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