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 struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) 106 { 107 const int offset = sk->sk_prot->ipv6_pinfo_offset; 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_assign_bit(IS_ICSK, sk, INET_PROTOSW_ICSK & answer_flags); 204 205 if (SOCK_RAW == sock->type) { 206 inet->inet_num = protocol; 207 if (IPPROTO_RAW == protocol) 208 inet_set_bit(HDRINCL, sk); 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_set_bit(MC_LOOP, sk); 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_test_bit(BIND_ADDRESS_NO_PORT, sk) || 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 int inet6_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len) 439 { 440 u32 flags = BIND_WITH_LOCK; 441 const struct proto *prot; 442 int err = 0; 443 444 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 445 prot = READ_ONCE(sk->sk_prot); 446 /* If the socket has its own bind function then use it. */ 447 if (prot->bind) 448 return prot->bind(sk, uaddr, addr_len); 449 450 if (addr_len < SIN6_LEN_RFC2133) 451 return -EINVAL; 452 453 /* BPF prog is run before any checks are done so that if the prog 454 * changes context in a wrong way it will be caught. 455 */ 456 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 457 CGROUP_INET6_BIND, &flags); 458 if (err) 459 return err; 460 461 return __inet6_bind(sk, uaddr, addr_len, flags); 462 } 463 464 /* bind for INET6 API */ 465 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 466 { 467 return inet6_bind_sk(sock->sk, uaddr, addr_len); 468 } 469 EXPORT_SYMBOL(inet6_bind); 470 471 int inet6_release(struct socket *sock) 472 { 473 struct sock *sk = sock->sk; 474 475 if (!sk) 476 return -EINVAL; 477 478 /* Free mc lists */ 479 ipv6_sock_mc_close(sk); 480 481 /* Free ac lists */ 482 ipv6_sock_ac_close(sk); 483 484 return inet_release(sock); 485 } 486 EXPORT_SYMBOL(inet6_release); 487 488 void inet6_cleanup_sock(struct sock *sk) 489 { 490 struct ipv6_pinfo *np = inet6_sk(sk); 491 struct sk_buff *skb; 492 struct ipv6_txoptions *opt; 493 494 /* Release rx options */ 495 496 skb = xchg(&np->pktoptions, NULL); 497 kfree_skb(skb); 498 499 skb = xchg(&np->rxpmtu, NULL); 500 kfree_skb(skb); 501 502 /* Free flowlabels */ 503 fl6_free_socklist(sk); 504 505 /* Free tx options */ 506 507 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 508 if (opt) { 509 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 510 txopt_put(opt); 511 } 512 } 513 EXPORT_SYMBOL_GPL(inet6_cleanup_sock); 514 515 /* 516 * This does both peername and sockname. 517 */ 518 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 519 int peer) 520 { 521 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 522 struct sock *sk = sock->sk; 523 struct inet_sock *inet = inet_sk(sk); 524 struct ipv6_pinfo *np = inet6_sk(sk); 525 526 sin->sin6_family = AF_INET6; 527 sin->sin6_flowinfo = 0; 528 sin->sin6_scope_id = 0; 529 lock_sock(sk); 530 if (peer) { 531 if (!inet->inet_dport || 532 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 533 peer == 1)) { 534 release_sock(sk); 535 return -ENOTCONN; 536 } 537 sin->sin6_port = inet->inet_dport; 538 sin->sin6_addr = sk->sk_v6_daddr; 539 if (np->sndflow) 540 sin->sin6_flowinfo = np->flow_label; 541 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 542 CGROUP_INET6_GETPEERNAME); 543 } else { 544 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 545 sin->sin6_addr = np->saddr; 546 else 547 sin->sin6_addr = sk->sk_v6_rcv_saddr; 548 sin->sin6_port = inet->inet_sport; 549 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 550 CGROUP_INET6_GETSOCKNAME); 551 } 552 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 553 sk->sk_bound_dev_if); 554 release_sock(sk); 555 return sizeof(*sin); 556 } 557 EXPORT_SYMBOL(inet6_getname); 558 559 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 560 { 561 void __user *argp = (void __user *)arg; 562 struct sock *sk = sock->sk; 563 struct net *net = sock_net(sk); 564 const struct proto *prot; 565 566 switch (cmd) { 567 case SIOCADDRT: 568 case SIOCDELRT: { 569 struct in6_rtmsg rtmsg; 570 571 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg))) 572 return -EFAULT; 573 return ipv6_route_ioctl(net, cmd, &rtmsg); 574 } 575 case SIOCSIFADDR: 576 return addrconf_add_ifaddr(net, argp); 577 case SIOCDIFADDR: 578 return addrconf_del_ifaddr(net, argp); 579 case SIOCSIFDSTADDR: 580 return addrconf_set_dstaddr(net, argp); 581 default: 582 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 583 prot = READ_ONCE(sk->sk_prot); 584 if (!prot->ioctl) 585 return -ENOIOCTLCMD; 586 return sk_ioctl(sk, cmd, (void __user *)arg); 587 } 588 /*NOTREACHED*/ 589 return 0; 590 } 591 EXPORT_SYMBOL(inet6_ioctl); 592 593 #ifdef CONFIG_COMPAT 594 struct compat_in6_rtmsg { 595 struct in6_addr rtmsg_dst; 596 struct in6_addr rtmsg_src; 597 struct in6_addr rtmsg_gateway; 598 u32 rtmsg_type; 599 u16 rtmsg_dst_len; 600 u16 rtmsg_src_len; 601 u32 rtmsg_metric; 602 u32 rtmsg_info; 603 u32 rtmsg_flags; 604 s32 rtmsg_ifindex; 605 }; 606 607 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 608 struct compat_in6_rtmsg __user *ur) 609 { 610 struct in6_rtmsg rt; 611 612 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst, 613 3 * sizeof(struct in6_addr)) || 614 get_user(rt.rtmsg_type, &ur->rtmsg_type) || 615 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) || 616 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) || 617 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) || 618 get_user(rt.rtmsg_info, &ur->rtmsg_info) || 619 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) || 620 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex)) 621 return -EFAULT; 622 623 624 return ipv6_route_ioctl(sock_net(sk), cmd, &rt); 625 } 626 627 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 628 { 629 void __user *argp = compat_ptr(arg); 630 struct sock *sk = sock->sk; 631 632 switch (cmd) { 633 case SIOCADDRT: 634 case SIOCDELRT: 635 return inet6_compat_routing_ioctl(sk, cmd, argp); 636 default: 637 return -ENOIOCTLCMD; 638 } 639 } 640 EXPORT_SYMBOL_GPL(inet6_compat_ioctl); 641 #endif /* CONFIG_COMPAT */ 642 643 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 644 size_t)); 645 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 646 { 647 struct sock *sk = sock->sk; 648 const struct proto *prot; 649 650 if (unlikely(inet_send_prepare(sk))) 651 return -EAGAIN; 652 653 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 654 prot = READ_ONCE(sk->sk_prot); 655 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 656 sk, msg, size); 657 } 658 659 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 660 size_t, int, int *)); 661 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 662 int flags) 663 { 664 struct sock *sk = sock->sk; 665 const struct proto *prot; 666 int addr_len = 0; 667 int err; 668 669 if (likely(!(flags & MSG_ERRQUEUE))) 670 sock_rps_record_flow(sk); 671 672 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 673 prot = READ_ONCE(sk->sk_prot); 674 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 675 sk, msg, size, flags, &addr_len); 676 if (err >= 0) 677 msg->msg_namelen = addr_len; 678 return err; 679 } 680 681 const struct proto_ops inet6_stream_ops = { 682 .family = PF_INET6, 683 .owner = THIS_MODULE, 684 .release = inet6_release, 685 .bind = inet6_bind, 686 .connect = inet_stream_connect, /* ok */ 687 .socketpair = sock_no_socketpair, /* a do nothing */ 688 .accept = inet_accept, /* ok */ 689 .getname = inet6_getname, 690 .poll = tcp_poll, /* ok */ 691 .ioctl = inet6_ioctl, /* must change */ 692 .gettstamp = sock_gettstamp, 693 .listen = inet_listen, /* ok */ 694 .shutdown = inet_shutdown, /* ok */ 695 .setsockopt = sock_common_setsockopt, /* ok */ 696 .getsockopt = sock_common_getsockopt, /* ok */ 697 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 698 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 699 #ifdef CONFIG_MMU 700 .mmap = tcp_mmap, 701 #endif 702 .splice_eof = inet_splice_eof, 703 .sendmsg_locked = tcp_sendmsg_locked, 704 .splice_read = tcp_splice_read, 705 .read_sock = tcp_read_sock, 706 .read_skb = tcp_read_skb, 707 .peek_len = tcp_peek_len, 708 #ifdef CONFIG_COMPAT 709 .compat_ioctl = inet6_compat_ioctl, 710 #endif 711 .set_rcvlowat = tcp_set_rcvlowat, 712 }; 713 714 const struct proto_ops inet6_dgram_ops = { 715 .family = PF_INET6, 716 .owner = THIS_MODULE, 717 .release = inet6_release, 718 .bind = inet6_bind, 719 .connect = inet_dgram_connect, /* ok */ 720 .socketpair = sock_no_socketpair, /* a do nothing */ 721 .accept = sock_no_accept, /* a do nothing */ 722 .getname = inet6_getname, 723 .poll = udp_poll, /* ok */ 724 .ioctl = inet6_ioctl, /* must change */ 725 .gettstamp = sock_gettstamp, 726 .listen = sock_no_listen, /* ok */ 727 .shutdown = inet_shutdown, /* ok */ 728 .setsockopt = sock_common_setsockopt, /* ok */ 729 .getsockopt = sock_common_getsockopt, /* ok */ 730 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 731 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 732 .read_skb = udp_read_skb, 733 .mmap = sock_no_mmap, 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 WRITE_ONCE(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 net->ipv6.sysctl.icmpv6_error_anycast_as_unicast = 0; 958 959 /* By default, rate limit error messages. 960 * Except for pmtu discovery, it would break it. 961 * proc_do_large_bitmap needs pointer to the bitmap. 962 */ 963 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 964 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 965 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 966 967 net->ipv6.sysctl.flowlabel_consistency = 1; 968 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 969 net->ipv6.sysctl.idgen_retries = 3; 970 net->ipv6.sysctl.idgen_delay = 1 * HZ; 971 net->ipv6.sysctl.flowlabel_state_ranges = 0; 972 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 973 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 974 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 975 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 976 net->ipv6.sysctl.fib_notify_on_flag_change = 0; 977 atomic_set(&net->ipv6.fib6_sernum, 1); 978 979 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID; 980 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE; 981 982 err = ipv6_init_mibs(net); 983 if (err) 984 return err; 985 #ifdef CONFIG_PROC_FS 986 err = udp6_proc_init(net); 987 if (err) 988 goto out; 989 err = tcp6_proc_init(net); 990 if (err) 991 goto proc_tcp6_fail; 992 err = ac6_proc_init(net); 993 if (err) 994 goto proc_ac6_fail; 995 #endif 996 return err; 997 998 #ifdef CONFIG_PROC_FS 999 proc_ac6_fail: 1000 tcp6_proc_exit(net); 1001 proc_tcp6_fail: 1002 udp6_proc_exit(net); 1003 out: 1004 ipv6_cleanup_mibs(net); 1005 return err; 1006 #endif 1007 } 1008 1009 static void __net_exit inet6_net_exit(struct net *net) 1010 { 1011 #ifdef CONFIG_PROC_FS 1012 udp6_proc_exit(net); 1013 tcp6_proc_exit(net); 1014 ac6_proc_exit(net); 1015 #endif 1016 ipv6_cleanup_mibs(net); 1017 } 1018 1019 static struct pernet_operations inet6_net_ops = { 1020 .init = inet6_net_init, 1021 .exit = inet6_net_exit, 1022 }; 1023 1024 static int ipv6_route_input(struct sk_buff *skb) 1025 { 1026 ip6_route_input(skb); 1027 return skb_dst(skb)->error; 1028 } 1029 1030 static const struct ipv6_stub ipv6_stub_impl = { 1031 .ipv6_sock_mc_join = ipv6_sock_mc_join, 1032 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 1033 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 1034 .ipv6_route_input = ipv6_route_input, 1035 .fib6_get_table = fib6_get_table, 1036 .fib6_table_lookup = fib6_table_lookup, 1037 .fib6_lookup = fib6_lookup, 1038 .fib6_select_path = fib6_select_path, 1039 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 1040 .fib6_nh_init = fib6_nh_init, 1041 .fib6_nh_release = fib6_nh_release, 1042 .fib6_nh_release_dsts = fib6_nh_release_dsts, 1043 .fib6_update_sernum = fib6_update_sernum_stub, 1044 .fib6_rt_update = fib6_rt_update, 1045 .ip6_del_rt = ip6_del_rt, 1046 .udpv6_encap_enable = udpv6_encap_enable, 1047 .ndisc_send_na = ndisc_send_na, 1048 #if IS_ENABLED(CONFIG_XFRM) 1049 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu, 1050 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv, 1051 .xfrm6_rcv_encap = xfrm6_rcv_encap, 1052 #endif 1053 .nd_tbl = &nd_tbl, 1054 .ipv6_fragment = ip6_fragment, 1055 .ipv6_dev_find = ipv6_dev_find, 1056 }; 1057 1058 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 1059 .inet6_bind = __inet6_bind, 1060 .udp6_lib_lookup = __udp6_lib_lookup, 1061 .ipv6_setsockopt = do_ipv6_setsockopt, 1062 .ipv6_getsockopt = do_ipv6_getsockopt, 1063 }; 1064 1065 static int __init inet6_init(void) 1066 { 1067 struct list_head *r; 1068 int err = 0; 1069 1070 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 1071 1072 /* Register the socket-side information for inet6_create. */ 1073 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 1074 INIT_LIST_HEAD(r); 1075 1076 raw_hashinfo_init(&raw_v6_hashinfo); 1077 1078 if (disable_ipv6_mod) { 1079 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 1080 goto out; 1081 } 1082 1083 err = proto_register(&tcpv6_prot, 1); 1084 if (err) 1085 goto out; 1086 1087 err = proto_register(&udpv6_prot, 1); 1088 if (err) 1089 goto out_unregister_tcp_proto; 1090 1091 err = proto_register(&udplitev6_prot, 1); 1092 if (err) 1093 goto out_unregister_udp_proto; 1094 1095 err = proto_register(&rawv6_prot, 1); 1096 if (err) 1097 goto out_unregister_udplite_proto; 1098 1099 err = proto_register(&pingv6_prot, 1); 1100 if (err) 1101 goto out_unregister_raw_proto; 1102 1103 /* We MUST register RAW sockets before we create the ICMP6, 1104 * IGMP6, or NDISC control sockets. 1105 */ 1106 err = rawv6_init(); 1107 if (err) 1108 goto out_unregister_ping_proto; 1109 1110 /* Register the family here so that the init calls below will 1111 * be able to create sockets. (?? is this dangerous ??) 1112 */ 1113 err = sock_register(&inet6_family_ops); 1114 if (err) 1115 goto out_sock_register_fail; 1116 1117 /* 1118 * ipngwg API draft makes clear that the correct semantics 1119 * for TCP and UDP is to consider one TCP and UDP instance 1120 * in a host available by both INET and INET6 APIs and 1121 * able to communicate via both network protocols. 1122 */ 1123 1124 err = register_pernet_subsys(&inet6_net_ops); 1125 if (err) 1126 goto register_pernet_fail; 1127 err = ip6_mr_init(); 1128 if (err) 1129 goto ipmr_fail; 1130 err = icmpv6_init(); 1131 if (err) 1132 goto icmp_fail; 1133 err = ndisc_init(); 1134 if (err) 1135 goto ndisc_fail; 1136 err = igmp6_init(); 1137 if (err) 1138 goto igmp_fail; 1139 1140 err = ipv6_netfilter_init(); 1141 if (err) 1142 goto netfilter_fail; 1143 /* Create /proc/foo6 entries. */ 1144 #ifdef CONFIG_PROC_FS 1145 err = -ENOMEM; 1146 if (raw6_proc_init()) 1147 goto proc_raw6_fail; 1148 if (udplite6_proc_init()) 1149 goto proc_udplite6_fail; 1150 if (ipv6_misc_proc_init()) 1151 goto proc_misc6_fail; 1152 if (if6_proc_init()) 1153 goto proc_if6_fail; 1154 #endif 1155 err = ip6_route_init(); 1156 if (err) 1157 goto ip6_route_fail; 1158 err = ndisc_late_init(); 1159 if (err) 1160 goto ndisc_late_fail; 1161 err = ip6_flowlabel_init(); 1162 if (err) 1163 goto ip6_flowlabel_fail; 1164 err = ipv6_anycast_init(); 1165 if (err) 1166 goto ipv6_anycast_fail; 1167 err = addrconf_init(); 1168 if (err) 1169 goto addrconf_fail; 1170 1171 /* Init v6 extension headers. */ 1172 err = ipv6_exthdrs_init(); 1173 if (err) 1174 goto ipv6_exthdrs_fail; 1175 1176 err = ipv6_frag_init(); 1177 if (err) 1178 goto ipv6_frag_fail; 1179 1180 /* Init v6 transport protocols. */ 1181 err = udpv6_init(); 1182 if (err) 1183 goto udpv6_fail; 1184 1185 err = udplitev6_init(); 1186 if (err) 1187 goto udplitev6_fail; 1188 1189 err = udpv6_offload_init(); 1190 if (err) 1191 goto udpv6_offload_fail; 1192 1193 err = tcpv6_init(); 1194 if (err) 1195 goto tcpv6_fail; 1196 1197 err = ipv6_packet_init(); 1198 if (err) 1199 goto ipv6_packet_fail; 1200 1201 err = pingv6_init(); 1202 if (err) 1203 goto pingv6_fail; 1204 1205 err = calipso_init(); 1206 if (err) 1207 goto calipso_fail; 1208 1209 err = seg6_init(); 1210 if (err) 1211 goto seg6_fail; 1212 1213 err = rpl_init(); 1214 if (err) 1215 goto rpl_fail; 1216 1217 err = ioam6_init(); 1218 if (err) 1219 goto ioam6_fail; 1220 1221 err = igmp6_late_init(); 1222 if (err) 1223 goto igmp6_late_err; 1224 1225 #ifdef CONFIG_SYSCTL 1226 err = ipv6_sysctl_register(); 1227 if (err) 1228 goto sysctl_fail; 1229 #endif 1230 1231 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1232 wmb(); 1233 ipv6_stub = &ipv6_stub_impl; 1234 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1235 out: 1236 return err; 1237 1238 #ifdef CONFIG_SYSCTL 1239 sysctl_fail: 1240 igmp6_late_cleanup(); 1241 #endif 1242 igmp6_late_err: 1243 ioam6_exit(); 1244 ioam6_fail: 1245 rpl_exit(); 1246 rpl_fail: 1247 seg6_exit(); 1248 seg6_fail: 1249 calipso_exit(); 1250 calipso_fail: 1251 pingv6_exit(); 1252 pingv6_fail: 1253 ipv6_packet_cleanup(); 1254 ipv6_packet_fail: 1255 tcpv6_exit(); 1256 tcpv6_fail: 1257 udpv6_offload_exit(); 1258 udpv6_offload_fail: 1259 udplitev6_exit(); 1260 udplitev6_fail: 1261 udpv6_exit(); 1262 udpv6_fail: 1263 ipv6_frag_exit(); 1264 ipv6_frag_fail: 1265 ipv6_exthdrs_exit(); 1266 ipv6_exthdrs_fail: 1267 addrconf_cleanup(); 1268 addrconf_fail: 1269 ipv6_anycast_cleanup(); 1270 ipv6_anycast_fail: 1271 ip6_flowlabel_cleanup(); 1272 ip6_flowlabel_fail: 1273 ndisc_late_cleanup(); 1274 ndisc_late_fail: 1275 ip6_route_cleanup(); 1276 ip6_route_fail: 1277 #ifdef CONFIG_PROC_FS 1278 if6_proc_exit(); 1279 proc_if6_fail: 1280 ipv6_misc_proc_exit(); 1281 proc_misc6_fail: 1282 udplite6_proc_exit(); 1283 proc_udplite6_fail: 1284 raw6_proc_exit(); 1285 proc_raw6_fail: 1286 #endif 1287 ipv6_netfilter_fini(); 1288 netfilter_fail: 1289 igmp6_cleanup(); 1290 igmp_fail: 1291 ndisc_cleanup(); 1292 ndisc_fail: 1293 icmpv6_cleanup(); 1294 icmp_fail: 1295 ip6_mr_cleanup(); 1296 ipmr_fail: 1297 unregister_pernet_subsys(&inet6_net_ops); 1298 register_pernet_fail: 1299 sock_unregister(PF_INET6); 1300 rtnl_unregister_all(PF_INET6); 1301 out_sock_register_fail: 1302 rawv6_exit(); 1303 out_unregister_ping_proto: 1304 proto_unregister(&pingv6_prot); 1305 out_unregister_raw_proto: 1306 proto_unregister(&rawv6_prot); 1307 out_unregister_udplite_proto: 1308 proto_unregister(&udplitev6_prot); 1309 out_unregister_udp_proto: 1310 proto_unregister(&udpv6_prot); 1311 out_unregister_tcp_proto: 1312 proto_unregister(&tcpv6_prot); 1313 goto out; 1314 } 1315 module_init(inet6_init); 1316 1317 MODULE_ALIAS_NETPROTO(PF_INET6); 1318