net/ipv4/ip_sockglue.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              The IP to API glue.
   7  *
   8  * Authors:     see ip.c
   9  *
  10  * Fixes:
  11  *              Many            :       Split from ip.c , see ip.c for history.
  12  *              Martin Mares    :       TOS setting fixed.
  13  *              Alan Cox        :       Fixed a couple of oopses in Martin's
  14  *                                      TOS tweaks.
  15  *              Mike McLagan    :       Routing by source
  16  */
  17
  18 #include <linux/module.h>
  19 #include <linux/types.h>
  20 #include <linux/mm.h>
  21 #include <linux/skbuff.h>
  22 #include <linux/ip.h>
  23 #include <linux/icmp.h>
  24 #include <linux/inetdevice.h>
  25 #include <linux/netdevice.h>
  26 #include <linux/slab.h>
  27 #include <net/sock.h>
  28 #include <net/ip.h>
  29 #include <net/icmp.h>
  30 #include <net/tcp_states.h>
  31 #include <linux/udp.h>
  32 #include <linux/igmp.h>
  33 #include <linux/netfilter.h>
  34 #include <linux/route.h>
  35 #include <linux/mroute.h>
  36 #include <net/inet_ecn.h>
  37 #include <net/route.h>
  38 #include <net/xfrm.h>
  39 #include <net/compat.h>
  40 #include <net/checksum.h>
  41 #if IS_ENABLED(CONFIG_IPV6)
  42 #include <net/transp_v6.h>
  43 #endif
  44 #include <net/ip_fib.h>
  45
  46 #include <linux/errqueue.h>
  47 #include <asm/uaccess.h>
  48
  49 /*
  50  *      SOL_IP control messages.
  51  */
  52
  53 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
  54 {
  55         struct in_pktinfo info = *PKTINFO_SKB_CB(skb);
  56
  57         info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
  58
  59         put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
  60 }
  61
  62 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
  63 {
  64         int ttl = ip_hdr(skb)->ttl;
  65         put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
  66 }
  67
  68 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
  69 {
  70         put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
  71 }
  72
  73 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
  74 {
  75         if (IPCB(skb)->opt.optlen == 0)
  76                 return;
  77
  78         put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
  79                  ip_hdr(skb) + 1);
  80 }
  81
  82
  83 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
  84 {
  85         unsigned char optbuf[sizeof(struct ip_options) + 40];
  86         struct ip_options *opt = (struct ip_options *)optbuf;
  87
  88         if (IPCB(skb)->opt.optlen == 0)
  89                 return;
  90
  91         if (ip_options_echo(opt, skb)) {
  92                 msg->msg_flags |= MSG_CTRUNC;
  93                 return;
  94         }
  95         ip_options_undo(opt);
  96
  97         put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
  98 }
  99
 100 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb,
 101                                   int offset)
 102 {
 103         __wsum csum = skb->csum;
 104
 105         if (skb->ip_summed != CHECKSUM_COMPLETE)
 106                 return;
 107
 108         if (offset != 0)
 109                 csum = csum_sub(csum, csum_partial(skb->data, offset, 0));
 110
 111         put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
 112 }
 113
 114 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
 115 {
 116         char *secdata;
 117         u32 seclen, secid;
 118         int err;
 119
 120         err = security_socket_getpeersec_dgram(NULL, skb, &secid);
 121         if (err)
 122                 return;
 123
 124         err = security_secid_to_secctx(secid, &secdata, &seclen);
 125         if (err)
 126                 return;
 127
 128         put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
 129         security_release_secctx(secdata, seclen);
 130 }
 131
 132 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
 133 {
 134         struct sockaddr_in sin;
 135         const struct iphdr *iph = ip_hdr(skb);
 136         __be16 *ports = (__be16 *)skb_transport_header(skb);
 137
 138         if (skb_transport_offset(skb) + 4 > skb->len)
 139                 return;
 140
 141         /* All current transport protocols have the port numbers in the
 142          * first four bytes of the transport header and this function is
 143          * written with this assumption in mind.
 144          */
 145
 146         sin.sin_family = AF_INET;
 147         sin.sin_addr.s_addr = iph->daddr;
 148         sin.sin_port = ports[1];
 149         memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 150
 151         put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
 152 }
 153
 154 void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb,
 155                          int offset)
 156 {
 157         struct inet_sock *inet = inet_sk(skb->sk);
 158         unsigned int flags = inet->cmsg_flags;
 159
 160         /* Ordered by supposed usage frequency */
 161         if (flags & IP_CMSG_PKTINFO) {
 162                 ip_cmsg_recv_pktinfo(msg, skb);
 163
 164                 flags &= ~IP_CMSG_PKTINFO;
 165                 if (!flags)
 166                         return;
 167         }
 168
 169         if (flags & IP_CMSG_TTL) {
 170                 ip_cmsg_recv_ttl(msg, skb);
 171
 172                 flags &= ~IP_CMSG_TTL;
 173                 if (!flags)
 174                         return;
 175         }
 176
 177         if (flags & IP_CMSG_TOS) {
 178                 ip_cmsg_recv_tos(msg, skb);
 179
 180                 flags &= ~IP_CMSG_TOS;
 181                 if (!flags)
 182                         return;
 183         }
 184
 185         if (flags & IP_CMSG_RECVOPTS) {
 186                 ip_cmsg_recv_opts(msg, skb);
 187
 188                 flags &= ~IP_CMSG_RECVOPTS;
 189                 if (!flags)
 190                         return;
 191         }
 192
 193         if (flags & IP_CMSG_RETOPTS) {
 194                 ip_cmsg_recv_retopts(msg, skb);
 195
 196                 flags &= ~IP_CMSG_RETOPTS;
 197                 if (!flags)
 198                         return;
 199         }
 200
 201         if (flags & IP_CMSG_PASSSEC) {
 202                 ip_cmsg_recv_security(msg, skb);
 203
 204                 flags &= ~IP_CMSG_PASSSEC;
 205                 if (!flags)
 206                         return;
 207         }
 208
 209         if (flags & IP_CMSG_ORIGDSTADDR) {
 210                 ip_cmsg_recv_dstaddr(msg, skb);
 211
 212                 flags &= ~IP_CMSG_ORIGDSTADDR;
 213                 if (!flags)
 214                         return;
 215         }
 216
 217         if (flags & IP_CMSG_CHECKSUM)
 218                 ip_cmsg_recv_checksum(msg, skb, offset);
 219 }
 220 EXPORT_SYMBOL(ip_cmsg_recv_offset);
 221
 222 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc,
 223                  bool allow_ipv6)
 224 {
 225         int err, val;
 226         struct cmsghdr *cmsg;
 227
 228         for_each_cmsghdr(cmsg, msg) {
 229                 if (!CMSG_OK(msg, cmsg))
 230                         return -EINVAL;
 231 #if IS_ENABLED(CONFIG_IPV6)
 232                 if (allow_ipv6 &&
 233                     cmsg->cmsg_level == SOL_IPV6 &&
 234                     cmsg->cmsg_type == IPV6_PKTINFO) {
 235                         struct in6_pktinfo *src_info;
 236
 237                         if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
 238                                 return -EINVAL;
 239                         src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
 240                         if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
 241                                 return -EINVAL;
 242                         ipc->oif = src_info->ipi6_ifindex;
 243                         ipc->addr = src_info->ipi6_addr.s6_addr32[3];
 244                         continue;
 245                 }
 246 #endif
 247                 if (cmsg->cmsg_level != SOL_IP)
 248                         continue;
 249                 switch (cmsg->cmsg_type) {
 250                 case IP_RETOPTS:
 251                         err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
 252
 253                         /* Our caller is responsible for freeing ipc->opt */
 254                         err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
 255                                              err < 40 ? err : 40);
 256                         if (err)
 257                                 return err;
 258                         break;
 259                 case IP_PKTINFO:
 260                 {
 261                         struct in_pktinfo *info;
 262                         if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
 263                                 return -EINVAL;
 264                         info = (struct in_pktinfo *)CMSG_DATA(cmsg);
 265                         ipc->oif = info->ipi_ifindex;
 266                         ipc->addr = info->ipi_spec_dst.s_addr;
 267                         break;
 268                 }
 269                 case IP_TTL:
 270                         if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
 271                                 return -EINVAL;
 272                         val = *(int *)CMSG_DATA(cmsg);
 273                         if (val < 1 || val > 255)
 274                                 return -EINVAL;
 275                         ipc->ttl = val;
 276                         break;
 277                 case IP_TOS:
 278                         if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
 279                                 return -EINVAL;
 280                         val = *(int *)CMSG_DATA(cmsg);
 281                         if (val < 0 || val > 255)
 282                                 return -EINVAL;
 283                         ipc->tos = val;
 284                         ipc->priority = rt_tos2priority(ipc->tos);
 285                         break;
 286
 287                 default:
 288                         return -EINVAL;
 289                 }
 290         }
 291         return 0;
 292 }
 293
 294
 295 /* Special input handler for packets caught by router alert option.
 296    They are selected only by protocol field, and then processed likely
 297    local ones; but only if someone wants them! Otherwise, router
 298    not running rsvpd will kill RSVP.
 299
 300    It is user level problem, what it will make with them.
 301    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
 302    but receiver should be enough clever f.e. to forward mtrace requests,
 303    sent to multicast group to reach destination designated router.
 304  */
 305 struct ip_ra_chain __rcu *ip_ra_chain;
 306 static DEFINE_SPINLOCK(ip_ra_lock);
 307
 308
 309 static void ip_ra_destroy_rcu(struct rcu_head *head)
 310 {
 311         struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
 312
 313         sock_put(ra->saved_sk);
 314         kfree(ra);
 315 }
 316
 317 int ip_ra_control(struct sock *sk, unsigned char on,
 318                   void (*destructor)(struct sock *))
 319 {
 320         struct ip_ra_chain *ra, *new_ra;
 321         struct ip_ra_chain __rcu **rap;
 322
 323         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
 324                 return -EINVAL;
 325
 326         new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
 327
 328         spin_lock_bh(&ip_ra_lock);
 329         for (rap = &ip_ra_chain;
 330              (ra = rcu_dereference_protected(*rap,
 331                         lockdep_is_held(&ip_ra_lock))) != NULL;
 332              rap = &ra->next) {
 333                 if (ra->sk == sk) {
 334                         if (on) {
 335                                 spin_unlock_bh(&ip_ra_lock);
 336                                 kfree(new_ra);
 337                                 return -EADDRINUSE;
 338                         }
 339                         /* dont let ip_call_ra_chain() use sk again */
 340                         ra->sk = NULL;
 341                         RCU_INIT_POINTER(*rap, ra->next);
 342                         spin_unlock_bh(&ip_ra_lock);
 343
 344                         if (ra->destructor)
 345                                 ra->destructor(sk);
 346                         /*
 347                          * Delay sock_put(sk) and kfree(ra) after one rcu grace
 348                          * period. This guarantee ip_call_ra_chain() dont need
 349                          * to mess with socket refcounts.
 350                          */
 351                         ra->saved_sk = sk;
 352                         call_rcu(&ra->rcu, ip_ra_destroy_rcu);
 353                         return 0;
 354                 }
 355         }
 356         if (!new_ra) {
 357                 spin_unlock_bh(&ip_ra_lock);
 358                 return -ENOBUFS;
 359         }
 360         new_ra->sk = sk;
 361         new_ra->destructor = destructor;
 362
 363         RCU_INIT_POINTER(new_ra->next, ra);
 364         rcu_assign_pointer(*rap, new_ra);
 365         sock_hold(sk);
 366         spin_unlock_bh(&ip_ra_lock);
 367
 368         return 0;
 369 }
 370
 371 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
 372                    __be16 port, u32 info, u8 *payload)
 373 {
 374         struct sock_exterr_skb *serr;
 375
 376         skb = skb_clone(skb, GFP_ATOMIC);
 377         if (!skb)
 378                 return;
 379
 380         serr = SKB_EXT_ERR(skb);
 381         serr->ee.ee_errno = err;
 382         serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
 383         serr->ee.ee_type = icmp_hdr(skb)->type;
 384         serr->ee.ee_code = icmp_hdr(skb)->code;
 385         serr->ee.ee_pad = 0;
 386         serr->ee.ee_info = info;
 387         serr->ee.ee_data = 0;
 388         serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
 389                                    skb_network_header(skb);
 390         serr->port = port;
 391
 392         if (skb_pull(skb, payload - skb->data)) {
 393                 skb_reset_transport_header(skb);
 394                 if (sock_queue_err_skb(sk, skb) == 0)
 395                         return;
 396         }
 397         kfree_skb(skb);
 398 }
 399
 400 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
 401 {
 402         struct inet_sock *inet = inet_sk(sk);
 403         struct sock_exterr_skb *serr;
 404         struct iphdr *iph;
 405         struct sk_buff *skb;
 406
 407         if (!inet->recverr)
 408                 return;
 409
 410         skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
 411         if (!skb)
 412                 return;
 413
 414         skb_put(skb, sizeof(struct iphdr));
 415         skb_reset_network_header(skb);
 416         iph = ip_hdr(skb);
 417         iph->daddr = daddr;
 418
 419         serr = SKB_EXT_ERR(skb);
 420         serr->ee.ee_errno = err;
 421         serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
 422         serr->ee.ee_type = 0;
 423         serr->ee.ee_code = 0;
 424         serr->ee.ee_pad = 0;
 425         serr->ee.ee_info = info;
 426         serr->ee.ee_data = 0;
 427         serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
 428         serr->port = port;
 429
 430         __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
 431         skb_reset_transport_header(skb);
 432
 433         if (sock_queue_err_skb(sk, skb))
 434                 kfree_skb(skb);
 435 }
 436
 437 /* For some errors we have valid addr_offset even with zero payload and
 438  * zero port. Also, addr_offset should be supported if port is set.
 439  */
 440 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
 441 {
 442         return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
 443                serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
 444 }
 445
 446 /* IPv4 supports cmsg on all imcp errors and some timestamps
 447  *
 448  * Timestamp code paths do not initialize the fields expected by cmsg:
 449  * the PKTINFO fields in skb->cb[]. Fill those in here.
 450  */
 451 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
 452                                        struct sk_buff *skb,
 453                                        int ee_origin)
 454 {
 455         struct in_pktinfo *info;
 456
 457         if (ee_origin == SO_EE_ORIGIN_ICMP)
 458                 return true;
 459
 460         if (ee_origin == SO_EE_ORIGIN_LOCAL)
 461                 return false;
 462
 463         /* Support IP_PKTINFO on tstamp packets if requested, to correlate
 464          * timestamp with egress dev. Not possible for packets without dev
 465          * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
 466          */
 467         if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
 468             (!skb->dev))
 469                 return false;
 470
 471         info = PKTINFO_SKB_CB(skb);
 472         info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
 473         info->ipi_ifindex = skb->dev->ifindex;
 474         return true;
 475 }
 476
 477 /*
 478  *      Handle MSG_ERRQUEUE
 479  */
 480 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 481 {
 482         struct sock_exterr_skb *serr;
 483         struct sk_buff *skb;
 484         DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 485         struct {
 486                 struct sock_extended_err ee;
 487                 struct sockaddr_in       offender;
 488         } errhdr;
 489         int err;
 490         int copied;
 491
 492         WARN_ON_ONCE(sk->sk_family == AF_INET6);
 493
 494         err = -EAGAIN;
 495         skb = sock_dequeue_err_skb(sk);
 496         if (!skb)
 497                 goto out;
 498
 499         copied = skb->len;
 500         if (copied > len) {
 501                 msg->msg_flags |= MSG_TRUNC;
 502                 copied = len;
 503         }
 504         err = skb_copy_datagram_msg(skb, 0, msg, copied);
 505         if (err)
 506                 goto out_free_skb;
 507
 508         sock_recv_timestamp(msg, sk, skb);
 509
 510         serr = SKB_EXT_ERR(skb);
 511
 512         if (sin && ipv4_datagram_support_addr(serr)) {
 513                 sin->sin_family = AF_INET;
 514                 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
 515                                                    serr->addr_offset);
 516                 sin->sin_port = serr->port;
 517                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 518                 *addr_len = sizeof(*sin);
 519         }
 520
 521         memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
 522         sin = &errhdr.offender;
 523         memset(sin, 0, sizeof(*sin));
 524
 525         if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
 526                 sin->sin_family = AF_INET;
 527                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 528                 if (inet_sk(sk)->cmsg_flags)
 529                         ip_cmsg_recv(msg, skb);
 530         }
 531
 532         put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
 533
 534         /* Now we could try to dump offended packet options */
 535
 536         msg->msg_flags |= MSG_ERRQUEUE;
 537         err = copied;
 538
 539 out_free_skb:
 540         kfree_skb(skb);
 541 out:
 542         return err;
 543 }
 544
 545
 546 /*
 547  *      Socket option code for IP. This is the end of the line after any
 548  *      TCP,UDP etc options on an IP socket.
 549  */
 550 static bool setsockopt_needs_rtnl(int optname)
 551 {
 552         switch (optname) {
 553         case IP_ADD_MEMBERSHIP:
 554         case IP_ADD_SOURCE_MEMBERSHIP:
 555         case IP_BLOCK_SOURCE:
 556         case IP_DROP_MEMBERSHIP:
 557         case IP_DROP_SOURCE_MEMBERSHIP:
 558         case IP_MSFILTER:
 559         case IP_UNBLOCK_SOURCE:
 560         case MCAST_BLOCK_SOURCE:
 561         case MCAST_MSFILTER:
 562         case MCAST_JOIN_GROUP:
 563         case MCAST_JOIN_SOURCE_GROUP:
 564         case MCAST_LEAVE_GROUP:
 565         case MCAST_LEAVE_SOURCE_GROUP:
 566         case MCAST_UNBLOCK_SOURCE:
 567                 return true;
 568         }
 569         return false;
 570 }
 571
 572 static int do_ip_setsockopt(struct sock *sk, int level,
 573                             int optname, char __user *optval, unsigned int optlen)
 574 {
 575         struct inet_sock *inet = inet_sk(sk);
 576         int val = 0, err;
 577         bool needs_rtnl = setsockopt_needs_rtnl(optname);
 578
 579         switch (optname) {
 580         case IP_PKTINFO:
 581         case IP_RECVTTL:
 582         case IP_RECVOPTS:
 583         case IP_RECVTOS:
 584         case IP_RETOPTS:
 585         case IP_TOS:
 586         case IP_TTL:
 587         case IP_HDRINCL:
 588         case IP_MTU_DISCOVER:
 589         case IP_RECVERR:
 590         case IP_ROUTER_ALERT:
 591         case IP_FREEBIND:
 592         case IP_PASSSEC:
 593         case IP_TRANSPARENT:
 594         case IP_MINTTL:
 595         case IP_NODEFRAG:
 596         case IP_UNICAST_IF:
 597         case IP_MULTICAST_TTL:
 598         case IP_MULTICAST_ALL:
 599         case IP_MULTICAST_LOOP:
 600         case IP_RECVORIGDSTADDR:
 601         case IP_CHECKSUM:
 602                 if (optlen >= sizeof(int)) {
 603                         if (get_user(val, (int __user *) optval))
 604                                 return -EFAULT;
 605                 } else if (optlen >= sizeof(char)) {
 606                         unsigned char ucval;
 607
 608                         if (get_user(ucval, (unsigned char __user *) optval))
 609                                 return -EFAULT;
 610                         val = (int) ucval;
 611                 }
 612         }
 613
 614         /* If optlen==0, it is equivalent to val == 0 */
 615
 616         if (ip_mroute_opt(optname))
 617                 return ip_mroute_setsockopt(sk, optname, optval, optlen);
 618
 619         err = 0;
 620         if (needs_rtnl)
 621                 rtnl_lock();
 622         lock_sock(sk);
 623
 624         switch (optname) {
 625         case IP_OPTIONS:
 626         {
 627                 struct ip_options_rcu *old, *opt = NULL;
 628
 629                 if (optlen > 40)
 630                         goto e_inval;
 631                 err = ip_options_get_from_user(sock_net(sk), &opt,
 632                                                optval, optlen);
 633                 if (err)
 634                         break;
 635                 old = rcu_dereference_protected(inet->inet_opt,
 636                                                 sock_owned_by_user(sk));
 637                 if (inet->is_icsk) {
 638                         struct inet_connection_sock *icsk = inet_csk(sk);
 639 #if IS_ENABLED(CONFIG_IPV6)
 640                         if (sk->sk_family == PF_INET ||
 641                             (!((1 << sk->sk_state) &
 642                                (TCPF_LISTEN | TCPF_CLOSE)) &&
 643                              inet->inet_daddr != LOOPBACK4_IPV6)) {
 644 #endif
 645                                 if (old)
 646                                         icsk->icsk_ext_hdr_len -= old->opt.optlen;
 647                                 if (opt)
 648                                         icsk->icsk_ext_hdr_len += opt->opt.optlen;
 649                                 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 650 #if IS_ENABLED(CONFIG_IPV6)
 651                         }
 652 #endif
 653                 }
 654                 rcu_assign_pointer(inet->inet_opt, opt);
 655                 if (old)
 656                         kfree_rcu(old, rcu);
 657                 break;
 658         }
 659         case IP_PKTINFO:
 660                 if (val)
 661                         inet->cmsg_flags |= IP_CMSG_PKTINFO;
 662                 else
 663                         inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
 664                 break;
 665         case IP_RECVTTL:
 666                 if (val)
 667                         inet->cmsg_flags |=  IP_CMSG_TTL;
 668                 else
 669                         inet->cmsg_flags &= ~IP_CMSG_TTL;
 670                 break;
 671         case IP_RECVTOS:
 672                 if (val)
 673                         inet->cmsg_flags |=  IP_CMSG_TOS;
 674                 else
 675                         inet->cmsg_flags &= ~IP_CMSG_TOS;
 676                 break;
 677         case IP_RECVOPTS:
 678                 if (val)
 679                         inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
 680                 else
 681                         inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
 682                 break;
 683         case IP_RETOPTS:
 684                 if (val)
 685                         inet->cmsg_flags |= IP_CMSG_RETOPTS;
 686                 else
 687                         inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
 688                 break;
 689         case IP_PASSSEC:
 690                 if (val)
 691                         inet->cmsg_flags |= IP_CMSG_PASSSEC;
 692                 else
 693                         inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
 694                 break;
 695         case IP_RECVORIGDSTADDR:
 696                 if (val)
 697                         inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
 698                 else
 699                         inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
 700                 break;
 701         case IP_CHECKSUM:
 702                 if (val) {
 703                         if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
 704                                 inet_inc_convert_csum(sk);
 705                                 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
 706                         }
 707                 } else {
 708                         if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
 709                                 inet_dec_convert_csum(sk);
 710                                 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
 711                         }
 712                 }
 713                 break;
 714         case IP_TOS:    /* This sets both TOS and Precedence */
 715                 if (sk->sk_type == SOCK_STREAM) {
 716                         val &= ~INET_ECN_MASK;
 717                         val |= inet->tos & INET_ECN_MASK;
 718                 }
 719                 if (inet->tos != val) {
 720                         inet->tos = val;
 721                         sk->sk_priority = rt_tos2priority(val);
 722                         sk_dst_reset(sk);
 723                 }
 724                 break;
 725         case IP_TTL:
 726                 if (optlen < 1)
 727                         goto e_inval;
 728                 if (val != -1 && (val < 1 || val > 255))
 729                         goto e_inval;
 730                 inet->uc_ttl = val;
 731                 break;
 732         case IP_HDRINCL:
 733                 if (sk->sk_type != SOCK_RAW) {
 734                         err = -ENOPROTOOPT;
 735                         break;
 736                 }
 737                 inet->hdrincl = val ? 1 : 0;
 738                 break;
 739         case IP_NODEFRAG:
 740                 if (sk->sk_type != SOCK_RAW) {
 741                         err = -ENOPROTOOPT;
 742                         break;
 743                 }
 744                 inet->nodefrag = val ? 1 : 0;
 745                 break;
 746         case IP_MTU_DISCOVER:
 747                 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
 748                         goto e_inval;
 749                 inet->pmtudisc = val;
 750                 break;
 751         case IP_RECVERR:
 752                 inet->recverr = !!val;
 753                 if (!val)
 754                         skb_queue_purge(&sk->sk_error_queue);
 755                 break;
 756         case IP_MULTICAST_TTL:
 757                 if (sk->sk_type == SOCK_STREAM)
 758                         goto e_inval;
 759                 if (optlen < 1)
 760                         goto e_inval;
 761                 if (val == -1)
 762                         val = 1;
 763                 if (val < 0 || val > 255)
 764                         goto e_inval;
 765                 inet->mc_ttl = val;
 766                 break;
 767         case IP_MULTICAST_LOOP:
 768                 if (optlen < 1)
 769                         goto e_inval;
 770                 inet->mc_loop = !!val;
 771                 break;
 772         case IP_UNICAST_IF:
 773         {
 774                 struct net_device *dev = NULL;
 775                 int ifindex;
 776
 777                 if (optlen != sizeof(int))
 778                         goto e_inval;
 779
 780                 ifindex = (__force int)ntohl((__force __be32)val);
 781                 if (ifindex == 0) {
 782                         inet->uc_index = 0;
 783                         err = 0;
 784                         break;
 785                 }
 786
 787                 dev = dev_get_by_index(sock_net(sk), ifindex);
 788                 err = -EADDRNOTAVAIL;
 789                 if (!dev)
 790                         break;
 791                 dev_put(dev);
 792
 793                 err = -EINVAL;
 794                 if (sk->sk_bound_dev_if)
 795                         break;
 796
 797                 inet->uc_index = ifindex;
 798                 err = 0;
 799                 break;
 800         }
 801         case IP_MULTICAST_IF:
 802         {
 803                 struct ip_mreqn mreq;
 804                 struct net_device *dev = NULL;
 805
 806                 if (sk->sk_type == SOCK_STREAM)
 807                         goto e_inval;
 808                 /*
 809                  *      Check the arguments are allowable
 810                  */
 811
 812                 if (optlen < sizeof(struct in_addr))
 813                         goto e_inval;
 814
 815                 err = -EFAULT;
 816                 if (optlen >= sizeof(struct ip_mreqn)) {
 817                         if (copy_from_user(&mreq, optval, sizeof(mreq)))
 818                                 break;
 819                 } else {
 820                         memset(&mreq, 0, sizeof(mreq));
 821                         if (optlen >= sizeof(struct ip_mreq)) {
 822                                 if (copy_from_user(&mreq, optval,
 823                                                    sizeof(struct ip_mreq)))
 824                                         break;
 825                         } else if (optlen >= sizeof(struct in_addr)) {
 826                                 if (copy_from_user(&mreq.imr_address, optval,
 827                                                    sizeof(struct in_addr)))
 828                                         break;
 829                         }
 830                 }
 831
 832                 if (!mreq.imr_ifindex) {
 833                         if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
 834                                 inet->mc_index = 0;
 835                                 inet->mc_addr  = 0;
 836                                 err = 0;
 837                                 break;
 838                         }
 839                         dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
 840                         if (dev)
 841                                 mreq.imr_ifindex = dev->ifindex;
 842                 } else
 843                         dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
 844
 845
 846                 err = -EADDRNOTAVAIL;
 847                 if (!dev)
 848                         break;
 849                 dev_put(dev);
 850
 851                 err = -EINVAL;
 852                 if (sk->sk_bound_dev_if &&
 853                     mreq.imr_ifindex != sk->sk_bound_dev_if)
 854                         break;
 855
 856                 inet->mc_index = mreq.imr_ifindex;
 857                 inet->mc_addr  = mreq.imr_address.s_addr;
 858                 err = 0;
 859                 break;
 860         }
 861
 862         case IP_ADD_MEMBERSHIP:
 863         case IP_DROP_MEMBERSHIP:
 864         {
 865                 struct ip_mreqn mreq;
 866
 867                 err = -EPROTO;
 868                 if (inet_sk(sk)->is_icsk)
 869                         break;
 870
 871                 if (optlen < sizeof(struct ip_mreq))
 872                         goto e_inval;
 873                 err = -EFAULT;
 874                 if (optlen >= sizeof(struct ip_mreqn)) {
 875                         if (copy_from_user(&mreq, optval, sizeof(mreq)))
 876                                 break;
 877                 } else {
 878                         memset(&mreq, 0, sizeof(mreq));
 879                         if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
 880                                 break;
 881                 }
 882
 883                 if (optname == IP_ADD_MEMBERSHIP)
 884                         err = ip_mc_join_group(sk, &mreq);
 885                 else
 886                         err = ip_mc_leave_group(sk, &mreq);
 887                 break;
 888         }
 889         case IP_MSFILTER:
 890         {
 891                 struct ip_msfilter *msf;
 892
 893                 if (optlen < IP_MSFILTER_SIZE(0))
 894                         goto e_inval;
 895                 if (optlen > sysctl_optmem_max) {
 896                         err = -ENOBUFS;
 897                         break;
 898                 }
 899                 msf = kmalloc(optlen, GFP_KERNEL);
 900                 if (!msf) {
 901                         err = -ENOBUFS;
 902                         break;
 903                 }
 904                 err = -EFAULT;
 905                 if (copy_from_user(msf, optval, optlen)) {
 906                         kfree(msf);
 907                         break;
 908                 }
 909                 /* numsrc >= (1G-4) overflow in 32 bits */
 910                 if (msf->imsf_numsrc >= 0x3ffffffcU ||
 911                     msf->imsf_numsrc > sysctl_igmp_max_msf) {
 912                         kfree(msf);
 913                         err = -ENOBUFS;
 914                         break;
 915                 }
 916                 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
 917                         kfree(msf);
 918                         err = -EINVAL;
 919                         break;
 920                 }
 921                 err = ip_mc_msfilter(sk, msf, 0);
 922                 kfree(msf);
 923                 break;
 924         }
 925         case IP_BLOCK_SOURCE:
 926         case IP_UNBLOCK_SOURCE:
 927         case IP_ADD_SOURCE_MEMBERSHIP:
 928         case IP_DROP_SOURCE_MEMBERSHIP:
 929         {
 930                 struct ip_mreq_source mreqs;
 931                 int omode, add;
 932
 933                 if (optlen != sizeof(struct ip_mreq_source))
 934                         goto e_inval;
 935                 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
 936                         err = -EFAULT;
 937                         break;
 938                 }
 939                 if (optname == IP_BLOCK_SOURCE) {
 940                         omode = MCAST_EXCLUDE;
 941                         add = 1;
 942                 } else if (optname == IP_UNBLOCK_SOURCE) {
 943                         omode = MCAST_EXCLUDE;
 944                         add = 0;
 945                 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
 946                         struct ip_mreqn mreq;
 947
 948                         mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
 949                         mreq.imr_address.s_addr = mreqs.imr_interface;
 950                         mreq.imr_ifindex = 0;
 951                         err = ip_mc_join_group(sk, &mreq);
 952                         if (err && err != -EADDRINUSE)
 953                                 break;
 954                         omode = MCAST_INCLUDE;
 955                         add = 1;
 956                 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
 957                         omode = MCAST_INCLUDE;
 958                         add = 0;
 959                 }
 960                 err = ip_mc_source(add, omode, sk, &mreqs, 0);
 961                 break;
 962         }
 963         case MCAST_JOIN_GROUP:
 964         case MCAST_LEAVE_GROUP:
 965         {
 966                 struct group_req greq;
 967                 struct sockaddr_in *psin;
 968                 struct ip_mreqn mreq;
 969
 970                 if (optlen < sizeof(struct group_req))
 971                         goto e_inval;
 972                 err = -EFAULT;
 973                 if (copy_from_user(&greq, optval, sizeof(greq)))
 974                         break;
 975                 psin = (struct sockaddr_in *)&greq.gr_group;
 976                 if (psin->sin_family != AF_INET)
 977                         goto e_inval;
 978                 memset(&mreq, 0, sizeof(mreq));
 979                 mreq.imr_multiaddr = psin->sin_addr;
 980                 mreq.imr_ifindex = greq.gr_interface;
 981
 982                 if (optname == MCAST_JOIN_GROUP)
 983                         err = ip_mc_join_group(sk, &mreq);
 984                 else
 985                         err = ip_mc_leave_group(sk, &mreq);
 986                 break;
 987         }
 988         case MCAST_JOIN_SOURCE_GROUP:
 989         case MCAST_LEAVE_SOURCE_GROUP:
 990         case MCAST_BLOCK_SOURCE:
 991         case MCAST_UNBLOCK_SOURCE:
 992         {
 993                 struct group_source_req greqs;
 994                 struct ip_mreq_source mreqs;
 995                 struct sockaddr_in *psin;
 996                 int omode, add;
 997
 998                 if (optlen != sizeof(struct group_source_req))
 999                         goto e_inval;
1000                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1001                         err = -EFAULT;
1002                         break;
1003                 }
1004                 if (greqs.gsr_group.ss_family != AF_INET ||
1005                     greqs.gsr_source.ss_family != AF_INET) {
1006                         err = -EADDRNOTAVAIL;
1007                         break;
1008                 }
1009                 psin = (struct sockaddr_in *)&greqs.gsr_group;
1010                 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1011                 psin = (struct sockaddr_in *)&greqs.gsr_source;
1012                 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1013                 mreqs.imr_interface = 0; /* use index for mc_source */
1014
1015                 if (optname == MCAST_BLOCK_SOURCE) {
1016                         omode = MCAST_EXCLUDE;
1017                         add = 1;
1018                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1019                         omode = MCAST_EXCLUDE;
1020                         add = 0;
1021                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1022                         struct ip_mreqn mreq;
1023
1024                         psin = (struct sockaddr_in *)&greqs.gsr_group;
1025                         mreq.imr_multiaddr = psin->sin_addr;
1026                         mreq.imr_address.s_addr = 0;
1027                         mreq.imr_ifindex = greqs.gsr_interface;
1028                         err = ip_mc_join_group(sk, &mreq);
1029                         if (err && err != -EADDRINUSE)
1030                                 break;
1031                         greqs.gsr_interface = mreq.imr_ifindex;
1032                         omode = MCAST_INCLUDE;
1033                         add = 1;
1034                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1035                         omode = MCAST_INCLUDE;
1036                         add = 0;
1037                 }
1038                 err = ip_mc_source(add, omode, sk, &mreqs,
1039                                    greqs.gsr_interface);
1040                 break;
1041         }
1042         case MCAST_MSFILTER:
1043         {
1044                 struct sockaddr_in *psin;
1045                 struct ip_msfilter *msf = NULL;
1046                 struct group_filter *gsf = NULL;
1047                 int msize, i, ifindex;
1048
1049                 if (optlen < GROUP_FILTER_SIZE(0))
1050                         goto e_inval;
1051                 if (optlen > sysctl_optmem_max) {
1052                         err = -ENOBUFS;
1053                         break;
1054                 }
1055                 gsf = kmalloc(optlen, GFP_KERNEL);
1056                 if (!gsf) {
1057                         err = -ENOBUFS;
1058                         break;
1059                 }
1060                 err = -EFAULT;
1061                 if (copy_from_user(gsf, optval, optlen))
1062                         goto mc_msf_out;
1063
1064                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1065                 if (gsf->gf_numsrc >= 0x1ffffff ||
1066                     gsf->gf_numsrc > sysctl_igmp_max_msf) {
1067                         err = -ENOBUFS;
1068                         goto mc_msf_out;
1069                 }
1070                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1071                         err = -EINVAL;
1072                         goto mc_msf_out;
1073                 }
1074                 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1075                 msf = kmalloc(msize, GFP_KERNEL);
1076                 if (!msf) {
1077                         err = -ENOBUFS;
1078                         goto mc_msf_out;
1079                 }
1080                 ifindex = gsf->gf_interface;
1081                 psin = (struct sockaddr_in *)&gsf->gf_group;
1082                 if (psin->sin_family != AF_INET) {
1083                         err = -EADDRNOTAVAIL;
1084                         goto mc_msf_out;
1085                 }
1086                 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1087                 msf->imsf_interface = 0;
1088                 msf->imsf_fmode = gsf->gf_fmode;
1089                 msf->imsf_numsrc = gsf->gf_numsrc;
1090                 err = -EADDRNOTAVAIL;
1091                 for (i = 0; i < gsf->gf_numsrc; ++i) {
1092                         psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1093
1094                         if (psin->sin_family != AF_INET)
1095                                 goto mc_msf_out;
1096                         msf->imsf_slist[i] = psin->sin_addr.s_addr;
1097                 }
1098                 kfree(gsf);
1099                 gsf = NULL;
1100
1101                 err = ip_mc_msfilter(sk, msf, ifindex);
1102 mc_msf_out:
1103                 kfree(msf);
1104                 kfree(gsf);
1105                 break;
1106         }
1107         case IP_MULTICAST_ALL:
1108                 if (optlen < 1)
1109                         goto e_inval;
1110                 if (val != 0 && val != 1)
1111                         goto e_inval;
1112                 inet->mc_all = val;
1113                 break;
1114         case IP_ROUTER_ALERT:
1115                 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1116                 break;
1117
1118         case IP_FREEBIND:
1119                 if (optlen < 1)
1120                         goto e_inval;
1121                 inet->freebind = !!val;
1122                 break;
1123
1124         case IP_IPSEC_POLICY:
1125         case IP_XFRM_POLICY:
1126                 err = -EPERM;
1127                 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1128                         break;
1129                 err = xfrm_user_policy(sk, optname, optval, optlen);
1130                 break;
1131
1132         case IP_TRANSPARENT:
1133                 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1134                     !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1135                         err = -EPERM;
1136                         break;
1137                 }
1138                 if (optlen < 1)
1139                         goto e_inval;
1140                 inet->transparent = !!val;
1141                 break;
1142
1143         case IP_MINTTL:
1144                 if (optlen < 1)
1145                         goto e_inval;
1146                 if (val < 0 || val > 255)
1147                         goto e_inval;
1148                 inet->min_ttl = val;
1149                 break;
1150
1151         default:
1152                 err = -ENOPROTOOPT;
1153                 break;
1154         }
1155         release_sock(sk);
1156         if (needs_rtnl)
1157                 rtnl_unlock();
1158         return err;
1159
1160 e_inval:
1161         release_sock(sk);
1162         if (needs_rtnl)
1163                 rtnl_unlock();
1164         return -EINVAL;
1165 }
1166
1167 /**
1168  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1169  * @sk: socket
1170  * @skb: buffer
1171  *
1172  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1173  * destination in skb->cb[] before dst drop.
1174  * This way, receiver doesn't make cache line misses to read rtable.
1175  */
1176 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1177 {
1178         struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1179         bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1180                        ipv6_sk_rxinfo(sk);
1181
1182         if (prepare && skb_rtable(skb)) {
1183                 pktinfo->ipi_ifindex = inet_iif(skb);
1184                 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1185         } else {
1186                 pktinfo->ipi_ifindex = 0;
1187                 pktinfo->ipi_spec_dst.s_addr = 0;
1188         }
1189         skb_dst_drop(skb);
1190 }
1191
1192 int ip_setsockopt(struct sock *sk, int level,
1193                 int optname, char __user *optval, unsigned int optlen)
1194 {
1195         int err;
1196
1197         if (level != SOL_IP)
1198                 return -ENOPROTOOPT;
1199
1200         err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1201 #ifdef CONFIG_NETFILTER
1202         /* we need to exclude all possible ENOPROTOOPTs except default case */
1203         if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1204                         optname != IP_IPSEC_POLICY &&
1205                         optname != IP_XFRM_POLICY &&
1206                         !ip_mroute_opt(optname)) {
1207                 lock_sock(sk);
1208                 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1209                 release_sock(sk);
1210         }
1211 #endif
1212         return err;
1213 }
1214 EXPORT_SYMBOL(ip_setsockopt);
1215
1216 #ifdef CONFIG_COMPAT
1217 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1218                          char __user *optval, unsigned int optlen)
1219 {
1220         int err;
1221
1222         if (level != SOL_IP)
1223                 return -ENOPROTOOPT;
1224
1225         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1226                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1227                         ip_setsockopt);
1228
1229         err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1230 #ifdef CONFIG_NETFILTER
1231         /* we need to exclude all possible ENOPROTOOPTs except default case */
1232         if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1233                         optname != IP_IPSEC_POLICY &&
1234                         optname != IP_XFRM_POLICY &&
1235                         !ip_mroute_opt(optname)) {
1236                 lock_sock(sk);
1237                 err = compat_nf_setsockopt(sk, PF_INET, optname,
1238                                            optval, optlen);
1239                 release_sock(sk);
1240         }
1241 #endif
1242         return err;
1243 }
1244 EXPORT_SYMBOL(compat_ip_setsockopt);
1245 #endif
1246
1247 /*
1248  *      Get the options. Note for future reference. The GET of IP options gets
1249  *      the _received_ ones. The set sets the _sent_ ones.
1250  */
1251
1252 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1253                             char __user *optval, int __user *optlen, unsigned int flags)
1254 {
1255         struct inet_sock *inet = inet_sk(sk);
1256         int val;
1257         int len;
1258
1259         if (level != SOL_IP)
1260                 return -EOPNOTSUPP;
1261
1262         if (ip_mroute_opt(optname))
1263                 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1264
1265         if (get_user(len, optlen))
1266                 return -EFAULT;
1267         if (len < 0)
1268                 return -EINVAL;
1269
1270         lock_sock(sk);
1271
1272         switch (optname) {
1273         case IP_OPTIONS:
1274         {
1275                 unsigned char optbuf[sizeof(struct ip_options)+40];
1276                 struct ip_options *opt = (struct ip_options *)optbuf;
1277                 struct ip_options_rcu *inet_opt;
1278
1279                 inet_opt = rcu_dereference_protected(inet->inet_opt,
1280                                                      sock_owned_by_user(sk));
1281                 opt->optlen = 0;
1282                 if (inet_opt)
1283                         memcpy(optbuf, &inet_opt->opt,
1284                                sizeof(struct ip_options) +
1285                                inet_opt->opt.optlen);
1286                 release_sock(sk);
1287
1288                 if (opt->optlen == 0)
1289                         return put_user(0, optlen);
1290
1291                 ip_options_undo(opt);
1292
1293                 len = min_t(unsigned int, len, opt->optlen);
1294                 if (put_user(len, optlen))
1295                         return -EFAULT;
1296                 if (copy_to_user(optval, opt->__data, len))
1297                         return -EFAULT;
1298                 return 0;
1299         }
1300         case IP_PKTINFO:
1301                 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1302                 break;
1303         case IP_RECVTTL:
1304                 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1305                 break;
1306         case IP_RECVTOS:
1307                 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1308                 break;
1309         case IP_RECVOPTS:
1310                 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1311                 break;
1312         case IP_RETOPTS:
1313                 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1314                 break;
1315         case IP_PASSSEC:
1316                 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1317                 break;
1318         case IP_RECVORIGDSTADDR:
1319                 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1320                 break;
1321         case IP_CHECKSUM:
1322                 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1323                 break;
1324         case IP_TOS:
1325                 val = inet->tos;
1326                 break;
1327         case IP_TTL:
1328                 val = (inet->uc_ttl == -1 ?
1329                        sysctl_ip_default_ttl :
1330                        inet->uc_ttl);
1331                 break;
1332         case IP_HDRINCL:
1333                 val = inet->hdrincl;
1334                 break;
1335         case IP_NODEFRAG:
1336                 val = inet->nodefrag;
1337                 break;
1338         case IP_MTU_DISCOVER:
1339                 val = inet->pmtudisc;
1340                 break;
1341         case IP_MTU:
1342         {
1343                 struct dst_entry *dst;
1344                 val = 0;
1345                 dst = sk_dst_get(sk);
1346                 if (dst) {
1347                         val = dst_mtu(dst);
1348                         dst_release(dst);
1349                 }
1350                 if (!val) {
1351                         release_sock(sk);
1352                         return -ENOTCONN;
1353                 }
1354                 break;
1355         }
1356         case IP_RECVERR:
1357                 val = inet->recverr;
1358                 break;
1359         case IP_MULTICAST_TTL:
1360                 val = inet->mc_ttl;
1361                 break;
1362         case IP_MULTICAST_LOOP:
1363                 val = inet->mc_loop;
1364                 break;
1365         case IP_UNICAST_IF:
1366                 val = (__force int)htonl((__u32) inet->uc_index);
1367                 break;
1368         case IP_MULTICAST_IF:
1369         {
1370                 struct in_addr addr;
1371                 len = min_t(unsigned int, len, sizeof(struct in_addr));
1372                 addr.s_addr = inet->mc_addr;
1373                 release_sock(sk);
1374
1375                 if (put_user(len, optlen))
1376                         return -EFAULT;
1377                 if (copy_to_user(optval, &addr, len))
1378                         return -EFAULT;
1379                 return 0;
1380         }
1381         case IP_MSFILTER:
1382         {
1383                 struct ip_msfilter msf;
1384                 int err;
1385
1386                 if (len < IP_MSFILTER_SIZE(0)) {
1387                         release_sock(sk);
1388                         return -EINVAL;
1389                 }
1390                 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1391                         release_sock(sk);
1392                         return -EFAULT;
1393                 }
1394                 err = ip_mc_msfget(sk, &msf,
1395                                    (struct ip_msfilter __user *)optval, optlen);
1396                 release_sock(sk);
1397                 return err;
1398         }
1399         case MCAST_MSFILTER:
1400         {
1401                 struct group_filter gsf;
1402                 int err;
1403
1404                 if (len < GROUP_FILTER_SIZE(0)) {
1405                         release_sock(sk);
1406                         return -EINVAL;
1407                 }
1408                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1409                         release_sock(sk);
1410                         return -EFAULT;
1411                 }
1412                 err = ip_mc_gsfget(sk, &gsf,
1413                                    (struct group_filter __user *)optval,
1414                                    optlen);
1415                 release_sock(sk);
1416                 return err;
1417         }
1418         case IP_MULTICAST_ALL:
1419                 val = inet->mc_all;
1420                 break;
1421         case IP_PKTOPTIONS:
1422         {
1423                 struct msghdr msg;
1424
1425                 release_sock(sk);
1426
1427                 if (sk->sk_type != SOCK_STREAM)
1428                         return -ENOPROTOOPT;
1429
1430                 msg.msg_control = (__force void *) optval;
1431                 msg.msg_controllen = len;
1432                 msg.msg_flags = flags;
1433
1434                 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1435                         struct in_pktinfo info;
1436
1437                         info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1438                         info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1439                         info.ipi_ifindex = inet->mc_index;
1440                         put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1441                 }
1442                 if (inet->cmsg_flags & IP_CMSG_TTL) {
1443                         int hlim = inet->mc_ttl;
1444                         put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1445                 }
1446                 if (inet->cmsg_flags & IP_CMSG_TOS) {
1447                         int tos = inet->rcv_tos;
1448                         put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1449                 }
1450                 len -= msg.msg_controllen;
1451                 return put_user(len, optlen);
1452         }
1453         case IP_FREEBIND:
1454                 val = inet->freebind;
1455                 break;
1456         case IP_TRANSPARENT:
1457                 val = inet->transparent;
1458                 break;
1459         case IP_MINTTL:
1460                 val = inet->min_ttl;
1461                 break;
1462         default:
1463                 release_sock(sk);
1464                 return -ENOPROTOOPT;
1465         }
1466         release_sock(sk);
1467
1468         if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1469                 unsigned char ucval = (unsigned char)val;
1470                 len = 1;
1471                 if (put_user(len, optlen))
1472                         return -EFAULT;
1473                 if (copy_to_user(optval, &ucval, 1))
1474                         return -EFAULT;
1475         } else {
1476                 len = min_t(unsigned int, sizeof(int), len);
1477                 if (put_user(len, optlen))
1478                         return -EFAULT;
1479                 if (copy_to_user(optval, &val, len))
1480                         return -EFAULT;
1481         }
1482         return 0;
1483 }
1484
1485 int ip_getsockopt(struct sock *sk, int level,
1486                   int optname, char __user *optval, int __user *optlen)
1487 {
1488         int err;
1489
1490         err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1491 #ifdef CONFIG_NETFILTER
1492         /* we need to exclude all possible ENOPROTOOPTs except default case */
1493         if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1494                         !ip_mroute_opt(optname)) {
1495                 int len;
1496
1497                 if (get_user(len, optlen))
1498                         return -EFAULT;
1499
1500                 lock_sock(sk);
1501                 err = nf_getsockopt(sk, PF_INET, optname, optval,
1502                                 &len);
1503                 release_sock(sk);
1504                 if (err >= 0)
1505                         err = put_user(len, optlen);
1506                 return err;
1507         }
1508 #endif
1509         return err;
1510 }
1511 EXPORT_SYMBOL(ip_getsockopt);
1512
1513 #ifdef CONFIG_COMPAT
1514 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1515                          char __user *optval, int __user *optlen)
1516 {
1517         int err;
1518
1519         if (optname == MCAST_MSFILTER)
1520                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1521                         ip_getsockopt);
1522
1523         err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1524                 MSG_CMSG_COMPAT);
1525
1526 #ifdef CONFIG_NETFILTER
1527         /* we need to exclude all possible ENOPROTOOPTs except default case */
1528         if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1529                         !ip_mroute_opt(optname)) {
1530                 int len;
1531
1532                 if (get_user(len, optlen))
1533                         return -EFAULT;
1534
1535                 lock_sock(sk);
1536                 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1537                 release_sock(sk);
1538                 if (err >= 0)
1539                         err = put_user(len, optlen);
1540                 return err;
1541         }
1542 #endif
1543         return err;
1544 }
1545 EXPORT_SYMBOL(compat_ip_getsockopt);
1546 #endif