock *inet = inet_sk(sk); 521 struct ipv6_pinfo *np = inet6_sk(sk); 522 523 sin->sin6_family = AF_INET6; 524 sin->sin6_flowinfo = 0; 525 sin->sin6_scope_id = 0; 526 lock_sock(sk); 527 if (peer) { 528 if (!inet->inet_dport || 529 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 530 peer == 1)) { 531 release_sock(sk); 532 return -ENOTCONN; 533 } 534 sin->sin6_port = inet->inet_dport; 535 sin->sin6_addr = sk->sk_v6_daddr; 536 if (np->sndflow) 537 sin->sin6_flowinfo = np->flow_label; 538 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 539 CGROUP_INET6_GETPEERNAME); 540 } else { 541 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 542 sin->sin6_addr = np->saddr; 543 else 544 sin->sin6_addr = sk->sk_v6_rcv_saddr; 545 sin->sin6_port = inet->inet_sport; 546 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 547 CGROUP_INET6_GETSOCKNAME); 548 } 549 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 550 sk->sk_bound_dev_if); 551 release_sock(sk); 552 return sizeof(*sin); 553 } 554 EXPORT_SYMBOL(inet6_getname); 555 556 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 557 { 558 void __user *argp = (void __user *)arg; 559 struct sock *sk = sock->sk; 560 struct net *net = sock_net(sk); 561 const struct proto *prot; 562 563 switch (cmd) { 564 case SIOCADDRT: 565 case SIOCDELRT: { 566 struct in6_rtmsg rtmsg; 567 568 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg))) 569 return -EFAULT; 570 return ipv6_route_ioctl(net, cmd, &rtmsg); 571 } 572 case SIOCSIFADDR: 573 return addrconf_add_ifaddr(net, argp); 574 case SIOCDIFADDR: 575 return addrconf_del_ifaddr(net, argp); 576 case SIOCSIFDSTADDR: 577 return addrconf_set_dstaddr(net, argp); 578 default: 579 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 580 prot = READ_ONCE(sk->sk_prot); 581 if (!prot->ioctl) 582 return -ENOIOCTLCMD; 583 return prot->ioctl(sk, cmd, arg); 584 } 585 /*NOTREACHED*/ 586 return 0; 587 } 588 EXPORT_SYMBOL(inet6_ioctl); 589 590 #ifdef CONFIG_COMPAT 591 struct compat_in6_rtmsg { 592 struct in6_addr rtmsg_dst; 593 struct in6_addr rtmsg_src; 594 struct in6_addr rtmsg_gateway; 595 u32 rtmsg_type; 596 u16 rtmsg_dst_len; 597 u16 rtmsg_src_len; 598 u32 rtmsg_metric; 599 u32 rtmsg_info; 600 u32 rtmsg_flags; 601 s32 rtmsg_ifindex; 602 }; 603 604 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 605 struct compat_in6_rtmsg __user *ur) 606 { 607 struct in6_rtmsg rt; 608 609 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst, 610 3 * sizeof(struct in6_addr)) || 611 get_user(rt.rtmsg_type, &ur->rtmsg_type) || 612 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) || 613 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) || 614 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) || 615 get_user(rt.rtmsg_info, &ur->rtmsg_info) || 616 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) || 617 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex)) 618 return -EFAULT; 619 620 621 return ipv6_route_ioctl(sock_net(sk), cmd, &rt); 622 } 623 624 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 625 { 626 void __user *argp = compat_ptr(arg); 627 struct sock *sk = sock->sk; 628 629 switch (cmd) { 630 case SIOCADDRT: 631 case SIOCDELRT: 632 return inet6_compat_routing_ioctl(sk, cmd, argp); 633 default: 634 return -ENOIOCTLCMD; 635 } 636 } 637 EXPORT_SYMBOL_GPL(inet6_compat_ioctl); 638 #endif /* CONFIG_COMPAT */ 639 640 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 641 size_t)); 642 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 643 { 644 struct sock *sk = sock->sk; 645 const struct proto *prot; 646 647 if (unlikely(inet_send_prepare(sk))) 648 return -EAGAIN; 649 650 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 651 prot = READ_ONCE(sk->sk_prot); 652 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 653 sk, msg, size); 654 } 655 656 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 657 size_t, int, int *)); 658 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 659 int flags) 660 { 661 struct sock *sk = sock->sk; 662 const struct proto *prot; 663 int addr_len = 0; 664 int err; 665 666 if (likely(!(flags & MSG_ERRQUEUE))) 667 sock_rps_record_flow(sk); 668 669 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 670 prot = READ_ONCE(sk->sk_prot); 671 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 672 sk, msg, size, flags, &addr_len); 673 if (err >= 0) 674 msg->msg_namelen = addr_len; 675 return err; 676 } 677 678 const struct proto_ops inet6_stream_ops = { 679 .family = PF_INET6, 680 .owner = THIS_MODULE, 681 .release = inet6_release, 682 .bind = inet6_bind, 683 .connect = inet_stream_connect, /* ok */ 684 .socketpair = sock_no_socketpair, /* a do nothing */ 685 .accept = inet_accept, /* ok */ 686 .getname = inet6_getname, 687 .poll = tcp_poll, /* ok */ 688 .ioctl = inet6_ioctl, /* must change */ 689 .gettstamp = sock_gettstamp, 690 .listen = inet_listen, /* ok */ 691 .shutdown = inet_shutdown, /* ok */ 692 .setsockopt = sock_common_setsockopt, /* ok */ 693 .getsockopt = sock_common_getsockopt, /* ok */ 694 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 695 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 696 #ifdef CONFIG_MMU 697 .mmap = tcp_mmap, 698 #endif 699 .sendpage = inet_sendpage, 700 .sendmsg_locked = tcp_sendmsg_locked, 701 .sendpage_locked = tcp_sendpage_locked, 702 .splice_read = tcp_splice_read, 703 .read_sock = tcp_read_sock, 704 .peek_len = tcp_peek_len, 705 #ifdef CONFIG_COMPAT 706 .compat_ioctl = inet6_compat_ioctl, 707 #endif 708 .set_rcvlowat = tcp_set_rcvlowat, 709 }; 710 711 const struct proto_ops inet6_dgram_ops = { 712 .family = PF_INET6, 713 .owner = THIS_MODULE, 714 .release = inet6_release, 715 .bind = inet6_bind, 716 .connect = inet_dgram_connect, /* ok */ 717 .socketpair = sock_no_socketpair, /* a do nothing */ 718 .accept = sock_no_accept, /* a do nothing */ 719 .getname = inet6_getname, 720 .poll = udp_poll, /* ok */ 721 .ioctl = inet6_ioctl, /* must change */ 722 .gettstamp = sock_gettstamp, 723 .listen = sock_no_listen, /* ok */ 724 .shutdown = inet_shutdown, /* ok */ 725 .setsockopt = sock_common_setsockopt, /* ok */ 726 .getsockopt = sock_common_getsockopt, /* ok */ 727 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 728 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 729 .read_sock = udp_read_sock, 730 .mmap = sock_no_mmap, 731 .sendpage = sock_no_sendpage, 732 .set_peek_off = sk_set_peek_off, 733 #ifdef CONFIG_COMPAT 734 .compat_ioctl = inet6_compat_ioctl, 735 #endif 736 }; 737 738 static const struct net_proto_family inet6_family_ops = { 739 .family = PF_INET6, 740 .create = inet6_create, 741 .owner = THIS_MODULE, 742 }; 743 744 int inet6_register_protosw(struct inet_protosw *p) 745 { 746 struct list_head *lh; 747 struct inet_protosw *answer; 748 struct list_head *last_perm; 749 int protocol = p->protocol; 750 int ret; 751 752 spin_lock_bh(&inetsw6_lock); 753 754 ret = -EINVAL; 755 if (p->type >= SOCK_MAX) 756 goto out_illegal; 757 758 /* If we are trying to override a permanent protocol, bail. */ 759 answer = NULL; 760 ret = -EPERM; 761 last_perm = &inetsw6[p->type]; 762 list_for_each(lh, &inetsw6[p->type]) { 763 answer = list_entry(lh, struct inet_protosw, list); 764 765 /* Check only the non-wild match. */ 766 if (INET_PROTOSW_PERMANENT & answer->flags) { 767 if (protocol == answer->protocol) 768 break; 769 last_perm = lh; 770 } 771 772 answer = NULL; 773 } 774 if (answer) 775 goto out_permanent; 776 777 /* Add the new entry after the last permanent entry if any, so that 778 * the new entry does not override a permanent entry when matched with 779 * a wild-card protocol. But it is allowed to override any existing 780 * non-permanent entry. This means that when we remove this entry, the 781 * system automatically returns to the old behavior. 782 */ 783 list_add_rcu(&p->list, last_perm); 784 ret = 0; 785 out: 786 spin_unlock_bh(&inetsw6_lock); 787 return ret; 788 789 out_permanent: 790 pr_err("Attempt to override permanent protocol %d\n", protocol); 791 goto out; 792 793 out_illegal: 794 pr_err("Ignoring attempt to register invalid socket type %d\n", 795 p->type); 796 goto out; 797 } 798 EXPORT_SYMBOL(inet6_register_protosw); 799 800 void 801 inet6_unregister_protosw(struct inet_protosw *p) 802 { 803 if (INET_PROTOSW_PERMANENT & p->flags) { 804 pr_err("Attempt to unregister permanent protocol %d\n", 805 p->protocol); 806 } else { 807 spin_lock_bh(&inetsw6_lock); 808 list_del_rcu(&p->list); 809 spin_unlock_bh(&inetsw6_lock); 810 811 synchronize_net(); 812 } 813 } 814 EXPORT_SYMBOL(inet6_unregister_protosw); 815 816 int inet6_sk_rebuild_header(struct sock *sk) 817 { 818 struct ipv6_pinfo *np = inet6_sk(sk); 819 struct dst_entry *dst; 820 821 dst = __sk_dst_check(sk, np->dst_cookie); 822 823 if (!dst) { 824 struct inet_sock *inet = inet_sk(sk); 825 struct in6_addr *final_p, final; 826 struct flowi6 fl6; 827 828 memset(&fl6, 0, sizeof(fl6)); 829 fl6.flowi6_proto = sk->sk_protocol; 830 fl6.daddr = sk->sk_v6_daddr; 831 fl6.saddr = np->saddr; 832 fl6.flowlabel = np->flow_label; 833 fl6.flowi6_oif = sk->sk_bound_dev_if; 834 fl6.flowi6_mark = sk->sk_mark; 835 fl6.fl6_dport = inet->inet_dport; 836 fl6.fl6_sport = inet->inet_sport; 837 fl6.flowi6_uid = sk->sk_uid; 838 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); 839 840 rcu_read_lock(); 841 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 842 &final); 843 rcu_read_unlock(); 844 845 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 846 if (IS_ERR(dst)) { 847 sk->sk_route_caps = 0; 848 sk->sk_err_soft = -PTR_ERR(dst); 849 return PTR_ERR(dst); 850 } 851 852 ip6_dst_store(sk, dst, NULL, NULL); 853 } 854 855 return 0; 856 } 857 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 858 859 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 860 const struct inet6_skb_parm *opt) 861 { 862 const struct ipv6_pinfo *np = inet6_sk(sk); 863 864 if (np->rxopt.all) { 865 if (((opt->flags & IP6SKB_HOPBYHOP) && 866 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 867 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 868 np->rxopt.bits.rxflow) || 869 (opt->srcrt && (np->rxopt.bits.srcrt || 870 np->rxopt.bits.osrcrt)) || 871 ((opt->dst1 || opt->dst0) && 872 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 873 return true; 874 } 875 return false; 876 } 877 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 878 879 static struct packet_type ipv6_packet_type __read_mostly = { 880 .type = cpu_to_be16(ETH_P_IPV6), 881 .func = ipv6_rcv, 882 .list_func = ipv6_list_rcv, 883 }; 884 885 static int __init ipv6_packet_init(void) 886 { 887 dev_add_pack(&ipv6_packet_type); 888 return 0; 889 } 890 891 static void ipv6_packet_cleanup(void) 892 { 893 dev_remove_pack(&ipv6_packet_type); 894 } 895 896 static int __net_init ipv6_init_mibs(struct net *net) 897 { 898 int i; 899 900 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 901 if (!net->mib.udp_stats_in6) 902 return -ENOMEM; 903 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 904 if (!net->mib.udplite_stats_in6) 905 goto err_udplite_mib; 906 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 907 if (!net->mib.ipv6_statistics) 908 goto err_ip_mib; 909 910 for_each_possible_cpu(i) { 911 struct ipstats_mib *af_inet6_stats; 912 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 913 u64_stats_init(&af_inet6_stats->syncp); 914 } 915 916 917 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 918 if (!net->mib.icmpv6_statistics) 919 goto err_icmp_mib; 920 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 921 GFP_KERNEL); 922 if (!net->mib.icmpv6msg_statistics) 923 goto err_icmpmsg_mib; 924 return 0; 925 926 err_icmpmsg_mib: 927 free_percpu(net->mib.icmpv6_statistics); 928 err_icmp_mib: 929 free_percpu(net->mib.ipv6_statistics); 930 err_ip_mib: 931 free_percpu(net->mib.udplite_stats_in6); 932 err_udplite_mib: 933 free_percpu(net->mib.udp_stats_in6); 934 return -ENOMEM; 935 } 936 937 static void ipv6_cleanup_mibs(struct net *net) 938 { 939 free_percpu(net->mib.udp_stats_in6); 940 free_percpu(net->mib.udplite_stats_in6); 941 free_percpu(net->mib.ipv6_statistics); 942 free_percpu(net->mib.icmpv6_statistics); 943 kfree(net->mib.icmpv6msg_statistics); 944 } 945 946 static int __net_init inet6_net_init(struct net *net) 947 { 948 int err = 0; 949 950 net->ipv6.sysctl.bindv6only = 0; 951 net->ipv6.sysctl.icmpv6_time = 1*HZ; 952 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 953 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0; 954 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0; 955 956 /* By default, rate limit error messages. 957 * Except for pmtu discovery, it would break it. 958 * proc_do_large_bitmap needs pointer to the bitmap. 959 */ 960 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 961 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 962 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 963 964 net->ipv6.sysctl.flowlabel_consistency = 1; 965 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 966 net->ipv6.sysctl.idgen_retries = 3; 967 net->ipv6.sysctl.idgen_delay = 1 * HZ; 968 net->ipv6.sysctl.flowlabel_state_ranges = 0; 969 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 970 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 971 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 972 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 973 net->ipv6.sysctl.fib_notify_on_flag_change = 0; 974 atomic_set(&net->ipv6.fib6_sernum, 1); 975 976 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID; 977 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE; 978 979 err = ipv6_init_mibs(net); 980 if (err) 981 return err; 982 #ifdef CONFIG_PROC_FS 983 err = udp6_proc_init(net); 984 if (err) 985 goto out; 986 err = tcp6_proc_init(net); 987 if (err) 988 goto proc_tcp6_fail; 989 err = ac6_proc_init(net); 990 if (err) 991 goto proc_ac6_fail; 992 #endif 993 return err; 994 995 #ifdef CONFIG_PROC_FS 996 proc_ac6_fail: 997 tcp6_proc_exit(net); 998 proc_tcp6_fail: 999 udp6_proc_exit(net); 1000 out: 1001 ipv6_cleanup_mibs(net); 1002 return err; 1003 #endif 1004 } 1005 1006 static void __net_exit inet6_net_exit(struct net *net) 1007 { 1008 #ifdef CONFIG_PROC_FS 1009 udp6_proc_exit(net); 1010 tcp6_proc_exit(net); 1011 ac6_proc_exit(net); 1012 #endif 1013 ipv6_cleanup_mibs(net); 1014 } 1015 1016 static struct pernet_operations inet6_net_ops = { 1017 .init = inet6_net_init, 1018 .exit = inet6_net_exit, 1019 }; 1020 1021 static int ipv6_route_input(struct sk_buff *skb) 1022 { 1023 ip6_route_input(skb); 1024 return skb_dst(skb)->error; 1025 } 1026 1027 static const struct ipv6_stub ipv6_stub_impl = { 1028 .ipv6_sock_mc_join = ipv6_sock_mc_join, 1029 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 1030 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 1031 .ipv6_route_input = ipv6_route_input, 1032 .fib6_get_table = fib6_get_table, 1033 .fib6_table_lookup = fib6_table_lookup, 1034 .fib6_lookup = fib6_lookup, 1035 .fib6_select_path = fib6_select_path, 1036 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 1037 .fib6_nh_init = fib6_nh_init, 1038 .fib6_nh_release = fib6_nh_release, 1039 .fib6_nh_release_dsts = fib6_nh_release_dsts, 1040 .fib6_update_sernum = fib6_update_sernum_stub, 1041 .fib6_rt_update = fib6_rt_update, 1042 .ip6_del_rt = ip6_del_rt, 1043 .udpv6_encap_enable = udpv6_encap_enable, 1044 .ndisc_send_na = ndisc_send_na, 1045 #if IS_ENABLED(CONFIG_XFRM) 1046 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu, 1047 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv, 1048 .xfrm6_rcv_encap = xfrm6_rcv_encap, 1049 #endif 1050 .nd_tbl = &nd_tbl, 1051 .ipv6_fragment = ip6_fragment, 1052 .ipv6_dev_find = ipv6_dev_find, 1053 }; 1054 1055 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 1056 .inet6_bind = __inet6_bind, 1057 .udp6_lib_lookup = __udp6_lib_lookup, 1058 }; 1059 1060 static int __init inet6_init(void) 1061 { 1062 struct list_head *r; 1063 int err = 0; 1064 1065 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 1066 1067 /* Register the socket-side information for inet6_create. */ 1068 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 1069 INIT_LIST_HEAD(r); 1070 1071 if (disable_ipv6_mod) { 1072 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 1073 goto out; 1074 } 1075 1076 err = proto_register(&tcpv6_prot, 1); 1077 if (err) 1078 goto out; 1079 1080 err = proto_register(&udpv6_prot, 1); 1081 if (err) 1082 goto out_unregister_tcp_proto; 1083 1084 err = proto_register(&udplitev6_prot, 1); 1085 if (err) 1086 goto out_unregister_udp_proto; 1087 1088 err = proto_register(&rawv6_prot, 1); 1089 if (err) 1090 goto out_unregister_udplite_proto; 1091 1092 err = proto_register(&pingv6_prot, 1); 1093 if (err) 1094 goto out_unregister_raw_proto; 1095 1096 /* We MUST register RAW sockets before we create the ICMP6, 1097 * IGMP6, or NDISC control sockets. 1098 */ 1099 err = rawv6_init(); 1100 if (err) 1101 goto out_unregister_ping_proto; 1102 1103 /* Register the family here so that the init calls below will 1104 * be able to create sockets. (?? is this dangerous ??) 1105 */ 1106 err = sock_register(&inet6_family_ops); 1107 if (err) 1108 goto out_sock_register_fail; 1109 1110 /* 1111 * ipngwg API draft makes clear that the correct semantics 1112 * for TCP and UDP is to consider one TCP and UDP instance 1113 * in a host available by both INET and INET6 APIs and 1114 * able to communicate via both network protocols. 1115 */ 1116 1117 err = register_pernet_subsys(&inet6_net_ops); 1118 if (err) 1119 goto register_pernet_fail; 1120 err = ip6_mr_init(); 1121 if (err) 1122 goto ipmr_fail; 1123 err = icmpv6_init(); 1124 if (err) 1125 goto icmp_fail; 1126 err = ndisc_init(); 1127 if (err) 1128 goto ndisc_fail; 1129 err = igmp6_init(); 1130 if (err) 1131 goto igmp_fail; 1132 1133 err = ipv6_netfilter_init(); 1134 if (err) 1135 goto netfilter_fail; 1136 /* Create /proc/foo6 entries. */ 1137 #ifdef CONFIG_PROC_FS 1138 err = -ENOMEM; 1139 if (raw6_proc_init()) 1140 goto proc_raw6_fail; 1141 if (udplite6_proc_init()) 1142 goto proc_udplite6_fail; 1143 if (ipv6_misc_proc_init()) 1144 goto proc_misc6_fail; 1145 if (if6_proc_init()) 1146 goto proc_if6_fail; 1147 #endif 1148 err = ip6_route_init(); 1149 if (err) 1150 goto ip6_route_fail; 1151 err = ndisc_late_init(); 1152 if (err) 1153 goto ndisc_late_fail; 1154 err = ip6_flowlabel_init(); 1155 if (err) 1156 goto ip6_flowlabel_fail; 1157 err = ipv6_anycast_init(); 1158 if (err) 1159 goto ipv6_anycast_fail; 1160 err = addrconf_init(); 1161 if (err) 1162 goto addrconf_fail; 1163 1164 /* Init v6 extension headers. */ 1165 err = ipv6_exthdrs_init(); 1166 if (err) 1167 goto ipv6_exthdrs_fail; 1168 1169 err = ipv6_frag_init(); 1170 if (err) 1171 goto ipv6_frag_fail; 1172 1173 /* Init v6 transport protocols. */ 1174 err = udpv6_init(); 1175 if (err) 1176 goto udpv6_fail; 1177 1178 err = udplitev6_init(); 1179 if (err) 1180 goto udplitev6_fail; 1181 1182 err = udpv6_offload_init(); 1183 if (err) 1184 goto udpv6_offload_fail; 1185 1186 err = tcpv6_init(); 1187 if (err) 1188 goto tcpv6_fail; 1189 1190 err = ipv6_packet_init(); 1191 if (err) 1192 goto ipv6_packet_fail; 1193 1194 err = pingv6_init(); 1195 if (err) 1196 goto pingv6_fail; 1197 1198 err = calipso_init(); 1199 if (err) 1200 goto calipso_fail; 1201 1202 err = seg6_init(); 1203 if (err) 1204 goto seg6_fail; 1205 1206 err = rpl_init(); 1207 if (err) 1208 goto rpl_fail; 1209 1210 err = ioam6_init(); 1211 if (err) 1212 goto ioam6_fail; 1213 1214 err = igmp6_late_init(); 1215 if (err) 1216 goto igmp6_late_err; 1217 1218 #ifdef CONFIG_SYSCTL 1219 err = ipv6_sysctl_register(); 1220 if (err) 1221 goto sysctl_fail; 1222 #endif 1223 1224 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1225 wmb(); 1226 ipv6_stub = &ipv6_stub_impl; 1227 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1228 out: 1229 return err; 1230 1231 #ifdef CONFIG_SYSCTL 1232 sysctl_fail: 1233 igmp6_late_cleanup(); 1234 #endif 1235 igmp6_late_err: 1236 ioam6_exit(); 1237 ioam6_fail: 1238 rpl_exit(); 1239 rpl_fail: 1240 seg6_exit(); 1241 seg6_fail: 1242 calipso_exit(); 1243 calipso_fail: 1244 pingv6_exit(); 1245 pingv6_fail: 1246 ipv6_packet_cleanup(); 1247 ipv6_packet_fail: 1248 tcpv6_exit(); 1249 tcpv6_fail: 1250 udpv6_offload_exit(); 1251 udpv6_offload_fail: 1252 udplitev6_exit(); 1253 udplitev6_fail: 1254 udpv6_exit(); 1255 udpv6_fail: 1256 ipv6_frag_exit(); 1257 ipv6_frag_fail: 1258 ipv6_exthdrs_exit(); 1259 ipv6_exthdrs_fail: 1260 addrconf_cleanup(); 1261 addrconf_fail: 1262 ipv6_anycast_cleanup(); 1263 ipv6_anycast_fail: 1264 ip6_flowlabel_cleanup(); 1265 ip6_flowlabel_fail: 1266 ndisc_late_cleanup(); 1267 ndisc_late_fail: 1268 ip6_route_cleanup(); 1269 ip6_route_fail: 1270 #ifdef CONFIG_PROC_FS 1271 if6_proc_exit(); 1272 proc_if6_fail: 1273 ipv6_misc_proc_exit(); 1274 proc_misc6_fail: 1275 udplite6_proc_exit(); 1276 proc_udplite6_fail: 1277 raw6_proc_exit(); 1278 proc_raw6_fail: 1279 #endif 1280 ipv6_netfilter_fini(); 1281 netfilter_fail: 1282 igmp6_cleanup(); 1283 igmp_fail: 1284 ndisc_cleanup(); 1285 ndisc_fail: 1286 icmpv6_cleanup(); 1287 icmp_fail: 1288 ip6_mr_cleanup(); 1289 ipmr_fail: 1290 unregister_pernet_subsys(&inet6_net_ops); 1291 register_pernet_fail: 1292 sock_unregister(PF_INET6); 1293 rtnl_unregister_all(PF_INET6); 1294 out_sock_register_fail: 1295 rawv6_exit(); 1296 out_unregister_ping_proto: 1297 proto_unregister(&pingv6_prot); 1298 out_unregister_raw_proto: 1299 proto_unregister(&rawv6_prot); 1300 out_unregister_udplite_proto: 1301 proto_unregister(&udplitev6_prot); 1302 out_unregister_udp_proto: 1303 proto_unregister(&udpv6_prot); 1304 out_unregister_tcp_proto: 1305 proto_unregister(&tcpv6_prot); 1306 goto out; 1307 } 1308 module_init(inet6_init); 1309 1310 MODULE_ALIAS_NETPROTO(PF_INET6); 1311