1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * PF_INET6 socket protocol family 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * 9 * Adapted from linux/net/ipv4/af_inet.c 10 * 11 * Fixes: 12 * piggy, Karl Knutson : Socket protocol table 13 * Hideaki YOSHIFUJI : sin6_scope_id support 14 * Arnaldo Melo : check proc_net_create return, cleanups 15 */ 16 17 #define pr_fmt(fmt) "IPv6: " fmt 18 19 #include <linux/module.h> 20 #include <linux/capability.h> 21 #include <linux/errno.h> 22 #include <linux/types.h> 23 #include <linux/socket.h> 24 #include <linux/in.h> 25 #include <linux/kernel.h> 26 #include <linux/timer.h> 27 #include <linux/string.h> 28 #include <linux/sockios.h> 29 #include <linux/net.h> 30 #include <linux/fcntl.h> 31 #include <linux/mm.h> 32 #include <linux/interrupt.h> 33 #include <linux/proc_fs.h> 34 #include <linux/stat.h> 35 #include <linux/init.h> 36 #include <linux/slab.h> 37 38 #include <linux/inet.h> 39 #include <linux/netdevice.h> 40 #include <linux/icmpv6.h> 41 #include <linux/netfilter_ipv6.h> 42 43 #include <net/ip.h> 44 #include <net/ipv6.h> 45 #include <net/udp.h> 46 #include <net/udplite.h> 47 #include <net/tcp.h> 48 #include <net/ping.h> 49 #include <net/protocol.h> 50 #include <net/inet_common.h> 51 #include <net/route.h> 52 #include <net/transp_v6.h> 53 #include <net/ip6_route.h> 54 #include <net/addrconf.h> 55 #include <net/ipv6_stubs.h> 56 #include <net/ndisc.h> 57 #ifdef CONFIG_IPV6_TUNNEL 58 #include <net/ip6_tunnel.h> 59 #endif 60 #include <net/calipso.h> 61 #include <net/seg6.h> 62 #include <net/rpl.h> 63 #include <net/compat.h> 64 #include <net/xfrm.h> 65 #include <net/ioam6.h> 66 #include <net/rawv6.h> 67 68 #include <linux/uaccess.h> 69 #include <linux/mroute6.h> 70 71 #include "ip6_offload.h" 72 73 MODULE_AUTHOR("Cast of dozens"); 74 MODULE_DESCRIPTION("IPv6 protocol stack for Linux"); 75 MODULE_LICENSE("GPL"); 76 77 /* The inetsw6 table contains everything that inet6_create needs to 78 * build a new socket. 79 */ 80 static struct list_head inetsw6[SOCK_MAX]; 81 static DEFINE_SPINLOCK(inetsw6_lock); 82 83 struct ipv6_params ipv6_defaults = { 84 .disable_ipv6 = 0, 85 .autoconf = 1, 86 }; 87 88 static int disable_ipv6_mod; 89 90 module_param_named(disable, disable_ipv6_mod, int, 0444); 91 MODULE_PARM_DESC(disable, "Disable IPv6 module such that it is non-functional"); 92 93 module_param_named(disable_ipv6, ipv6_defaults.disable_ipv6, int, 0444); 94 MODULE_PARM_DESC(disable_ipv6, "Disable IPv6 on all interfaces"); 95 96 module_param_named(autoconf, ipv6_defaults.autoconf, int, 0444); 97 MODULE_PARM_DESC(autoconf, "Enable IPv6 address autoconfiguration on all interfaces"); 98 99 bool ipv6_mod_enabled(void) 100 { 101 return disable_ipv6_mod == 0; 102 } 103 EXPORT_SYMBOL_GPL(ipv6_mod_enabled); 104 105 static __inline__ struct ipv6_pinfo *inet6_sk_generic(struct sock *sk) 106 { 107 const int offset = sk->sk_prot->obj_size - sizeof(struct ipv6_pinfo); 108 109 return (struct ipv6_pinfo *)(((u8 *)sk) + offset); 110 } 111 112 static int inet6_create(struct net *net, struct socket *sock, int protocol, 113 int kern) 114 { 115 struct inet_sock *inet; 116 struct ipv6_pinfo *np; 117 struct sock *sk; 118 struct inet_protosw *answer; 119 struct proto *answer_prot; 120 unsigned char answer_flags; 121 int try_loading_module = 0; 122 int err; 123 124 if (protocol < 0 || protocol >= IPPROTO_MAX) 125 return -EINVAL; 126 127 /* Look for the requested type/protocol pair. */ 128 lookup_protocol: 129 err = -ESOCKTNOSUPPORT; 130 rcu_read_lock(); 131 list_for_each_entry_rcu(answer, &inetsw6[sock->type], list) { 132 133 err = 0; 134 /* Check the non-wild match. */ 135 if (protocol == answer->protocol) { 136 if (protocol != IPPROTO_IP) 137 break; 138 } else { 139 /* Check for the two wild cases. */ 140 if (IPPROTO_IP == protocol) { 141 protocol = answer->protocol; 142 break; 143 } 144 if (IPPROTO_IP == answer->protocol) 145 break; 146 } 147 err = -EPROTONOSUPPORT; 148 } 149 150 if (err) { 151 if (try_loading_module < 2) { 152 rcu_read_unlock(); 153 /* 154 * Be more specific, e.g. net-pf-10-proto-132-type-1 155 * (net-pf-PF_INET6-proto-IPPROTO_SCTP-type-SOCK_STREAM) 156 */ 157 if (++try_loading_module == 1) 158 request_module("net-pf-%d-proto-%d-type-%d", 159 PF_INET6, protocol, sock->type); 160 /* 161 * Fall back to generic, e.g. net-pf-10-proto-132 162 * (net-pf-PF_INET6-proto-IPPROTO_SCTP) 163 */ 164 else 165 request_module("net-pf-%d-proto-%d", 166 PF_INET6, protocol); 167 goto lookup_protocol; 168 } else 169 goto out_rcu_unlock; 170 } 171 172 err = -EPERM; 173 if (sock->type == SOCK_RAW && !kern && 174 !ns_capable(net->user_ns, CAP_NET_RAW)) 175 goto out_rcu_unlock; 176 177 sock->ops = answer->ops; 178 answer_prot = answer->prot; 179 answer_flags = answer->flags; 180 rcu_read_unlock(); 181 182 WARN_ON(!answer_prot->slab); 183 184 err = -ENOBUFS; 185 sk = sk_alloc(net, PF_INET6, GFP_KERNEL, answer_prot, kern); 186 if (!sk) 187 goto out; 188 189 sock_init_data(sock, sk); 190 191 err = 0; 192 if (INET_PROTOSW_REUSE & answer_flags) 193 sk->sk_reuse = SK_CAN_REUSE; 194 195 inet = inet_sk(sk); 196 inet->is_icsk = (INET_PROTOSW_ICSK & answer_flags) != 0; 197 198 if (SOCK_RAW == sock->type) { 199 inet->inet_num = protocol; 200 if (IPPROTO_RAW == protocol) 201 inet->hdrincl = 1; 202 } 203 204 sk->sk_destruct = inet_sock_destruct; 205 sk->sk_family = PF_INET6; 206 sk->sk_protocol = protocol; 207 208 sk->sk_backlog_rcv = answer->prot->backlog_rcv; 209 210 inet_sk(sk)->pinet6 = np = inet6_sk_generic(sk); 211 np->hop_limit = -1; 212 np->mcast_hops = IPV6_DEFAULT_MCASTHOPS; 213 np->mc_loop = 1; 214 np->mc_all = 1; 215 np->pmtudisc = IPV6_PMTUDISC_WANT; 216 np->repflow = net->ipv6.sysctl.flowlabel_reflect & FLOWLABEL_REFLECT_ESTABLISHED; 217 sk->sk_ipv6only = net->ipv6.sysctl.bindv6only; 218 219 /* Init the ipv4 part of the socket since we can have sockets 220 * using v6 API for ipv4. 221 */ 222 inet->uc_ttl = -1; 223 224 inet->mc_loop = 1; 225 inet->mc_ttl = 1; 226 inet->mc_index = 0; 227 RCU_INIT_POINTER(inet->mc_list, NULL); 228 inet->rcv_tos = 0; 229 230 if (net->ipv4.sysctl_ip_no_pmtu_disc) 231 inet->pmtudisc = IP_PMTUDISC_DONT; 232 else 233 inet->pmtudisc = IP_PMTUDISC_WANT; 234 /* 235 * Increment only the relevant sk_prot->socks debug field, this changes 236 * the previous behaviour of incrementing both the equivalent to 237 * answer->prot->socks (inet6_sock_nr) and inet_sock_nr. 238 * 239 * This allows better debug granularity as we'll know exactly how many 240 * UDPv6, TCPv6, etc socks were allocated, not the sum of all IPv6 241 * transport protocol socks. -acme 242 */ 243 sk_refcnt_debug_inc(sk); 244 245 if (inet->inet_num) { 246 /* It assumes that any protocol which allows 247 * the user to assign a number at socket 248 * creation time automatically shares. 249 */ 250 inet->inet_sport = htons(inet->inet_num); 251 err = sk->sk_prot->hash(sk); 252 if (err) { 253 sk_common_release(sk); 254 goto out; 255 } 256 } 257 if (sk->sk_prot->init) { 258 err = sk->sk_prot->init(sk); 259 if (err) { 260 sk_common_release(sk); 261 goto out; 262 } 263 } 264 265 if (!kern) { 266 err = BPF_CGROUP_RUN_PROG_INET_SOCK(sk); 267 if (err) { 268 sk_common_release(sk); 269 goto out; 270 } 271 } 272 out: 273 return err; 274 out_rcu_unlock: 275 rcu_read_unlock(); 276 goto out; 277 } 278 279 static int __inet6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len, 280 u32 flags) 281 { 282 struct sockaddr_in6 *addr = (struct sockaddr_in6 *)uaddr; 283 struct inet_sock *inet = inet_sk(sk); 284 struct ipv6_pinfo *np = inet6_sk(sk); 285 struct net *net = sock_net(sk); 286 __be32 v4addr = 0; 287 unsigned short snum; 288 bool saved_ipv6only; 289 int addr_type = 0; 290 int err = 0; 291 292 if (addr->sin6_family != AF_INET6) 293 return -EAFNOSUPPORT; 294 295 addr_type = ipv6_addr_type(&addr->sin6_addr); 296 if ((addr_type & IPV6_ADDR_MULTICAST) && sk->sk_type == SOCK_STREAM) 297 return -EINVAL; 298 299 snum = ntohs(addr->sin6_port); 300 if (!(flags & BIND_NO_CAP_NET_BIND_SERVICE) && 301 snum && inet_port_requires_bind_service(net, snum) && 302 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) 303 return -EACCES; 304 305 if (flags & BIND_WITH_LOCK) 306 lock_sock(sk); 307 308 /* Check these errors (active socket, double bind). */ 309 if (sk->sk_state != TCP_CLOSE || inet->inet_num) { 310 err = -EINVAL; 311 goto out; 312 } 313 314 /* Check if the address belongs to the host. */ 315 if (addr_type == IPV6_ADDR_MAPPED) { 316 struct net_device *dev = NULL; 317 int chk_addr_ret; 318 319 /* Binding to v4-mapped address on a v6-only socket 320 * makes no sense 321 */ 322 if (ipv6_only_sock(sk)) { 323 err = -EINVAL; 324 goto out; 325 } 326 327 rcu_read_lock(); 328 if (sk->sk_bound_dev_if) { 329 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 330 if (!dev) { 331 err = -ENODEV; 332 goto out_unlock; 333 } 334 } 335 336 /* Reproduce AF_INET checks to make the bindings consistent */ 337 v4addr = addr->sin6_addr.s6_addr32[3]; 338 chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr); 339 rcu_read_unlock(); 340 341 if (!inet_addr_valid_or_nonlocal(net, inet, v4addr, 342 chk_addr_ret)) { 343 err = -EADDRNOTAVAIL; 344 goto out; 345 } 346 } else { 347 if (addr_type != IPV6_ADDR_ANY) { 348 struct net_device *dev = NULL; 349 350 rcu_read_lock(); 351 if (__ipv6_addr_needs_scope_id(addr_type)) { 352 if (addr_len >= sizeof(struct sockaddr_in6) && 353 addr->sin6_scope_id) { 354 /* Override any existing binding, if another one 355 * is supplied by user. 356 */ 357 sk->sk_bound_dev_if = addr->sin6_scope_id; 358 } 359 360 /* Binding to link-local address requires an interface */ 361 if (!sk->sk_bound_dev_if) { 362 err = -EINVAL; 363 goto out_unlock; 364 } 365 } 366 367 if (sk->sk_bound_dev_if) { 368 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if); 369 if (!dev) { 370 err = -ENODEV; 371 goto out_unlock; 372 } 373 } 374 375 /* ipv4 addr of the socket is invalid. Only the 376 * unspecified and mapped address have a v4 equivalent. 377 */ 378 v4addr = LOOPBACK4_IPV6; 379 if (!(addr_type & IPV6_ADDR_MULTICAST)) { 380 if (!ipv6_can_nonlocal_bind(net, inet) && 381 !ipv6_chk_addr(net, &addr->sin6_addr, 382 dev, 0)) { 383 err = -EADDRNOTAVAIL; 384 goto out_unlock; 385 } 386 } 387 rcu_read_unlock(); 388 } 389 } 390 391 inet->inet_rcv_saddr = v4addr; 392 inet->inet_saddr = v4addr; 393 394 sk->sk_v6_rcv_saddr = addr->sin6_addr; 395 396 if (!(addr_type & IPV6_ADDR_MULTICAST)) 397 np->saddr = addr->sin6_addr; 398 399 saved_ipv6only = sk->sk_ipv6only; 400 if (addr_type != IPV6_ADDR_ANY && addr_type != IPV6_ADDR_MAPPED) 401 sk->sk_ipv6only = 1; 402 403 /* Make sure we are allowed to bind here. */ 404 if (snum || !(inet->bind_address_no_port || 405 (flags & BIND_FORCE_ADDRESS_NO_PORT))) { 406 if (sk->sk_prot->get_port(sk, snum)) { 407 sk->sk_ipv6only = saved_ipv6only; 408 inet_reset_saddr(sk); 409 err = -EADDRINUSE; 410 goto out; 411 } 412 if (!(flags & BIND_FROM_BPF)) { 413 err = BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk); 414 if (err) { 415 sk->sk_ipv6only = saved_ipv6only; 416 inet_reset_saddr(sk); 417 if (sk->sk_prot->put_port) 418 sk->sk_prot->put_port(sk); 419 goto out; 420 } 421 } 422 } 423 424 if (addr_type != IPV6_ADDR_ANY) 425 sk->sk_userlocks |= SOCK_BINDADDR_LOCK; 426 if (snum) 427 sk->sk_userlocks |= SOCK_BINDPORT_LOCK; 428 inet->inet_sport = htons(inet->inet_num); 429 inet->inet_dport = 0; 430 inet->inet_daddr = 0; 431 out: 432 if (flags & BIND_WITH_LOCK) 433 release_sock(sk); 434 return err; 435 out_unlock: 436 rcu_read_unlock(); 437 goto out; 438 } 439 440 /* bind for INET6 API */ 441 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) 442 { 443 struct sock *sk = sock->sk; 444 u32 flags = BIND_WITH_LOCK; 445 const struct proto *prot; 446 int err = 0; 447 448 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 449 prot = READ_ONCE(sk->sk_prot); 450 /* If the socket has its own bind function then use it. */ 451 if (prot->bind) 452 return prot->bind(sk, uaddr, addr_len); 453 454 if (addr_len < SIN6_LEN_RFC2133) 455 return -EINVAL; 456 457 /* BPF prog is run before any checks are done so that if the prog 458 * changes context in a wrong way it will be caught. 459 */ 460 err = BPF_CGROUP_RUN_PROG_INET_BIND_LOCK(sk, uaddr, 461 CGROUP_INET6_BIND, &flags); 462 if (err) 463 return err; 464 465 return __inet6_bind(sk, uaddr, addr_len, flags); 466 } 467 EXPORT_SYMBOL(inet6_bind); 468 469 int inet6_release(struct socket *sock) 470 { 471 struct sock *sk = sock->sk; 472 473 if (!sk) 474 return -EINVAL; 475 476 /* Free mc lists */ 477 ipv6_sock_mc_close(sk); 478 479 /* Free ac lists */ 480 ipv6_sock_ac_close(sk); 481 482 return inet_release(sock); 483 } 484 EXPORT_SYMBOL(inet6_release); 485 486 void inet6_destroy_sock(struct sock *sk) 487 { 488 struct ipv6_pinfo *np = inet6_sk(sk); 489 struct sk_buff *skb; 490 struct ipv6_txoptions *opt; 491 492 /* Release rx options */ 493 494 skb = xchg(&np->pktoptions, NULL); 495 kfree_skb(skb); 496 497 skb = xchg(&np->rxpmtu, NULL); 498 kfree_skb(skb); 499 500 /* Free flowlabels */ 501 fl6_free_socklist(sk); 502 503 /* Free tx options */ 504 505 opt = xchg((__force struct ipv6_txoptions **)&np->opt, NULL); 506 if (opt) { 507 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 508 txopt_put(opt); 509 } 510 } 511 EXPORT_SYMBOL_GPL(inet6_destroy_sock); 512 513 /* 514 * This does both peername and sockname. 515 */ 516 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 517 int peer) 518 { 519 struct sockaddr_in6 *sin = (struct sockaddr_in6 *)uaddr; 520 struct sock *sk = sock->sk; 521 struct inet_sock *inet = inet_sk(sk); 522 struct ipv6_pinfo *np = inet6_sk(sk); 523 524 sin->sin6_family = AF_INET6; 525 sin->sin6_flowinfo = 0; 526 sin->sin6_scope_id = 0; 527 lock_sock(sk); 528 if (peer) { 529 if (!inet->inet_dport || 530 (((1 << sk->sk_state) & (TCPF_CLOSE | TCPF_SYN_SENT)) && 531 peer == 1)) { 532 release_sock(sk); 533 return -ENOTCONN; 534 } 535 sin->sin6_port = inet->inet_dport; 536 sin->sin6_addr = sk->sk_v6_daddr; 537 if (np->sndflow) 538 sin->sin6_flowinfo = np->flow_label; 539 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 540 CGROUP_INET6_GETPEERNAME); 541 } else { 542 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 543 sin->sin6_addr = np->saddr; 544 else 545 sin->sin6_addr = sk->sk_v6_rcv_saddr; 546 sin->sin6_port = inet->inet_sport; 547 BPF_CGROUP_RUN_SA_PROG(sk, (struct sockaddr *)sin, 548 CGROUP_INET6_GETSOCKNAME); 549 } 550 sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, 551 sk->sk_bound_dev_if); 552 release_sock(sk); 553 return sizeof(*sin); 554 } 555 EXPORT_SYMBOL(inet6_getname); 556 557 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 558 { 559 void __user *argp = (void __user *)arg; 560 struct sock *sk = sock->sk; 561 struct net *net = sock_net(sk); 562 const struct proto *prot; 563 564 switch (cmd) { 565 case SIOCADDRT: 566 case SIOCDELRT: { 567 struct in6_rtmsg rtmsg; 568 569 if (copy_from_user(&rtmsg, argp, sizeof(rtmsg))) 570 return -EFAULT; 571 return ipv6_route_ioctl(net, cmd, &rtmsg); 572 } 573 case SIOCSIFADDR: 574 return addrconf_add_ifaddr(net, argp); 575 case SIOCDIFADDR: 576 return addrconf_del_ifaddr(net, argp); 577 case SIOCSIFDSTADDR: 578 return addrconf_set_dstaddr(net, argp); 579 default: 580 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 581 prot = READ_ONCE(sk->sk_prot); 582 if (!prot->ioctl) 583 return -ENOIOCTLCMD; 584 return prot->ioctl(sk, cmd, arg); 585 } 586 /*NOTREACHED*/ 587 return 0; 588 } 589 EXPORT_SYMBOL(inet6_ioctl); 590 591 #ifdef CONFIG_COMPAT 592 struct compat_in6_rtmsg { 593 struct in6_addr rtmsg_dst; 594 struct in6_addr rtmsg_src; 595 struct in6_addr rtmsg_gateway; 596 u32 rtmsg_type; 597 u16 rtmsg_dst_len; 598 u16 rtmsg_src_len; 599 u32 rtmsg_metric; 600 u32 rtmsg_info; 601 u32 rtmsg_flags; 602 s32 rtmsg_ifindex; 603 }; 604 605 static int inet6_compat_routing_ioctl(struct sock *sk, unsigned int cmd, 606 struct compat_in6_rtmsg __user *ur) 607 { 608 struct in6_rtmsg rt; 609 610 if (copy_from_user(&rt.rtmsg_dst, &ur->rtmsg_dst, 611 3 * sizeof(struct in6_addr)) || 612 get_user(rt.rtmsg_type, &ur->rtmsg_type) || 613 get_user(rt.rtmsg_dst_len, &ur->rtmsg_dst_len) || 614 get_user(rt.rtmsg_src_len, &ur->rtmsg_src_len) || 615 get_user(rt.rtmsg_metric, &ur->rtmsg_metric) || 616 get_user(rt.rtmsg_info, &ur->rtmsg_info) || 617 get_user(rt.rtmsg_flags, &ur->rtmsg_flags) || 618 get_user(rt.rtmsg_ifindex, &ur->rtmsg_ifindex)) 619 return -EFAULT; 620 621 622 return ipv6_route_ioctl(sock_net(sk), cmd, &rt); 623 } 624 625 int inet6_compat_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) 626 { 627 void __user *argp = compat_ptr(arg); 628 struct sock *sk = sock->sk; 629 630 switch (cmd) { 631 case SIOCADDRT: 632 case SIOCDELRT: 633 return inet6_compat_routing_ioctl(sk, cmd, argp); 634 default: 635 return -ENOIOCTLCMD; 636 } 637 } 638 EXPORT_SYMBOL_GPL(inet6_compat_ioctl); 639 #endif /* CONFIG_COMPAT */ 640 641 INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, 642 size_t)); 643 int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) 644 { 645 struct sock *sk = sock->sk; 646 const struct proto *prot; 647 648 if (unlikely(inet_send_prepare(sk))) 649 return -EAGAIN; 650 651 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 652 prot = READ_ONCE(sk->sk_prot); 653 return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, 654 sk, msg, size); 655 } 656 657 INDIRECT_CALLABLE_DECLARE(int udpv6_recvmsg(struct sock *, struct msghdr *, 658 size_t, int, int *)); 659 int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, 660 int flags) 661 { 662 struct sock *sk = sock->sk; 663 const struct proto *prot; 664 int addr_len = 0; 665 int err; 666 667 if (likely(!(flags & MSG_ERRQUEUE))) 668 sock_rps_record_flow(sk); 669 670 /* IPV6_ADDRFORM can change sk->sk_prot under us. */ 671 prot = READ_ONCE(sk->sk_prot); 672 err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, 673 sk, msg, size, flags, &addr_len); 674 if (err >= 0) 675 msg->msg_namelen = addr_len; 676 return err; 677 } 678 679 const struct proto_ops inet6_stream_ops = { 680 .family = PF_INET6, 681 .owner = THIS_MODULE, 682 .release = inet6_release, 683 .bind = inet6_bind, 684 .connect = inet_stream_connect, /* ok */ 685 .socketpair = sock_no_socketpair, /* a do nothing */ 686 .accept = inet_accept, /* ok */ 687 .getname = inet6_getname, 688 .poll = tcp_poll, /* ok */ 689 .ioctl = inet6_ioctl, /* must change */ 690 .gettstamp = sock_gettstamp, 691 .listen = inet_listen, /* ok */ 692 .shutdown = inet_shutdown, /* ok */ 693 .setsockopt = sock_common_setsockopt, /* ok */ 694 .getsockopt = sock_common_getsockopt, /* ok */ 695 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 696 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 697 #ifdef CONFIG_MMU 698 .mmap = tcp_mmap, 699 #endif 700 .sendpage = inet_sendpage, 701 .sendmsg_locked = tcp_sendmsg_locked, 702 .sendpage_locked = tcp_sendpage_locked, 703 .splice_read = tcp_splice_read, 704 .read_sock = tcp_read_sock, 705 .peek_len = tcp_peek_len, 706 #ifdef CONFIG_COMPAT 707 .compat_ioctl = inet6_compat_ioctl, 708 #endif 709 .set_rcvlowat = tcp_set_rcvlowat, 710 }; 711 712 const struct proto_ops inet6_dgram_ops = { 713 .family = PF_INET6, 714 .owner = THIS_MODULE, 715 .release = inet6_release, 716 .bind = inet6_bind, 717 .connect = inet_dgram_connect, /* ok */ 718 .socketpair = sock_no_socketpair, /* a do nothing */ 719 .accept = sock_no_accept, /* a do nothing */ 720 .getname = inet6_getname, 721 .poll = udp_poll, /* ok */ 722 .ioctl = inet6_ioctl, /* must change */ 723 .gettstamp = sock_gettstamp, 724 .listen = sock_no_listen, /* ok */ 725 .shutdown = inet_shutdown, /* ok */ 726 .setsockopt = sock_common_setsockopt, /* ok */ 727 .getsockopt = sock_common_getsockopt, /* ok */ 728 .sendmsg = inet6_sendmsg, /* retpoline's sake */ 729 .recvmsg = inet6_recvmsg, /* retpoline's sake */ 730 .read_sock = udp_read_sock, 731 .mmap = sock_no_mmap, 732 .sendpage = sock_no_sendpage, 733 .set_peek_off = sk_set_peek_off, 734 #ifdef CONFIG_COMPAT 735 .compat_ioctl = inet6_compat_ioctl, 736 #endif 737 }; 738 739 static const struct net_proto_family inet6_family_ops = { 740 .family = PF_INET6, 741 .create = inet6_create, 742 .owner = THIS_MODULE, 743 }; 744 745 int inet6_register_protosw(struct inet_protosw *p) 746 { 747 struct list_head *lh; 748 struct inet_protosw *answer; 749 struct list_head *last_perm; 750 int protocol = p->protocol; 751 int ret; 752 753 spin_lock_bh(&inetsw6_lock); 754 755 ret = -EINVAL; 756 if (p->type >= SOCK_MAX) 757 goto out_illegal; 758 759 /* If we are trying to override a permanent protocol, bail. */ 760 answer = NULL; 761 ret = -EPERM; 762 last_perm = &inetsw6[p->type]; 763 list_for_each(lh, &inetsw6[p->type]) { 764 answer = list_entry(lh, struct inet_protosw, list); 765 766 /* Check only the non-wild match. */ 767 if (INET_PROTOSW_PERMANENT & answer->flags) { 768 if (protocol == answer->protocol) 769 break; 770 last_perm = lh; 771 } 772 773 answer = NULL; 774 } 775 if (answer) 776 goto out_permanent; 777 778 /* Add the new entry after the last permanent entry if any, so that 779 * the new entry does not override a permanent entry when matched with 780 * a wild-card protocol. But it is allowed to override any existing 781 * non-permanent entry. This means that when we remove this entry, the 782 * system automatically returns to the old behavior. 783 */ 784 list_add_rcu(&p->list, last_perm); 785 ret = 0; 786 out: 787 spin_unlock_bh(&inetsw6_lock); 788 return ret; 789 790 out_permanent: 791 pr_err("Attempt to override permanent protocol %d\n", protocol); 792 goto out; 793 794 out_illegal: 795 pr_err("Ignoring attempt to register invalid socket type %d\n", 796 p->type); 797 goto out; 798 } 799 EXPORT_SYMBOL(inet6_register_protosw); 800 801 void 802 inet6_unregister_protosw(struct inet_protosw *p) 803 { 804 if (INET_PROTOSW_PERMANENT & p->flags) { 805 pr_err("Attempt to unregister permanent protocol %d\n", 806 p->protocol); 807 } else { 808 spin_lock_bh(&inetsw6_lock); 809 list_del_rcu(&p->list); 810 spin_unlock_bh(&inetsw6_lock); 811 812 synchronize_net(); 813 } 814 } 815 EXPORT_SYMBOL(inet6_unregister_protosw); 816 817 int inet6_sk_rebuild_header(struct sock *sk) 818 { 819 struct ipv6_pinfo *np = inet6_sk(sk); 820 struct dst_entry *dst; 821 822 dst = __sk_dst_check(sk, np->dst_cookie); 823 824 if (!dst) { 825 struct inet_sock *inet = inet_sk(sk); 826 struct in6_addr *final_p, final; 827 struct flowi6 fl6; 828 829 memset(&fl6, 0, sizeof(fl6)); 830 fl6.flowi6_proto = sk->sk_protocol; 831 fl6.daddr = sk->sk_v6_daddr; 832 fl6.saddr = np->saddr; 833 fl6.flowlabel = np->flow_label; 834 fl6.flowi6_oif = sk->sk_bound_dev_if; 835 fl6.flowi6_mark = sk->sk_mark; 836 fl6.fl6_dport = inet->inet_dport; 837 fl6.fl6_sport = inet->inet_sport; 838 fl6.flowi6_uid = sk->sk_uid; 839 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); 840 841 rcu_read_lock(); 842 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), 843 &final); 844 rcu_read_unlock(); 845 846 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 847 if (IS_ERR(dst)) { 848 sk->sk_route_caps = 0; 849 sk->sk_err_soft = -PTR_ERR(dst); 850 return PTR_ERR(dst); 851 } 852 853 ip6_dst_store(sk, dst, NULL, NULL); 854 } 855 856 return 0; 857 } 858 EXPORT_SYMBOL_GPL(inet6_sk_rebuild_header); 859 860 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 861 const struct inet6_skb_parm *opt) 862 { 863 const struct ipv6_pinfo *np = inet6_sk(sk); 864 865 if (np->rxopt.all) { 866 if (((opt->flags & IP6SKB_HOPBYHOP) && 867 (np->rxopt.bits.hopopts || np->rxopt.bits.ohopopts)) || 868 (ip6_flowinfo((struct ipv6hdr *) skb_network_header(skb)) && 869 np->rxopt.bits.rxflow) || 870 (opt->srcrt && (np->rxopt.bits.srcrt || 871 np->rxopt.bits.osrcrt)) || 872 ((opt->dst1 || opt->dst0) && 873 (np->rxopt.bits.dstopts || np->rxopt.bits.odstopts))) 874 return true; 875 } 876 return false; 877 } 878 EXPORT_SYMBOL_GPL(ipv6_opt_accepted); 879 880 static struct packet_type ipv6_packet_type __read_mostly = { 881 .type = cpu_to_be16(ETH_P_IPV6), 882 .func = ipv6_rcv, 883 .list_func = ipv6_list_rcv, 884 }; 885 886 static int __init ipv6_packet_init(void) 887 { 888 dev_add_pack(&ipv6_packet_type); 889 return 0; 890 } 891 892 static void ipv6_packet_cleanup(void) 893 { 894 dev_remove_pack(&ipv6_packet_type); 895 } 896 897 static int __net_init ipv6_init_mibs(struct net *net) 898 { 899 int i; 900 901 net->mib.udp_stats_in6 = alloc_percpu(struct udp_mib); 902 if (!net->mib.udp_stats_in6) 903 return -ENOMEM; 904 net->mib.udplite_stats_in6 = alloc_percpu(struct udp_mib); 905 if (!net->mib.udplite_stats_in6) 906 goto err_udplite_mib; 907 net->mib.ipv6_statistics = alloc_percpu(struct ipstats_mib); 908 if (!net->mib.ipv6_statistics) 909 goto err_ip_mib; 910 911 for_each_possible_cpu(i) { 912 struct ipstats_mib *af_inet6_stats; 913 af_inet6_stats = per_cpu_ptr(net->mib.ipv6_statistics, i); 914 u64_stats_init(&af_inet6_stats->syncp); 915 } 916 917 918 net->mib.icmpv6_statistics = alloc_percpu(struct icmpv6_mib); 919 if (!net->mib.icmpv6_statistics) 920 goto err_icmp_mib; 921 net->mib.icmpv6msg_statistics = kzalloc(sizeof(struct icmpv6msg_mib), 922 GFP_KERNEL); 923 if (!net->mib.icmpv6msg_statistics) 924 goto err_icmpmsg_mib; 925 return 0; 926 927 err_icmpmsg_mib: 928 free_percpu(net->mib.icmpv6_statistics); 929 err_icmp_mib: 930 free_percpu(net->mib.ipv6_statistics); 931 err_ip_mib: 932 free_percpu(net->mib.udplite_stats_in6); 933 err_udplite_mib: 934 free_percpu(net->mib.udp_stats_in6); 935 return -ENOMEM; 936 } 937 938 static void ipv6_cleanup_mibs(struct net *net) 939 { 940 free_percpu(net->mib.udp_stats_in6); 941 free_percpu(net->mib.udplite_stats_in6); 942 free_percpu(net->mib.ipv6_statistics); 943 free_percpu(net->mib.icmpv6_statistics); 944 kfree(net->mib.icmpv6msg_statistics); 945 } 946 947 static int __net_init inet6_net_init(struct net *net) 948 { 949 int err = 0; 950 951 net->ipv6.sysctl.bindv6only = 0; 952 net->ipv6.sysctl.icmpv6_time = 1*HZ; 953 net->ipv6.sysctl.icmpv6_echo_ignore_all = 0; 954 net->ipv6.sysctl.icmpv6_echo_ignore_multicast = 0; 955 net->ipv6.sysctl.icmpv6_echo_ignore_anycast = 0; 956 957 /* By default, rate limit error messages. 958 * Except for pmtu discovery, it would break it. 959 * proc_do_large_bitmap needs pointer to the bitmap. 960 */ 961 bitmap_set(net->ipv6.sysctl.icmpv6_ratemask, 0, ICMPV6_ERRMSG_MAX + 1); 962 bitmap_clear(net->ipv6.sysctl.icmpv6_ratemask, ICMPV6_PKT_TOOBIG, 1); 963 net->ipv6.sysctl.icmpv6_ratemask_ptr = net->ipv6.sysctl.icmpv6_ratemask; 964 965 net->ipv6.sysctl.flowlabel_consistency = 1; 966 net->ipv6.sysctl.auto_flowlabels = IP6_DEFAULT_AUTO_FLOW_LABELS; 967 net->ipv6.sysctl.idgen_retries = 3; 968 net->ipv6.sysctl.idgen_delay = 1 * HZ; 969 net->ipv6.sysctl.flowlabel_state_ranges = 0; 970 net->ipv6.sysctl.max_dst_opts_cnt = IP6_DEFAULT_MAX_DST_OPTS_CNT; 971 net->ipv6.sysctl.max_hbh_opts_cnt = IP6_DEFAULT_MAX_HBH_OPTS_CNT; 972 net->ipv6.sysctl.max_dst_opts_len = IP6_DEFAULT_MAX_DST_OPTS_LEN; 973 net->ipv6.sysctl.max_hbh_opts_len = IP6_DEFAULT_MAX_HBH_OPTS_LEN; 974 net->ipv6.sysctl.fib_notify_on_flag_change = 0; 975 atomic_set(&net->ipv6.fib6_sernum, 1); 976 977 net->ipv6.sysctl.ioam6_id = IOAM6_DEFAULT_ID; 978 net->ipv6.sysctl.ioam6_id_wide = IOAM6_DEFAULT_ID_WIDE; 979 980 err = ipv6_init_mibs(net); 981 if (err) 982 return err; 983 #ifdef CONFIG_PROC_FS 984 err = udp6_proc_init(net); 985 if (err) 986 goto out; 987 err = tcp6_proc_init(net); 988 if (err) 989 goto proc_tcp6_fail; 990 err = ac6_proc_init(net); 991 if (err) 992 goto proc_ac6_fail; 993 #endif 994 return err; 995 996 #ifdef CONFIG_PROC_FS 997 proc_ac6_fail: 998 tcp6_proc_exit(net); 999 proc_tcp6_fail: 1000 udp6_proc_exit(net); 1001 out: 1002 ipv6_cleanup_mibs(net); 1003 return err; 1004 #endif 1005 } 1006 1007 static void __net_exit inet6_net_exit(struct net *net) 1008 { 1009 #ifdef CONFIG_PROC_FS 1010 udp6_proc_exit(net); 1011 tcp6_proc_exit(net); 1012 ac6_proc_exit(net); 1013 #endif 1014 ipv6_cleanup_mibs(net); 1015 } 1016 1017 static struct pernet_operations inet6_net_ops = { 1018 .init = inet6_net_init, 1019 .exit = inet6_net_exit, 1020 }; 1021 1022 static int ipv6_route_input(struct sk_buff *skb) 1023 { 1024 ip6_route_input(skb); 1025 return skb_dst(skb)->error; 1026 } 1027 1028 static const struct ipv6_stub ipv6_stub_impl = { 1029 .ipv6_sock_mc_join = ipv6_sock_mc_join, 1030 .ipv6_sock_mc_drop = ipv6_sock_mc_drop, 1031 .ipv6_dst_lookup_flow = ip6_dst_lookup_flow, 1032 .ipv6_route_input = ipv6_route_input, 1033 .fib6_get_table = fib6_get_table, 1034 .fib6_table_lookup = fib6_table_lookup, 1035 .fib6_lookup = fib6_lookup, 1036 .fib6_select_path = fib6_select_path, 1037 .ip6_mtu_from_fib6 = ip6_mtu_from_fib6, 1038 .fib6_nh_init = fib6_nh_init, 1039 .fib6_nh_release = fib6_nh_release, 1040 .fib6_nh_release_dsts = fib6_nh_release_dsts, 1041 .fib6_update_sernum = fib6_update_sernum_stub, 1042 .fib6_rt_update = fib6_rt_update, 1043 .ip6_del_rt = ip6_del_rt, 1044 .udpv6_encap_enable = udpv6_encap_enable, 1045 .ndisc_send_na = ndisc_send_na, 1046 #if IS_ENABLED(CONFIG_XFRM) 1047 .xfrm6_local_rxpmtu = xfrm6_local_rxpmtu, 1048 .xfrm6_udp_encap_rcv = xfrm6_udp_encap_rcv, 1049 .xfrm6_rcv_encap = xfrm6_rcv_encap, 1050 #endif 1051 .nd_tbl = &nd_tbl, 1052 .ipv6_fragment = ip6_fragment, 1053 .ipv6_dev_find = ipv6_dev_find, 1054 }; 1055 1056 static const struct ipv6_bpf_stub ipv6_bpf_stub_impl = { 1057 .inet6_bind = __inet6_bind, 1058 .udp6_lib_lookup = __udp6_lib_lookup, 1059 }; 1060 1061 static int __init inet6_init(void) 1062 { 1063 struct list_head *r; 1064 int err = 0; 1065 1066 sock_skb_cb_check_size(sizeof(struct inet6_skb_parm)); 1067 1068 /* Register the socket-side information for inet6_create. */ 1069 for (r = &inetsw6[0]; r < &inetsw6[SOCK_MAX]; ++r) 1070 INIT_LIST_HEAD(r); 1071 1072 if (disable_ipv6_mod) { 1073 pr_info("Loaded, but administratively disabled, reboot required to enable\n"); 1074 goto out; 1075 } 1076 1077 raw_hashinfo_init(&raw_v6_hashinfo); 1078 1079 err = proto_register(&tcpv6_prot, 1); 1080 if (err) 1081 goto out; 1082 1083 err = proto_register(&udpv6_prot, 1); 1084 if (err) 1085 goto out_unregister_tcp_proto; 1086 1087 err = proto_register(&udplitev6_prot, 1); 1088 if (err) 1089 goto out_unregister_udp_proto; 1090 1091 err = proto_register(&rawv6_prot, 1); 1092 if (err) 1093 goto out_unregister_udplite_proto; 1094 1095 err = proto_register(&pingv6_prot, 1); 1096 if (err) 1097 goto out_unregister_raw_proto; 1098 1099 /* We MUST register RAW sockets before we create the ICMP6, 1100 * IGMP6, or NDISC control sockets. 1101 */ 1102 err = rawv6_init(); 1103 if (err) 1104 goto out_unregister_ping_proto; 1105 1106 /* Register the family here so that the init calls below will 1107 * be able to create sockets. (?? is this dangerous ??) 1108 */ 1109 err = sock_register(&inet6_family_ops); 1110 if (err) 1111 goto out_sock_register_fail; 1112 1113 /* 1114 * ipngwg API draft makes clear that the correct semantics 1115 * for TCP and UDP is to consider one TCP and UDP instance 1116 * in a host available by both INET and INET6 APIs and 1117 * able to communicate via both network protocols. 1118 */ 1119 1120 err = register_pernet_subsys(&inet6_net_ops); 1121 if (err) 1122 goto register_pernet_fail; 1123 err = ip6_mr_init(); 1124 if (err) 1125 goto ipmr_fail; 1126 err = icmpv6_init(); 1127 if (err) 1128 goto icmp_fail; 1129 err = ndisc_init(); 1130 if (err) 1131 goto ndisc_fail; 1132 err = igmp6_init(); 1133 if (err) 1134 goto igmp_fail; 1135 1136 err = ipv6_netfilter_init(); 1137 if (err) 1138 goto netfilter_fail; 1139 /* Create /proc/foo6 entries. */ 1140 #ifdef CONFIG_PROC_FS 1141 err = -ENOMEM; 1142 if (raw6_proc_init()) 1143 goto proc_raw6_fail; 1144 if (udplite6_proc_init()) 1145 goto proc_udplite6_fail; 1146 if (ipv6_misc_proc_init()) 1147 goto proc_misc6_fail; 1148 if (if6_proc_init()) 1149 goto proc_if6_fail; 1150 #endif 1151 err = ip6_route_init(); 1152 if (err) 1153 goto ip6_route_fail; 1154 err = ndisc_late_init(); 1155 if (err) 1156 goto ndisc_late_fail; 1157 err = ip6_flowlabel_init(); 1158 if (err) 1159 goto ip6_flowlabel_fail; 1160 err = ipv6_anycast_init(); 1161 if (err) 1162 goto ipv6_anycast_fail; 1163 err = addrconf_init(); 1164 if (err) 1165 goto addrconf_fail; 1166 1167 /* Init v6 extension headers. */ 1168 err = ipv6_exthdrs_init(); 1169 if (err) 1170 goto ipv6_exthdrs_fail; 1171 1172 err = ipv6_frag_init(); 1173 if (err) 1174 goto ipv6_frag_fail; 1175 1176 /* Init v6 transport protocols. */ 1177 err = udpv6_init(); 1178 if (err) 1179 goto udpv6_fail; 1180 1181 err = udplitev6_init(); 1182 if (err) 1183 goto udplitev6_fail; 1184 1185 err = udpv6_offload_init(); 1186 if (err) 1187 goto udpv6_offload_fail; 1188 1189 err = tcpv6_init(); 1190 if (err) 1191 goto tcpv6_fail; 1192 1193 err = ipv6_packet_init(); 1194 if (err) 1195 goto ipv6_packet_fail; 1196 1197 err = pingv6_init(); 1198 if (err) 1199 goto pingv6_fail; 1200 1201 err = calipso_init(); 1202 if (err) 1203 goto calipso_fail; 1204 1205 err = seg6_init(); 1206 if (err) 1207 goto seg6_fail; 1208 1209 err = rpl_init(); 1210 if (err) 1211 goto rpl_fail; 1212 1213 err = ioam6_init(); 1214 if (err) 1215 goto ioam6_fail; 1216 1217 err = igmp6_late_init(); 1218 if (err) 1219 goto igmp6_late_err; 1220 1221 #ifdef CONFIG_SYSCTL 1222 err = ipv6_sysctl_register(); 1223 if (err) 1224 goto sysctl_fail; 1225 #endif 1226 1227 /* ensure that ipv6 stubs are visible only after ipv6 is ready */ 1228 wmb(); 1229 ipv6_stub = &ipv6_stub_impl; 1230 ipv6_bpf_stub = &ipv6_bpf_stub_impl; 1231 out: 1232 return err; 1233 1234 #ifdef CONFIG_SYSCTL 1235 sysctl_fail: 1236 igmp6_late_cleanup(); 1237 #endif 1238 igmp6_late_err: 1239 ioam6_exit(); 1240 ioam6_fail: 1241 rpl_exit(); 1242 rpl_fail: 1243 seg6_exit(); 1244 seg6_fail: 1245 calipso_exit(); 1246 calipso_fail: 1247 pingv6_exit(); 1248 pingv6_fail: 1249 ipv6_packet_cleanup(); 1250 ipv6_packet_fail: 1251 tcpv6_exit(); 1252 tcpv6_fail: 1253 udpv6_offload_exit(); 1254 udpv6_offload_fail: 1255 udplitev6_exit(); 1256 udplitev6_fail: 1257 udpv6_exit(); 1258 udpv6_fail: 1259 ipv6_frag_exit(); 1260 ipv6_frag_fail: 1261 ipv6_exthdrs_exit(); 1262 ipv6_exthdrs_fail: 1263 addrconf_cleanup(); 1264 addrconf_fail: 1265 ipv6_anycast_cleanup(); 1266 ipv6_anycast_fail: 1267 ip6_flowlabel_cleanup(); 1268 ip6_flowlabel_fail: 1269 ndisc_late_cleanup(); 1270 ndisc_late_fail: 1271 ip6_route_cleanup(); 1272 ip6_route_fail: 1273 #ifdef CONFIG_PROC_FS 1274 if6_proc_exit(); 1275 proc_if6_fail: 1276 ipv6_misc_proc_exit(); 1277 proc_misc6_fail: 1278 udplite6_proc_exit(); 1279 proc_udplite6_fail: 1280 raw6_proc_exit(); 1281 proc_raw6_fail: 1282 #endif 1283 ipv6_netfilter_fini(); 1284 netfilter_fail: 1285 igmp6_cleanup(); 1286 igmp_fail: 1287 ndisc_cleanup(); 1288 ndisc_fail: 1289 icmpv6_cleanup(); 1290 icmp_fail: 1291 ip6_mr_cleanup(); 1292 ipmr_fail: 1293 unregister_pernet_subsys(&inet6_net_ops); 1294 register_pernet_fail: 1295 sock_unregister(PF_INET6); 1296 rtnl_unregister_all(PF_INET6); 1297 out_sock_register_fail: 1298 rawv6_exit(); 1299 out_unregister_ping_proto: 1300 proto_unregister(&pingv6_prot); 1301 out_unregister_raw_proto: 1302 proto_unregister(&rawv6_prot); 1303 out_unregister_udplite_proto: 1304 proto_unregister(&udplitev6_prot); 1305 out_unregister_udp_proto: 1306 proto_unregister(&udpv6_prot); 1307 out_unregister_tcp_proto: 1308 proto_unregister(&tcpv6_prot); 1309 goto out; 1310 } 1311 module_init(inet6_init); 1312 1313 MODULE_ALIAS_NETPROTO(PF_INET6); 1314