1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * PF_INET6 socket protocol family 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * 9 * Adapted from linux/net/ipv4/af_inet.c 10 * 11 * Fixes: 12 * piggy, Karl Knutson : Socket protocol table 13 * Hideaki YOSHIFUJI : sin6_scope_id support 14 * Arnaldo Melo : check proc_net_create return, cleanups 15 */ 16 17 #define pr_fmt(fmt) "IPv6: " fmt 18 19 #include <linux/module.h> 20 #include <linux/capability.h> 21 #include <linux/errno.h> 22 #include <linux/types.h> 23 #include <linux/socket.h> 24 #include <linux/in.h> 25 #include <linux/kernel.h> 26 #include <linux/timer.h> 27 #include <linux/string.h> 28 #include <linux/sockios.h> 29 #include <linux/net.h> 30 #include <linux/fcntl.h> 31 #include <linux/mm.h> 32 #include <linux/interrupt.h> 33 #include <linux/proc_fs.h> 34 #include <linux/stat.h> 35 #include <linux/init.h> 36 #include <linux/slab.h> 37 38 #include <linux/inet.h> 39 #include <linux/netdevice.h> 40 #include <linux/icmpv6.h> 41 #include <linux/netfilter_ipv6.h> 42 43 #include <net/ip.h> 44 #include <net/ipv6.h> 45 #include <net/udp.h> 46 #include <net/udplite.h> 47 #include <net/tcp.h> 48 #include <net/ping.h> 49 #include <net/protocol.h> 50 #include <net/inet_common.h> 51 #include <net/route.h> 52 #include <net/transp_v6.h> 53 #include <net/ip6_route.h> 54 #include <net/addrconf.h> 55 #include <net/ipv6_stubs.h> 56 #include <net/ndisc.h> 57 #ifdef CONFIG_IPV6_TUNNEL 58 #include <net/ip6_tunnel.h> 59 #endif 60 #include <net/calipso.h> 61 #include <net/seg6.h> 62 #include <net/rpl.h> 63 #include <net/compat.h> 64 #include <net/xfrm.h> 65 #include <net/ioam6.h> 66 #include <net/rawv6.h> 67 68 #include <linux/uaccess.h> 69 #include <linux/mroute6.h> 70 71 #include "ip6_offload.h" 72 73 MODULE_AUTHOR("Cast of dozens"); 74 MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); 75 MODULE_LICENSE("GPL"); 76 77 /* The inetsw6 table contains everything that inet6_create needs to 78 * build a new socket. 79 */ 80 static struct list_head inetsw6[SOCK_MAX]; 81 static DEFINE_SPINLOCK(inetsw6_lock); 82 83 struct ipv6_params ipv6_defaults = { 84 .disable_ipv6 = 0, 85 .autoconf = 1, 86 }; 87 88 static int disable_ipv6_mod; 89 90 module_param_named(disable, disable_ipv6_mod, int, 0444); 91 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional"); 92 93 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444); 94 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces"); 95 96 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444); 97 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces"); 98 99 bool ipv6_mod_enabled(void) 100 { 101 return disable_ipv6_mod == 0; 102 } 103 EXPORT_SYMBOL_GPL(ipv6_mod_enabled); 104 105 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) 106 { 107 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); 108 109 return (struct ipv6_pinfo *)(((u8 *)sk) + offset); 110 } 111 112 void inet6_sock_destruct(struct sock *sk) 113 { 114 inet6_cleanup_sock(sk); 115 inet_sock_destruct(sk); 116 } 117 EXPORT_SYMBOL_GPL(inet6_sock_destruct); 118 119 static int inet6_create(struct net *net, struct socket *sock, int protocol, 120 int kern) 121 { 122 struct inet_sock *inet; 123 struct ipv6_pinfo *np; 124 struct sock *sk; 125 struct inet_protosw *answer; 126 struct proto *answer_prot; 127 unsigned char answer_flags; 128 int try_loading_module = 0; 129 int err; 130 131 if (protocol < 0 || protocol >= IPPROTO_MAX) 132 return -EINVAL; 133 134 /* Look for the requested type/protocol pair. */ 135 lookup_protocol: 136 err = -ESOCKTNOSUPPORT; 137 rcu_read_lock(); 138 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) { 139 140 err = 0; 141 /* Check the non-wild match. */ 142 if (protocol == answer->protocol) { 143 if (protocol != IPPROTO_IP) 144 break; 145 } else { 146 /* Check for the two wild cases. */ 147 if (IPPROTO_IP == protocol) { 148 protocol = answer->protocol; 149 break; 150 } 151 if (IPPROTO_IP == answer->protocol) 152 break; 153 } 154 err = -EPROTONOSUPPORT; 155 } 156 157 if (err) { 158 if (try_loading_module < 2) { 159 rcu_read_unlock(); 160 /* 161 * Be more specific, e.g. net-pf-10-proto-132-type-1 162 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) 163 */ 164 if (++try_loading_module == 1) 165 request_module("net-pf-%d-proto-%d-type-%d", 166 PF_INET6, protocol, sock->type); 167 /* 168 * Fall back to generic, e.g. net-pf-10-proto-132 169 * (net-pf-PF_INET6-proto-IPPROTO_SCTP) 170 */ 171 else 172 request_module("net-pf-%d-proto-%d", 173 PF_INET6, protocol); 174 goto lookup_protocol; 175 } else 176 goto out_rcu_unlock; 177 } 178 179 err = -EPERM; 180 if (sock->type == SOCK_RAW && !kern && 181 !ns_capable(net->user_ns, CAP_NET_RAW)) 182 goto out_rcu_unlock; 183 184 sock->ops = answer->ops; 185 answer_prot = answer->prot; 186 answer_flags = answer->flags; 187 rcu_read_unlock(); 188 189 WARN_ON(!answer_prot->slab); 190 191 err = -ENOBUFS; 192 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern); 193 if (!sk) 194 goto out; 195 196 sock_init_data(sock, sk); 197 198 err = 0; 199 if (INET_PROTOSW_REUSE & answer_flags) 200 sk->sk_reuse = SK_CAN_REUSE; 201 202 inet = inet_sk(sk); 203 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; 204 205 if (SOCK_RAW == sock->type) { 206 inet->inet_num = protocol; 207 if (IPPROTO_RAW == protocol) 208 inet->hdrincl = 1; 209 } 210 211 sk->sk_destruct = inet6_sock_destruct; 212 sk->sk_family = PF_INET6; 213 sk->sk_protocol = protocol; 214 215 sk->sk_backlog_rcv = answer->prot->backlog_rcv; 216 217 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); 218 np->hop_limit = -1; 219 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS; 220 np->mc_loop = 1; 221 np->mc_all = 1; 222 np->pmtudisc = IPV6_PMTUDISC_WANT; 223 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED; 224 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only; 225 226 /* Init the ipv4 part of the socket since we can have sockets 227 * using v6 API for ipv4. 228 */ 229 inet->uc_ttl = -1; 230 231 inet->mc_loop = 1; 232 inet->mc_ttl = 1; 233 inet->mc_index = 0; 234 RCU_INIT_POINTER(inet->mc_list, NULL); 235 inet->rcv_tos = 0; 236 237 if (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc)) 238 inet->pmtudisc = IP_PMTUDISC_DONT; 239 else 240 inet->pmtudisc = IP_PMTUDISC_WANT; 241 /* 242 * Increment only the relevant sk_prot->socks debug field, this changes 243 * the previous behaviour of incrementing both the equivalent to 244 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr. 245 * 246 * This allows better debug granularity as we'll know exactly how many 247 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6 248 * transport protocol socks. -acme 249 */ 250 sk_refcnt_debug_inc(sk); 251 252 if (inet->inet_num) { 253 /* It assumes that any protocol which allows 254 * the user to assign a number at socket 255 * creation time automatically shares. 256 */ 257 inet->inet_sport = htons(inet->inet_num); 258 err = sk->sk_prot->hash(sk); 259 if (err) { 260 sk_common_release(sk); 261 goto out; 262 } 263 } 264 if (sk->sk_prot->init) { 265 err = sk->sk_prot->init(sk); 266 if (err) { 267 sk_common_release(sk); 268 goto out; 269 } 270 } 271 272 if (!kern) { 273 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); 274 if (err) { 275 sk_common_release(sk); 276 goto out; 277 } 278 } 279 out: 280 return err; 281 out_rcu_unlock: 282 rcu_read_unlock(); 283 goto out; 284 } 285 286 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, 287 u32 flags) 288 { 289 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr; 290 struct inet_sock *inet = inet_sk(sk); 291 struct ipv6_pinfo *np = inet6_sk(sk); 292 struct net *net = sock_net(sk); 293 __be32 v4addr = 0; 294 unsigned short snum; 295 bool saved_ipv6only; 296 int addr_type = 0; 297 int err = 0; 298 299 if (addr->sin6_family != AF_INET6) 300 return -EAFNOSUPPORT; 301 302 addr_type = ipv6_addr_type(&addr->sin6_addr); 303 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM) 304 return -EINVAL; 305 306 snum = ntohs(addr->sin6_port); 307 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) && 308 snum && inet_port_requires_bind_service(net, snum) && 309 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 310 return -EACCES; 311 312 if (flags & BIND_WITH_LOCK) 313 lock_sock(sk); 314 315 /* Check these errors (active socket, double bind). */ 316 if (sk->sk_state != TCP_CLOSE || inet->inet_num) { 317 err = -EINVAL; 318 goto out; 319 } 320 321 /* Check if the address belongs to the host. */ 322 if (addr_type == IPV6_ADDR_MAPPED) { 323 struct net_device *dev = NULL; 324 int chk_addr_ret; 325 326 /* Binding to v4-mapped address on a v6-only socket 327 * makes no sense 328 */ 329 if (ipv6_only_sock(sk)) { 330 err = -EINVAL; 331 goto out; 332 } 333 334 rcu_read_lock(); 335 if (sk->sk_bound_dev_if) { 336 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 337 if (!dev) { 338 err = -ENODEV; 339 goto out_unlock; 340 } 341 } 342 343 /* Reproduce AF_INET checks to make the bindings consistent */ 344 v4addr = addr->sin6_addr.s6_addr32[3]; 345 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr); 346 rcu_read_unlock(); 347 348 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr, 349 chk_addr_ret)) { 350 err = -EADDRNOTAVAIL; 351 goto out; 352 } 353 } else { 354 if (addr_type != IPV6_ADDR_ANY) { 355 struct net_device *dev = NULL; 356 357 rcu_read_lock(); 358 if (__ipv6_addr_needs_scope_id(addr_type)) { 359 if (addr_len >= sizeof(struct sockaddr_in6) && 360 addr->sin6_scope_id) { 361 /* Override any existing binding, if another one 362 * is supplied by user. 363 */ 364 sk->sk_bound_dev_if = addr->sin6_scope_id; 365 } 366 367 /* Binding to link-local address requires an interface */ 368 if (!sk->sk_bound_dev_if) { 369 err = -EINVAL; 370 goto out_unlock; 371 } 372 } 373 374 if (sk->sk_bound_dev_if) { 375 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 376 if (!dev) { 377 err = -ENODEV; 378 goto out_unlock; 379 } 380 } 381 382 /* ipv4 addr of the socket is invalid. Only the 383 * unspecified and mapped address have a v4 equivalent. 384 */ 385 v4addr = LOOPBACK4_IPV6; 386 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 387 if (!ipv6_can_nonlocal_bind(net, inet) && 388 !ipv6_chk_addr(net, &addr->sin6_addr, 389 dev, 0)) { 390 err = -EADDRNOTAVAIL; 391 goto out_unlock; 392 } 393 } 394 rcu_read_unlock(); 395 } 396 } 397 398 inet->inet_rcv_saddr = v4addr; 399 inet->inet_saddr = v4addr; 400 401 sk->sk_v6_rcv_saddr = addr->sin6_addr; 402 403 if (!(addr_type & IPV6_ADDR_MULTICAST)) 404 np->saddr = addr->sin6_addr; 405 406 saved_ipv6only = sk->sk_ipv6only; 407 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED) 408 sk->sk_ipv6only = 1; 409 410 /* Make sure we are allowed to bind here. */ 411 if (snum || !(inet->bind_address_no_port || 412 (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 413 err = sk->sk_prot->get_port(sk, snum); 414 if (err) { 415 sk->sk_ipv6only = saved_ipv6only; 416 inet_reset_saddr(sk); 417 goto out; 418 } 419 if (!(flags & BIND_FROM_BPF)) { 420 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk); 421 if (err) { 422 sk->sk_ipv6only = saved_ipv6only; 423 inet_reset_saddr(sk); 424 if (sk->sk_prot->put_port) 425 sk->sk_prot->put_port(sk); 426 goto out; 427 } 428 } 429 } 430 431 if (addr_type != IPV6_ADDR_ANY) 432 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 433 if (snum) 434 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 435 inet->inet_sport = htons(inet->inet_num); 436 inet->inet_dport = 0; 437 inet->inet_daddr = 0; 438 out: 439 if (flags & BIND_WITH_LOCK) 440 release_sock(sk); 441 return err; 442 out_unlock: 443 rcu_read_unlock(); 444 goto out; 445 } 446 447 /* bind for INET6 API */ 448 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 449 { 450 struct sock *sk = sock->sk; 451 u32 flags = BIND_WITH_LOCK; 452 const struct proto *prot; 453 int err = 0; 454 455 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 456 prot = READ_ONCE(sk->sk_prot); 457 /* If the socket has its own bind function then use it. */ 458 if (prot->bind) 459 return prot->bind(sk, uaddr, addr_len); 460 461 if (addr_len < SIN6_LEN_RFC2133) 462 return -EINVAL; 463 464 /* BPF prog is run before any checks are done so that if the prog 465 * changes context in a wrong way it will be caught. 466 */ 467 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 468 CGROUP_INET6_BIND, &flags); 469 if (err) 470 return err; 471 472 return __inet6_bind(sk, uaddr, addr_len, flags); 473 } 474 EXPORT_SYMBOL(inet6_bind); 475 476 int inet6_release(struct socket *sock) 477 { 478 struct sock *sk = sock->sk; 479 480 if (!sk) 481 return -EINVAL; 482 483 /* Free mc lists */ 484 ipv6_sock_mc_close(sk); 485 486 /* Free ac lists */ 487 ipv6_sock_ac_close(sk); 488 489 return inet_release(sock); 490 } 491 EXPORT_SYMBOL(inet6_release); 492 493 void inet6_cleanup_sock(struct sock *sk) 494 { 495 struct ipv6_pinfo *np = inet6_sk(sk); 496 struct sk_buff *skb; 497 struct ipv6_txoptions *opt; 498 499 /* Release rx options */ 500 501 skb = xchg(&np->pktoptions, NULL); 502 kfree_skb(skb); 503 504 skb = xchg(&np->rxpmtu, NULL); 505 kfree_skb(skb); 506 507 /* Free flowlabels */ 508 fl6_free_socklist(sk); 509 510 /* Free tx options */ 511 512 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 513 if (opt) { 514 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 515 txopt_put(opt); 516 } 517 } 518 EXPORT_SYMBOL_GPL(inet6_cleanup_sock); 519 520 /* 521 * This does both peername and sockname. 522 */ 523 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 524 int peer) 525 { 526 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 527 struct sock *sk = sock->sk; 528 struct inet_sock *inet = inet_sk(sk); 529 struct ipv6_pinfo *np = inet6_sk(sk); 530 531 sin->sin6_family = AF_INET6; 532 sin->sin6_flowinfo = 0; 533 sin->sin6_scope_id = 0; 534 lock_sock(sk); 535 if (peer) { 536 if (!inet->inet_dport || 537 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 538 peer == 1)) { 539 release_sock(sk); 540 return -ENOTCONN; 541 } 542 sin->sin6_port = inet->inet_dport; 543 sin->sin6_addr = sk->sk_v6_daddr; 544 if (np->sndflow) 545 sin->sin6_flowinfo = np->flow_label; 546 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 547 CGROUP_INET6_GETPEERNAME); 548 } else { 549 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 550 sin->sin6_addr = np->saddr; 551 else 552 sin->sin6_addr = sk->sk_v6_rcv_saddr; 553 sin->sin6_port = inet->inet_sport; 554 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 555 CGROUP_INET6_GETSOCKNAME); 556 } 557 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 558 sk->sk_bound_dev_if); 559 release_sock(sk); 560 return sizeof(*sin); 561 } 562 EXPORT_SYMBOL(inet6_getname); 563 564 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 565 { 566 void __user *argp = (void __user *)arg; 567 struct sock *sk = sock->sk; 568 struct net *net = sock_net(sk); 569 const struct proto *prot; 570 571 switch (cmd) { 572 case SIOCADDRT: 573 case SIOCDELRT: { 574 struct in6_rtmsg rtmsg; 575 576 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg))) 577 return -EFAULT; 578 return ipv6_route_ioctl(net, cmd, &rtmsg); 579 } 580 case SIOCSIFADDR: 581 return addrconf_add_ifaddr(net, argp); 582 case SIOCDIFADDR: 583 return addrconf_del_ifaddr(net, argp); 584 case SIOCSIFDSTADDR: 585 return addrconf_set_dstaddr(net, argp); 586 default: 587 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 588 prot = READ_ONCE(sk->sk_prot); 589 if (!prot->ioctl) 590 return -ENOIOCTLCMD; 591 return prot->ioctl(sk, cmd, arg); 592 } 593 /*NOTREACHED*/ 594 return 0; 595 } 596 EXPORT_SYMBOL(inet6_ioctl); 597 598 #ifdef CONFIG_COMPAT 599 struct compat_in6_rtmsg { 600 struct in6_addr rtmsg_dst; 601 struct in6_addr rtmsg_src; 602 struct in6_addr rtmsg_gateway; 603 u32 rtmsg_type; 604 u16 rtmsg_dst_len; 605 u16 rtmsg_src_len; 606 u32 rtmsg_metric; 607 u32 rtmsg_info; 608 u32 rtmsg_flags; 609 s32 rtmsg_ifindex; 610 }; 611 612 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 613 struct compat_in6_rtmsg __user *ur) 614 { 615 struct in6_rtmsg rt; 616 617 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst, 618 3 * sizeof(struct in6_addr)) || 619 get_user(rt.rtmsg_type, &ur->rtmsg_type) || 620 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) || 621 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) || 622 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) || 623 get_user(rt.rtmsg_info, &ur->rtmsg_info) || 624 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) || 625 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex)) 626 return -EFAULT; 627 628 629 return ipv6_route_ioctl(sock_net(sk), cmd, &rt); 630 } 631 632 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 633 { 634 void __user *argp = compat_ptr(arg); 635 struct sock *sk = sock->sk; 636 637 switch (cmd) { 638 case SIOCADDRT: 639 case SIOCDELRT: 640 return inet6_compat_routing_ioctl(sk, cmd, argp); 641 default: 642 return -ENOIOCTLCMD; 643 } 644 } 645 EXPORT_SYMBOL_GPL(inet6_compat_ioctl); 646 #endif /* CONFIG_COMPAT */ 647 648 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 649 size_t)); 650 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 651 { 652 struct sock *sk = sock->sk; 653 const struct proto *prot; 654 655 if (unlikely(inet_send_prepare(sk))) 656 return -EAGAIN; 657 658 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 659 prot = READ_ONCE(sk->sk_prot); 660 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 661 sk, msg, size); 662 } 663 664 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 665 size_t, int, int *)); 666 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 667 int flags) 668 { 669 struct sock *sk = sock->sk; 670 const struct proto *prot; 671 int addr_len = 0; 672 int err; 673 674 if (likely(!(flags & MSG_ERRQUEUE))) 675 sock_rps_record_flow(sk); 676 677 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 678 prot = READ_ONCE(sk->sk_prot); 679 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 680 sk, msg, size, flags, &addr_len); 681 if (err >= 0) 682 msg->msg_namelen = addr_len; 683 return err; 684 } 685 686 const struct proto_ops inet6_stream_ops = { 687 .family = PF_INET6, 688 .owner = THIS_MODULE, 689 .release = inet6_release, 690 .bind = inet6_bind, 691 .connect = inet_stream_connect, /* ok */ 692 .socketpair = sock_no_socketpair, /* a do nothing */ 693 .accept = inet_accept, /* ok */ 694 .getname = inet6_getname, 695 .poll = tcp_poll, /* ok */ 696 .ioctl = inet6_ioctl, /* must change */ 697 .gettstamp = sock_gettstamp, 698 .listen = inet_listen, /* ok */ 699 .shutdown = inet_shutdown, /* ok */ 700 .setsockopt = sock_common_setsockopt, /* ok */ 701 .getsockopt = sock_common_getsockopt, /* ok */ 702 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 703 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 704 #ifdef CONFIG_MMU 705 .mmap = tcp_mmap, 706 #endif 707 .sendpage = inet_sendpage, 708 .sendmsg_locked = tcp_sendmsg_locked, 709 .sendpage_locked = tcp_sendpage_locked, 710 .splice_read = tcp_splice_read, 711 .read_sock = tcp_read_sock, 712 .read_skb = tcp_read_skb, 713 .peek_len = tcp_peek_len, 714 #ifdef CONFIG_COMPAT 715 .compat_ioctl = inet6_compat_ioctl, 716 #endif 717 .set_rcvlowat = tcp_set_rcvlowat, 718 }; 719 720 const struct proto_ops inet6_dgram_ops = { 721 .family = PF_INET6, 722 .owner = THIS_MODULE, 723 .release = inet6_release, 724 .bind = inet6_bind, 725 .connect = inet_dgram_connect, /* ok */ 726 .socketpair = sock_no_socketpair, /* a do nothing */ 727 .accept = sock_no_accept, /* a do nothing */ 728 .getname = inet6_getname, 729 .poll = udp_poll, /* ok */ 730 .ioctl = inet6_ioctl, /* must change */ 731 .gettstamp = sock_gettstamp, 732 .listen = sock_no_listen, /* ok */ 733 .shutdown = inet_shutdown, /* ok */ 734 .setsockopt = sock_common_setsockopt, /* ok */ 735 .getsockopt = sock_common_getsockopt, /* ok */ 736 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 737 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 738 .read_skb = udp_read_skb, 739 .mmap = sock_no_mmap, 740 .sendpage = sock_no_sendpage, 741 .set_peek_off = sk_set_peek_off, 742 #ifdef CONFIG_COMPAT 743 .compat_ioctl = inet6_compat_ioctl, 744 #endif 745 }; 746 747 static const struct net_proto_family inet6_family_ops = { 748 .family = PF_INET6, 749 .create = inet6_create, 750 .owner = THIS_MODULE, 751 }; 752 753 int inet6_register_protosw(struct inet_protosw *p) 754 { 755 struct list_head *lh; 756 struct inet_protosw *answer; 757 struct list_head *last_perm; 758 int protocol = p->protocol; 759 int ret; 760 761 spin_lock_bh(&inetsw6_lock); 762 763 ret = -EINVAL; 764 if (p->type >= SOCK_MAX) 765 goto out_illegal; 766 767 /* If we are trying to override a permanent protocol, bail. */ 768 answer = NULL; 769 ret = -EPERM; 770 last_perm = &inetsw6[p->type]; 771 list_for_each(lh, &inetsw6[p->type]) { 772 answer = list_entry(lh, struct inet_protosw, list); 773 774 /* Check only the non-wild match. */ 775 if (INET_PROTOSW_PERMANENT & answer->flags) { 776 if (protocol == answer->protocol) 777 break; 778 last_perm = lh; 779 } 780 781 answer = NULL; 782 } 783 if (answer) 784 goto out_permanent; 785 786 /* Add the new entry after the last permanent entry if any, so that 787 * the new entry does not override a permanent entry when matched with 788 * a wild-card protocol. But it is allowed to override any existing 789 * non-permanent entry. This means that when we remove this entry, the 790 * system automatically returns to the old behavior. 791 */ 792 list_add_rcu(&p->list, last_perm); 793 ret = 0; 794 out: 795 spin_unlock_bh(&inetsw6_lock); 796 return ret; 797 798 out_permanent: 799 pr_err("Attempt to override permanent protocol %d\n", protocol); 800 goto out; 801 802 out_illegal: 803 pr_err("Ignoring attempt to register invalid socket type %d\n", 804 p->type); 805 goto out; 806 } 807 EXPORT_SYMBOL(inet6_register_protosw); 808 809 void 810 inet6_unregister_protosw(struct inet_protosw *p) 811 { 812 if (INET_PROTOSW_PERMANENT & p->flags) { 813 pr_err("Attempt to unregister permanent protocol %d\n", 814 p->protocol); 815 } else { 816 spin_lock_bh(&inetsw6_lock); 817 list_del_rcu(&p->list); 818 spin_unlock_bh(&inetsw6_lock); 819 820 synchronize_net(); 821 } 822 } 823 EXPORT_SYMBOL(inet6_unregister_protosw); 824 825 int inet6_sk_rebuild_header(struct sock *sk) 826 { 827 struct ipv6_pinfo *np = inet6_sk(sk); 828 struct dst_entry *dst; 829 830 dst = __sk_dst_check(sk, np->dst_cookie); 831 832 if (!dst) { 833 struct inet_sock *inet = inet_sk(sk); 834 struct in6_addr *final_p, final; 835 struct flowi6 fl6; 836 837 memset(&fl6, 0, sizeof(fl6)); 838 fl6.flowi6_proto = sk->sk_protocol; 839 fl6.daddr = sk->sk_v6_daddr; 840 fl6.saddr = np->saddr; 841 fl6.flowlabel = np->flow_label; 842 fl6.flowi6_oif = sk->sk_bound_dev_if; 843 fl6.flowi6_mark = sk->sk_mark; 844 fl6.fl6_dport = inet->inet_dport; 845 fl6.fl6_sport = inet->inet_sport; 846 fl6.flowi6_uid = sk->sk_uid; 847 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); 848 849 rcu_read_lock(); 850 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 851 &final); 852 rcu_read_unlock(); 853 854 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 855 if (IS_ERR(dst)) { 856 sk->sk_route_caps = 0; 857 sk->sk_err_soft = -PTR_ERR(dst); 858 return PTR_ERR(dst); 859 } 860 861 ip6_dst_store(sk, dst, NULL, NULL); 862 } 863 864 return 0; 865 } 866 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 867 868 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 869 const struct inet6_skb_parm *opt) 870 { 871 const struct ipv6_pinfo *np = inet6_sk(sk); 872 873 if (np->rxopt.all) { 874 if (((opt->flags & IP6SKB_HOPBYHOP) && 875 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 876 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 877 np->rxopt.bits.rxflow) || 878 (opt->srcrt && (np->rxopt.bits.srcrt || 879 np->rxopt.bits.osrcrt)) || 880 ((opt->dst1 || opt->dst0) && 881 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 882 return true; 883 } 884 return false; 885 } 886 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 887 888 static struct packet_type ipv6_packet_type __read_mostly = { 889 .type = cpu_to_be16(ETH_P_IPV6), 890 .func = ipv6_rcv, 891 .list_func = ipv6_list_rcv, 892 }; 893 894 static int __init ipv6_packet_init(void) 895 { 896 dev_add_pack(&ipv6_packet_type); 897 return 0; 898 } 899 900 static void ipv6_packet_cleanup(void) 901 { 902 dev_remove_pack(&ipv6_packet_type); 903 } 904 905 static int __net_init ipv6_init_mibs(struct net *net) 906 { 907 int i; 908 909 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 910 if (!net->mib.udp_stats_in6) 911 return -ENOMEM; 912 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 913 if (!net->mib.udplite_stats_in6) 914 goto err_udplite_mib; 915 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 916 if (!net->mib.ipv6_statistics) 917 goto err_ip_mib; 918 919 for_each_possible_cpu(i) { 920 struct ipstats_mib *af_inet6_stats; 921 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 922 u64_stats_init(&af_inet6_stats->syncp); 923 } 924 925 926 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 927 if (!net->mib.icmpv6_statistics) 928 goto err_icmp_mib; 929 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 930 GFP_KERNEL); 931 if (!net->mib.icmpv6msg_statistics) 932 goto err_icmpmsg_mib; 933 return 0; 934 935 err_icmpmsg_mib: 936 free_percpu(net->mib.icmpv6_statistics); 937 err_icmp_mib: 938 free_percpu(net->mib.ipv6_statistics); 939 err_ip_mib: 940 free_percpu(net->mib.udplite_stats_in6); 941 err_udplite_mib: 942 free_percpu(net->mib.udp_stats_in6); 943 return -ENOMEM; 944 } 945 946 static void ipv6_cleanup_mibs(struct net *net) 947 { 948 free_percpu(net->mib.udp_stats_in6); 949 free_percpu(net->mib.udplite_stats_in6); 950 free_percpu(net->mib.ipv6_statistics); 951 free_percpu(net->mib.icmpv6_statistics); 952 kfree(net->mib.icmpv6msg_statistics); 953 } 954 955 static int __net_init inet6_net_init(struct net *net) 956 { 957 int err = 0; 958 959 net->ipv6.sysctl.bindv6only = 0; 960 net->ipv6.sysctl.icmpv6_time = 1*HZ; 961 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 962 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0; 963 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0; 964 965 /* By default, rate limit error messages. 966 * Except for pmtu discovery, it would break it. 967 * proc_do_large_bitmap needs pointer to the bitmap. 968 */ 969 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 970 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 971 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 972 973 net->ipv6.sysctl.flowlabel_consistency = 1; 974 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 975 net->ipv6.sysctl.idgen_retries = 3; 976 net->ipv6.sysctl.idgen_delay = 1 * HZ; 977 net->ipv6.sysctl.flowlabel_state_ranges = 0; 978 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 979 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 980 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 981 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 982 net->ipv6.sysctl.fib_notify_on_flag_change = 0; 983 atomic_set(&net->ipv6.fib6_sernum, 1); 984 985 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID; 986 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE; 987 988 err = ipv6_init_mibs(net); 989 if (err) 990 return err; 991 #ifdef CONFIG_PROC_FS 992 err = udp6_proc_init(net); 993 if (err) 994 goto out; 995 err = tcp6_proc_init(net); 996 if (err) 997 goto proc_tcp6_fail; 998 err = ac6_proc_init(net); 999 if (err) 1000 goto proc_ac6_fail; 1001 #endif 1002 return err; 1003 1004 #ifdef CONFIG_PROC_FS 1005 proc_ac6_fail: 1006 tcp6_proc_exit(net); 1007 proc_tcp6_fail: 1008 udp6_proc_exit(net); 1009 out: 1010 ipv6_cleanup_mibs(net); 1011 return err; 1012 #endif 1013 } 1014 1015 static void __net_exit inet6_net_exit(struct net *net) 1016 { 1017 #ifdef CONFIG_PROC_FS 1018 udp6_proc_exit(net); 1019 tcp6_proc_exit(net); 1020 ac6_proc_exit(net); 1021 #endif 1022 ipv6_cleanup_mibs(net); 1023 } 1024 1025 static struct pernet_operations inet6_net_ops = { 1026 .init = inet6_net_init, 1027 .exit = inet6_net_exit, 1028 }; 1029 1030 static int ipv6_route_input(struct sk_buff *skb) 1031 { 1032 ip6_route_input(skb); 1033 return skb_dst(skb)->error; 1034 } 1035 1036 static const struct ipv6_stub ipv6_stub_impl = { 1037 .ipv6_sock_mc_join = ipv6_sock_mc_join, 1038 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 1039 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 1040 .ipv6_route_input = ipv6_route_input, 1041 .fib6_get_table = fib6_get_table, 1042 .fib6_table_lookup = fib6_table_lookup, 1043 .fib6_lookup = fib6_lookup, 1044 .fib6_select_path = fib6_select_path, 1045 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 1046 .fib6_nh_init = fib6_nh_init, 1047 .fib6_nh_release = fib6_nh_release, 1048 .fib6_nh_release_dsts = fib6_nh_release_dsts, 1049 .fib6_update_sernum = fib6_update_sernum_stub, 1050 .fib6_rt_update = fib6_rt_update, 1051 .ip6_del_rt = ip6_del_rt, 1052 .udpv6_encap_enable = udpv6_encap_enable, 1053 .ndisc_send_na = ndisc_send_na, 1054 #if IS_ENABLED(CONFIG_XFRM) 1055 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu, 1056 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv, 1057 .xfrm6_rcv_encap = xfrm6_rcv_encap, 1058 #endif 1059 .nd_tbl = &nd_tbl, 1060 .ipv6_fragment = ip6_fragment, 1061 .ipv6_dev_find = ipv6_dev_find, 1062 }; 1063 1064 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 1065 .inet6_bind = __inet6_bind, 1066 .udp6_lib_lookup = __udp6_lib_lookup, 1067 .ipv6_setsockopt = do_ipv6_setsockopt, 1068 .ipv6_getsockopt = do_ipv6_getsockopt, 1069 }; 1070 1071 static int __init inet6_init(void) 1072 { 1073 struct list_head *r; 1074 int err = 0; 1075 1076 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 1077 1078 /* Register the socket-side information for inet6_create. */ 1079 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 1080 INIT_LIST_HEAD(r); 1081 1082 raw_hashinfo_init(&raw_v6_hashinfo); 1083 1084 if (disable_ipv6_mod) { 1085 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 1086 goto out; 1087 } 1088 1089 err = proto_register(&tcpv6_prot, 1); 1090 if (err) 1091 goto out; 1092 1093 err = proto_register(&udpv6_prot, 1); 1094 if (err) 1095 goto out_unregister_tcp_proto; 1096 1097 err = proto_register(&udplitev6_prot, 1); 1098 if (err) 1099 goto out_unregister_udp_proto; 1100 1101 err = proto_register(&rawv6_prot, 1); 1102 if (err) 1103 goto out_unregister_udplite_proto; 1104 1105 err = proto_register(&pingv6_prot, 1); 1106 if (err) 1107 goto out_unregister_raw_proto; 1108 1109 /* We MUST register RAW sockets before we create the ICMP6, 1110 * IGMP6, or NDISC control sockets. 1111 */ 1112 err = rawv6_init(); 1113 if (err) 1114 goto out_unregister_ping_proto; 1115 1116 /* Register the family here so that the init calls below will 1117 * be able to create sockets. (?? is this dangerous ??) 1118 */ 1119 err = sock_register(&inet6_family_ops); 1120 if (err) 1121 goto out_sock_register_fail; 1122 1123 /* 1124 * ipngwg API draft makes clear that the correct semantics 1125 * for TCP and UDP is to consider one TCP and UDP instance 1126 * in a host available by both INET and INET6 APIs and 1127 * able to communicate via both network protocols. 1128 */ 1129 1130 err = register_pernet_subsys(&inet6_net_ops); 1131 if (err) 1132 goto register_pernet_fail; 1133 err = ip6_mr_init(); 1134 if (err) 1135 goto ipmr_fail; 1136 err = icmpv6_init(); 1137 if (err) 1138 goto icmp_fail; 1139 err = ndisc_init(); 1140 if (err) 1141 goto ndisc_fail; 1142 err = igmp6_init(); 1143 if (err) 1144 goto igmp_fail; 1145 1146 err = ipv6_netfilter_init(); 1147 if (err) 1148 goto netfilter_fail; 1149 /* Create /proc/foo6 entries. */ 1150 #ifdef CONFIG_PROC_FS 1151 err = -ENOMEM; 1152 if (raw6_proc_init()) 1153 goto proc_raw6_fail; 1154 if (udplite6_proc_init()) 1155 goto proc_udplite6_fail; 1156 if (ipv6_misc_proc_init()) 1157 goto proc_misc6_fail; 1158 if (if6_proc_init()) 1159 goto proc_if6_fail; 1160 #endif 1161 err = ip6_route_init(); 1162 if (err) 1163 goto ip6_route_fail; 1164 err = ndisc_late_init(); 1165 if (err) 1166 goto ndisc_late_fail; 1167 err = ip6_flowlabel_init(); 1168 if (err) 1169 goto ip6_flowlabel_fail; 1170 err = ipv6_anycast_init(); 1171 if (err) 1172 goto ipv6_anycast_fail; 1173 err = addrconf_init(); 1174 if (err) 1175 goto addrconf_fail; 1176 1177 /* Init v6 extension headers. */ 1178 err = ipv6_exthdrs_init(); 1179 if (err) 1180 goto ipv6_exthdrs_fail; 1181 1182 err = ipv6_frag_init(); 1183 if (err) 1184 goto ipv6_frag_fail; 1185 1186 /* Init v6 transport protocols. */ 1187 err = udpv6_init(); 1188 if (err) 1189 goto udpv6_fail; 1190 1191 err = udplitev6_init(); 1192 if (err) 1193 goto udplitev6_fail; 1194 1195 err = udpv6_offload_init(); 1196 if (err) 1197 goto udpv6_offload_fail; 1198 1199 err = tcpv6_init(); 1200 if (err) 1201 goto tcpv6_fail; 1202 1203 err = ipv6_packet_init(); 1204 if (err) 1205 goto ipv6_packet_fail; 1206 1207 err = pingv6_init(); 1208 if (err) 1209 goto pingv6_fail; 1210 1211 err = calipso_init(); 1212 if (err) 1213 goto calipso_fail; 1214 1215 err = seg6_init(); 1216 if (err) 1217 goto seg6_fail; 1218 1219 err = rpl_init(); 1220 if (err) 1221 goto rpl_fail; 1222 1223 err = ioam6_init(); 1224 if (err) 1225 goto ioam6_fail; 1226 1227 err = igmp6_late_init(); 1228 if (err) 1229 goto igmp6_late_err; 1230 1231 #ifdef CONFIG_SYSCTL 1232 err = ipv6_sysctl_register(); 1233 if (err) 1234 goto sysctl_fail; 1235 #endif 1236 1237 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1238 wmb(); 1239 ipv6_stub = &ipv6_stub_impl; 1240 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1241 out: 1242 return err; 1243 1244 #ifdef CONFIG_SYSCTL 1245 sysctl_fail: 1246 igmp6_late_cleanup(); 1247 #endif 1248 igmp6_late_err: 1249 ioam6_exit(); 1250 ioam6_fail: 1251 rpl_exit(); 1252 rpl_fail: 1253 seg6_exit(); 1254 seg6_fail: 1255 calipso_exit(); 1256 calipso_fail: 1257 pingv6_exit(); 1258 pingv6_fail: 1259 ipv6_packet_cleanup(); 1260 ipv6_packet_fail: 1261 tcpv6_exit(); 1262 tcpv6_fail: 1263 udpv6_offload_exit(); 1264 udpv6_offload_fail: 1265 udplitev6_exit(); 1266 udplitev6_fail: 1267 udpv6_exit(); 1268 udpv6_fail: 1269 ipv6_frag_exit(); 1270 ipv6_frag_fail: 1271 ipv6_exthdrs_exit(); 1272 ipv6_exthdrs_fail: 1273 addrconf_cleanup(); 1274 addrconf_fail: 1275 ipv6_anycast_cleanup(); 1276 ipv6_anycast_fail: 1277 ip6_flowlabel_cleanup(); 1278 ip6_flowlabel_fail: 1279 ndisc_late_cleanup(); 1280 ndisc_late_fail: 1281 ip6_route_cleanup(); 1282 ip6_route_fail: 1283 #ifdef CONFIG_PROC_FS 1284 if6_proc_exit(); 1285 proc_if6_fail: 1286 ipv6_misc_proc_exit(); 1287 proc_misc6_fail: 1288 udplite6_proc_exit(); 1289 proc_udplite6_fail: 1290 raw6_proc_exit(); 1291 proc_raw6_fail: 1292 #endif 1293 ipv6_netfilter_fini(); 1294 netfilter_fail: 1295 igmp6_cleanup(); 1296 igmp_fail: 1297 ndisc_cleanup(); 1298 ndisc_fail: 1299 icmpv6_cleanup(); 1300 icmp_fail: 1301 ip6_mr_cleanup(); 1302 ipmr_fail: 1303 unregister_pernet_subsys(&inet6_net_ops); 1304 register_pernet_fail: 1305 sock_unregister(PF_INET6); 1306 rtnl_unregister_all(PF_INET6); 1307 out_sock_register_fail: 1308 rawv6_exit(); 1309 out_unregister_ping_proto: 1310 proto_unregister(&pingv6_prot); 1311 out_unregister_raw_proto: 1312 proto_unregister(&rawv6_prot); 1313 out_unregister_udplite_proto: 1314 proto_unregister(&udplitev6_prot); 1315 out_unregister_udp_proto: 1316 proto_unregister(&udpv6_prot); 1317 out_unregister_tcp_proto: 1318 proto_unregister(&tcpv6_prot); 1319 goto out; 1320 } 1321 module_init(inet6_init); 1322 1323 MODULE_ALIAS_NETPROTO(PF_INET6); 1324