net/ipv6/ipv6_sockglue.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  *      IPv6 BSD socket options interface
   4  *      Linux INET6 implementation
   5  *
   6  *      Authors:
   7  *      Pedro Roque             <roque@di.fc.ul.pt>
   8  *
   9  *      Based on linux/net/ipv4/ip_sockglue.c
  10  *
  11  *      FIXME: Make the setsockopt code POSIX compliant: That is
  12  *
  13  *      o       Truncate getsockopt returns
  14  *      o       Return an optlen of the truncated length if need be
  15  *
  16  *      Changes:
  17  *      David L Stevens <dlstevens@us.ibm.com>:
  18  *              - added multicast source filtering API for MLDv2
  19  */
  20
  21 #include <linux/module.h>
  22 #include <linux/capability.h>
  23 #include <linux/errno.h>
  24 #include <linux/types.h>
  25 #include <linux/socket.h>
  26 #include <linux/sockios.h>
  27 #include <linux/net.h>
  28 #include <linux/in6.h>
  29 #include <linux/mroute6.h>
  30 #include <linux/netdevice.h>
  31 #include <linux/if_arp.h>
  32 #include <linux/init.h>
  33 #include <linux/sysctl.h>
  34 #include <linux/netfilter.h>
  35 #include <linux/slab.h>
  36
  37 #include <net/sock.h>
  38 #include <net/snmp.h>
  39 #include <net/ipv6.h>
  40 #include <net/ndisc.h>
  41 #include <net/protocol.h>
  42 #include <net/transp_v6.h>
  43 #include <net/ip6_route.h>
  44 #include <net/addrconf.h>
  45 #include <net/inet_common.h>
  46 #include <net/tcp.h>
  47 #include <net/udp.h>
  48 #include <net/udplite.h>
  49 #include <net/xfrm.h>
  50 #include <net/compat.h>
  51 #include <net/seg6.h>
  52
  53 #include <linux/uaccess.h>
  54
  55 struct ip6_ra_chain *ip6_ra_chain;
  56 DEFINE_RWLOCK(ip6_ra_lock);
  57
  58 DEFINE_STATIC_KEY_FALSE(ip6_min_hopcount);
  59
  60 int ip6_ra_control(struct sock *sk, int sel)
  61 {
  62         struct ip6_ra_chain *ra, *new_ra, **rap;
  63
  64         /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
  65         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
  66                 return -ENOPROTOOPT;
  67
  68         new_ra = (sel >= 0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
  69         if (sel >= 0 && !new_ra)
  70                 return -ENOMEM;
  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 copy_group_source_from_sockptr(struct group_source_req *greqs,
 142                 sockptr_t optval, int optlen)
 143 {
 144         if (in_compat_syscall()) {
 145                 struct compat_group_source_req gr32;
 146
 147                 if (optlen < sizeof(gr32))
 148                         return -EINVAL;
 149                 if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
 150                         return -EFAULT;
 151                 greqs->gsr_interface = gr32.gsr_interface;
 152                 greqs->gsr_group = gr32.gsr_group;
 153                 greqs->gsr_source = gr32.gsr_source;
 154         } else {
 155                 if (optlen < sizeof(*greqs))
 156                         return -EINVAL;
 157                 if (copy_from_sockptr(greqs, optval, sizeof(*greqs)))
 158                         return -EFAULT;
 159         }
 160
 161         return 0;
 162 }
 163
 164 static int do_ipv6_mcast_group_source(struct sock *sk, int optname,
 165                 sockptr_t optval, int optlen)
 166 {
 167         struct group_source_req greqs;
 168         int omode, add;
 169         int ret;
 170
 171         ret = copy_group_source_from_sockptr(&greqs, optval, optlen);
 172         if (ret)
 173                 return ret;
 174
 175         if (greqs.gsr_group.ss_family != AF_INET6 ||
 176             greqs.gsr_source.ss_family != AF_INET6)
 177                 return -EADDRNOTAVAIL;
 178
 179         if (optname == MCAST_BLOCK_SOURCE) {
 180                 omode = MCAST_EXCLUDE;
 181                 add = 1;
 182         } else if (optname == MCAST_UNBLOCK_SOURCE) {
 183                 omode = MCAST_EXCLUDE;
 184                 add = 0;
 185         } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 186                 struct sockaddr_in6 *psin6;
 187                 int retv;
 188
 189                 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
 190                 retv = ipv6_sock_mc_join_ssm(sk, greqs.gsr_interface,
 191                                              &psin6->sin6_addr,
 192                                              MCAST_INCLUDE);
 193                 /* prior join w/ different source is ok */
 194                 if (retv && retv != -EADDRINUSE)
 195                         return retv;
 196                 omode = MCAST_INCLUDE;
 197                 add = 1;
 198         } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 199                 omode = MCAST_INCLUDE;
 200                 add = 0;
 201         }
 202         return ip6_mc_source(add, omode, sk, &greqs);
 203 }
 204
 205 static int ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
 206                 int optlen)
 207 {
 208         struct group_filter *gsf;
 209         int ret;
 210
 211         if (optlen < GROUP_FILTER_SIZE(0))
 212                 return -EINVAL;
 213         if (optlen > READ_ONCE(sysctl_optmem_max))
 214                 return -ENOBUFS;
 215
 216         gsf = memdup_sockptr(optval, optlen);
 217         if (IS_ERR(gsf))
 218                 return PTR_ERR(gsf);
 219
 220         /* numsrc >= (4G-140)/128 overflow in 32 bits */
 221         ret = -ENOBUFS;
 222         if (gsf->gf_numsrc >= 0x1ffffffU ||
 223             gsf->gf_numsrc > sysctl_mld_max_msf)
 224                 goto out_free_gsf;
 225
 226         ret = -EINVAL;
 227         if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen)
 228                 goto out_free_gsf;
 229
 230         ret = ip6_mc_msfilter(sk, gsf, gsf->gf_slist_flex);
 231 out_free_gsf:
 232         kfree(gsf);
 233         return ret;
 234 }
 235
 236 static int compat_ipv6_set_mcast_msfilter(struct sock *sk, sockptr_t optval,
 237                 int optlen)
 238 {
 239         const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
 240         struct compat_group_filter *gf32;
 241         void *p;
 242         int ret;
 243         int n;
 244
 245         if (optlen < size0)
 246                 return -EINVAL;
 247         if (optlen > READ_ONCE(sysctl_optmem_max) - 4)
 248                 return -ENOBUFS;
 249
 250         p = kmalloc(optlen + 4, GFP_KERNEL);
 251         if (!p)
 252                 return -ENOMEM;
 253
 254         gf32 = p + 4; /* we want ->gf_group and ->gf_slist_flex aligned */
 255         ret = -EFAULT;
 256         if (copy_from_sockptr(gf32, optval, optlen))
 257                 goto out_free_p;
 258
 259         /* numsrc >= (4G-140)/128 overflow in 32 bits */
 260         ret = -ENOBUFS;
 261         n = gf32->gf_numsrc;
 262         if (n >= 0x1ffffffU || n > sysctl_mld_max_msf)
 263                 goto out_free_p;
 264
 265         ret = -EINVAL;
 266         if (offsetof(struct compat_group_filter, gf_slist_flex[n]) > optlen)
 267                 goto out_free_p;
 268
 269         ret = ip6_mc_msfilter(sk, &(struct group_filter){
 270                         .gf_interface = gf32->gf_interface,
 271                         .gf_group = gf32->gf_group,
 272                         .gf_fmode = gf32->gf_fmode,
 273                         .gf_numsrc = gf32->gf_numsrc}, gf32->gf_slist_flex);
 274
 275 out_free_p:
 276         kfree(p);
 277         return ret;
 278 }
 279
 280 static int ipv6_mcast_join_leave(struct sock *sk, int optname,
 281                 sockptr_t optval, int optlen)
 282 {
 283         struct sockaddr_in6 *psin6;
 284         struct group_req greq;
 285
 286         if (optlen < sizeof(greq))
 287                 return -EINVAL;
 288         if (copy_from_sockptr(&greq, optval, sizeof(greq)))
 289                 return -EFAULT;
 290
 291         if (greq.gr_group.ss_family != AF_INET6)
 292                 return -EADDRNOTAVAIL;
 293         psin6 = (struct sockaddr_in6 *)&greq.gr_group;
 294         if (optname == MCAST_JOIN_GROUP)
 295                 return ipv6_sock_mc_join(sk, greq.gr_interface,
 296                                          &psin6->sin6_addr);
 297         return ipv6_sock_mc_drop(sk, greq.gr_interface, &psin6->sin6_addr);
 298 }
 299
 300 static int compat_ipv6_mcast_join_leave(struct sock *sk, int optname,
 301                 sockptr_t optval, int optlen)
 302 {
 303         struct compat_group_req gr32;
 304         struct sockaddr_in6 *psin6;
 305
 306         if (optlen < sizeof(gr32))
 307                 return -EINVAL;
 308         if (copy_from_sockptr(&gr32, optval, sizeof(gr32)))
 309                 return -EFAULT;
 310
 311         if (gr32.gr_group.ss_family != AF_INET6)
 312                 return -EADDRNOTAVAIL;
 313         psin6 = (struct sockaddr_in6 *)&gr32.gr_group;
 314         if (optname == MCAST_JOIN_GROUP)
 315                 return ipv6_sock_mc_join(sk, gr32.gr_interface,
 316                                         &psin6->sin6_addr);
 317         return ipv6_sock_mc_drop(sk, gr32.gr_interface, &psin6->sin6_addr);
 318 }
 319
 320 static int ipv6_set_opt_hdr(struct sock *sk, int optname, sockptr_t optval,
 321                 int optlen)
 322 {
 323         struct ipv6_pinfo *np = inet6_sk(sk);
 324         struct ipv6_opt_hdr *new = NULL;
 325         struct net *net = sock_net(sk);
 326         struct ipv6_txoptions *opt;
 327         int err;
 328
 329         /* hop-by-hop / destination options are privileged option */
 330         if (optname != IPV6_RTHDR && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW))
 331                 return -EPERM;
 332
 333         /* remove any sticky options header with a zero option
 334          * length, per RFC3542.
 335          */
 336         if (optlen > 0) {
 337                 if (sockptr_is_null(optval))
 338                         return -EINVAL;
 339                 if (optlen < sizeof(struct ipv6_opt_hdr) ||
 340                     optlen & 0x7 ||
 341                     optlen > 8 * 255)
 342                         return -EINVAL;
 343
 344                 new = memdup_sockptr(optval, optlen);
 345                 if (IS_ERR(new))
 346                         return PTR_ERR(new);
 347                 if (unlikely(ipv6_optlen(new) > optlen)) {
 348                         kfree(new);
 349                         return -EINVAL;
 350                 }
 351         }
 352
 353         opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
 354         opt = ipv6_renew_options(sk, opt, optname, new);
 355         kfree(new);
 356         if (IS_ERR(opt))
 357                 return PTR_ERR(opt);
 358
 359         /* routing header option needs extra check */
 360         err = -EINVAL;
 361         if (optname == IPV6_RTHDR && opt && opt->srcrt) {
 362                 struct ipv6_rt_hdr *rthdr = opt->srcrt;
 363                 switch (rthdr->type) {
 364 #if IS_ENABLED(CONFIG_IPV6_MIP6)
 365                 case IPV6_SRCRT_TYPE_2:
 366                         if (rthdr->hdrlen != 2 || rthdr->segments_left != 1)
 367                                 goto sticky_done;
 368                         break;
 369 #endif
 370                 case IPV6_SRCRT_TYPE_4:
 371                 {
 372                         struct ipv6_sr_hdr *srh =
 373                                 (struct ipv6_sr_hdr *)opt->srcrt;
 374
 375                         if (!seg6_validate_srh(srh, optlen, false))
 376                                 goto sticky_done;
 377                         break;
 378                 }
 379                 default:
 380                         goto sticky_done;
 381                 }
 382         }
 383
 384         err = 0;
 385         opt = ipv6_update_options(sk, opt);
 386 sticky_done:
 387         if (opt) {
 388                 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 389                 txopt_put(opt);
 390         }
 391         return err;
 392 }
 393
 394 int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
 395                        sockptr_t optval, unsigned int optlen)
 396 {
 397         struct ipv6_pinfo *np = inet6_sk(sk);
 398         struct net *net = sock_net(sk);
 399         int val, valbool;
 400         int retv = -ENOPROTOOPT;
 401         bool needs_rtnl = setsockopt_needs_rtnl(optname);
 402
 403         if (sockptr_is_null(optval))
 404                 val = 0;
 405         else {
 406                 if (optlen >= sizeof(int)) {
 407                         if (copy_from_sockptr(&val, optval, sizeof(val)))
 408                                 return -EFAULT;
 409                 } else
 410                         val = 0;
 411         }
 412
 413         valbool = (val != 0);
 414
 415         if (ip6_mroute_opt(optname))
 416                 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
 417
 418         if (needs_rtnl)
 419                 rtnl_lock();
 420         sockopt_lock_sock(sk);
 421
 422         /* Another thread has converted the socket into IPv4 with
 423          * IPV6_ADDRFORM concurrently.
 424          */
 425         if (unlikely(sk->sk_family != AF_INET6))
 426                 goto unlock;
 427
 428         switch (optname) {
 429
 430         case IPV6_ADDRFORM:
 431                 if (optlen < sizeof(int))
 432                         goto e_inval;
 433                 if (val == PF_INET) {
 434                         if (sk->sk_type == SOCK_RAW)
 435                                 break;
 436
 437                         if (sk->sk_protocol == IPPROTO_UDP ||
 438                             sk->sk_protocol == IPPROTO_UDPLITE) {
 439                                 struct udp_sock *up = udp_sk(sk);
 440                                 if (up->pending == AF_INET6) {
 441                                         retv = -EBUSY;
 442                                         break;
 443                                 }
 444                         } else if (sk->sk_protocol == IPPROTO_TCP) {
 445                                 if (sk->sk_prot != &tcpv6_prot) {
 446                                         retv = -EBUSY;
 447                                         break;
 448                                 }
 449                         } else {
 450                                 break;
 451                         }
 452
 453                         if (sk->sk_state != TCP_ESTABLISHED) {
 454                                 retv = -ENOTCONN;
 455                                 break;
 456                         }
 457
 458                         if (ipv6_only_sock(sk) ||
 459                             !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
 460                                 retv = -EADDRNOTAVAIL;
 461                                 break;
 462                         }
 463
 464                         __ipv6_sock_mc_close(sk);
 465                         __ipv6_sock_ac_close(sk);
 466
 467                         /*
 468                          * Sock is moving from IPv6 to IPv4 (sk_prot), so
 469                          * remove it from the refcnt debug socks count in the
 470                          * original family...
 471                          */
 472                         sk_refcnt_debug_dec(sk);
 473
 474                         if (sk->sk_protocol == IPPROTO_TCP) {
 475                                 struct inet_connection_sock *icsk = inet_csk(sk);
 476
 477                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 478                                 sock_prot_inuse_add(net, &tcp_prot, 1);
 479
 480                                 /* Paired with READ_ONCE(sk->sk_prot) in inet6_stream_ops */
 481                                 WRITE_ONCE(sk->sk_prot, &tcp_prot);
 482                                 icsk->icsk_af_ops = &ipv4_specific;
 483                                 sk->sk_socket->ops = &inet_stream_ops;
 484                                 sk->sk_family = PF_INET;
 485                                 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
 486                         } else {
 487                                 struct proto *prot = &udp_prot;
 488
 489                                 if (sk->sk_protocol == IPPROTO_UDPLITE)
 490                                         prot = &udplite_prot;
 491
 492                                 sock_prot_inuse_add(net, sk->sk_prot, -1);
 493                                 sock_prot_inuse_add(net, prot, 1);
 494
 495                                 /* Paired with READ_ONCE(sk->sk_prot) in inet6_dgram_ops */
 496                                 WRITE_ONCE(sk->sk_prot, prot);
 497                                 sk->sk_socket->ops = &inet_dgram_ops;
 498                                 sk->sk_family = PF_INET;
 499                         }
 500
 501                         /* Disable all options not to allocate memory anymore,
 502                          * but there is still a race.  See the lockless path
 503                          * in udpv6_sendmsg() and ipv6_local_rxpmtu().
 504                          */
 505                         np->rxopt.all = 0;
 506
 507                         inet6_cleanup_sock(sk);
 508
 509                         /*
 510                          * ... and add it to the refcnt debug socks count
 511                          * in the new family. -acme
 512                          */
 513                         sk_refcnt_debug_inc(sk);
 514                         module_put(THIS_MODULE);
 515                         retv = 0;
 516                         break;
 517                 }
 518                 goto e_inval;
 519
 520         case IPV6_V6ONLY:
 521                 if (optlen < sizeof(int) ||
 522                     inet_sk(sk)->inet_num)
 523                         goto e_inval;
 524                 sk->sk_ipv6only = valbool;
 525                 retv = 0;
 526                 break;
 527
 528         case IPV6_RECVPKTINFO:
 529                 if (optlen < sizeof(int))
 530                         goto e_inval;
 531                 np->rxopt.bits.rxinfo = valbool;
 532                 retv = 0;
 533                 break;
 534
 535         case IPV6_2292PKTINFO:
 536                 if (optlen < sizeof(int))
 537                         goto e_inval;
 538                 np->rxopt.bits.rxoinfo = valbool;
 539                 retv = 0;
 540                 break;
 541
 542         case IPV6_RECVHOPLIMIT:
 543                 if (optlen < sizeof(int))
 544                         goto e_inval;
 545                 np->rxopt.bits.rxhlim = valbool;
 546                 retv = 0;
 547                 break;
 548
 549         case IPV6_2292HOPLIMIT:
 550                 if (optlen < sizeof(int))
 551                         goto e_inval;
 552                 np->rxopt.bits.rxohlim = valbool;
 553                 retv = 0;
 554                 break;
 555
 556         case IPV6_RECVRTHDR:
 557                 if (optlen < sizeof(int))
 558                         goto e_inval;
 559                 np->rxopt.bits.srcrt = valbool;
 560                 retv = 0;
 561                 break;
 562
 563         case IPV6_2292RTHDR:
 564                 if (optlen < sizeof(int))
 565                         goto e_inval;
 566                 np->rxopt.bits.osrcrt = valbool;
 567                 retv = 0;
 568                 break;
 569
 570         case IPV6_RECVHOPOPTS:
 571                 if (optlen < sizeof(int))
 572                         goto e_inval;
 573                 np->rxopt.bits.hopopts = valbool;
 574                 retv = 0;
 575                 break;
 576
 577         case IPV6_2292HOPOPTS:
 578                 if (optlen < sizeof(int))
 579                         goto e_inval;
 580                 np->rxopt.bits.ohopopts = valbool;
 581                 retv = 0;
 582                 break;
 583
 584         case IPV6_RECVDSTOPTS:
 585                 if (optlen < sizeof(int))
 586                         goto e_inval;
 587                 np->rxopt.bits.dstopts = valbool;
 588                 retv = 0;
 589                 break;
 590
 591         case IPV6_2292DSTOPTS:
 592                 if (optlen < sizeof(int))
 593                         goto e_inval;
 594                 np->rxopt.bits.odstopts = valbool;
 595                 retv = 0;
 596                 break;
 597
 598         case IPV6_TCLASS:
 599                 if (optlen < sizeof(int))
 600                         goto e_inval;
 601                 if (val < -1 || val > 0xff)
 602                         goto e_inval;
 603                 /* RFC 3542, 6.5: default traffic class of 0x0 */
 604                 if (val == -1)
 605                         val = 0;
 606                 if (sk->sk_type == SOCK_STREAM) {
 607                         val &= ~INET_ECN_MASK;
 608                         val |= np->tclass & INET_ECN_MASK;
 609                 }
 610                 if (np->tclass != val) {
 611                         np->tclass = val;
 612                         sk_dst_reset(sk);
 613                 }
 614                 retv = 0;
 615                 break;
 616
 617         case IPV6_RECVTCLASS:
 618                 if (optlen < sizeof(int))
 619                         goto e_inval;
 620                 np->rxopt.bits.rxtclass = valbool;
 621                 retv = 0;
 622                 break;
 623
 624         case IPV6_FLOWINFO:
 625                 if (optlen < sizeof(int))
 626                         goto e_inval;
 627                 np->rxopt.bits.rxflow = valbool;
 628                 retv = 0;
 629                 break;
 630
 631         case IPV6_RECVPATHMTU:
 632                 if (optlen < sizeof(int))
 633                         goto e_inval;
 634                 np->rxopt.bits.rxpmtu = valbool;
 635                 retv = 0;
 636                 break;
 637
 638         case IPV6_TRANSPARENT:
 639                 if (valbool && !sockopt_ns_capable(net->user_ns, CAP_NET_RAW) &&
 640                     !sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN)) {
 641                         retv = -EPERM;
 642                         break;
 643                 }
 644                 if (optlen < sizeof(int))
 645                         goto e_inval;
 646                 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
 647                 inet_sk(sk)->transparent = valbool;
 648                 retv = 0;
 649                 break;
 650
 651         case IPV6_FREEBIND:
 652                 if (optlen < sizeof(int))
 653                         goto e_inval;
 654                 /* we also don't have a separate freebind bit for IPV6 */
 655                 inet_sk(sk)->freebind = valbool;
 656                 retv = 0;
 657                 break;
 658
 659         case IPV6_RECVORIGDSTADDR:
 660                 if (optlen < sizeof(int))
 661                         goto e_inval;
 662                 np->rxopt.bits.rxorigdstaddr = valbool;
 663                 retv = 0;
 664                 break;
 665
 666         case IPV6_HOPOPTS:
 667         case IPV6_RTHDRDSTOPTS:
 668         case IPV6_RTHDR:
 669         case IPV6_DSTOPTS:
 670                 retv = ipv6_set_opt_hdr(sk, optname, optval, optlen);
 671                 break;
 672
 673         case IPV6_PKTINFO:
 674         {
 675                 struct in6_pktinfo pkt;
 676
 677                 if (optlen == 0)
 678                         goto e_inval;
 679                 else if (optlen < sizeof(struct in6_pktinfo) ||
 680                          sockptr_is_null(optval))
 681                         goto e_inval;
 682
 683                 if (copy_from_sockptr(&pkt, optval, sizeof(pkt))) {
 684                         retv = -EFAULT;
 685                         break;
 686                 }
 687                 if (!sk_dev_equal_l3scope(sk, pkt.ipi6_ifindex))
 688                         goto e_inval;
 689
 690                 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
 691                 np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
 692                 retv = 0;
 693                 break;
 694         }
 695
 696         case IPV6_2292PKTOPTIONS:
 697         {
 698                 struct ipv6_txoptions *opt = NULL;
 699                 struct msghdr msg;
 700                 struct flowi6 fl6;
 701                 struct ipcm6_cookie ipc6;
 702
 703                 memset(&fl6, 0, sizeof(fl6));
 704                 fl6.flowi6_oif = sk->sk_bound_dev_if;
 705                 fl6.flowi6_mark = sk->sk_mark;
 706
 707                 if (optlen == 0)
 708                         goto update;
 709
 710                 /* 1K is probably excessive
 711                  * 1K is surely not enough, 2K per standard header is 16K.
 712                  */
 713                 retv = -EINVAL;
 714                 if (optlen > 64*1024)
 715                         break;
 716
 717                 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
 718                 retv = -ENOBUFS;
 719                 if (!opt)
 720                         break;
 721
 722                 memset(opt, 0, sizeof(*opt));
 723                 refcount_set(&opt->refcnt, 1);
 724                 opt->tot_len = sizeof(*opt) + optlen;
 725                 retv = -EFAULT;
 726                 if (copy_from_sockptr(opt + 1, optval, optlen))
 727                         goto done;
 728
 729                 msg.msg_controllen = optlen;
 730                 msg.msg_control = (void *)(opt+1);
 731                 ipc6.opt = opt;
 732
 733                 retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6);
 734                 if (retv)
 735                         goto done;
 736 update:
 737                 retv = 0;
 738                 opt = ipv6_update_options(sk, opt);
 739 done:
 740                 if (opt) {
 741                         atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
 742                         txopt_put(opt);
 743                 }
 744                 break;
 745         }
 746         case IPV6_UNICAST_HOPS:
 747                 if (optlen < sizeof(int))
 748                         goto e_inval;
 749                 if (val > 255 || val < -1)
 750                         goto e_inval;
 751                 np->hop_limit = val;
 752                 retv = 0;
 753                 break;
 754
 755         case IPV6_MULTICAST_HOPS:
 756                 if (sk->sk_type == SOCK_STREAM)
 757                         break;
 758                 if (optlen < sizeof(int))
 759                         goto e_inval;
 760                 if (val > 255 || val < -1)
 761                         goto e_inval;
 762                 np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
 763                 retv = 0;
 764                 break;
 765
 766         case IPV6_MULTICAST_LOOP:
 767                 if (optlen < sizeof(int))
 768                         goto e_inval;
 769                 if (val != valbool)
 770                         goto e_inval;
 771                 np->mc_loop = valbool;
 772                 retv = 0;
 773                 break;
 774
 775         case IPV6_UNICAST_IF:
 776         {
 777                 struct net_device *dev = NULL;
 778                 int ifindex;
 779
 780                 if (optlen != sizeof(int))
 781                         goto e_inval;
 782
 783                 ifindex = (__force int)ntohl((__force __be32)val);
 784                 if (ifindex == 0) {
 785                         np->ucast_oif = 0;
 786                         retv = 0;
 787                         break;
 788                 }
 789
 790                 dev = dev_get_by_index(net, ifindex);
 791                 retv = -EADDRNOTAVAIL;
 792                 if (!dev)
 793                         break;
 794                 dev_put(dev);
 795
 796                 retv = -EINVAL;
 797                 if (sk->sk_bound_dev_if)
 798                         break;
 799
 800                 np->ucast_oif = ifindex;
 801                 retv = 0;
 802                 break;
 803         }
 804
 805         case IPV6_MULTICAST_IF:
 806                 if (sk->sk_type == SOCK_STREAM)
 807                         break;
 808                 if (optlen < sizeof(int))
 809                         goto e_inval;
 810
 811                 if (val) {
 812                         struct net_device *dev;
 813                         int midx;
 814
 815                         rcu_read_lock();
 816
 817                         dev = dev_get_by_index_rcu(net, val);
 818                         if (!dev) {
 819                                 rcu_read_unlock();
 820                                 retv = -ENODEV;
 821                                 break;
 822                         }
 823                         midx = l3mdev_master_ifindex_rcu(dev);
 824
 825                         rcu_read_unlock();
 826
 827                         if (sk->sk_bound_dev_if &&
 828                             sk->sk_bound_dev_if != val &&
 829                             (!midx || midx != sk->sk_bound_dev_if))
 830                                 goto e_inval;
 831                 }
 832                 np->mcast_oif = val;
 833                 retv = 0;
 834                 break;
 835         case IPV6_ADD_MEMBERSHIP:
 836         case IPV6_DROP_MEMBERSHIP:
 837         {
 838                 struct ipv6_mreq mreq;
 839
 840                 if (optlen < sizeof(struct ipv6_mreq))
 841                         goto e_inval;
 842
 843                 retv = -EPROTO;
 844                 if (inet_sk(sk)->is_icsk)
 845                         break;
 846
 847                 retv = -EFAULT;
 848                 if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
 849                         break;
 850
 851                 if (optname == IPV6_ADD_MEMBERSHIP)
 852                         retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 853                 else
 854                         retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
 855                 break;
 856         }
 857         case IPV6_JOIN_ANYCAST:
 858         case IPV6_LEAVE_ANYCAST:
 859         {
 860                 struct ipv6_mreq mreq;
 861
 862                 if (optlen < sizeof(struct ipv6_mreq))
 863                         goto e_inval;
 864
 865                 retv = -EFAULT;
 866                 if (copy_from_sockptr(&mreq, optval, sizeof(struct ipv6_mreq)))
 867                         break;
 868
 869                 if (optname == IPV6_JOIN_ANYCAST)
 870                         retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 871                 else
 872                         retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
 873                 break;
 874         }
 875         case IPV6_MULTICAST_ALL:
 876                 if (optlen < sizeof(int))
 877                         goto e_inval;
 878                 np->mc_all = valbool;
 879                 retv = 0;
 880                 break;
 881
 882         case MCAST_JOIN_GROUP:
 883         case MCAST_LEAVE_GROUP:
 884                 if (in_compat_syscall())
 885                         retv = compat_ipv6_mcast_join_leave(sk, optname, optval,
 886                                                             optlen);
 887                 else
 888                         retv = ipv6_mcast_join_leave(sk, optname, optval,
 889                                                      optlen);
 890                 break;
 891         case MCAST_JOIN_SOURCE_GROUP:
 892         case MCAST_LEAVE_SOURCE_GROUP:
 893         case MCAST_BLOCK_SOURCE:
 894         case MCAST_UNBLOCK_SOURCE:
 895                 retv = do_ipv6_mcast_group_source(sk, optname, optval, optlen);
 896                 break;
 897         case MCAST_MSFILTER:
 898                 if (in_compat_syscall())
 899                         retv = compat_ipv6_set_mcast_msfilter(sk, optval,
 900                                                               optlen);
 901                 else
 902                         retv = ipv6_set_mcast_msfilter(sk, optval, optlen);
 903                 break;
 904         case IPV6_ROUTER_ALERT:
 905                 if (optlen < sizeof(int))
 906                         goto e_inval;
 907                 retv = ip6_ra_control(sk, val);
 908                 break;
 909         case IPV6_ROUTER_ALERT_ISOLATE:
 910                 if (optlen < sizeof(int))
 911                         goto e_inval;
 912                 np->rtalert_isolate = valbool;
 913                 retv = 0;
 914                 break;
 915         case IPV6_MTU_DISCOVER:
 916                 if (optlen < sizeof(int))
 917                         goto e_inval;
 918                 if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
 919                         goto e_inval;
 920                 np->pmtudisc = val;
 921                 retv = 0;
 922                 break;
 923         case IPV6_MTU:
 924                 if (optlen < sizeof(int))
 925                         goto e_inval;
 926                 if (val && val < IPV6_MIN_MTU)
 927                         goto e_inval;
 928                 np->frag_size = val;
 929                 retv = 0;
 930                 break;
 931         case IPV6_RECVERR:
 932                 if (optlen < sizeof(int))
 933                         goto e_inval;
 934                 np->recverr = valbool;
 935                 if (!val)
 936                         skb_queue_purge(&sk->sk_error_queue);
 937                 retv = 0;
 938                 break;
 939         case IPV6_FLOWINFO_SEND:
 940                 if (optlen < sizeof(int))
 941                         goto e_inval;
 942                 np->sndflow = valbool;
 943                 retv = 0;
 944                 break;
 945         case IPV6_FLOWLABEL_MGR:
 946                 retv = ipv6_flowlabel_opt(sk, optval, optlen);
 947                 break;
 948         case IPV6_IPSEC_POLICY:
 949         case IPV6_XFRM_POLICY:
 950                 retv = -EPERM;
 951                 if (!sockopt_ns_capable(net->user_ns, CAP_NET_ADMIN))
 952                         break;
 953                 retv = xfrm_user_policy(sk, optname, optval, optlen);
 954                 break;
 955
 956         case IPV6_ADDR_PREFERENCES:
 957                 if (optlen < sizeof(int))
 958                         goto e_inval;
 959                 retv = __ip6_sock_set_addr_preferences(sk, val);
 960                 break;
 961         case IPV6_MINHOPCOUNT:
 962                 if (optlen < sizeof(int))
 963                         goto e_inval;
 964                 if (val < 0 || val > 255)
 965                         goto e_inval;
 966
 967                 if (val)
 968                         static_branch_enable(&ip6_min_hopcount);
 969
 970                 /* tcp_v6_err() and tcp_v6_rcv() might read min_hopcount
 971                  * while we are changing it.
 972                  */
 973                 WRITE_ONCE(np->min_hopcount, val);
 974                 retv = 0;
 975                 break;
 976         case IPV6_DONTFRAG:
 977                 np->dontfrag = valbool;
 978                 retv = 0;
 979                 break;
 980         case IPV6_AUTOFLOWLABEL:
 981                 np->autoflowlabel = valbool;
 982                 np->autoflowlabel_set = 1;
 983                 retv = 0;
 984                 break;
 985         case IPV6_RECVFRAGSIZE:
 986                 np->rxopt.bits.recvfragsize = valbool;
 987                 retv = 0;
 988                 break;
 989         case IPV6_RECVERR_RFC4884:
 990                 if (optlen < sizeof(int))
 991                         goto e_inval;
 992                 if (val < 0 || val > 1)
 993                         goto e_inval;
 994                 np->recverr_rfc4884 = valbool;
 995                 retv = 0;
 996                 break;
 997         }
 998
 999 unlock:
1000         sockopt_release_sock(sk);
1001         if (needs_rtnl)
1002                 rtnl_unlock();
1003
1004         return retv;
1005
1006 e_inval:
1007         sockopt_release_sock(sk);
1008         if (needs_rtnl)
1009                 rtnl_unlock();
1010         return -EINVAL;
1011 }
1012
1013 int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1014                     unsigned int optlen)
1015 {
1016         int err;
1017
1018         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1019                 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
1020
1021         if (level != SOL_IPV6)
1022                 return -ENOPROTOOPT;
1023
1024         err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
1025 #ifdef CONFIG_NETFILTER
1026         /* we need to exclude all possible ENOPROTOOPTs except default case */
1027         if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
1028                         optname != IPV6_XFRM_POLICY)
1029                 err = nf_setsockopt(sk, PF_INET6, optname, optval, optlen);
1030 #endif
1031         return err;
1032 }
1033 EXPORT_SYMBOL(ipv6_setsockopt);
1034
1035 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
1036                                   int optname, sockptr_t optval, int len)
1037 {
1038         struct ipv6_opt_hdr *hdr;
1039
1040         if (!opt)
1041                 return 0;
1042
1043         switch (optname) {
1044         case IPV6_HOPOPTS:
1045                 hdr = opt->hopopt;
1046                 break;
1047         case IPV6_RTHDRDSTOPTS:
1048                 hdr = opt->dst0opt;
1049                 break;
1050         case IPV6_RTHDR:
1051                 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
1052                 break;
1053         case IPV6_DSTOPTS:
1054                 hdr = opt->dst1opt;
1055                 break;
1056         default:
1057                 return -EINVAL; /* should not happen */
1058         }
1059
1060         if (!hdr)
1061                 return 0;
1062
1063         len = min_t(unsigned int, len, ipv6_optlen(hdr));
1064         if (copy_to_sockptr(optval, hdr, len))
1065                 return -EFAULT;
1066         return len;
1067 }
1068
1069 static int ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1070                              sockptr_t optlen, int len)
1071 {
1072         const int size0 = offsetof(struct group_filter, gf_slist_flex);
1073         struct group_filter gsf;
1074         int num;
1075         int err;
1076
1077         if (len < size0)
1078                 return -EINVAL;
1079         if (copy_from_sockptr(&gsf, optval, size0))
1080                 return -EFAULT;
1081         if (gsf.gf_group.ss_family != AF_INET6)
1082                 return -EADDRNOTAVAIL;
1083         num = gsf.gf_numsrc;
1084         sockopt_lock_sock(sk);
1085         err = ip6_mc_msfget(sk, &gsf, optval, size0);
1086         if (!err) {
1087                 if (num > gsf.gf_numsrc)
1088                         num = gsf.gf_numsrc;
1089                 len = GROUP_FILTER_SIZE(num);
1090                 if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1091                     copy_to_sockptr(optval, &gsf, size0))
1092                         err = -EFAULT;
1093         }
1094         sockopt_release_sock(sk);
1095         return err;
1096 }
1097
1098 static int compat_ipv6_get_msfilter(struct sock *sk, sockptr_t optval,
1099                                     sockptr_t optlen, int len)
1100 {
1101         const int size0 = offsetof(struct compat_group_filter, gf_slist_flex);
1102         struct compat_group_filter gf32;
1103         struct group_filter gf;
1104         int err;
1105         int num;
1106
1107         if (len < size0)
1108                 return -EINVAL;
1109
1110         if (copy_from_sockptr(&gf32, optval, size0))
1111                 return -EFAULT;
1112         gf.gf_interface = gf32.gf_interface;
1113         gf.gf_fmode = gf32.gf_fmode;
1114         num = gf.gf_numsrc = gf32.gf_numsrc;
1115         gf.gf_group = gf32.gf_group;
1116
1117         if (gf.gf_group.ss_family != AF_INET6)
1118                 return -EADDRNOTAVAIL;
1119
1120         sockopt_lock_sock(sk);
1121         err = ip6_mc_msfget(sk, &gf, optval, size0);
1122         sockopt_release_sock(sk);
1123         if (err)
1124                 return err;
1125         if (num > gf.gf_numsrc)
1126                 num = gf.gf_numsrc;
1127         len = GROUP_FILTER_SIZE(num) - (sizeof(gf)-sizeof(gf32));
1128         if (copy_to_sockptr(optlen, &len, sizeof(int)) ||
1129             copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_fmode),
1130                                    &gf.gf_fmode, sizeof(gf32.gf_fmode)) ||
1131             copy_to_sockptr_offset(optval, offsetof(struct compat_group_filter, gf_numsrc),
1132                                    &gf.gf_numsrc, sizeof(gf32.gf_numsrc)))
1133                 return -EFAULT;
1134         return 0;
1135 }
1136
1137 int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1138                        sockptr_t optval, sockptr_t optlen)
1139 {
1140         struct ipv6_pinfo *np = inet6_sk(sk);
1141         int len;
1142         int val;
1143
1144         if (ip6_mroute_opt(optname))
1145                 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1146
1147         if (copy_from_sockptr(&len, optlen, sizeof(int)))
1148                 return -EFAULT;
1149         switch (optname) {
1150         case IPV6_ADDRFORM:
1151                 if (sk->sk_protocol != IPPROTO_UDP &&
1152                     sk->sk_protocol != IPPROTO_UDPLITE &&
1153                     sk->sk_protocol != IPPROTO_TCP)
1154                         return -ENOPROTOOPT;
1155                 if (sk->sk_state != TCP_ESTABLISHED)
1156                         return -ENOTCONN;
1157                 val = sk->sk_family;
1158                 break;
1159         case MCAST_MSFILTER:
1160                 if (in_compat_syscall())
1161                         return compat_ipv6_get_msfilter(sk, optval, optlen, len);
1162                 return ipv6_get_msfilter(sk, optval, optlen, len);
1163         case IPV6_2292PKTOPTIONS:
1164         {
1165                 struct msghdr msg;
1166                 struct sk_buff *skb;
1167
1168                 if (sk->sk_type != SOCK_STREAM)
1169                         return -ENOPROTOOPT;
1170
1171                 if (optval.is_kernel) {
1172                         msg.msg_control_is_user = false;
1173                         msg.msg_control = optval.kernel;
1174                 } else {
1175                         msg.msg_control_is_user = true;
1176                         msg.msg_control_user = optval.user;
1177                 }
1178                 msg.msg_controllen = len;
1179                 msg.msg_flags = 0;
1180
1181                 sockopt_lock_sock(sk);
1182                 skb = np->pktoptions;
1183                 if (skb)
1184                         ip6_datagram_recv_ctl(sk, &msg, skb);
1185                 sockopt_release_sock(sk);
1186                 if (!skb) {
1187                         if (np->rxopt.bits.rxinfo) {
1188                                 struct in6_pktinfo src_info;
1189                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1190                                         np->sticky_pktinfo.ipi6_ifindex;
1191                                 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1192                                 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1193                         }
1194                         if (np->rxopt.bits.rxhlim) {
1195                                 int hlim = np->mcast_hops;
1196                                 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1197                         }
1198                         if (np->rxopt.bits.rxtclass) {
1199                                 int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1200
1201                                 put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1202                         }
1203                         if (np->rxopt.bits.rxoinfo) {
1204                                 struct in6_pktinfo src_info;
1205                                 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1206                                         np->sticky_pktinfo.ipi6_ifindex;
1207                                 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1208                                                                      np->sticky_pktinfo.ipi6_addr;
1209                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1210                         }
1211                         if (np->rxopt.bits.rxohlim) {
1212                                 int hlim = np->mcast_hops;
1213                                 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1214                         }
1215                         if (np->rxopt.bits.rxflow) {
1216                                 __be32 flowinfo = np->rcv_flowinfo;
1217
1218                                 put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1219                         }
1220                 }
1221                 len -= msg.msg_controllen;
1222                 return copy_to_sockptr(optlen, &len, sizeof(int));
1223         }
1224         case IPV6_MTU:
1225         {
1226                 struct dst_entry *dst;
1227
1228                 val = 0;
1229                 rcu_read_lock();
1230                 dst = __sk_dst_get(sk);
1231                 if (dst)
1232                         val = dst_mtu(dst);
1233                 rcu_read_unlock();
1234                 if (!val)
1235                         return -ENOTCONN;
1236                 break;
1237         }
1238
1239         case IPV6_V6ONLY:
1240                 val = sk->sk_ipv6only;
1241                 break;
1242
1243         case IPV6_RECVPKTINFO:
1244                 val = np->rxopt.bits.rxinfo;
1245                 break;
1246
1247         case IPV6_2292PKTINFO:
1248                 val = np->rxopt.bits.rxoinfo;
1249                 break;
1250
1251         case IPV6_RECVHOPLIMIT:
1252                 val = np->rxopt.bits.rxhlim;
1253                 break;
1254
1255         case IPV6_2292HOPLIMIT:
1256                 val = np->rxopt.bits.rxohlim;
1257                 break;
1258
1259         case IPV6_RECVRTHDR:
1260                 val = np->rxopt.bits.srcrt;
1261                 break;
1262
1263         case IPV6_2292RTHDR:
1264                 val = np->rxopt.bits.osrcrt;
1265                 break;
1266
1267         case IPV6_HOPOPTS:
1268         case IPV6_RTHDRDSTOPTS:
1269         case IPV6_RTHDR:
1270         case IPV6_DSTOPTS:
1271         {
1272                 struct ipv6_txoptions *opt;
1273
1274                 sockopt_lock_sock(sk);
1275                 opt = rcu_dereference_protected(np->opt,
1276                                                 lockdep_sock_is_held(sk));
1277                 len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1278                 sockopt_release_sock(sk);
1279                 /* check if ipv6_getsockopt_sticky() returns err code */
1280                 if (len < 0)
1281                         return len;
1282                 return copy_to_sockptr(optlen, &len, sizeof(int));
1283         }
1284
1285         case IPV6_RECVHOPOPTS:
1286                 val = np->rxopt.bits.hopopts;
1287                 break;
1288
1289         case IPV6_2292HOPOPTS:
1290                 val = np->rxopt.bits.ohopopts;
1291                 break;
1292
1293         case IPV6_RECVDSTOPTS:
1294                 val = np->rxopt.bits.dstopts;
1295                 break;
1296
1297         case IPV6_2292DSTOPTS:
1298                 val = np->rxopt.bits.odstopts;
1299                 break;
1300
1301         case IPV6_TCLASS:
1302                 val = np->tclass;
1303                 break;
1304
1305         case IPV6_RECVTCLASS:
1306                 val = np->rxopt.bits.rxtclass;
1307                 break;
1308
1309         case IPV6_FLOWINFO:
1310                 val = np->rxopt.bits.rxflow;
1311                 break;
1312
1313         case IPV6_RECVPATHMTU:
1314                 val = np->rxopt.bits.rxpmtu;
1315                 break;
1316
1317         case IPV6_PATHMTU:
1318         {
1319                 struct dst_entry *dst;
1320                 struct ip6_mtuinfo mtuinfo;
1321
1322                 if (len < sizeof(mtuinfo))
1323                         return -EINVAL;
1324
1325                 len = sizeof(mtuinfo);
1326                 memset(&mtuinfo, 0, sizeof(mtuinfo));
1327
1328                 rcu_read_lock();
1329                 dst = __sk_dst_get(sk);
1330                 if (dst)
1331                         mtuinfo.ip6m_mtu = dst_mtu(dst);
1332                 rcu_read_unlock();
1333                 if (!mtuinfo.ip6m_mtu)
1334                         return -ENOTCONN;
1335
1336                 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1337                         return -EFAULT;
1338                 if (copy_to_sockptr(optval, &mtuinfo, len))
1339                         return -EFAULT;
1340
1341                 return 0;
1342         }
1343
1344         case IPV6_TRANSPARENT:
1345                 val = inet_sk(sk)->transparent;
1346                 break;
1347
1348         case IPV6_FREEBIND:
1349                 val = inet_sk(sk)->freebind;
1350                 break;
1351
1352         case IPV6_RECVORIGDSTADDR:
1353                 val = np->rxopt.bits.rxorigdstaddr;
1354                 break;
1355
1356         case IPV6_UNICAST_HOPS:
1357         case IPV6_MULTICAST_HOPS:
1358         {
1359                 struct dst_entry *dst;
1360
1361                 if (optname == IPV6_UNICAST_HOPS)
1362                         val = np->hop_limit;
1363                 else
1364                         val = np->mcast_hops;
1365
1366                 if (val < 0) {
1367                         rcu_read_lock();
1368                         dst = __sk_dst_get(sk);
1369                         if (dst)
1370                                 val = ip6_dst_hoplimit(dst);
1371                         rcu_read_unlock();
1372                 }
1373
1374                 if (val < 0)
1375                         val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1376                 break;
1377         }
1378
1379         case IPV6_MULTICAST_LOOP:
1380                 val = np->mc_loop;
1381                 break;
1382
1383         case IPV6_MULTICAST_IF:
1384                 val = np->mcast_oif;
1385                 break;
1386
1387         case IPV6_MULTICAST_ALL:
1388                 val = np->mc_all;
1389                 break;
1390
1391         case IPV6_UNICAST_IF:
1392                 val = (__force int)htonl((__u32) np->ucast_oif);
1393                 break;
1394
1395         case IPV6_MTU_DISCOVER:
1396                 val = np->pmtudisc;
1397                 break;
1398
1399         case IPV6_RECVERR:
1400                 val = np->recverr;
1401                 break;
1402
1403         case IPV6_FLOWINFO_SEND:
1404                 val = np->sndflow;
1405                 break;
1406
1407         case IPV6_FLOWLABEL_MGR:
1408         {
1409                 struct in6_flowlabel_req freq;
1410                 int flags;
1411
1412                 if (len < sizeof(freq))
1413                         return -EINVAL;
1414
1415                 if (copy_from_sockptr(&freq, optval, sizeof(freq)))
1416                         return -EFAULT;
1417
1418                 if (freq.flr_action != IPV6_FL_A_GET)
1419                         return -EINVAL;
1420
1421                 len = sizeof(freq);
1422                 flags = freq.flr_flags;
1423
1424                 memset(&freq, 0, sizeof(freq));
1425
1426                 val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1427                 if (val < 0)
1428                         return val;
1429
1430                 if (copy_to_sockptr(optlen, &len, sizeof(int)))
1431                         return -EFAULT;
1432                 if (copy_to_sockptr(optval, &freq, len))
1433                         return -EFAULT;
1434
1435                 return 0;
1436         }
1437
1438         case IPV6_ADDR_PREFERENCES:
1439                 val = 0;
1440
1441                 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1442                         val |= IPV6_PREFER_SRC_TMP;
1443                 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1444                         val |= IPV6_PREFER_SRC_PUBLIC;
1445                 else {
1446                         /* XXX: should we return system default? */
1447                         val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1448                 }
1449
1450                 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1451                         val |= IPV6_PREFER_SRC_COA;
1452                 else
1453                         val |= IPV6_PREFER_SRC_HOME;
1454                 break;
1455
1456         case IPV6_MINHOPCOUNT:
1457                 val = np->min_hopcount;
1458                 break;
1459
1460         case IPV6_DONTFRAG:
1461                 val = np->dontfrag;
1462                 break;
1463
1464         case IPV6_AUTOFLOWLABEL:
1465                 val = ip6_autoflowlabel(sock_net(sk), np);
1466                 break;
1467
1468         case IPV6_RECVFRAGSIZE:
1469                 val = np->rxopt.bits.recvfragsize;
1470                 break;
1471
1472         case IPV6_ROUTER_ALERT_ISOLATE:
1473                 val = np->rtalert_isolate;
1474                 break;
1475
1476         case IPV6_RECVERR_RFC4884:
1477                 val = np->recverr_rfc4884;
1478                 break;
1479
1480         default:
1481                 return -ENOPROTOOPT;
1482         }
1483         len = min_t(unsigned int, sizeof(int), len);
1484         if (copy_to_sockptr(optlen, &len, sizeof(int)))
1485                 return -EFAULT;
1486         if (copy_to_sockptr(optval, &val, len))
1487                 return -EFAULT;
1488         return 0;
1489 }
1490
1491 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1492                     char __user *optval, int __user *optlen)
1493 {
1494         int err;
1495
1496         if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1497                 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1498
1499         if (level != SOL_IPV6)
1500                 return -ENOPROTOOPT;
1501
1502         err = do_ipv6_getsockopt(sk, level, optname,
1503                                  USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
1504 #ifdef CONFIG_NETFILTER
1505         /* we need to exclude all possible ENOPROTOOPTs except default case */
1506         if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1507                 int len;
1508
1509                 if (get_user(len, optlen))
1510                         return -EFAULT;
1511
1512                 err = nf_getsockopt(sk, PF_INET6, optname, optval, &len);
1513                 if (err >= 0)
1514                         err = put_user(len, optlen);
1515         }
1516 #endif
1517         return err;
1518 }
1519 EXPORT_SYMBOL(ipv6_getsockopt);