1 /* 2 * IPv6 BSD socket options interface 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 * 8 * Based on linux/net/ipv4/ip_sockglue.c 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public License 12 * as published by the Free Software Foundation; either version 13 * 2 of the License, or (at your option) any later version. 14 * 15 * FIXME: Make the setsockopt code POSIX compliant: That is 16 * 17 * o Truncate getsockopt returns 18 * o Return an optlen of the truncated length if need be 19 * 20 * Changes: 21 * David L Stevens <dlstevens@us.ibm.com>: 22 * - added multicast source filtering API for MLDv2 23 */ 24 25 #include <linux/module.h> 26 #include <linux/capability.h> 27 #include <linux/errno.h> 28 #include <linux/types.h> 29 #include <linux/socket.h> 30 #include <linux/sockios.h> 31 #include <linux/net.h> 32 #include <linux/in6.h> 33 #include <linux/mroute6.h> 34 #include <linux/netdevice.h> 35 #include <linux/if_arp.h> 36 #include <linux/init.h> 37 #include <linux/sysctl.h> 38 #include <linux/netfilter.h> 39 #include <linux/slab.h> 40 41 #include <net/sock.h> 42 #include <net/snmp.h> 43 #include <net/ipv6.h> 44 #include <net/ndisc.h> 45 #include <net/protocol.h> 46 #include <net/transp_v6.h> 47 #include <net/ip6_route.h> 48 #include <net/addrconf.h> 49 #include <net/inet_common.h> 50 #include <net/tcp.h> 51 #include <net/udp.h> 52 #include <net/udplite.h> 53 #include <net/xfrm.h> 54 #include <net/compat.h> 55 #include <net/seg6.h> 56 57 #include <linux/uaccess.h> 58 59 struct ip6_ra_chain *ip6_ra_chain; 60 DEFINE_RWLOCK(ip6_ra_lock); 61 62 int ip6_ra_control(struct sock *sk, int sel) 63 { 64 struct ip6_ra_chain *ra, *new_ra, **rap; 65 66 /* RA packet may be delivered ONLY to IPPROTO_RAW socket */ 67 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW) 68 return -ENOPROTOOPT; 69 70 new_ra = (sel >= 0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL; 71 72 write_lock_bh(&ip6_ra_lock); 73 for (rap = &ip6_ra_chain; (ra = *rap) != NULL; rap = &ra->next) { 74 if (ra->sk == sk) { 75 if (sel >= 0) { 76 write_unlock_bh(&ip6_ra_lock); 77 kfree(new_ra); 78 return -EADDRINUSE; 79 } 80 81 *rap = ra->next; 82 write_unlock_bh(&ip6_ra_lock); 83 84 sock_put(sk); 85 kfree(ra); 86 return 0; 87 } 88 } 89 if (!new_ra) { 90 write_unlock_bh(&ip6_ra_lock); 91 return -ENOBUFS; 92 } 93 new_ra->sk = sk; 94 new_ra->sel = sel; 95 new_ra->next = ra; 96 *rap = new_ra; 97 sock_hold(sk); 98 write_unlock_bh(&ip6_ra_lock); 99 return 0; 100 } 101 102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk, 103 struct ipv6_txoptions *opt) 104 { 105 if (inet_sk(sk)->is_icsk) { 106 if (opt && 107 !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) && 108 inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) { 109 struct inet_connection_sock *icsk = inet_csk(sk); 110 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen; 111 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 112 } 113 } 114 opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt, 115 opt); 116 sk_dst_reset(sk); 117 118 return opt; 119 } 120 121 static bool setsockopt_needs_rtnl(int optname) 122 { 123 switch (optname) { 124 case IPV6_ADDRFORM: 125 case IPV6_ADD_MEMBERSHIP: 126 case IPV6_DROP_MEMBERSHIP: 127 case IPV6_JOIN_ANYCAST: 128 case IPV6_LEAVE_ANYCAST: 129 case MCAST_JOIN_GROUP: 130 case MCAST_LEAVE_GROUP: 131 case MCAST_JOIN_SOURCE_GROUP: 132 case MCAST_LEAVE_SOURCE_GROUP: 133 case MCAST_BLOCK_SOURCE: 134 case MCAST_UNBLOCK_SOURCE: 135 case MCAST_MSFILTER: 136 return true; 137 } 138 return false; 139 } 140 141 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname, 142 char __user *optval, unsigned int optlen) 143 { 144 struct ipv6_pinfo *np = inet6_sk(sk); 145 struct net *net = sock_net(sk); 146 int val, valbool; 147 int retv = -ENOPROTOOPT; 148 bool needs_rtnl = setsockopt_needs_rtnl(optname); 149 150 if (!optval) 151 val = 0; 152 else { 153 if (optlen >= sizeof(int)) { 154 if (get_user(val, (int __user *) optval)) 155 return -EFAULT; 156 } else 157 val = 0; 158 } 159 160 valbool = (val != 0); 161 162 if (ip6_mroute_opt(optname)) 163 return ip6_mroute_setsockopt(sk, optname, optval, optlen); 164 165 if (needs_rtnl) 166 rtnl_lock(); 167 lock_sock(sk); 168 169 switch (optname) { 170 171 case IPV6_ADDRFORM: 172 if (optlen < sizeof(int)) 173 goto e_inval; 174 if (val == PF_INET) { 175 struct ipv6_txoptions *opt; 176 struct sk_buff *pktopt; 177 178 if (sk->sk_type == SOCK_RAW) 179 break; 180 181 if (sk->sk_protocol == IPPROTO_UDP || 182 sk->sk_protocol == IPPROTO_UDPLITE) { 183 struct udp_sock *up = udp_sk(sk); 184 if (up->pending == AF_INET6) { 185 retv = -EBUSY; 186 break; 187 } 188 } else if (sk->sk_protocol != IPPROTO_TCP) 189 break; 190 191 if (sk->sk_state != TCP_ESTABLISHED) { 192 retv = -ENOTCONN; 193 break; 194 } 195 196 if (ipv6_only_sock(sk) || 197 !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) { 198 retv = -EADDRNOTAVAIL; 199 break; 200 } 201 202 fl6_free_socklist(sk); 203 __ipv6_sock_mc_close(sk); 204 205 /* 206 * Sock is moving from IPv6 to IPv4 (sk_prot), so 207 * remove it from the refcnt debug socks count in the 208 * original family... 209 */ 210 sk_refcnt_debug_dec(sk); 211 212 if (sk->sk_protocol == IPPROTO_TCP) { 213 struct inet_connection_sock *icsk = inet_csk(sk); 214 local_bh_disable(); 215 sock_prot_inuse_add(net, sk->sk_prot, -1); 216 sock_prot_inuse_add(net, &tcp_prot, 1); 217 local_bh_enable(); 218 sk->sk_prot = &tcp_prot; 219 icsk->icsk_af_ops = &ipv4_specific; 220 sk->sk_socket->ops = &inet_stream_ops; 221 sk->sk_family = PF_INET; 222 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie); 223 } else { 224 struct proto *prot = &udp_prot; 225 226 if (sk->sk_protocol == IPPROTO_UDPLITE) 227 prot = &udplite_prot; 228 local_bh_disable(); 229 sock_prot_inuse_add(net, sk->sk_prot, -1); 230 sock_prot_inuse_add(net, prot, 1); 231 local_bh_enable(); 232 sk->sk_prot = prot; 233 sk->sk_socket->ops = &inet_dgram_ops; 234 sk->sk_family = PF_INET; 235 } 236 opt = xchg((__force struct ipv6_txoptions **)&np->opt, 237 NULL); 238 if (opt) { 239 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 240 txopt_put(opt); 241 } 242 pktopt = xchg(&np->pktoptions, NULL); 243 kfree_skb(pktopt); 244 245 /* 246 * ... and add it to the refcnt debug socks count 247 * in the new family. -acme 248 */ 249 sk_refcnt_debug_inc(sk); 250 module_put(THIS_MODULE); 251 retv = 0; 252 break; 253 } 254 goto e_inval; 255 256 case IPV6_V6ONLY: 257 if (optlen < sizeof(int) || 258 inet_sk(sk)->inet_num) 259 goto e_inval; 260 sk->sk_ipv6only = valbool; 261 retv = 0; 262 break; 263 264 case IPV6_RECVPKTINFO: 265 if (optlen < sizeof(int)) 266 goto e_inval; 267 np->rxopt.bits.rxinfo = valbool; 268 retv = 0; 269 break; 270 271 case IPV6_2292PKTINFO: 272 if (optlen < sizeof(int)) 273 goto e_inval; 274 np->rxopt.bits.rxoinfo = valbool; 275 retv = 0; 276 break; 277 278 case IPV6_RECVHOPLIMIT: 279 if (optlen < sizeof(int)) 280 goto e_inval; 281 np->rxopt.bits.rxhlim = valbool; 282 retv = 0; 283 break; 284 285 case IPV6_2292HOPLIMIT: 286 if (optlen < sizeof(int)) 287 goto e_inval; 288 np->rxopt.bits.rxohlim = valbool; 289 retv = 0; 290 break; 291 292 case IPV6_RECVRTHDR: 293 if (optlen < sizeof(int)) 294 goto e_inval; 295 np->rxopt.bits.srcrt = valbool; 296 retv = 0; 297 break; 298 299 case IPV6_2292RTHDR: 300 if (optlen < sizeof(int)) 301 goto e_inval; 302 np->rxopt.bits.osrcrt = valbool; 303 retv = 0; 304 break; 305 306 case IPV6_RECVHOPOPTS: 307 if (optlen < sizeof(int)) 308 goto e_inval; 309 np->rxopt.bits.hopopts = valbool; 310 retv = 0; 311 break; 312 313 case IPV6_2292HOPOPTS: 314 if (optlen < sizeof(int)) 315 goto e_inval; 316 np->rxopt.bits.ohopopts = valbool; 317 retv = 0; 318 break; 319 320 case IPV6_RECVDSTOPTS: 321 if (optlen < sizeof(int)) 322 goto e_inval; 323 np->rxopt.bits.dstopts = valbool; 324 retv = 0; 325 break; 326 327 case IPV6_2292DSTOPTS: 328 if (optlen < sizeof(int)) 329 goto e_inval; 330 np->rxopt.bits.odstopts = valbool; 331 retv = 0; 332 break; 333 334 case IPV6_TCLASS: 335 if (optlen < sizeof(int)) 336 goto e_inval; 337 if (val < -1 || val > 0xff) 338 goto e_inval; 339 /* RFC 3542, 6.5: default traffic class of 0x0 */ 340 if (val == -1) 341 val = 0; 342 np->tclass = val; 343 retv = 0; 344 break; 345 346 case IPV6_RECVTCLASS: 347 if (optlen < sizeof(int)) 348 goto e_inval; 349 np->rxopt.bits.rxtclass = valbool; 350 retv = 0; 351 break; 352 353 case IPV6_FLOWINFO: 354 if (optlen < sizeof(int)) 355 goto e_inval; 356 np->rxopt.bits.rxflow = valbool; 357 retv = 0; 358 break; 359 360 case IPV6_RECVPATHMTU: 361 if (optlen < sizeof(int)) 362 goto e_inval; 363 np->rxopt.bits.rxpmtu = valbool; 364 retv = 0; 365 break; 366 367 case IPV6_TRANSPARENT: 368 if (valbool && !ns_capable(net->user_ns, CAP_NET_ADMIN) && 369 !ns_capable(net->user_ns, CAP_NET_RAW)) { 370 retv = -EPERM; 371 break; 372 } 373 if (optlen < sizeof(int)) 374 goto e_inval; 375 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */ 376 inet_sk(sk)->transparent = valbool; 377 retv = 0; 378 break; 379 380 case IPV6_FREEBIND: 381 if (optlen < sizeof(int)) 382 goto e_inval; 383 /* we also don't have a separate freebind bit for IPV6 */ 384 inet_sk(sk)->freebind = valbool; 385 retv = 0; 386 break; 387 388 case IPV6_RECVORIGDSTADDR: 389 if (optlen < sizeof(int)) 390 goto e_inval; 391 np->rxopt.bits.rxorigdstaddr = valbool; 392 retv = 0; 393 break; 394 395 case IPV6_HOPOPTS: 396 case IPV6_RTHDRDSTOPTS: 397 case IPV6_RTHDR: 398 case IPV6_DSTOPTS: 399 { 400 struct ipv6_txoptions *opt; 401 struct ipv6_opt_hdr *new = NULL; 402 403 /* hop-by-hop / destination options are privileged option */ 404 retv = -EPERM; 405 if (optname != IPV6_RTHDR && !ns_capable(net->user_ns, CAP_NET_RAW)) 406 break; 407 408 /* remove any sticky options header with a zero option 409 * length, per RFC3542. 410 */ 411 if (optlen == 0) 412 optval = NULL; 413 else if (!optval) 414 goto e_inval; 415 else if (optlen < sizeof(struct ipv6_opt_hdr) || 416 optlen & 0x7 || optlen > 8 * 255) 417 goto e_inval; 418 else { 419 new = memdup_user(optval, optlen); 420 if (IS_ERR(new)) { 421 retv = PTR_ERR(new); 422 break; 423 } 424 if (unlikely(ipv6_optlen(new) > optlen)) { 425 kfree(new); 426 goto e_inval; 427 } 428 } 429 430 opt = rcu_dereference_protected(np->opt, 431 lockdep_sock_is_held(sk)); 432 opt = ipv6_renew_options(sk, opt, optname, new); 433 kfree(new); 434 if (IS_ERR(opt)) { 435 retv = PTR_ERR(opt); 436 break; 437 } 438 439 /* routing header option needs extra check */ 440 retv = -EINVAL; 441 if (optname == IPV6_RTHDR && opt && opt->srcrt) { 442 struct ipv6_rt_hdr *rthdr = opt->srcrt; 443 switch (rthdr->type) { 444 #if IS_ENABLED(CONFIG_IPV6_MIP6) 445 case IPV6_SRCRT_TYPE_2: 446 if (rthdr->hdrlen != 2 || 447 rthdr->segments_left != 1) 448 goto sticky_done; 449 450 break; 451 #endif 452 case IPV6_SRCRT_TYPE_4: 453 { 454 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *) 455 opt->srcrt; 456 457 if (!seg6_validate_srh(srh, optlen)) 458 goto sticky_done; 459 break; 460 } 461 default: 462 goto sticky_done; 463 } 464 } 465 466 retv = 0; 467 opt = ipv6_update_options(sk, opt); 468 sticky_done: 469 if (opt) { 470 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 471 txopt_put(opt); 472 } 473 break; 474 } 475 476 case IPV6_PKTINFO: 477 { 478 struct in6_pktinfo pkt; 479 480 if (optlen == 0) 481 goto e_inval; 482 else if (optlen < sizeof(struct in6_pktinfo) || !optval) 483 goto e_inval; 484 485 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) { 486 retv = -EFAULT; 487 break; 488 } 489 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if) 490 goto e_inval; 491 492 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex; 493 np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr; 494 retv = 0; 495 break; 496 } 497 498 case IPV6_2292PKTOPTIONS: 499 { 500 struct ipv6_txoptions *opt = NULL; 501 struct msghdr msg; 502 struct flowi6 fl6; 503 struct sockcm_cookie sockc_junk; 504 struct ipcm6_cookie ipc6; 505 506 memset(&fl6, 0, sizeof(fl6)); 507 fl6.flowi6_oif = sk->sk_bound_dev_if; 508 fl6.flowi6_mark = sk->sk_mark; 509 510 if (optlen == 0) 511 goto update; 512 513 /* 1K is probably excessive 514 * 1K is surely not enough, 2K per standard header is 16K. 515 */ 516 retv = -EINVAL; 517 if (optlen > 64*1024) 518 break; 519 520 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL); 521 retv = -ENOBUFS; 522 if (!opt) 523 break; 524 525 memset(opt, 0, sizeof(*opt)); 526 refcount_set(&opt->refcnt, 1); 527 opt->tot_len = sizeof(*opt) + optlen; 528 retv = -EFAULT; 529 if (copy_from_user(opt+1, optval, optlen)) 530 goto done; 531 532 msg.msg_controllen = optlen; 533 msg.msg_control = (void *)(opt+1); 534 ipc6.opt = opt; 535 536 retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6, &sockc_junk); 537 if (retv) 538 goto done; 539 update: 540 retv = 0; 541 opt = ipv6_update_options(sk, opt); 542 done: 543 if (opt) { 544 atomic_sub(opt->tot_len, &sk->sk_omem_alloc); 545 txopt_put(opt); 546 } 547 break; 548 } 549 case IPV6_UNICAST_HOPS: 550 if (optlen < sizeof(int)) 551 goto e_inval; 552 if (val > 255 || val < -1) 553 goto e_inval; 554 np->hop_limit = val; 555 retv = 0; 556 break; 557 558 case IPV6_MULTICAST_HOPS: 559 if (sk->sk_type == SOCK_STREAM) 560 break; 561 if (optlen < sizeof(int)) 562 goto e_inval; 563 if (val > 255 || val < -1) 564 goto e_inval; 565 np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val); 566 retv = 0; 567 break; 568 569 case IPV6_MULTICAST_LOOP: 570 if (optlen < sizeof(int)) 571 goto e_inval; 572 if (val != valbool) 573 goto e_inval; 574 np->mc_loop = valbool; 575 retv = 0; 576 break; 577 578 case IPV6_UNICAST_IF: 579 { 580 struct net_device *dev = NULL; 581 int ifindex; 582 583 if (optlen != sizeof(int)) 584 goto e_inval; 585 586 ifindex = (__force int)ntohl((__force __be32)val); 587 if (ifindex == 0) { 588 np->ucast_oif = 0; 589 retv = 0; 590 break; 591 } 592 593 dev = dev_get_by_index(net, ifindex); 594 retv = -EADDRNOTAVAIL; 595 if (!dev) 596 break; 597 dev_put(dev); 598 599 retv = -EINVAL; 600 if (sk->sk_bound_dev_if) 601 break; 602 603 np->ucast_oif = ifindex; 604 retv = 0; 605 break; 606 } 607 608 case IPV6_MULTICAST_IF: 609 if (sk->sk_type == SOCK_STREAM) 610 break; 611 if (optlen < sizeof(int)) 612 goto e_inval; 613 614 if (val) { 615 struct net_device *dev; 616 int midx; 617 618 rcu_read_lock(); 619 620 dev = dev_get_by_index_rcu(net, val); 621 if (!dev) { 622 rcu_read_unlock(); 623 retv = -ENODEV; 624 break; 625 } 626 midx = l3mdev_master_ifindex_rcu(dev); 627 628 rcu_read_unlock(); 629 630 if (sk->sk_bound_dev_if && 631 sk->sk_bound_dev_if != val && 632 (!midx || midx != sk->sk_bound_dev_if)) 633 goto e_inval; 634 } 635 np->mcast_oif = val; 636 retv = 0; 637 break; 638 case IPV6_ADD_MEMBERSHIP: 639 case IPV6_DROP_MEMBERSHIP: 640 { 641 struct ipv6_mreq mreq; 642 643 if (optlen < sizeof(struct ipv6_mreq)) 644 goto e_inval; 645 646 retv = -EPROTO; 647 if (inet_sk(sk)->is_icsk) 648 break; 649 650 retv = -EFAULT; 651 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq))) 652 break; 653 654 if (optname == IPV6_ADD_MEMBERSHIP) 655 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr); 656 else 657 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr); 658 break; 659 } 660 case IPV6_JOIN_ANYCAST: 661 case IPV6_LEAVE_ANYCAST: 662 { 663 struct ipv6_mreq mreq; 664 665 if (optlen < sizeof(struct ipv6_mreq)) 666 goto e_inval; 667 668 retv = -EFAULT; 669 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq))) 670 break; 671 672 if (optname == IPV6_JOIN_ANYCAST) 673 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr); 674 else 675 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr); 676 break; 677 } 678 case MCAST_JOIN_GROUP: 679 case MCAST_LEAVE_GROUP: 680 { 681 struct group_req greq; 682 struct sockaddr_in6 *psin6; 683 684 if (optlen < sizeof(struct group_req)) 685 goto e_inval; 686 687 retv = -EFAULT; 688 if (copy_from_user(&greq, optval, sizeof(struct group_req))) 689 break; 690 if (greq.gr_group.ss_family != AF_INET6) { 691 retv = -EADDRNOTAVAIL; 692 break; 693 } 694 psin6 = (struct sockaddr_in6 *)&greq.gr_group; 695 if (optname == MCAST_JOIN_GROUP) 696 retv = ipv6_sock_mc_join(sk, greq.gr_interface, 697 &psin6->sin6_addr); 698 else 699 retv = ipv6_sock_mc_drop(sk, greq.gr_interface, 700 &psin6->sin6_addr); 701 break; 702 } 703 case MCAST_JOIN_SOURCE_GROUP: 704 case MCAST_LEAVE_SOURCE_GROUP: 705 case MCAST_BLOCK_SOURCE: 706 case MCAST_UNBLOCK_SOURCE: 707 { 708 struct group_source_req greqs; 709 int omode, add; 710 711 if (optlen < sizeof(struct group_source_req)) 712 goto e_inval; 713 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 714 retv = -EFAULT; 715 break; 716 } 717 if (greqs.gsr_group.ss_family != AF_INET6 || 718 greqs.gsr_source.ss_family != AF_INET6) { 719 retv = -EADDRNOTAVAIL; 720 break; 721 } 722 if (optname == MCAST_BLOCK_SOURCE) { 723 omode = MCAST_EXCLUDE; 724 add = 1; 725 } else if (optname == MCAST_UNBLOCK_SOURCE) { 726 omode = MCAST_EXCLUDE; 727 add = 0; 728 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 729 struct sockaddr_in6 *psin6; 730 731 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group; 732 retv = ipv6_sock_mc_join_ssm(sk, greqs.gsr_interface, 733 &psin6->sin6_addr, 734 MCAST_INCLUDE); 735 /* prior join w/ different source is ok */ 736 if (retv && retv != -EADDRINUSE) 737 break; 738 omode = MCAST_INCLUDE; 739 add = 1; 740 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 741 omode = MCAST_INCLUDE; 742 add = 0; 743 } 744 retv = ip6_mc_source(add, omode, sk, &greqs); 745 break; 746 } 747 case MCAST_MSFILTER: 748 { 749 struct group_filter *gsf; 750 751 if (optlen < GROUP_FILTER_SIZE(0)) 752 goto e_inval; 753 if (optlen > sysctl_optmem_max) { 754 retv = -ENOBUFS; 755 break; 756 } 757 gsf = memdup_user(optval, optlen); 758 if (IS_ERR(gsf)) { 759 retv = PTR_ERR(gsf); 760 break; 761 } 762 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 763 if (gsf->gf_numsrc >= 0x1ffffffU || 764 gsf->gf_numsrc > sysctl_mld_max_msf) { 765 kfree(gsf); 766 retv = -ENOBUFS; 767 break; 768 } 769 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 770 kfree(gsf); 771 retv = -EINVAL; 772 break; 773 } 774 retv = ip6_mc_msfilter(sk, gsf); 775 kfree(gsf); 776 777 break; 778 } 779 case IPV6_ROUTER_ALERT: 780 if (optlen < sizeof(int)) 781 goto e_inval; 782 retv = ip6_ra_control(sk, val); 783 break; 784 case IPV6_MTU_DISCOVER: 785 if (optlen < sizeof(int)) 786 goto e_inval; 787 if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT) 788 goto e_inval; 789 np->pmtudisc = val; 790 retv = 0; 791 break; 792 case IPV6_MTU: 793 if (optlen < sizeof(int)) 794 goto e_inval; 795 if (val && val < IPV6_MIN_MTU) 796 goto e_inval; 797 np->frag_size = val; 798 retv = 0; 799 break; 800 case IPV6_RECVERR: 801 if (optlen < sizeof(int)) 802 goto e_inval; 803 np->recverr = valbool; 804 if (!val) 805 skb_queue_purge(&sk->sk_error_queue); 806 retv = 0; 807 break; 808 case IPV6_FLOWINFO_SEND: 809 if (optlen < sizeof(int)) 810 goto e_inval; 811 np->sndflow = valbool; 812 retv = 0; 813 break; 814 case IPV6_FLOWLABEL_MGR: 815 retv = ipv6_flowlabel_opt(sk, optval, optlen); 816 break; 817 case IPV6_IPSEC_POLICY: 818 case IPV6_XFRM_POLICY: 819 retv = -EPERM; 820 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 821 break; 822 retv = xfrm_user_policy(sk, optname, optval, optlen); 823 break; 824 825 case IPV6_ADDR_PREFERENCES: 826 { 827 unsigned int pref = 0; 828 unsigned int prefmask = ~0; 829 830 if (optlen < sizeof(int)) 831 goto e_inval; 832 833 retv = -EINVAL; 834 835 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */ 836 switch (val & (IPV6_PREFER_SRC_PUBLIC| 837 IPV6_PREFER_SRC_TMP| 838 IPV6_PREFER_SRC_PUBTMP_DEFAULT)) { 839 case IPV6_PREFER_SRC_PUBLIC: 840 pref |= IPV6_PREFER_SRC_PUBLIC; 841 break; 842 case IPV6_PREFER_SRC_TMP: 843 pref |= IPV6_PREFER_SRC_TMP; 844 break; 845 case IPV6_PREFER_SRC_PUBTMP_DEFAULT: 846 break; 847 case 0: 848 goto pref_skip_pubtmp; 849 default: 850 goto e_inval; 851 } 852 853 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC| 854 IPV6_PREFER_SRC_TMP); 855 pref_skip_pubtmp: 856 857 /* check HOME/COA conflicts */ 858 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) { 859 case IPV6_PREFER_SRC_HOME: 860 break; 861 case IPV6_PREFER_SRC_COA: 862 pref |= IPV6_PREFER_SRC_COA; 863 case 0: 864 goto pref_skip_coa; 865 default: 866 goto e_inval; 867 } 868 869 prefmask &= ~IPV6_PREFER_SRC_COA; 870 pref_skip_coa: 871 872 /* check CGA/NONCGA conflicts */ 873 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) { 874 case IPV6_PREFER_SRC_CGA: 875 case IPV6_PREFER_SRC_NONCGA: 876 case 0: 877 break; 878 default: 879 goto e_inval; 880 } 881 882 np->srcprefs = (np->srcprefs & prefmask) | pref; 883 retv = 0; 884 885 break; 886 } 887 case IPV6_MINHOPCOUNT: 888 if (optlen < sizeof(int)) 889 goto e_inval; 890 if (val < 0 || val > 255) 891 goto e_inval; 892 np->min_hopcount = val; 893 retv = 0; 894 break; 895 case IPV6_DONTFRAG: 896 np->dontfrag = valbool; 897 retv = 0; 898 break; 899 case IPV6_AUTOFLOWLABEL: 900 np->autoflowlabel = valbool; 901 np->autoflowlabel_set = 1; 902 retv = 0; 903 break; 904 case IPV6_RECVFRAGSIZE: 905 np->rxopt.bits.recvfragsize = valbool; 906 retv = 0; 907 break; 908 } 909 910 release_sock(sk); 911 if (needs_rtnl) 912 rtnl_unlock(); 913 914 return retv; 915 916 e_inval: 917 release_sock(sk); 918 if (needs_rtnl) 919 rtnl_unlock(); 920 return -EINVAL; 921 } 922 923 int ipv6_setsockopt(struct sock *sk, int level, int optname, 924 char __user *optval, unsigned int optlen) 925 { 926 int err; 927 928 if (level == SOL_IP && sk->sk_type != SOCK_RAW) 929 return udp_prot.setsockopt(sk, level, optname, optval, optlen); 930 931 if (level != SOL_IPV6) 932 return -ENOPROTOOPT; 933 934 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen); 935 #ifdef CONFIG_NETFILTER 936 /* we need to exclude all possible ENOPROTOOPTs except default case */ 937 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY && 938 optname != IPV6_XFRM_POLICY) 939 err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen); 940 #endif 941 return err; 942 } 943 EXPORT_SYMBOL(ipv6_setsockopt); 944 945 #ifdef CONFIG_COMPAT 946 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname, 947 char __user *optval, unsigned int optlen) 948 { 949 int err; 950 951 if (level == SOL_IP && sk->sk_type != SOCK_RAW) { 952 if (udp_prot.compat_setsockopt != NULL) 953 return udp_prot.compat_setsockopt(sk, level, optname, 954 optval, optlen); 955 return udp_prot.setsockopt(sk, level, optname, optval, optlen); 956 } 957 958 if (level != SOL_IPV6) 959 return -ENOPROTOOPT; 960 961 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 962 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 963 ipv6_setsockopt); 964 965 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen); 966 #ifdef CONFIG_NETFILTER 967 /* we need to exclude all possible ENOPROTOOPTs except default case */ 968 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY && 969 optname != IPV6_XFRM_POLICY) 970 err = compat_nf_setsockopt(sk, PF_INET6, optname, optval, 971 optlen); 972 #endif 973 return err; 974 } 975 EXPORT_SYMBOL(compat_ipv6_setsockopt); 976 #endif 977 978 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt, 979 int optname, char __user *optval, int len) 980 { 981 struct ipv6_opt_hdr *hdr; 982 983 if (!opt) 984 return 0; 985 986 switch (optname) { 987 case IPV6_HOPOPTS: 988 hdr = opt->hopopt; 989 break; 990 case IPV6_RTHDRDSTOPTS: 991 hdr = opt->dst0opt; 992 break; 993 case IPV6_RTHDR: 994 hdr = (struct ipv6_opt_hdr *)opt->srcrt; 995 break; 996 case IPV6_DSTOPTS: 997 hdr = opt->dst1opt; 998 break; 999 default: 1000 return -EINVAL; /* should not happen */ 1001 } 1002 1003 if (!hdr) 1004 return 0; 1005 1006 len = min_t(unsigned int, len, ipv6_optlen(hdr)); 1007 if (copy_to_user(optval, hdr, len)) 1008 return -EFAULT; 1009 return len; 1010 } 1011 1012 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname, 1013 char __user *optval, int __user *optlen, unsigned int flags) 1014 { 1015 struct ipv6_pinfo *np = inet6_sk(sk); 1016 int len; 1017 int val; 1018 1019 if (ip6_mroute_opt(optname)) 1020 return ip6_mroute_getsockopt(sk, optname, optval, optlen); 1021 1022 if (get_user(len, optlen)) 1023 return -EFAULT; 1024 switch (optname) { 1025 case IPV6_ADDRFORM: 1026 if (sk->sk_protocol != IPPROTO_UDP && 1027 sk->sk_protocol != IPPROTO_UDPLITE && 1028 sk->sk_protocol != IPPROTO_TCP) 1029 return -ENOPROTOOPT; 1030 if (sk->sk_state != TCP_ESTABLISHED) 1031 return -ENOTCONN; 1032 val = sk->sk_family; 1033 break; 1034 case MCAST_MSFILTER: 1035 { 1036 struct group_filter gsf; 1037 int err; 1038 1039 if (len < GROUP_FILTER_SIZE(0)) 1040 return -EINVAL; 1041 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) 1042 return -EFAULT; 1043 if (gsf.gf_group.ss_family != AF_INET6) 1044 return -EADDRNOTAVAIL; 1045 lock_sock(sk); 1046 err = ip6_mc_msfget(sk, &gsf, 1047 (struct group_filter __user *)optval, optlen); 1048 release_sock(sk); 1049 return err; 1050 } 1051 1052 case IPV6_2292PKTOPTIONS: 1053 { 1054 struct msghdr msg; 1055 struct sk_buff *skb; 1056 1057 if (sk->sk_type != SOCK_STREAM) 1058 return -ENOPROTOOPT; 1059 1060 msg.msg_control = optval; 1061 msg.msg_controllen = len; 1062 msg.msg_flags = flags; 1063 1064 lock_sock(sk); 1065 skb = np->pktoptions; 1066 if (skb) 1067 ip6_datagram_recv_ctl(sk, &msg, skb); 1068 release_sock(sk); 1069 if (!skb) { 1070 if (np->rxopt.bits.rxinfo) { 1071 struct in6_pktinfo src_info; 1072 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif : 1073 np->sticky_pktinfo.ipi6_ifindex; 1074 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr; 1075 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info); 1076 } 1077 if (np->rxopt.bits.rxhlim) { 1078 int hlim = np->mcast_hops; 1079 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim); 1080 } 1081 if (np->rxopt.bits.rxtclass) { 1082 int tclass = (int)ip6_tclass(np->rcv_flowinfo); 1083 1084 put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass); 1085 } 1086 if (np->rxopt.bits.rxoinfo) { 1087 struct in6_pktinfo src_info; 1088 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif : 1089 np->sticky_pktinfo.ipi6_ifindex; 1090 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : 1091 np->sticky_pktinfo.ipi6_addr; 1092 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info); 1093 } 1094 if (np->rxopt.bits.rxohlim) { 1095 int hlim = np->mcast_hops; 1096 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim); 1097 } 1098 if (np->rxopt.bits.rxflow) { 1099 __be32 flowinfo = np->rcv_flowinfo; 1100 1101 put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo); 1102 } 1103 } 1104 len -= msg.msg_controllen; 1105 return put_user(len, optlen); 1106 } 1107 case IPV6_MTU: 1108 { 1109 struct dst_entry *dst; 1110 1111 val = 0; 1112 rcu_read_lock(); 1113 dst = __sk_dst_get(sk); 1114 if (dst) 1115 val = dst_mtu(dst); 1116 rcu_read_unlock(); 1117 if (!val) 1118 return -ENOTCONN; 1119 break; 1120 } 1121 1122 case IPV6_V6ONLY: 1123 val = sk->sk_ipv6only; 1124 break; 1125 1126 case IPV6_RECVPKTINFO: 1127 val = np->rxopt.bits.rxinfo; 1128 break; 1129 1130 case IPV6_2292PKTINFO: 1131 val = np->rxopt.bits.rxoinfo; 1132 break; 1133 1134 case IPV6_RECVHOPLIMIT: 1135 val = np->rxopt.bits.rxhlim; 1136 break; 1137 1138 case IPV6_2292HOPLIMIT: 1139 val = np->rxopt.bits.rxohlim; 1140 break; 1141 1142 case IPV6_RECVRTHDR: 1143 val = np->rxopt.bits.srcrt; 1144 break; 1145 1146 case IPV6_2292RTHDR: 1147 val = np->rxopt.bits.osrcrt; 1148 break; 1149 1150 case IPV6_HOPOPTS: 1151 case IPV6_RTHDRDSTOPTS: 1152 case IPV6_RTHDR: 1153 case IPV6_DSTOPTS: 1154 { 1155 struct ipv6_txoptions *opt; 1156 1157 lock_sock(sk); 1158 opt = rcu_dereference_protected(np->opt, 1159 lockdep_sock_is_held(sk)); 1160 len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len); 1161 release_sock(sk); 1162 /* check if ipv6_getsockopt_sticky() returns err code */ 1163 if (len < 0) 1164 return len; 1165 return put_user(len, optlen); 1166 } 1167 1168 case IPV6_RECVHOPOPTS: 1169 val = np->rxopt.bits.hopopts; 1170 break; 1171 1172 case IPV6_2292HOPOPTS: 1173 val = np->rxopt.bits.ohopopts; 1174 break; 1175 1176 case IPV6_RECVDSTOPTS: 1177 val = np->rxopt.bits.dstopts; 1178 break; 1179 1180 case IPV6_2292DSTOPTS: 1181 val = np->rxopt.bits.odstopts; 1182 break; 1183 1184 case IPV6_TCLASS: 1185 val = np->tclass; 1186 break; 1187 1188 case IPV6_RECVTCLASS: 1189 val = np->rxopt.bits.rxtclass; 1190 break; 1191 1192 case IPV6_FLOWINFO: 1193 val = np->rxopt.bits.rxflow; 1194 break; 1195 1196 case IPV6_RECVPATHMTU: 1197 val = np->rxopt.bits.rxpmtu; 1198 break; 1199 1200 case IPV6_PATHMTU: 1201 { 1202 struct dst_entry *dst; 1203 struct ip6_mtuinfo mtuinfo; 1204 1205 if (len < sizeof(mtuinfo)) 1206 return -EINVAL; 1207 1208 len = sizeof(mtuinfo); 1209 memset(&mtuinfo, 0, sizeof(mtuinfo)); 1210 1211 rcu_read_lock(); 1212 dst = __sk_dst_get(sk); 1213 if (dst) 1214 mtuinfo.ip6m_mtu = dst_mtu(dst); 1215 rcu_read_unlock(); 1216 if (!mtuinfo.ip6m_mtu) 1217 return -ENOTCONN; 1218 1219 if (put_user(len, optlen)) 1220 return -EFAULT; 1221 if (copy_to_user(optval, &mtuinfo, len)) 1222 return -EFAULT; 1223 1224 return 0; 1225 } 1226 1227 case IPV6_TRANSPARENT: 1228 val = inet_sk(sk)->transparent; 1229 break; 1230 1231 case IPV6_FREEBIND: 1232 val = inet_sk(sk)->freebind; 1233 break; 1234 1235 case IPV6_RECVORIGDSTADDR: 1236 val = np->rxopt.bits.rxorigdstaddr; 1237 break; 1238 1239 case IPV6_UNICAST_HOPS: 1240 case IPV6_MULTICAST_HOPS: 1241 { 1242 struct dst_entry *dst; 1243 1244 if (optname == IPV6_UNICAST_HOPS) 1245 val = np->hop_limit; 1246 else 1247 val = np->mcast_hops; 1248 1249 if (val < 0) { 1250 rcu_read_lock(); 1251 dst = __sk_dst_get(sk); 1252 if (dst) 1253 val = ip6_dst_hoplimit(dst); 1254 rcu_read_unlock(); 1255 } 1256 1257 if (val < 0) 1258 val = sock_net(sk)->ipv6.devconf_all->hop_limit; 1259 break; 1260 } 1261 1262 case IPV6_MULTICAST_LOOP: 1263 val = np->mc_loop; 1264 break; 1265 1266 case IPV6_MULTICAST_IF: 1267 val = np->mcast_oif; 1268 break; 1269 1270 case IPV6_UNICAST_IF: 1271 val = (__force int)htonl((__u32) np->ucast_oif); 1272 break; 1273 1274 case IPV6_MTU_DISCOVER: 1275 val = np->pmtudisc; 1276 break; 1277 1278 case IPV6_RECVERR: 1279 val = np->recverr; 1280 break; 1281 1282 case IPV6_FLOWINFO_SEND: 1283 val = np->sndflow; 1284 break; 1285 1286 case IPV6_FLOWLABEL_MGR: 1287 { 1288 struct in6_flowlabel_req freq; 1289 int flags; 1290 1291 if (len < sizeof(freq)) 1292 return -EINVAL; 1293 1294 if (copy_from_user(&freq, optval, sizeof(freq))) 1295 return -EFAULT; 1296 1297 if (freq.flr_action != IPV6_FL_A_GET) 1298 return -EINVAL; 1299 1300 len = sizeof(freq); 1301 flags = freq.flr_flags; 1302 1303 memset(&freq, 0, sizeof(freq)); 1304 1305 val = ipv6_flowlabel_opt_get(sk, &freq, flags); 1306 if (val < 0) 1307 return val; 1308 1309 if (put_user(len, optlen)) 1310 return -EFAULT; 1311 if (copy_to_user(optval, &freq, len)) 1312 return -EFAULT; 1313 1314 return 0; 1315 } 1316 1317 case IPV6_ADDR_PREFERENCES: 1318 val = 0; 1319 1320 if (np->srcprefs & IPV6_PREFER_SRC_TMP) 1321 val |= IPV6_PREFER_SRC_TMP; 1322 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC) 1323 val |= IPV6_PREFER_SRC_PUBLIC; 1324 else { 1325 /* XXX: should we return system default? */ 1326 val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT; 1327 } 1328 1329 if (np->srcprefs & IPV6_PREFER_SRC_COA) 1330 val |= IPV6_PREFER_SRC_COA; 1331 else 1332 val |= IPV6_PREFER_SRC_HOME; 1333 break; 1334 1335 case IPV6_MINHOPCOUNT: 1336 val = np->min_hopcount; 1337 break; 1338 1339 case IPV6_DONTFRAG: 1340 val = np->dontfrag; 1341 break; 1342 1343 case IPV6_AUTOFLOWLABEL: 1344 val = ip6_autoflowlabel(sock_net(sk), np); 1345 break; 1346 1347 case IPV6_RECVFRAGSIZE: 1348 val = np->rxopt.bits.recvfragsize; 1349 break; 1350 1351 default: 1352 return -ENOPROTOOPT; 1353 } 1354 len = min_t(unsigned int, sizeof(int), len); 1355 if (put_user(len, optlen)) 1356 return -EFAULT; 1357 if (copy_to_user(optval, &val, len)) 1358 return -EFAULT; 1359 return 0; 1360 } 1361 1362 int ipv6_getsockopt(struct sock *sk, int level, int optname, 1363 char __user *optval, int __user *optlen) 1364 { 1365 int err; 1366 1367 if (level == SOL_IP && sk->sk_type != SOCK_RAW) 1368 return udp_prot.getsockopt(sk, level, optname, optval, optlen); 1369 1370 if (level != SOL_IPV6) 1371 return -ENOPROTOOPT; 1372 1373 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 0); 1374 #ifdef CONFIG_NETFILTER 1375 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1376 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) { 1377 int len; 1378 1379 if (get_user(len, optlen)) 1380 return -EFAULT; 1381 1382 err = nf_getsockopt(sk, PF_INET6, optname, optval, &len); 1383 if (err >= 0) 1384 err = put_user(len, optlen); 1385 } 1386 #endif 1387 return err; 1388 } 1389 EXPORT_SYMBOL(ipv6_getsockopt); 1390 1391 #ifdef CONFIG_COMPAT 1392 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname, 1393 char __user *optval, int __user *optlen) 1394 { 1395 int err; 1396 1397 if (level == SOL_IP && sk->sk_type != SOCK_RAW) { 1398 if (udp_prot.compat_getsockopt != NULL) 1399 return udp_prot.compat_getsockopt(sk, level, optname, 1400 optval, optlen); 1401 return udp_prot.getsockopt(sk, level, optname, optval, optlen); 1402 } 1403 1404 if (level != SOL_IPV6) 1405 return -ENOPROTOOPT; 1406 1407 if (optname == MCAST_MSFILTER) 1408 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1409 ipv6_getsockopt); 1410 1411 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 1412 MSG_CMSG_COMPAT); 1413 #ifdef CONFIG_NETFILTER 1414 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1415 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) { 1416 int len; 1417 1418 if (get_user(len, optlen)) 1419 return -EFAULT; 1420 1421 err = compat_nf_getsockopt(sk, PF_INET6, optname, optval, &len); 1422 if (err >= 0) 1423 err = put_user(len, optlen); 1424 } 1425 #endif 1426 return err; 1427 } 1428 EXPORT_SYMBOL(compat_ipv6_getsockopt); 1429 #endif 1430