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 <net/sock.h>
  27 #include <net/ip.h>
  28 #include <net/icmp.h>
  29 #include <net/tcp_states.h>
  30 #include <linux/udp.h>
  31 #include <linux/igmp.h>
  32 #include <linux/netfilter.h>
  33 #include <linux/route.h>
  34 #include <linux/mroute.h>
  35 #include <net/route.h>
  36 #include <net/xfrm.h>
  37 #include <net/compat.h>
  38 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  39 #include <net/transp_v6.h>
  40 #endif
  41
  42 #include <linux/errqueue.h>
  43 #include <asm/uaccess.h>
  44
  45 #define IP_CMSG_PKTINFO         1
  46 #define IP_CMSG_TTL             2
  47 #define IP_CMSG_TOS             4
  48 #define IP_CMSG_RECVOPTS        8
  49 #define IP_CMSG_RETOPTS         16
  50 #define IP_CMSG_PASSSEC         32
  51 #define IP_CMSG_ORIGDSTADDR     64
  52
  53 /*
  54  *      SOL_IP control messages.
  55  */
  56
  57 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb)
  58 {
  59         struct in_pktinfo info;
  60         struct rtable *rt = skb_rtable(skb);
  61
  62         info.ipi_addr.s_addr = ip_hdr(skb)->daddr;
  63         if (rt) {
  64                 info.ipi_ifindex = rt->rt_iif;
  65                 info.ipi_spec_dst.s_addr = rt->rt_spec_dst;
  66         } else {
  67                 info.ipi_ifindex = 0;
  68                 info.ipi_spec_dst.s_addr = 0;
  69         }
  70
  71         put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
  72 }
  73
  74 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb)
  75 {
  76         int ttl = ip_hdr(skb)->ttl;
  77         put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl);
  78 }
  79
  80 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb)
  81 {
  82         put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos);
  83 }
  84
  85 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb)
  86 {
  87         if (IPCB(skb)->opt.optlen == 0)
  88                 return;
  89
  90         put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen,
  91                  ip_hdr(skb) + 1);
  92 }
  93
  94
  95 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb)
  96 {
  97         unsigned char optbuf[sizeof(struct ip_options) + 40];
  98         struct ip_options * opt = (struct ip_options *)optbuf;
  99
 100         if (IPCB(skb)->opt.optlen == 0)
 101                 return;
 102
 103         if (ip_options_echo(opt, skb)) {
 104                 msg->msg_flags |= MSG_CTRUNC;
 105                 return;
 106         }
 107         ip_options_undo(opt);
 108
 109         put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data);
 110 }
 111
 112 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb)
 113 {
 114         char *secdata;
 115         u32 seclen, secid;
 116         int err;
 117
 118         err = security_socket_getpeersec_dgram(NULL, skb, &secid);
 119         if (err)
 120                 return;
 121
 122         err = security_secid_to_secctx(secid, &secdata, &seclen);
 123         if (err)
 124                 return;
 125
 126         put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata);
 127         security_release_secctx(secdata, seclen);
 128 }
 129
 130 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb)
 131 {
 132         struct sockaddr_in sin;
 133         struct iphdr *iph = ip_hdr(skb);
 134         __be16 *ports = (__be16 *)skb_transport_header(skb);
 135
 136         if (skb_transport_offset(skb) + 4 > skb->len)
 137                 return;
 138
 139         /* All current transport protocols have the port numbers in the
 140          * first four bytes of the transport header and this function is
 141          * written with this assumption in mind.
 142          */
 143
 144         sin.sin_family = AF_INET;
 145         sin.sin_addr.s_addr = iph->daddr;
 146         sin.sin_port = ports[1];
 147         memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 148
 149         put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin);
 150 }
 151
 152 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
 153 {
 154         struct inet_sock *inet = inet_sk(skb->sk);
 155         unsigned flags = inet->cmsg_flags;
 156
 157         /* Ordered by supposed usage frequency */
 158         if (flags & 1)
 159                 ip_cmsg_recv_pktinfo(msg, skb);
 160         if ((flags >>= 1) == 0)
 161                 return;
 162
 163         if (flags & 1)
 164                 ip_cmsg_recv_ttl(msg, skb);
 165         if ((flags >>= 1) == 0)
 166                 return;
 167
 168         if (flags & 1)
 169                 ip_cmsg_recv_tos(msg, skb);
 170         if ((flags >>= 1) == 0)
 171                 return;
 172
 173         if (flags & 1)
 174                 ip_cmsg_recv_opts(msg, skb);
 175         if ((flags >>= 1) == 0)
 176                 return;
 177
 178         if (flags & 1)
 179                 ip_cmsg_recv_retopts(msg, skb);
 180         if ((flags >>= 1) == 0)
 181                 return;
 182
 183         if (flags & 1)
 184                 ip_cmsg_recv_security(msg, skb);
 185
 186         if ((flags >>= 1) == 0)
 187                 return;
 188         if (flags & 1)
 189                 ip_cmsg_recv_dstaddr(msg, skb);
 190
 191 }
 192 EXPORT_SYMBOL(ip_cmsg_recv);
 193
 194 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
 195 {
 196         int err;
 197         struct cmsghdr *cmsg;
 198
 199         for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
 200                 if (!CMSG_OK(msg, cmsg))
 201                         return -EINVAL;
 202                 if (cmsg->cmsg_level != SOL_IP)
 203                         continue;
 204                 switch (cmsg->cmsg_type) {
 205                 case IP_RETOPTS:
 206                         err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
 207                         err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg),
 208                                              err < 40 ? err : 40);
 209                         if (err)
 210                                 return err;
 211                         break;
 212                 case IP_PKTINFO:
 213                 {
 214                         struct in_pktinfo *info;
 215                         if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
 216                                 return -EINVAL;
 217                         info = (struct in_pktinfo *)CMSG_DATA(cmsg);
 218                         ipc->oif = info->ipi_ifindex;
 219                         ipc->addr = info->ipi_spec_dst.s_addr;
 220                         break;
 221                 }
 222                 default:
 223                         return -EINVAL;
 224                 }
 225         }
 226         return 0;
 227 }
 228
 229
 230 /* Special input handler for packets caught by router alert option.
 231    They are selected only by protocol field, and then processed likely
 232    local ones; but only if someone wants them! Otherwise, router
 233    not running rsvpd will kill RSVP.
 234
 235    It is user level problem, what it will make with them.
 236    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
 237    but receiver should be enough clever f.e. to forward mtrace requests,
 238    sent to multicast group to reach destination designated router.
 239  */
 240 struct ip_ra_chain *ip_ra_chain;
 241 DEFINE_RWLOCK(ip_ra_lock);
 242
 243 int ip_ra_control(struct sock *sk, unsigned char on,
 244                   void (*destructor)(struct sock *))
 245 {
 246         struct ip_ra_chain *ra, *new_ra, **rap;
 247
 248         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num == IPPROTO_RAW)
 249                 return -EINVAL;
 250
 251         new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
 252
 253         write_lock_bh(&ip_ra_lock);
 254         for (rap = &ip_ra_chain; (ra = *rap) != NULL; rap = &ra->next) {
 255                 if (ra->sk == sk) {
 256                         if (on) {
 257                                 write_unlock_bh(&ip_ra_lock);
 258                                 kfree(new_ra);
 259                                 return -EADDRINUSE;
 260                         }
 261                         *rap = ra->next;
 262                         write_unlock_bh(&ip_ra_lock);
 263
 264                         if (ra->destructor)
 265                                 ra->destructor(sk);
 266                         sock_put(sk);
 267                         kfree(ra);
 268                         return 0;
 269                 }
 270         }
 271         if (new_ra == NULL) {
 272                 write_unlock_bh(&ip_ra_lock);
 273                 return -ENOBUFS;
 274         }
 275         new_ra->sk = sk;
 276         new_ra->destructor = destructor;
 277
 278         new_ra->next = ra;
 279         *rap = new_ra;
 280         sock_hold(sk);
 281         write_unlock_bh(&ip_ra_lock);
 282
 283         return 0;
 284 }
 285
 286 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
 287                    __be16 port, u32 info, u8 *payload)
 288 {
 289         struct inet_sock *inet = inet_sk(sk);
 290         struct sock_exterr_skb *serr;
 291
 292         if (!inet->recverr)
 293                 return;
 294
 295         skb = skb_clone(skb, GFP_ATOMIC);
 296         if (!skb)
 297                 return;
 298
 299         serr = SKB_EXT_ERR(skb);
 300         serr->ee.ee_errno = err;
 301         serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
 302         serr->ee.ee_type = icmp_hdr(skb)->type;
 303         serr->ee.ee_code = icmp_hdr(skb)->code;
 304         serr->ee.ee_pad = 0;
 305         serr->ee.ee_info = info;
 306         serr->ee.ee_data = 0;
 307         serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
 308                                    skb_network_header(skb);
 309         serr->port = port;
 310
 311         if (skb_pull(skb, payload - skb->data) != NULL) {
 312                 skb_reset_transport_header(skb);
 313                 if (sock_queue_err_skb(sk, skb) == 0)
 314                         return;
 315         }
 316         kfree_skb(skb);
 317 }
 318
 319 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
 320 {
 321         struct inet_sock *inet = inet_sk(sk);
 322         struct sock_exterr_skb *serr;
 323         struct iphdr *iph;
 324         struct sk_buff *skb;
 325
 326         if (!inet->recverr)
 327                 return;
 328
 329         skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
 330         if (!skb)
 331                 return;
 332
 333         skb_put(skb, sizeof(struct iphdr));
 334         skb_reset_network_header(skb);
 335         iph = ip_hdr(skb);
 336         iph->daddr = daddr;
 337
 338         serr = SKB_EXT_ERR(skb);
 339         serr->ee.ee_errno = err;
 340         serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
 341         serr->ee.ee_type = 0;
 342         serr->ee.ee_code = 0;
 343         serr->ee.ee_pad = 0;
 344         serr->ee.ee_info = info;
 345         serr->ee.ee_data = 0;
 346         serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
 347         serr->port = port;
 348
 349         __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
 350         skb_reset_transport_header(skb);
 351
 352         if (sock_queue_err_skb(sk, skb))
 353                 kfree_skb(skb);
 354 }
 355
 356 /*
 357  *      Handle MSG_ERRQUEUE
 358  */
 359 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len)
 360 {
 361         struct sock_exterr_skb *serr;
 362         struct sk_buff *skb, *skb2;
 363         struct sockaddr_in *sin;
 364         struct {
 365                 struct sock_extended_err ee;
 366                 struct sockaddr_in       offender;
 367         } errhdr;
 368         int err;
 369         int copied;
 370
 371         err = -EAGAIN;
 372         skb = skb_dequeue(&sk->sk_error_queue);
 373         if (skb == NULL)
 374                 goto out;
 375
 376         copied = skb->len;
 377         if (copied > len) {
 378                 msg->msg_flags |= MSG_TRUNC;
 379                 copied = len;
 380         }
 381         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 382         if (err)
 383                 goto out_free_skb;
 384
 385         sock_recv_timestamp(msg, sk, skb);
 386
 387         serr = SKB_EXT_ERR(skb);
 388
 389         sin = (struct sockaddr_in *)msg->msg_name;
 390         if (sin) {
 391                 sin->sin_family = AF_INET;
 392                 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
 393                                                    serr->addr_offset);
 394                 sin->sin_port = serr->port;
 395                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 396         }
 397
 398         memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
 399         sin = &errhdr.offender;
 400         sin->sin_family = AF_UNSPEC;
 401         if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
 402                 struct inet_sock *inet = inet_sk(sk);
 403
 404                 sin->sin_family = AF_INET;
 405                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 406                 sin->sin_port = 0;
 407                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 408                 if (inet->cmsg_flags)
 409                         ip_cmsg_recv(msg, skb);
 410         }
 411
 412         put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
 413
 414         /* Now we could try to dump offended packet options */
 415
 416         msg->msg_flags |= MSG_ERRQUEUE;
 417         err = copied;
 418
 419         /* Reset and regenerate socket error */
 420         spin_lock_bh(&sk->sk_error_queue.lock);
 421         sk->sk_err = 0;
 422         skb2 = skb_peek(&sk->sk_error_queue);
 423         if (skb2 != NULL) {
 424                 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
 425                 spin_unlock_bh(&sk->sk_error_queue.lock);
 426                 sk->sk_error_report(sk);
 427         } else
 428                 spin_unlock_bh(&sk->sk_error_queue.lock);
 429
 430 out_free_skb:
 431         kfree_skb(skb);
 432 out:
 433         return err;
 434 }
 435
 436
 437 /*
 438  *      Socket option code for IP. This is the end of the line after any
 439  *      TCP,UDP etc options on an IP socket.
 440  */
 441
 442 static int do_ip_setsockopt(struct sock *sk, int level,
 443                             int optname, char __user *optval, int optlen)
 444 {
 445         struct inet_sock *inet = inet_sk(sk);
 446         int val = 0, err;
 447
 448         if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) |
 449                              (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) |
 450                              (1<<IP_RETOPTS) | (1<<IP_TOS) |
 451                              (1<<IP_TTL) | (1<<IP_HDRINCL) |
 452                              (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) |
 453                              (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) |
 454                              (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT))) ||
 455             optname == IP_MULTICAST_TTL ||
 456             optname == IP_MULTICAST_ALL ||
 457             optname == IP_MULTICAST_LOOP ||
 458             optname == IP_RECVORIGDSTADDR) {
 459                 if (optlen >= sizeof(int)) {
 460                         if (get_user(val, (int __user *) optval))
 461                                 return -EFAULT;
 462                 } else if (optlen >= sizeof(char)) {
 463                         unsigned char ucval;
 464
 465                         if (get_user(ucval, (unsigned char __user *) optval))
 466                                 return -EFAULT;
 467                         val = (int) ucval;
 468                 }
 469         }
 470
 471         /* If optlen==0, it is equivalent to val == 0 */
 472
 473         if (ip_mroute_opt(optname))
 474                 return ip_mroute_setsockopt(sk, optname, optval, optlen);
 475
 476         err = 0;
 477         lock_sock(sk);
 478
 479         switch (optname) {
 480         case IP_OPTIONS:
 481         {
 482                 struct ip_options *opt = NULL;
 483                 if (optlen > 40 || optlen < 0)
 484                         goto e_inval;
 485                 err = ip_options_get_from_user(sock_net(sk), &opt,
 486                                                optval, optlen);
 487                 if (err)
 488                         break;
 489                 if (inet->is_icsk) {
 490                         struct inet_connection_sock *icsk = inet_csk(sk);
 491 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 492                         if (sk->sk_family == PF_INET ||
 493                             (!((1 << sk->sk_state) &
 494                                (TCPF_LISTEN | TCPF_CLOSE)) &&
 495                              inet->daddr != LOOPBACK4_IPV6)) {
 496 #endif
 497                                 if (inet->opt)
 498                                         icsk->icsk_ext_hdr_len -= inet->opt->optlen;
 499                                 if (opt)
 500                                         icsk->icsk_ext_hdr_len += opt->optlen;
 501                                 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 502 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 503                         }
 504 #endif
 505                 }
 506                 opt = xchg(&inet->opt, opt);
 507                 kfree(opt);
 508                 break;
 509         }
 510         case IP_PKTINFO:
 511                 if (val)
 512                         inet->cmsg_flags |= IP_CMSG_PKTINFO;
 513                 else
 514                         inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
 515                 break;
 516         case IP_RECVTTL:
 517                 if (val)
 518                         inet->cmsg_flags |=  IP_CMSG_TTL;
 519                 else
 520                         inet->cmsg_flags &= ~IP_CMSG_TTL;
 521                 break;
 522         case IP_RECVTOS:
 523                 if (val)
 524                         inet->cmsg_flags |=  IP_CMSG_TOS;
 525                 else
 526                         inet->cmsg_flags &= ~IP_CMSG_TOS;
 527                 break;
 528         case IP_RECVOPTS:
 529                 if (val)
 530                         inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
 531                 else
 532                         inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
 533                 break;
 534         case IP_RETOPTS:
 535                 if (val)
 536                         inet->cmsg_flags |= IP_CMSG_RETOPTS;
 537                 else
 538                         inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
 539                 break;
 540         case IP_PASSSEC:
 541                 if (val)
 542                         inet->cmsg_flags |= IP_CMSG_PASSSEC;
 543                 else
 544                         inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
 545                 break;
 546         case IP_RECVORIGDSTADDR:
 547                 if (val)
 548                         inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR;
 549                 else
 550                         inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR;
 551                 break;
 552         case IP_TOS:    /* This sets both TOS and Precedence */
 553                 if (sk->sk_type == SOCK_STREAM) {
 554                         val &= ~3;
 555                         val |= inet->tos & 3;
 556                 }
 557                 if (inet->tos != val) {
 558                         inet->tos = val;
 559                         sk->sk_priority = rt_tos2priority(val);
 560                         sk_dst_reset(sk);
 561                 }
 562                 break;
 563         case IP_TTL:
 564                 if (optlen < 1)
 565                         goto e_inval;
 566                 if (val != -1 && (val < 0 || val > 255))
 567                         goto e_inval;
 568                 inet->uc_ttl = val;
 569                 break;
 570         case IP_HDRINCL:
 571                 if (sk->sk_type != SOCK_RAW) {
 572                         err = -ENOPROTOOPT;
 573                         break;
 574                 }
 575                 inet->hdrincl = val ? 1 : 0;
 576                 break;
 577         case IP_MTU_DISCOVER:
 578                 if (val < 0 || val > 3)
 579                         goto e_inval;
 580                 inet->pmtudisc = val;
 581                 break;
 582         case IP_RECVERR:
 583                 inet->recverr = !!val;
 584                 if (!val)
 585                         skb_queue_purge(&sk->sk_error_queue);
 586                 break;
 587         case IP_MULTICAST_TTL:
 588                 if (sk->sk_type == SOCK_STREAM)
 589                         goto e_inval;
 590                 if (optlen < 1)
 591                         goto e_inval;
 592                 if (val == -1)
 593                         val = 1;
 594                 if (val < 0 || val > 255)
 595                         goto e_inval;
 596                 inet->mc_ttl = val;
 597                 break;
 598         case IP_MULTICAST_LOOP:
 599                 if (optlen < 1)
 600                         goto e_inval;
 601                 inet->mc_loop = !!val;
 602                 break;
 603         case IP_MULTICAST_IF:
 604         {
 605                 struct ip_mreqn mreq;
 606                 struct net_device *dev = NULL;
 607
 608                 if (sk->sk_type == SOCK_STREAM)
 609                         goto e_inval;
 610                 /*
 611                  *      Check the arguments are allowable
 612                  */
 613
 614                 err = -EFAULT;
 615                 if (optlen >= sizeof(struct ip_mreqn)) {
 616                         if (copy_from_user(&mreq, optval, sizeof(mreq)))
 617                                 break;
 618                 } else {
 619                         memset(&mreq, 0, sizeof(mreq));
 620                         if (optlen >= sizeof(struct in_addr) &&
 621                             copy_from_user(&mreq.imr_address, optval,
 622                                            sizeof(struct in_addr)))
 623                                 break;
 624                 }
 625
 626                 if (!mreq.imr_ifindex) {
 627                         if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
 628                                 inet->mc_index = 0;
 629                                 inet->mc_addr  = 0;
 630                                 err = 0;
 631                                 break;
 632                         }
 633                         dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
 634                         if (dev) {
 635                                 mreq.imr_ifindex = dev->ifindex;
 636                                 dev_put(dev);
 637                         }
 638                 } else
 639                         dev = __dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
 640
 641
 642                 err = -EADDRNOTAVAIL;
 643                 if (!dev)
 644                         break;
 645
 646                 err = -EINVAL;
 647                 if (sk->sk_bound_dev_if &&
 648                     mreq.imr_ifindex != sk->sk_bound_dev_if)
 649                         break;
 650
 651                 inet->mc_index = mreq.imr_ifindex;
 652                 inet->mc_addr  = mreq.imr_address.s_addr;
 653                 err = 0;
 654                 break;
 655         }
 656
 657         case IP_ADD_MEMBERSHIP:
 658         case IP_DROP_MEMBERSHIP:
 659         {
 660                 struct ip_mreqn mreq;
 661
 662                 err = -EPROTO;
 663                 if (inet_sk(sk)->is_icsk)
 664                         break;
 665
 666                 if (optlen < sizeof(struct ip_mreq))
 667                         goto e_inval;
 668                 err = -EFAULT;
 669                 if (optlen >= sizeof(struct ip_mreqn)) {
 670                         if (copy_from_user(&mreq, optval, sizeof(mreq)))
 671                                 break;
 672                 } else {
 673                         memset(&mreq, 0, sizeof(mreq));
 674                         if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq)))
 675                                 break;
 676                 }
 677
 678                 if (optname == IP_ADD_MEMBERSHIP)
 679                         err = ip_mc_join_group(sk, &mreq);
 680                 else
 681                         err = ip_mc_leave_group(sk, &mreq);
 682                 break;
 683         }
 684         case IP_MSFILTER:
 685         {
 686                 struct ip_msfilter *msf;
 687
 688                 if (optlen < IP_MSFILTER_SIZE(0))
 689                         goto e_inval;
 690                 if (optlen > sysctl_optmem_max) {
 691                         err = -ENOBUFS;
 692                         break;
 693                 }
 694                 msf = kmalloc(optlen, GFP_KERNEL);
 695                 if (!msf) {
 696                         err = -ENOBUFS;
 697                         break;
 698                 }
 699                 err = -EFAULT;
 700                 if (copy_from_user(msf, optval, optlen)) {
 701                         kfree(msf);
 702                         break;
 703                 }
 704                 /* numsrc >= (1G-4) overflow in 32 bits */
 705                 if (msf->imsf_numsrc >= 0x3ffffffcU ||
 706                     msf->imsf_numsrc > sysctl_igmp_max_msf) {
 707                         kfree(msf);
 708                         err = -ENOBUFS;
 709                         break;
 710                 }
 711                 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
 712                         kfree(msf);
 713                         err = -EINVAL;
 714                         break;
 715                 }
 716                 err = ip_mc_msfilter(sk, msf, 0);
 717                 kfree(msf);
 718                 break;
 719         }
 720         case IP_BLOCK_SOURCE:
 721         case IP_UNBLOCK_SOURCE:
 722         case IP_ADD_SOURCE_MEMBERSHIP:
 723         case IP_DROP_SOURCE_MEMBERSHIP:
 724         {
 725                 struct ip_mreq_source mreqs;
 726                 int omode, add;
 727
 728                 if (optlen != sizeof(struct ip_mreq_source))
 729                         goto e_inval;
 730                 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
 731                         err = -EFAULT;
 732                         break;
 733                 }
 734                 if (optname == IP_BLOCK_SOURCE) {
 735                         omode = MCAST_EXCLUDE;
 736                         add = 1;
 737                 } else if (optname == IP_UNBLOCK_SOURCE) {
 738                         omode = MCAST_EXCLUDE;
 739                         add = 0;
 740                 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
 741                         struct ip_mreqn mreq;
 742
 743                         mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
 744                         mreq.imr_address.s_addr = mreqs.imr_interface;
 745                         mreq.imr_ifindex = 0;
 746                         err = ip_mc_join_group(sk, &mreq);
 747                         if (err && err != -EADDRINUSE)
 748                                 break;
 749                         omode = MCAST_INCLUDE;
 750                         add = 1;
 751                 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
 752                         omode = MCAST_INCLUDE;
 753                         add = 0;
 754                 }
 755                 err = ip_mc_source(add, omode, sk, &mreqs, 0);
 756                 break;
 757         }
 758         case MCAST_JOIN_GROUP:
 759         case MCAST_LEAVE_GROUP:
 760         {
 761                 struct group_req greq;
 762                 struct sockaddr_in *psin;
 763                 struct ip_mreqn mreq;
 764
 765                 if (optlen < sizeof(struct group_req))
 766                         goto e_inval;
 767                 err = -EFAULT;
 768                 if (copy_from_user(&greq, optval, sizeof(greq)))
 769                         break;
 770                 psin = (struct sockaddr_in *)&greq.gr_group;
 771                 if (psin->sin_family != AF_INET)
 772                         goto e_inval;
 773                 memset(&mreq, 0, sizeof(mreq));
 774                 mreq.imr_multiaddr = psin->sin_addr;
 775                 mreq.imr_ifindex = greq.gr_interface;
 776
 777                 if (optname == MCAST_JOIN_GROUP)
 778                         err = ip_mc_join_group(sk, &mreq);
 779                 else
 780                         err = ip_mc_leave_group(sk, &mreq);
 781                 break;
 782         }
 783         case MCAST_JOIN_SOURCE_GROUP:
 784         case MCAST_LEAVE_SOURCE_GROUP:
 785         case MCAST_BLOCK_SOURCE:
 786         case MCAST_UNBLOCK_SOURCE:
 787         {
 788                 struct group_source_req greqs;
 789                 struct ip_mreq_source mreqs;
 790                 struct sockaddr_in *psin;
 791                 int omode, add;
 792
 793                 if (optlen != sizeof(struct group_source_req))
 794                         goto e_inval;
 795                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 796                         err = -EFAULT;
 797                         break;
 798                 }
 799                 if (greqs.gsr_group.ss_family != AF_INET ||
 800                     greqs.gsr_source.ss_family != AF_INET) {
 801                         err = -EADDRNOTAVAIL;
 802                         break;
 803                 }
 804                 psin = (struct sockaddr_in *)&greqs.gsr_group;
 805                 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
 806                 psin = (struct sockaddr_in *)&greqs.gsr_source;
 807                 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
 808                 mreqs.imr_interface = 0; /* use index for mc_source */
 809
 810                 if (optname == MCAST_BLOCK_SOURCE) {
 811                         omode = MCAST_EXCLUDE;
 812                         add = 1;
 813                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 814                         omode = MCAST_EXCLUDE;
 815                         add = 0;
 816                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 817                         struct ip_mreqn mreq;
 818
 819                         psin = (struct sockaddr_in *)&greqs.gsr_group;
 820                         mreq.imr_multiaddr = psin->sin_addr;
 821                         mreq.imr_address.s_addr = 0;
 822                         mreq.imr_ifindex = greqs.gsr_interface;
 823                         err = ip_mc_join_group(sk, &mreq);
 824                         if (err && err != -EADDRINUSE)
 825                                 break;
 826                         greqs.gsr_interface = mreq.imr_ifindex;
 827                         omode = MCAST_INCLUDE;
 828                         add = 1;
 829                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 830                         omode = MCAST_INCLUDE;
 831                         add = 0;
 832                 }
 833                 err = ip_mc_source(add, omode, sk, &mreqs,
 834                                    greqs.gsr_interface);
 835                 break;
 836         }
 837         case MCAST_MSFILTER:
 838         {
 839                 struct sockaddr_in *psin;
 840                 struct ip_msfilter *msf = NULL;
 841                 struct group_filter *gsf = NULL;
 842                 int msize, i, ifindex;
 843
 844                 if (optlen < GROUP_FILTER_SIZE(0))
 845                         goto e_inval;
 846                 if (optlen > sysctl_optmem_max) {
 847                         err = -ENOBUFS;
 848                         break;
 849                 }
 850                 gsf = kmalloc(optlen, GFP_KERNEL);
 851                 if (!gsf) {
 852                         err = -ENOBUFS;
 853                         break;
 854                 }
 855                 err = -EFAULT;
 856                 if (copy_from_user(gsf, optval, optlen))
 857                         goto mc_msf_out;
 858
 859                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 860                 if (gsf->gf_numsrc >= 0x1ffffff ||
 861                     gsf->gf_numsrc > sysctl_igmp_max_msf) {
 862                         err = -ENOBUFS;
 863                         goto mc_msf_out;
 864                 }
 865                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 866                         err = -EINVAL;
 867                         goto mc_msf_out;
 868                 }
 869                 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
 870                 msf = kmalloc(msize, GFP_KERNEL);
 871                 if (!msf) {
 872                         err = -ENOBUFS;
 873                         goto mc_msf_out;
 874                 }
 875                 ifindex = gsf->gf_interface;
 876                 psin = (struct sockaddr_in *)&gsf->gf_group;
 877                 if (psin->sin_family != AF_INET) {
 878                         err = -EADDRNOTAVAIL;
 879                         goto mc_msf_out;
 880                 }
 881                 msf->imsf_multiaddr = psin->sin_addr.s_addr;
 882                 msf->imsf_interface = 0;
 883                 msf->imsf_fmode = gsf->gf_fmode;
 884                 msf->imsf_numsrc = gsf->gf_numsrc;
 885                 err = -EADDRNOTAVAIL;
 886                 for (i = 0; i < gsf->gf_numsrc; ++i) {
 887                         psin = (struct sockaddr_in *)&gsf->gf_slist[i];
 888
 889                         if (psin->sin_family != AF_INET)
 890                                 goto mc_msf_out;
 891                         msf->imsf_slist[i] = psin->sin_addr.s_addr;
 892                 }
 893                 kfree(gsf);
 894                 gsf = NULL;
 895
 896                 err = ip_mc_msfilter(sk, msf, ifindex);
 897 mc_msf_out:
 898                 kfree(msf);
 899                 kfree(gsf);
 900                 break;
 901         }
 902         case IP_MULTICAST_ALL:
 903                 if (optlen < 1)
 904                         goto e_inval;
 905                 if (val != 0 && val != 1)
 906                         goto e_inval;
 907                 inet->mc_all = val;
 908                 break;
 909         case IP_ROUTER_ALERT:
 910                 err = ip_ra_control(sk, val ? 1 : 0, NULL);
 911                 break;
 912
 913         case IP_FREEBIND:
 914                 if (optlen < 1)
 915                         goto e_inval;
 916                 inet->freebind = !!val;
 917                 break;
 918
 919         case IP_IPSEC_POLICY:
 920         case IP_XFRM_POLICY:
 921                 err = -EPERM;
 922                 if (!capable(CAP_NET_ADMIN))
 923                         break;
 924                 err = xfrm_user_policy(sk, optname, optval, optlen);
 925                 break;
 926
 927         case IP_TRANSPARENT:
 928                 if (!capable(CAP_NET_ADMIN)) {
 929                         err = -EPERM;
 930                         break;
 931                 }
 932                 if (optlen < 1)
 933                         goto e_inval;
 934                 inet->transparent = !!val;
 935                 break;
 936
 937         default:
 938                 err = -ENOPROTOOPT;
 939                 break;
 940         }
 941         release_sock(sk);
 942         return err;
 943
 944 e_inval:
 945         release_sock(sk);
 946         return -EINVAL;
 947 }
 948
 949 int ip_setsockopt(struct sock *sk, int level,
 950                 int optname, char __user *optval, int optlen)
 951 {
 952         int err;
 953
 954         if (level != SOL_IP)
 955                 return -ENOPROTOOPT;
 956
 957         err = do_ip_setsockopt(sk, level, optname, optval, optlen);
 958 #ifdef CONFIG_NETFILTER
 959         /* we need to exclude all possible ENOPROTOOPTs except default case */
 960         if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
 961                         optname != IP_IPSEC_POLICY &&
 962                         optname != IP_XFRM_POLICY &&
 963                         !ip_mroute_opt(optname)) {
 964                 lock_sock(sk);
 965                 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
 966                 release_sock(sk);
 967         }
 968 #endif
 969         return err;
 970 }
 971 EXPORT_SYMBOL(ip_setsockopt);
 972
 973 #ifdef CONFIG_COMPAT
 974 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
 975                          char __user *optval, int optlen)
 976 {
 977         int err;
 978
 979         if (level != SOL_IP)
 980                 return -ENOPROTOOPT;
 981
 982         if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
 983                 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
 984                         ip_setsockopt);
 985
 986         err = do_ip_setsockopt(sk, level, optname, optval, optlen);
 987 #ifdef CONFIG_NETFILTER
 988         /* we need to exclude all possible ENOPROTOOPTs except default case */
 989         if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
 990                         optname != IP_IPSEC_POLICY &&
 991                         optname != IP_XFRM_POLICY &&
 992                         !ip_mroute_opt(optname)) {
 993                 lock_sock(sk);
 994                 err = compat_nf_setsockopt(sk, PF_INET, optname,
 995                                            optval, optlen);
 996                 release_sock(sk);
 997         }
 998 #endif
 999         return err;
1000 }
1001 EXPORT_SYMBOL(compat_ip_setsockopt);
1002 #endif
1003
1004 /*
1005  *      Get the options. Note for future reference. The GET of IP options gets
1006  *      the _received_ ones. The set sets the _sent_ ones.
1007  */
1008
1009 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
1010                             char __user *optval, int __user *optlen)
1011 {
1012         struct inet_sock *inet = inet_sk(sk);
1013         int val;
1014         int len;
1015
1016         if (level != SOL_IP)
1017                 return -EOPNOTSUPP;
1018
1019         if (ip_mroute_opt(optname))
1020                 return ip_mroute_getsockopt(sk, optname, optval, optlen);
1021
1022         if (get_user(len, optlen))
1023                 return -EFAULT;
1024         if (len < 0)
1025                 return -EINVAL;
1026
1027         lock_sock(sk);
1028
1029         switch (optname) {
1030         case IP_OPTIONS:
1031         {
1032                 unsigned char optbuf[sizeof(struct ip_options)+40];
1033                 struct ip_options * opt = (struct ip_options *)optbuf;
1034                 opt->optlen = 0;
1035                 if (inet->opt)
1036                         memcpy(optbuf, inet->opt,
1037                                sizeof(struct ip_options)+
1038                                inet->opt->optlen);
1039                 release_sock(sk);
1040
1041                 if (opt->optlen == 0)
1042                         return put_user(0, optlen);
1043
1044                 ip_options_undo(opt);
1045
1046                 len = min_t(unsigned int, len, opt->optlen);
1047                 if (put_user(len, optlen))
1048                         return -EFAULT;
1049                 if (copy_to_user(optval, opt->__data, len))
1050                         return -EFAULT;
1051                 return 0;
1052         }
1053         case IP_PKTINFO:
1054                 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
1055                 break;
1056         case IP_RECVTTL:
1057                 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
1058                 break;
1059         case IP_RECVTOS:
1060                 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1061                 break;
1062         case IP_RECVOPTS:
1063                 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1064                 break;
1065         case IP_RETOPTS:
1066                 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1067                 break;
1068         case IP_PASSSEC:
1069                 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1070                 break;
1071         case IP_RECVORIGDSTADDR:
1072                 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0;
1073                 break;
1074         case IP_TOS:
1075                 val = inet->tos;
1076                 break;
1077         case IP_TTL:
1078                 val = (inet->uc_ttl == -1 ?
1079                        sysctl_ip_default_ttl :
1080                        inet->uc_ttl);
1081                 break;
1082         case IP_HDRINCL:
1083                 val = inet->hdrincl;
1084                 break;
1085         case IP_MTU_DISCOVER:
1086                 val = inet->pmtudisc;
1087                 break;
1088         case IP_MTU:
1089         {
1090                 struct dst_entry *dst;
1091                 val = 0;
1092                 dst = sk_dst_get(sk);
1093                 if (dst) {
1094                         val = dst_mtu(dst);
1095                         dst_release(dst);
1096                 }
1097                 if (!val) {
1098                         release_sock(sk);
1099                         return -ENOTCONN;
1100                 }
1101                 break;
1102         }
1103         case IP_RECVERR:
1104                 val = inet->recverr;
1105                 break;
1106         case IP_MULTICAST_TTL:
1107                 val = inet->mc_ttl;
1108                 break;
1109         case IP_MULTICAST_LOOP:
1110                 val = inet->mc_loop;
1111                 break;
1112         case IP_MULTICAST_IF:
1113         {
1114                 struct in_addr addr;
1115                 len = min_t(unsigned int, len, sizeof(struct in_addr));
1116                 addr.s_addr = inet->mc_addr;
1117                 release_sock(sk);
1118
1119                 if (put_user(len, optlen))
1120                         return -EFAULT;
1121                 if (copy_to_user(optval, &addr, len))
1122                         return -EFAULT;
1123                 return 0;
1124         }
1125         case IP_MSFILTER:
1126         {
1127                 struct ip_msfilter msf;
1128                 int err;
1129
1130                 if (len < IP_MSFILTER_SIZE(0)) {
1131                         release_sock(sk);
1132                         return -EINVAL;
1133                 }
1134                 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1135                         release_sock(sk);
1136                         return -EFAULT;
1137                 }
1138                 err = ip_mc_msfget(sk, &msf,
1139                                    (struct ip_msfilter __user *)optval, optlen);
1140                 release_sock(sk);
1141                 return err;
1142         }
1143         case MCAST_MSFILTER:
1144         {
1145                 struct group_filter gsf;
1146                 int err;
1147
1148                 if (len < GROUP_FILTER_SIZE(0)) {
1149                         release_sock(sk);
1150                         return -EINVAL;
1151                 }
1152                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1153                         release_sock(sk);
1154                         return -EFAULT;
1155                 }
1156                 err = ip_mc_gsfget(sk, &gsf,
1157                                    (struct group_filter __user *)optval,
1158                                    optlen);
1159                 release_sock(sk);
1160                 return err;
1161         }
1162         case IP_MULTICAST_ALL:
1163                 val = inet->mc_all;
1164                 break;
1165         case IP_PKTOPTIONS:
1166         {
1167                 struct msghdr msg;
1168
1169                 release_sock(sk);
1170
1171                 if (sk->sk_type != SOCK_STREAM)
1172                         return -ENOPROTOOPT;
1173
1174                 msg.msg_control = optval;
1175                 msg.msg_controllen = len;
1176                 msg.msg_flags = 0;
1177
1178                 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1179                         struct in_pktinfo info;
1180
1181                         info.ipi_addr.s_addr = inet->rcv_saddr;
1182                         info.ipi_spec_dst.s_addr = inet->rcv_saddr;
1183                         info.ipi_ifindex = inet->mc_index;
1184                         put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1185                 }
1186                 if (inet->cmsg_flags & IP_CMSG_TTL) {
1187                         int hlim = inet->mc_ttl;
1188                         put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1189                 }
1190                 len -= msg.msg_controllen;
1191                 return put_user(len, optlen);
1192         }
1193         case IP_FREEBIND:
1194                 val = inet->freebind;
1195                 break;
1196         case IP_TRANSPARENT:
1197                 val = inet->transparent;
1198                 break;
1199         default:
1200                 release_sock(sk);
1201                 return -ENOPROTOOPT;
1202         }
1203         release_sock(sk);
1204
1205         if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1206                 unsigned char ucval = (unsigned char)val;
1207                 len = 1;
1208                 if (put_user(len, optlen))
1209                         return -EFAULT;
1210                 if (copy_to_user(optval, &ucval, 1))
1211                         return -EFAULT;
1212         } else {
1213                 len = min_t(unsigned int, sizeof(int), len);
1214                 if (put_user(len, optlen))
1215                         return -EFAULT;
1216                 if (copy_to_user(optval, &val, len))
1217                         return -EFAULT;
1218         }
1219         return 0;
1220 }
1221
1222 int ip_getsockopt(struct sock *sk, int level,
1223                   int optname, char __user *optval, int __user *optlen)
1224 {
1225         int err;
1226
1227         err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1228 #ifdef CONFIG_NETFILTER
1229         /* we need to exclude all possible ENOPROTOOPTs except default case */
1230         if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1231                         !ip_mroute_opt(optname)) {
1232                 int len;
1233
1234                 if (get_user(len, optlen))
1235                         return -EFAULT;
1236
1237                 lock_sock(sk);
1238                 err = nf_getsockopt(sk, PF_INET, optname, optval,
1239                                 &len);
1240                 release_sock(sk);
1241                 if (err >= 0)
1242                         err = put_user(len, optlen);
1243                 return err;
1244         }
1245 #endif
1246         return err;
1247 }
1248 EXPORT_SYMBOL(ip_getsockopt);
1249
1250 #ifdef CONFIG_COMPAT
1251 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1252                          char __user *optval, int __user *optlen)
1253 {
1254         int err;
1255
1256         if (optname == MCAST_MSFILTER)
1257                 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1258                         ip_getsockopt);
1259
1260         err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1261
1262 #ifdef CONFIG_NETFILTER
1263         /* we need to exclude all possible ENOPROTOOPTs except default case */
1264         if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1265                         !ip_mroute_opt(optname)) {
1266                 int len;
1267
1268                 if (get_user(len, optlen))
1269                         return -EFAULT;
1270
1271                 lock_sock(sk);
1272                 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1273                 release_sock(sk);
1274                 if (err >= 0)
1275                         err = put_user(len, optlen);
1276                 return err;
1277         }
1278 #endif
1279         return err;
1280 }
1281 EXPORT_SYMBOL(compat_ip_getsockopt);
1282 #endif