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_transport_header(skb),
 110                                                    offset, 0));
 111
 112         put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum);
 113 }
 114
 115 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
 116 {
 117         char *secdata;
 118         u32 seclen, secid;
 119         int err;
 120
 121         err = security_socket_getpeersec_dgram(NULL, skb, &secid);
 122         if (err)
 123                 return;
 124
 125         err = security_secid_to_secctx(secid, &secdata, &seclen);
 126         if (err)
 127                 return;
 128
 129         put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
 130         security_release_secctx(secdata, seclen);
 131 }
 132
 133 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
 134 {
 135         struct sockaddr_in sin;
 136         const struct iphdr *iph = ip_hdr(skb);
 137         __be16 *ports = (__be16 *)skb_transport_header(skb);
 138
 139         if (skb_transport_offset(skb) + 4 > skb->len)
 140                 return;
 141
 142         /* All current transport protocols have the port numbers in the
 143          * first four bytes of the transport header and this function is
 144          * written with this assumption in mind.
 145          */
 146
 147         sin.sin_family = AF_INET;
 148         sin.sin_addr.s_addr = iph->daddr;
 149         sin.sin_port = ports[1];
 150         memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 151
 152         put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
 153 }
 154
 155 void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb,
 156                          int offset)
 157 {
 158         struct inet_sock *inet = inet_sk(skb->sk);
 159         unsigned int flags = inet->cmsg_flags;
 160
 161         /* Ordered by supposed usage frequency */
 162         if (flags & IP_CMSG_PKTINFO) {
 163                 ip_cmsg_recv_pktinfo(msg, skb);
 164
 165                 flags &= ~IP_CMSG_PKTINFO;
 166                 if (!flags)
 167                         return;
 168         }
 169
 170         if (flags & IP_CMSG_TTL) {
 171                 ip_cmsg_recv_ttl(msg, skb);
 172
 173                 flags &= ~IP_CMSG_TTL;
 174                 if (!flags)
 175                         return;
 176         }
 177
 178         if (flags & IP_CMSG_TOS) {
 179                 ip_cmsg_recv_tos(msg, skb);
 180
 181                 flags &= ~IP_CMSG_TOS;
 182                 if (!flags)
 183                         return;
 184         }
 185
 186         if (flags & IP_CMSG_RECVOPTS) {
 187                 ip_cmsg_recv_opts(msg, skb);
 188
 189                 flags &= ~IP_CMSG_RECVOPTS;
 190                 if (!flags)
 191                         return;
 192         }
 193
 194         if (flags & IP_CMSG_RETOPTS) {
 195                 ip_cmsg_recv_retopts(msg, skb);
 196
 197                 flags &= ~IP_CMSG_RETOPTS;
 198                 if (!flags)
 199                         return;
 200         }
 201
 202         if (flags & IP_CMSG_PASSSEC) {
 203                 ip_cmsg_recv_security(msg, skb);
 204
 205                 flags &= ~IP_CMSG_PASSSEC;
 206                 if (!flags)
 207                         return;
 208         }
 209
 210         if (flags & IP_CMSG_ORIGDSTADDR) {
 211                 ip_cmsg_recv_dstaddr(msg, skb);
 212
 213                 flags &= ~IP_CMSG_ORIGDSTADDR;
 214                 if (!flags)
 215                         return;
 216         }
 217
 218         if (flags & IP_CMSG_CHECKSUM)
 219                 ip_cmsg_recv_checksum(msg, skb, offset);
 220 }
 221 EXPORT_SYMBOL(ip_cmsg_recv_offset);
 222
 223 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc,
 224                  bool allow_ipv6)
 225 {
 226         int err, val;
 227         struct cmsghdr *cmsg;
 228         struct net *net = sock_net(sk);
 229
 230         for_each_cmsghdr(cmsg, msg) {
 231                 if (!CMSG_OK(msg, cmsg))
 232                         return -EINVAL;
 233 #if IS_ENABLED(CONFIG_IPV6)
 234                 if (allow_ipv6 &&
 235                     cmsg->cmsg_level == SOL_IPV6 &&
 236                     cmsg->cmsg_type == IPV6_PKTINFO) {
 237                         struct in6_pktinfo *src_info;
 238
 239                         if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info)))
 240                                 return -EINVAL;
 241                         src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg);
 242                         if (!ipv6_addr_v4mapped(&src_info->ipi6_addr))
 243                                 return -EINVAL;
 244                         ipc->oif = src_info->ipi6_ifindex;
 245                         ipc->addr = src_info->ipi6_addr.s6_addr32[3];
 246                         continue;
 247                 }
 248 #endif
 249                 if (cmsg->cmsg_level == SOL_SOCKET) {
 250                         err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc);
 251                         if (err)
 252                                 return err;
 253                         continue;
 254                 }
 255
 256                 if (cmsg->cmsg_level != SOL_IP)
 257                         continue;
 258                 switch (cmsg->cmsg_type) {
 259                 case IP_RETOPTS:
 260                         err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
 261
 262                         /* Our caller is responsible for freeing ipc->opt */
 263                         err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
 264                                              err < 40 ? err : 40);
 265                         if (err)
 266                                 return err;
 267                         break;
 268                 case IP_PKTINFO:
 269                 {
 270                         struct in_pktinfo *info;
 271                         if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
 272                                 return -EINVAL;
 273                         info = (struct in_pktinfo *)CMSG_DATA(cmsg);
 274                         ipc->oif = info->ipi_ifindex;
 275                         ipc->addr = info->ipi_spec_dst.s_addr;
 276                         break;
 277                 }
 278                 case IP_TTL:
 279                         if (cmsg->cmsg_len != CMSG_LEN(sizeof(int)))
 280                                 return -EINVAL;
 281                         val = *(int *)CMSG_DATA(cmsg);
 282                         if (val < 1 || val > 255)
 283                                 return -EINVAL;
 284                         ipc->ttl = val;
 285                         break;
 286                 case IP_TOS:
 287                         if (cmsg->cmsg_len == CMSG_LEN(sizeof(int)))
 288                                 val = *(int *)CMSG_DATA(cmsg);
 289                         else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8)))
 290                                 val = *(u8 *)CMSG_DATA(cmsg);
 291                         else
 292                                 return -EINVAL;
 293                         if (val < 0 || val > 255)
 294                                 return -EINVAL;
 295                         ipc->tos = val;
 296                         ipc->priority = rt_tos2priority(ipc->tos);
 297                         break;
 298
 299                 default:
 300                         return -EINVAL;
 301                 }
 302         }
 303         return 0;
 304 }
 305
 306
 307 /* Special input handler for packets caught by router alert option.
 308    They are selected only by protocol field, and then processed likely
 309    local ones; but only if someone wants them! Otherwise, router
 310    not running rsvpd will kill RSVP.
 311
 312    It is user level problem, what it will make with them.
 313    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
 314    but receiver should be enough clever f.e. to forward mtrace requests,
 315    sent to multicast group to reach destination designated router.
 316  */
 317 struct ip_ra_chain __rcu *ip_ra_chain;
 318 static DEFINE_SPINLOCK(ip_ra_lock);
 319
 320
 321 static void ip_ra_destroy_rcu(struct rcu_head *head)
 322 {
 323         struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu);
 324
 325         sock_put(ra->saved_sk);
 326         kfree(ra);
 327 }
 328
 329 int ip_ra_control(struct sock *sk, unsigned char on,
 330                   void (*destructor)(struct sock *))
 331 {
 332         struct ip_ra_chain *ra, *new_ra;
 333         struct ip_ra_chain __rcu **rap;
 334
 335         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW)
 336                 return -EINVAL;
 337
 338         new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
 339
 340         spin_lock_bh(&ip_ra_lock);
 341         for (rap = &ip_ra_chain;
 342              (ra = rcu_dereference_protected(*rap,
 343                         lockdep_is_held(&ip_ra_lock))) != NULL;
 344              rap = &ra->next) {
 345                 if (ra->sk == sk) {
 346                         if (on) {
 347                                 spin_unlock_bh(&ip_ra_lock);
 348                                 kfree(new_ra);
 349                                 return -EADDRINUSE;
 350                         }
 351                         /* dont let ip_call_ra_chain() use sk again */
 352                         ra->sk = NULL;
 353                         RCU_INIT_POINTER(*rap, ra->next);
 354                         spin_unlock_bh(&ip_ra_lock);
 355
 356                         if (ra->destructor)
 357                                 ra->destructor(sk);
 358                         /*
 359                          * Delay sock_put(sk) and kfree(ra) after one rcu grace
 360                          * period. This guarantee ip_call_ra_chain() dont need
 361                          * to mess with socket refcounts.
 362                          */
 363                         ra->saved_sk = sk;
 364                         call_rcu(&ra->rcu, ip_ra_destroy_rcu);
 365                         return 0;
 366                 }
 367         }
 368         if (!new_ra) {
 369                 spin_unlock_bh(&ip_ra_lock);
 370                 return -ENOBUFS;
 371         }
 372         new_ra->sk = sk;
 373         new_ra->destructor = destructor;
 374
 375         RCU_INIT_POINTER(new_ra->next, ra);
 376         rcu_assign_pointer(*rap, new_ra);
 377         sock_hold(sk);
 378         spin_unlock_bh(&ip_ra_lock);
 379
 380         return 0;
 381 }
 382
 383 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
 384                    __be16 port, u32 info, u8 *payload)
 385 {
 386         struct sock_exterr_skb *serr;
 387
 388         skb = skb_clone(skb, GFP_ATOMIC);
 389         if (!skb)
 390                 return;
 391
 392         serr = SKB_EXT_ERR(skb);
 393         serr->ee.ee_errno = err;
 394         serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
 395         serr->ee.ee_type = icmp_hdr(skb)->type;
 396         serr->ee.ee_code = icmp_hdr(skb)->code;
 397         serr->ee.ee_pad = 0;
 398         serr->ee.ee_info = info;
 399         serr->ee.ee_data = 0;
 400         serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
 401                                    skb_network_header(skb);
 402         serr->port = port;
 403
 404         if (skb_pull(skb, payload - skb->data)) {
 405                 skb_reset_transport_header(skb);
 406                 if (sock_queue_err_skb(sk, skb) == 0)
 407                         return;
 408         }
 409         kfree_skb(skb);
 410 }
 411
 412 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
 413 {
 414         struct inet_sock *inet = inet_sk(sk);
 415         struct sock_exterr_skb *serr;
 416         struct iphdr *iph;
 417         struct sk_buff *skb;
 418
 419         if (!inet->recverr)
 420                 return;
 421
 422         skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
 423         if (!skb)
 424                 return;
 425
 426         skb_put(skb, sizeof(struct iphdr));
 427         skb_reset_network_header(skb);
 428         iph = ip_hdr(skb);
 429         iph->daddr = daddr;
 430
 431         serr = SKB_EXT_ERR(skb);
 432         serr->ee.ee_errno = err;
 433         serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
 434         serr->ee.ee_type = 0;
 435         serr->ee.ee_code = 0;
 436         serr->ee.ee_pad = 0;
 437         serr->ee.ee_info = info;
 438         serr->ee.ee_data = 0;
 439         serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
 440         serr->port = port;
 441
 442         __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
 443         skb_reset_transport_header(skb);
 444
 445         if (sock_queue_err_skb(sk, skb))
 446                 kfree_skb(skb);
 447 }
 448
 449 /* For some errors we have valid addr_offset even with zero payload and
 450  * zero port. Also, addr_offset should be supported if port is set.
 451  */
 452 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr)
 453 {
 454         return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP ||
 455                serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port;
 456 }
 457
 458 /* IPv4 supports cmsg on all imcp errors and some timestamps
 459  *
 460  * Timestamp code paths do not initialize the fields expected by cmsg:
 461  * the PKTINFO fields in skb->cb[]. Fill those in here.
 462  */
 463 static bool ipv4_datagram_support_cmsg(const struct sock *sk,
 464                                        struct sk_buff *skb,
 465                                        int ee_origin)
 466 {
 467         struct in_pktinfo *info;
 468
 469         if (ee_origin == SO_EE_ORIGIN_ICMP)
 470                 return true;
 471
 472         if (ee_origin == SO_EE_ORIGIN_LOCAL)
 473                 return false;
 474
 475         /* Support IP_PKTINFO on tstamp packets if requested, to correlate
 476          * timestamp with egress dev. Not possible for packets without dev
 477          * or without payload (SOF_TIMESTAMPING_OPT_TSONLY).
 478          */
 479         if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) ||
 480             (!skb->dev))
 481                 return false;
 482
 483         info = PKTINFO_SKB_CB(skb);
 484         info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr;
 485         info->ipi_ifindex = skb->dev->ifindex;
 486         return true;
 487 }
 488
 489 /*
 490  *      Handle MSG_ERRQUEUE
 491  */
 492 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len)
 493 {
 494         struct sock_exterr_skb *serr;
 495         struct sk_buff *skb;
 496         DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 497         struct {
 498                 struct sock_extended_err ee;
 499                 struct sockaddr_in       offender;
 500         } errhdr;
 501         int err;
 502         int copied;
 503
 504         WARN_ON_ONCE(sk->sk_family == AF_INET6);
 505
 506         err = -EAGAIN;
 507         skb = sock_dequeue_err_skb(sk);
 508         if (!skb)
 509                 goto out;
 510
 511         copied = skb->len;
 512         if (copied > len) {
 513                 msg->msg_flags |= MSG_TRUNC;
 514                 copied = len;
 515         }
 516         err = skb_copy_datagram_msg(skb, 0, msg, copied);
 517         if (unlikely(err)) {
 518                 kfree_skb(skb);
 519                 return err;
 520         }
 521         sock_recv_timestamp(msg, sk, skb);
 522
 523         serr = SKB_EXT_ERR(skb);
 524
 525         if (sin && ipv4_datagram_support_addr(serr)) {
 526                 sin->sin_family = AF_INET;
 527                 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
 528                                                    serr->addr_offset);
 529                 sin->sin_port = serr->port;
 530                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 531                 *addr_len = sizeof(*sin);
 532         }
 533
 534         memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
 535         sin = &errhdr.offender;
 536         memset(sin, 0, sizeof(*sin));
 537
 538         if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) {
 539                 sin->sin_family = AF_INET;
 540                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 541                 if (inet_sk(sk)->cmsg_flags)
 542                         ip_cmsg_recv(msg, skb);
 543         }
 544
 545         put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
 546
 547         /* Now we could try to dump offended packet options */
 548
 549         msg->msg_flags |= MSG_ERRQUEUE;
 550         err = copied;
 551
 552         consume_skb(skb);
 553 out:
 554         return err;
 555 }
 556
 557
 558 /*
 559  *      Socket option code for IP. This is the end of the line after any
 560  *      TCP,UDP etc options on an IP socket.
 561  */
 562 static bool setsockopt_needs_rtnl(int optname)
 563 {
 564         switch (optname) {
 565         case IP_ADD_MEMBERSHIP:
 566         case IP_ADD_SOURCE_MEMBERSHIP:
 567         case IP_BLOCK_SOURCE:
 568         case IP_DROP_MEMBERSHIP:
 569         case IP_DROP_SOURCE_MEMBERSHIP:
 570         case IP_MSFILTER:
 571         case IP_UNBLOCK_SOURCE:
 572         case MCAST_BLOCK_SOURCE:
 573         case MCAST_MSFILTER:
 574         case MCAST_JOIN_GROUP:
 575         case MCAST_JOIN_SOURCE_GROUP:
 576         case MCAST_LEAVE_GROUP:
 577         case MCAST_LEAVE_SOURCE_GROUP:
 578         case MCAST_UNBLOCK_SOURCE:
 579                 return true;
 580         }
 581         return false;
 582 }
 583
 584 static int do_ip_setsockopt(struct sock *sk, int level,
 585                             int optname, char __user *optval, unsigned int optlen)
 586 {
 587         struct inet_sock *inet = inet_sk(sk);
 588         struct net *net = sock_net(sk);
 589         int val = 0, err;
 590         bool needs_rtnl = setsockopt_needs_rtnl(optname);
 591
 592         switch (optname) {
 593         case IP_PKTINFO:
 594         case IP_RECVTTL:
 595         case IP_RECVOPTS:
 596         case IP_RECVTOS:
 597         case IP_RETOPTS:
 598         case IP_TOS:
 599         case IP_TTL:
 600         case IP_HDRINCL:
 601         case IP_MTU_DISCOVER:
 602         case IP_RECVERR:
 603         case IP_ROUTER_ALERT:
 604         case IP_FREEBIND:
 605         case IP_PASSSEC:
 606         case IP_TRANSPARENT:
 607         case IP_MINTTL:
 608         case IP_NODEFRAG:
 609         case IP_BIND_ADDRESS_NO_PORT:
 610         case IP_UNICAST_IF:
 611         case IP_MULTICAST_TTL:
 612         case IP_MULTICAST_ALL:
 613         case IP_MULTICAST_LOOP:
 614         case IP_RECVORIGDSTADDR:
 615         case IP_CHECKSUM:
 616                 if (optlen >= sizeof(int)) {
 617                         if (get_user(val, (int __user *) optval))
 618                                 return -EFAULT;
 619                 } else if (optlen >= sizeof(char)) {
 620                         unsigned char ucval;
 621
 622                         if (get_user(ucval, (unsigned char __user *) optval))
 623                                 return -EFAULT;
 624                         val = (int) ucval;
 625                 }
 626         }
 627
 628         /* If optlen==0, it is equivalent to val == 0 */
 629
 630         if (ip_mroute_opt(optname))
 631                 return ip_mroute_setsockopt(sk, optname, optval, optlen);
 632
 633         err = 0;
 634         if (needs_rtnl)
 635                 rtnl_lock();
 636         lock_sock(sk);
 637
 638         switch (optname) {
 639         case IP_OPTIONS:
 640         {
 641                 struct ip_options_rcu *old, *opt = NULL;
 642
 643                 if (optlen > 40)
 644                         goto e_inval;
 645                 err = ip_options_get_from_user(sock_net(sk), &opt,
 646                                                optval, optlen);
 647                 if (err)
 648                         break;
 649                 old = rcu_dereference_protected(inet->inet_opt,
 650                                                 lockdep_sock_is_held(sk));
 651                 if (inet->is_icsk) {
 652                         struct inet_connection_sock *icsk = inet_csk(sk);
 653 #if IS_ENABLED(CONFIG_IPV6)
 654                         if (sk->sk_family == PF_INET ||
 655                             (!((1 << sk->sk_state) &
 656                                (TCPF_LISTEN | TCPF_CLOSE)) &&
 657                              inet->inet_daddr != LOOPBACK4_IPV6)) {
 658 #endif
 659                                 if (old)
 660                                         icsk->icsk_ext_hdr_len -= old->opt.optlen;
 661                                 if (opt)
 662                                         icsk->icsk_ext_hdr_len += opt->opt.optlen;
 663                                 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 664 #if IS_ENABLED(CONFIG_IPV6)
 665                         }
 666 #endif
 667                 }
 668                 rcu_assign_pointer(inet->inet_opt, opt);
 669                 if (old)
 670                         kfree_rcu(old, rcu);
 671                 break;
 672         }
 673         case IP_PKTINFO:
 674                 if (val)
 675                         inet->cmsg_flags |= IP_CMSG_PKTINFO;
 676                 else
 677                         inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
 678                 break;
 679         case IP_RECVTTL:
 680                 if (val)
 681                         inet->cmsg_flags |=  IP_CMSG_TTL;
 682                 else
 683                         inet->cmsg_flags &= ~IP_CMSG_TTL;
 684                 break;
 685         case IP_RECVTOS:
 686                 if (val)
 687                         inet->cmsg_flags |=  IP_CMSG_TOS;
 688                 else
 689                         inet->cmsg_flags &= ~IP_CMSG_TOS;
 690                 break;
 691         case IP_RECVOPTS:
 692                 if (val)
 693                         inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
 694                 else
 695                         inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
 696                 break;
 697         case IP_RETOPTS:
 698                 if (val)
 699                         inet->cmsg_flags |= IP_CMSG_RETOPTS;
 700                 else
 701                         inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
 702                 break;
 703         case IP_PASSSEC:
 704                 if (val)
 705                         inet->cmsg_flags |= IP_CMSG_PASSSEC;
 706                 else
 707                         inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
 708                 break;
 709         case IP_RECVORIGDSTADDR:
 710                 if (val)
 711                         inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
 712                 else
 713                         inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
 714                 break;
 715         case IP_CHECKSUM:
 716                 if (val) {
 717                         if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) {
 718                                 inet_inc_convert_csum(sk);
 719                                 inet->cmsg_flags |= IP_CMSG_CHECKSUM;
 720                         }
 721                 } else {
 722                         if (inet->cmsg_flags & IP_CMSG_CHECKSUM) {
 723                                 inet_dec_convert_csum(sk);
 724                                 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM;
 725                         }
 726                 }
 727                 break;
 728         case IP_TOS:    /* This sets both TOS and Precedence */
 729                 if (sk->sk_type == SOCK_STREAM) {
 730                         val &= ~INET_ECN_MASK;
 731                         val |= inet->tos & INET_ECN_MASK;
 732                 }
 733                 if (inet->tos != val) {
 734                         inet->tos = val;
 735                         sk->sk_priority = rt_tos2priority(val);
 736                         sk_dst_reset(sk);
 737                 }
 738                 break;
 739         case IP_TTL:
 740                 if (optlen < 1)
 741                         goto e_inval;
 742                 if (val != -1 && (val < 1 || val > 255))
 743                         goto e_inval;
 744                 inet->uc_ttl = val;
 745                 break;
 746         case IP_HDRINCL:
 747                 if (sk->sk_type != SOCK_RAW) {
 748                         err = -ENOPROTOOPT;
 749                         break;
 750                 }
 751                 inet->hdrincl = val ? 1 : 0;
 752                 break;
 753         case IP_NODEFRAG:
 754                 if (sk->sk_type != SOCK_RAW) {
 755                         err = -ENOPROTOOPT;
 756                         break;
 757                 }
 758                 inet->nodefrag = val ? 1 : 0;
 759                 break;
 760         case IP_BIND_ADDRESS_NO_PORT:
 761                 inet->bind_address_no_port = val ? 1 : 0;
 762                 break;
 763         case IP_MTU_DISCOVER:
 764                 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT)
 765                         goto e_inval;
 766                 inet->pmtudisc = val;
 767                 break;
 768         case IP_RECVERR:
 769                 inet->recverr = !!val;
 770                 if (!val)
 771                         skb_queue_purge(&sk->sk_error_queue);
 772                 break;
 773         case IP_MULTICAST_TTL:
 774                 if (sk->sk_type == SOCK_STREAM)
 775                         goto e_inval;
 776                 if (optlen < 1)
 777                         goto e_inval;
 778                 if (val == -1)
 779                         val = 1;
 780                 if (val < 0 || val > 255)
 781                         goto e_inval;
 782                 inet->mc_ttl = val;
 783                 break;
 784         case IP_MULTICAST_LOOP:
 785                 if (optlen < 1)
 786                         goto e_inval;
 787                 inet->mc_loop = !!val;
 788                 break;
 789         case IP_UNICAST_IF:
 790         {
 791                 struct net_device *dev = NULL;
 792                 int ifindex;
 793
 794                 if (optlen != sizeof(int))
 795                         goto e_inval;
 796
 797                 ifindex = (__force int)ntohl((__force __be32)val);
 798                 if (ifindex == 0) {
 799                         inet->uc_index = 0;
 800                         err = 0;
 801                         break;
 802                 }
 803
 804                 dev = dev_get_by_index(sock_net(sk), ifindex);
 805                 err = -EADDRNOTAVAIL;
 806                 if (!dev)
 807                         break;
 808                 dev_put(dev);
 809
 810                 err = -EINVAL;
 811                 if (sk->sk_bound_dev_if)
 812                         break;
 813
 814                 inet->uc_index = ifindex;
 815                 err = 0;
 816                 break;
 817         }
 818         case IP_MULTICAST_IF:
 819         {
 820                 struct ip_mreqn mreq;
 821                 struct net_device *dev = NULL;
 822
 823                 if (sk->sk_type == SOCK_STREAM)
 824                         goto e_inval;
 825                 /*
 826                  *      Check the arguments are allowable
 827                  */
 828
 829                 if (optlen < sizeof(struct in_addr))
 830                         goto e_inval;
 831
 832                 err = -EFAULT;
 833                 if (optlen >= sizeof(struct ip_mreqn)) {
 834                         if (copy_from_user(&mreq, optval, sizeof(mreq)))
 835                                 break;
 836                 } else {
 837                         memset(&mreq, 0, sizeof(mreq));
 838                         if (optlen >= sizeof(struct ip_mreq)) {
 839                                 if (copy_from_user(&mreq, optval,
 840                                                    sizeof(struct ip_mreq)))
 841                                         break;
 842                         } else if (optlen >= sizeof(struct in_addr)) {
 843                                 if (copy_from_user(&mreq.imr_address, optval,
 844                                                    sizeof(struct in_addr)))
 845                                         break;
 846                         }
 847                 }
 848
 849                 if (!mreq.imr_ifindex) {
 850                         if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
 851                                 inet->mc_index = 0;
 852                                 inet->mc_addr  = 0;
 853                                 err = 0;
 854                                 break;
 855                         }
 856                         dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
 857                         if (dev)
 858                                 mreq.imr_ifindex = dev->ifindex;
 859                 } else
 860                         dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
 861
 862
 863                 err = -EADDRNOTAVAIL;
 864                 if (!dev)
 865                         break;
 866                 dev_put(dev);
 867
 868                 err = -EINVAL;
 869                 if (sk->sk_bound_dev_if &&
 870                     mreq.imr_ifindex != sk->sk_bound_dev_if)
 871                         break;
 872
 873                 inet->mc_index = mreq.imr_ifindex;
 874                 inet->mc_addr  = mreq.imr_address.s_addr;
 875                 err = 0;
 876                 break;
 877         }
 878
 879         case IP_ADD_MEMBERSHIP:
 880         case IP_DROP_MEMBERSHIP:
 881         {
 882                 struct ip_mreqn mreq;
 883
 884                 err = -EPROTO;
 885                 if (inet_sk(sk)->is_icsk)
 886                         break;
 887
 888                 if (optlen < sizeof(struct ip_mreq))
 889                         goto e_inval;
 890                 err = -EFAULT;
 891                 if (optlen >= sizeof(struct ip_mreqn)) {
 892                         if (copy_from_user(&mreq, optval, sizeof(mreq)))
 893                                 break;
 894                 } else {
 895                         memset(&mreq, 0, sizeof(mreq));
 896                         if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
 897                                 break;
 898                 }
 899
 900                 if (optname == IP_ADD_MEMBERSHIP)
 901                         err = ip_mc_join_group(sk, &mreq);
 902                 else
 903                         err = ip_mc_leave_group(sk, &mreq);
 904                 break;
 905         }
 906         case IP_MSFILTER:
 907         {
 908                 struct ip_msfilter *msf;
 909
 910                 if (optlen < IP_MSFILTER_SIZE(0))
 911                         goto e_inval;
 912                 if (optlen > sysctl_optmem_max) {
 913                         err = -ENOBUFS;
 914                         break;
 915                 }
 916                 msf = kmalloc(optlen, GFP_KERNEL);
 917                 if (!msf) {
 918                         err = -ENOBUFS;
 919                         break;
 920                 }
 921                 err = -EFAULT;
 922                 if (copy_from_user(msf, optval, optlen)) {
 923                         kfree(msf);
 924                         break;
 925                 }
 926                 /* numsrc >= (1G-4) overflow in 32 bits */
 927                 if (msf->imsf_numsrc >= 0x3ffffffcU ||
 928                     msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
 929                         kfree(msf);
 930                         err = -ENOBUFS;
 931                         break;
 932                 }
 933                 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
 934                         kfree(msf);
 935                         err = -EINVAL;
 936                         break;
 937                 }
 938                 err = ip_mc_msfilter(sk, msf, 0);
 939                 kfree(msf);
 940                 break;
 941         }
 942         case IP_BLOCK_SOURCE:
 943         case IP_UNBLOCK_SOURCE:
 944         case IP_ADD_SOURCE_MEMBERSHIP:
 945         case IP_DROP_SOURCE_MEMBERSHIP:
 946         {
 947                 struct ip_mreq_source mreqs;
 948                 int omode, add;
 949
 950                 if (optlen != sizeof(struct ip_mreq_source))
 951                         goto e_inval;
 952                 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
 953                         err = -EFAULT;
 954                         break;
 955                 }
 956                 if (optname == IP_BLOCK_SOURCE) {
 957                         omode = MCAST_EXCLUDE;
 958                         add = 1;
 959                 } else if (optname == IP_UNBLOCK_SOURCE) {
 960                         omode = MCAST_EXCLUDE;
 961                         add = 0;
 962                 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
 963                         struct ip_mreqn mreq;
 964
 965                         mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
 966                         mreq.imr_address.s_addr = mreqs.imr_interface;
 967                         mreq.imr_ifindex = 0;
 968                         err = ip_mc_join_group(sk, &mreq);
 969                         if (err && err != -EADDRINUSE)
 970                                 break;
 971                         omode = MCAST_INCLUDE;
 972                         add = 1;
 973                 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
 974                         omode = MCAST_INCLUDE;
 975                         add = 0;
 976                 }
 977                 err = ip_mc_source(add, omode, sk, &mreqs, 0);
 978                 break;
 979         }
 980         case MCAST_JOIN_GROUP:
 981         case MCAST_LEAVE_GROUP:
 982         {
 983                 struct group_req greq;
 984                 struct sockaddr_in *psin;
 985                 struct ip_mreqn mreq;
 986
 987                 if (optlen < sizeof(struct group_req))
 988                         goto e_inval;
 989                 err = -EFAULT;
 990                 if (copy_from_user(&greq, optval, sizeof(greq)))
 991                         break;
 992                 psin = (struct sockaddr_in *)&greq.gr_group;
 993                 if (psin->sin_family != AF_INET)
 994                         goto e_inval;
 995                 memset(&mreq, 0, sizeof(mreq));
 996                 mreq.imr_multiaddr = psin->sin_addr;
 997                 mreq.imr_ifindex = greq.gr_interface;
 998
 999                 if (optname == MCAST_JOIN_GROUP)
1000                         err = ip_mc_join_group(sk, &mreq);
1001                 else
1002                         err = ip_mc_leave_group(sk, &mreq);
1003                 break;
1004         }
1005         case MCAST_JOIN_SOURCE_GROUP:
1006         case MCAST_LEAVE_SOURCE_GROUP:
1007         case MCAST_BLOCK_SOURCE:
1008         case MCAST_UNBLOCK_SOURCE:
1009         {
1010                 struct group_source_req greqs;
1011                 struct ip_mreq_source mreqs;
1012                 struct sockaddr_in *psin;
1013                 int omode, add;
1014
1015                 if (optlen != sizeof(struct group_source_req))
1016                         goto e_inval;
1017                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
1018                         err = -EFAULT;
1019                         break;
1020                 }
1021                 if (greqs.gsr_group.ss_family != AF_INET ||
1022                     greqs.gsr_source.ss_family != AF_INET) {
1023                         err = -EADDRNOTAVAIL;
1024                         break;
1025                 }
1026                 psin = (struct sockaddr_in *)&greqs.gsr_group;
1027                 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
1028                 psin = (struct sockaddr_in *)&greqs.gsr_source;
1029                 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
1030                 mreqs.imr_interface = 0; /* use index for mc_source */
1031
1032                 if (optname == MCAST_BLOCK_SOURCE) {
1033                         omode = MCAST_EXCLUDE;
1034                         add = 1;
1035                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
1036                         omode = MCAST_EXCLUDE;
1037                         add = 0;
1038                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
1039                         struct ip_mreqn mreq;
1040
1041                         psin = (struct sockaddr_in *)&greqs.gsr_group;
1042                         mreq.imr_multiaddr = psin->sin_addr;
1043                         mreq.imr_address.s_addr = 0;
1044                         mreq.imr_ifindex = greqs.gsr_interface;
1045                         err = ip_mc_join_group(sk, &mreq);
1046                         if (err && err != -EADDRINUSE)
1047                                 break;
1048                         greqs.gsr_interface = mreq.imr_ifindex;
1049                         omode = MCAST_INCLUDE;
1050                         add = 1;
1051                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
1052                         omode = MCAST_INCLUDE;
1053                         add = 0;
1054                 }
1055                 err = ip_mc_source(add, omode, sk, &mreqs,
1056                                    greqs.gsr_interface);
1057                 break;
1058         }
1059         case MCAST_MSFILTER:
1060         {
1061                 struct sockaddr_in *psin;
1062                 struct ip_msfilter *msf = NULL;
1063                 struct group_filter *gsf = NULL;
1064                 int msize, i, ifindex;
1065
1066                 if (optlen < GROUP_FILTER_SIZE(0))
1067                         goto e_inval;
1068                 if (optlen > sysctl_optmem_max) {
1069                         err = -ENOBUFS;
1070                         break;
1071                 }
1072                 gsf = kmalloc(optlen, GFP_KERNEL);
1073                 if (!gsf) {
1074                         err = -ENOBUFS;
1075                         break;
1076                 }
1077                 err = -EFAULT;
1078                 if (copy_from_user(gsf, optval, optlen))
1079                         goto mc_msf_out;
1080
1081                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
1082                 if (gsf->gf_numsrc >= 0x1ffffff ||
1083                     gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) {
1084                         err = -ENOBUFS;
1085                         goto mc_msf_out;
1086                 }
1087                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
1088                         err = -EINVAL;
1089                         goto mc_msf_out;
1090                 }
1091                 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
1092                 msf = kmalloc(msize, GFP_KERNEL);
1093                 if (!msf) {
1094                         err = -ENOBUFS;
1095                         goto mc_msf_out;
1096                 }
1097                 ifindex = gsf->gf_interface;
1098                 psin = (struct sockaddr_in *)&gsf->gf_group;
1099                 if (psin->sin_family != AF_INET) {
1100                         err = -EADDRNOTAVAIL;
1101                         goto mc_msf_out;
1102                 }
1103                 msf->imsf_multiaddr = psin->sin_addr.s_addr;
1104                 msf->imsf_interface = 0;
1105                 msf->imsf_fmode = gsf->gf_fmode;
1106                 msf->imsf_numsrc = gsf->gf_numsrc;
1107                 err = -EADDRNOTAVAIL;
1108                 for (i = 0; i < gsf->gf_numsrc; ++i) {
1109                         psin = (struct sockaddr_in *)&gsf->gf_slist[i];
1110
1111                         if (psin->sin_family != AF_INET)
1112                                 goto mc_msf_out;
1113                         msf->imsf_slist[i] = psin->sin_addr.s_addr;
1114                 }
1115                 kfree(gsf);
1116                 gsf = NULL;
1117
1118                 err = ip_mc_msfilter(sk, msf, ifindex);
1119 mc_msf_out:
1120                 kfree(msf);
1121                 kfree(gsf);
1122                 break;
1123         }
1124         case IP_MULTICAST_ALL:
1125                 if (optlen < 1)
1126                         goto e_inval;
1127                 if (val != 0 && val != 1)
1128                         goto e_inval;
1129                 inet->mc_all = val;
1130                 break;
1131         case IP_ROUTER_ALERT:
1132                 err = ip_ra_control(sk, val ? 1 : 0, NULL);
1133                 break;
1134
1135         case IP_FREEBIND:
1136                 if (optlen < 1)
1137                         goto e_inval;
1138                 inet->freebind = !!val;
1139                 break;
1140
1141         case IP_IPSEC_POLICY:
1142         case IP_XFRM_POLICY:
1143                 err = -EPERM;
1144                 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1145                         break;
1146                 err = xfrm_user_policy(sk, optname, optval, optlen);
1147                 break;
1148
1149         case IP_TRANSPARENT:
1150                 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) &&
1151                     !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1152                         err = -EPERM;
1153                         break;
1154                 }
1155                 if (optlen < 1)
1156                         goto e_inval;
1157                 inet->transparent = !!val;
1158                 break;
1159
1160         case IP_MINTTL:
1161                 if (optlen < 1)
1162                         goto e_inval;
1163                 if (val < 0 || val > 255)
1164                         goto e_inval;
1165                 inet->min_ttl = val;
1166                 break;
1167
1168         default:
1169                 err = -ENOPROTOOPT;
1170                 break;
1171         }
1172         release_sock(sk);
1173         if (needs_rtnl)
1174                 rtnl_unlock();
1175         return err;
1176
1177 e_inval:
1178         release_sock(sk);
1179         if (needs_rtnl)
1180                 rtnl_unlock();
1181         return -EINVAL;
1182 }
1183
1184 /**
1185  * ipv4_pktinfo_prepare - transfer some info from rtable to skb
1186  * @sk: socket
1187  * @skb: buffer
1188  *
1189  * To support IP_CMSG_PKTINFO option, we store rt_iif and specific
1190  * destination in skb->cb[] before dst drop.
1191  * This way, receiver doesn't make cache line misses to read rtable.
1192  */
1193 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb)
1194 {
1195         struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb);
1196         bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) ||
1197                        ipv6_sk_rxinfo(sk);
1198
1199         if (prepare && skb_rtable(skb)) {
1200                 /* skb->cb is overloaded: prior to this point it is IP{6}CB
1201                  * which has interface index (iif) as the first member of the
1202                  * underlying inet{6}_skb_parm struct. This code then overlays
1203                  * PKTINFO_SKB_CB and in_pktinfo also has iif as the first
1204                  * element so the iif is picked up from the prior IPCB
1205                  */
1206                 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb);
1207         } else {
1208                 pktinfo->ipi_ifindex = 0;
1209                 pktinfo->ipi_spec_dst.s_addr = 0;
1210         }
1211         skb_dst_drop(skb);
1212 }
1213
1214 int ip_setsockopt(struct sock *sk, int level,
1215                 int optname, char __user *optval, unsigned int optlen)
1216 {
1217         int err;
1218
1219         if (level != SOL_IP)
1220                 return -ENOPROTOOPT;
1221
1222         err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1223 #ifdef CONFIG_NETFILTER
1224         /* we need to exclude all possible ENOPROTOOPTs except default case */
1225         if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1226                         optname != IP_IPSEC_POLICY &&
1227                         optname != IP_XFRM_POLICY &&
1228                         !ip_mroute_opt(optname)) {
1229                 lock_sock(sk);
1230                 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
1231                 release_sock(sk);
1232         }
1233 #endif
1234         return err;
1235 }
1236 EXPORT_SYMBOL(ip_setsockopt);
1237
1238 #ifdef CONFIG_COMPAT
1239 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
1240                          char __user *optval, unsigned int optlen)
1241 {
1242         int err;
1243
1244         if (level != SOL_IP)
1245                 return -ENOPROTOOPT;
1246
1247         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
1248                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
1249                         ip_setsockopt);
1250
1251         err = do_ip_setsockopt(sk, level, optname, optval, optlen);
1252 #ifdef CONFIG_NETFILTER
1253         /* we need to exclude all possible ENOPROTOOPTs except default case */
1254         if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
1255                         optname != IP_IPSEC_POLICY &&
1256                         optname != IP_XFRM_POLICY &&
1257                         !ip_mroute_opt(optname)) {
1258                 lock_sock(sk);
1259                 err = compat_nf_setsockopt(sk, PF_INET, optname,
1260                                            optval, optlen);
1261                 release_sock(sk);
1262         }
1263 #endif
1264         return err;
1265 }
1266 EXPORT_SYMBOL(compat_ip_setsockopt);
1267 #endif
1268
1269 /*
1270  *      Get the options. Note for future reference. The GET of IP options gets
1271  *      the _received_ ones. The set sets the _sent_ ones.
1272  */
1273
1274 static bool getsockopt_needs_rtnl(int optname)
1275 {
1276         switch (optname) {
1277         case IP_MSFILTER:
1278         case MCAST_MSFILTER:
1279                 return true;
1280         }
1281         return false;
1282 }
1283
1284 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1285                             char __user *optval, int __user *optlen, unsigned int flags)
1286 {
1287         struct inet_sock *inet = inet_sk(sk);
1288         bool needs_rtnl = getsockopt_needs_rtnl(optname);
1289         int val, err = 0;
1290         int len;
1291
1292         if (level != SOL_IP)
1293                 return -EOPNOTSUPP;
1294
1295         if (ip_mroute_opt(optname))
1296                 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1297
1298         if (get_user(len, optlen))
1299                 return -EFAULT;
1300         if (len < 0)
1301                 return -EINVAL;
1302
1303         if (needs_rtnl)
1304                 rtnl_lock();
1305         lock_sock(sk);
1306
1307         switch (optname) {
1308         case IP_OPTIONS:
1309         {
1310                 unsigned char optbuf[sizeof(struct ip_options)+40];
1311                 struct ip_options *opt = (struct ip_options *)optbuf;
1312                 struct ip_options_rcu *inet_opt;
1313
1314                 inet_opt = rcu_dereference_protected(inet->inet_opt,
1315                                                      lockdep_sock_is_held(sk));
1316                 opt->optlen = 0;
1317                 if (inet_opt)
1318                         memcpy(optbuf, &inet_opt->opt,
1319                                sizeof(struct ip_options) +
1320                                inet_opt->opt.optlen);
1321                 release_sock(sk);
1322
1323                 if (opt->optlen == 0)
1324                         return put_user(0, optlen);
1325
1326                 ip_options_undo(opt);
1327
1328                 len = min_t(unsigned int, len, opt->optlen);
1329                 if (put_user(len, optlen))
1330                         return -EFAULT;
1331                 if (copy_to_user(optval, opt->__data, len))
1332                         return -EFAULT;
1333                 return 0;
1334         }
1335         case IP_PKTINFO:
1336                 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1337                 break;
1338         case IP_RECVTTL:
1339                 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1340                 break;
1341         case IP_RECVTOS:
1342                 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1343                 break;
1344         case IP_RECVOPTS:
1345                 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1346                 break;
1347         case IP_RETOPTS:
1348                 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1349                 break;
1350         case IP_PASSSEC:
1351                 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1352                 break;
1353         case IP_RECVORIGDSTADDR:
1354                 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1355                 break;
1356         case IP_CHECKSUM:
1357                 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0;
1358                 break;
1359         case IP_TOS:
1360                 val = inet->tos;
1361                 break;
1362         case IP_TTL:
1363         {
1364                 struct net *net = sock_net(sk);
1365                 val = (inet->uc_ttl == -1 ?
1366                        net->ipv4.sysctl_ip_default_ttl :
1367                        inet->uc_ttl);
1368                 break;
1369         }
1370         case IP_HDRINCL:
1371                 val = inet->hdrincl;
1372                 break;
1373         case IP_NODEFRAG:
1374                 val = inet->nodefrag;
1375                 break;
1376         case IP_BIND_ADDRESS_NO_PORT:
1377                 val = inet->bind_address_no_port;
1378                 break;
1379         case IP_MTU_DISCOVER:
1380                 val = inet->pmtudisc;
1381                 break;
1382         case IP_MTU:
1383         {
1384                 struct dst_entry *dst;
1385                 val = 0;
1386                 dst = sk_dst_get(sk);
1387                 if (dst) {
1388                         val = dst_mtu(dst);
1389                         dst_release(dst);
1390                 }
1391                 if (!val) {
1392                         release_sock(sk);
1393                         return -ENOTCONN;
1394                 }
1395                 break;
1396         }
1397         case IP_RECVERR:
1398                 val = inet->recverr;
1399                 break;
1400         case IP_MULTICAST_TTL:
1401                 val = inet->mc_ttl;
1402                 break;
1403         case IP_MULTICAST_LOOP:
1404                 val = inet->mc_loop;
1405                 break;
1406         case IP_UNICAST_IF:
1407                 val = (__force int)htonl((__u32) inet->uc_index);
1408                 break;
1409         case IP_MULTICAST_IF:
1410         {
1411                 struct in_addr addr;
1412                 len = min_t(unsigned int, len, sizeof(struct in_addr));
1413                 addr.s_addr = inet->mc_addr;
1414                 release_sock(sk);
1415
1416                 if (put_user(len, optlen))
1417                         return -EFAULT;
1418                 if (copy_to_user(optval, &addr, len))
1419                         return -EFAULT;
1420                 return 0;
1421         }
1422         case IP_MSFILTER:
1423         {
1424                 struct ip_msfilter msf;
1425
1426                 if (len < IP_MSFILTER_SIZE(0)) {
1427                         err = -EINVAL;
1428                         goto out;
1429                 }
1430                 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1431                         err = -EFAULT;
1432                         goto out;
1433                 }
1434                 err = ip_mc_msfget(sk, &msf,
1435                                    (struct ip_msfilter __user *)optval, optlen);
1436                 goto out;
1437         }
1438         case MCAST_MSFILTER:
1439         {
1440                 struct group_filter gsf;
1441
1442                 if (len < GROUP_FILTER_SIZE(0)) {
1443                         err = -EINVAL;
1444                         goto out;
1445                 }
1446                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1447                         err = -EFAULT;
1448                         goto out;
1449                 }
1450                 err = ip_mc_gsfget(sk, &gsf,
1451                                    (struct group_filter __user *)optval,
1452                                    optlen);
1453                 goto out;
1454         }
1455         case IP_MULTICAST_ALL:
1456                 val = inet->mc_all;
1457                 break;
1458         case IP_PKTOPTIONS:
1459         {
1460                 struct msghdr msg;
1461
1462                 release_sock(sk);
1463
1464                 if (sk->sk_type != SOCK_STREAM)
1465                         return -ENOPROTOOPT;
1466
1467                 msg.msg_control = (__force void *) optval;
1468                 msg.msg_controllen = len;
1469                 msg.msg_flags = flags;
1470
1471                 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1472                         struct in_pktinfo info;
1473
1474                         info.ipi_addr.s_addr = inet->inet_rcv_saddr;
1475                         info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr;
1476                         info.ipi_ifindex = inet->mc_index;
1477                         put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1478                 }
1479                 if (inet->cmsg_flags & IP_CMSG_TTL) {
1480                         int hlim = inet->mc_ttl;
1481                         put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1482                 }
1483                 if (inet->cmsg_flags & IP_CMSG_TOS) {
1484                         int tos = inet->rcv_tos;
1485                         put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos);
1486                 }
1487                 len -= msg.msg_controllen;
1488                 return put_user(len, optlen);
1489         }
1490         case IP_FREEBIND:
1491                 val = inet->freebind;
1492                 break;
1493         case IP_TRANSPARENT:
1494                 val = inet->transparent;
1495                 break;
1496         case IP_MINTTL:
1497                 val = inet->min_ttl;
1498                 break;
1499         default:
1500                 release_sock(sk);
1501                 return -ENOPROTOOPT;
1502         }
1503         release_sock(sk);
1504
1505         if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1506                 unsigned char ucval = (unsigned char)val;
1507                 len = 1;
1508                 if (put_user(len, optlen))
1509                         return -EFAULT;
1510                 if (copy_to_user(optval, &ucval, 1))
1511                         return -EFAULT;
1512         } else {
1513                 len = min_t(unsigned int, sizeof(int), len);
1514                 if (put_user(len, optlen))
1515                         return -EFAULT;
1516                 if (copy_to_user(optval, &val, len))
1517                         return -EFAULT;
1518         }
1519         return 0;
1520
1521 out:
1522         release_sock(sk);
1523         if (needs_rtnl)
1524                 rtnl_unlock();
1525         return err;
1526 }
1527
1528 int ip_getsockopt(struct sock *sk, int level,
1529                   int optname, char __user *optval, int __user *optlen)
1530 {
1531         int err;
1532
1533         err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0);
1534 #ifdef CONFIG_NETFILTER
1535         /* we need to exclude all possible ENOPROTOOPTs except default case */
1536         if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1537                         !ip_mroute_opt(optname)) {
1538                 int len;
1539
1540                 if (get_user(len, optlen))
1541                         return -EFAULT;
1542
1543                 lock_sock(sk);
1544                 err = nf_getsockopt(sk, PF_INET, optname, optval,
1545                                 &len);
1546                 release_sock(sk);
1547                 if (err >= 0)
1548                         err = put_user(len, optlen);
1549                 return err;
1550         }
1551 #endif
1552         return err;
1553 }
1554 EXPORT_SYMBOL(ip_getsockopt);
1555
1556 #ifdef CONFIG_COMPAT
1557 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1558                          char __user *optval, int __user *optlen)
1559 {
1560         int err;
1561
1562         if (optname == MCAST_MSFILTER)
1563                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1564                         ip_getsockopt);
1565
1566         err = do_ip_getsockopt(sk, level, optname, optval, optlen,
1567                 MSG_CMSG_COMPAT);
1568
1569 #ifdef CONFIG_NETFILTER
1570         /* we need to exclude all possible ENOPROTOOPTs except default case */
1571         if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1572                         !ip_mroute_opt(optname)) {
1573                 int len;
1574
1575                 if (get_user(len, optlen))
1576                         return -EFAULT;
1577
1578                 lock_sock(sk);
1579                 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1580                 release_sock(sk);
1581                 if (err >= 0)
1582                         err = put_user(len, optlen);
1583                 return err;
1584         }
1585 #endif
1586         return err;
1587 }
1588 EXPORT_SYMBOL(compat_ip_getsockopt);
1589 #endif