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  * Version:     $Id: ip_sockglue.c,v 1.62 2002/02/01 22:01:04 davem Exp $
   9  *
  10  * Authors:     see ip.c
  11  *
  12  * Fixes:
  13  *              Many            :       Split from ip.c , see ip.c for history.
  14  *              Martin Mares    :       TOS setting fixed.
  15  *              Alan Cox        :       Fixed a couple of oopses in Martin's
  16  *                                      TOS tweaks.
  17  *              Mike McLagan    :       Routing by source
  18  */
  19
  20 #include <linux/module.h>
  21 #include <linux/types.h>
  22 #include <linux/mm.h>
  23 #include <linux/skbuff.h>
  24 #include <linux/ip.h>
  25 #include <linux/icmp.h>
  26 #include <linux/inetdevice.h>
  27 #include <linux/netdevice.h>
  28 #include <net/sock.h>
  29 #include <net/ip.h>
  30 #include <net/icmp.h>
  31 #include <net/tcp_states.h>
  32 #include <linux/udp.h>
  33 #include <linux/igmp.h>
  34 #include <linux/netfilter.h>
  35 #include <linux/route.h>
  36 #include <linux/mroute.h>
  37 #include <net/route.h>
  38 #include <net/xfrm.h>
  39 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
  40 #include <net/transp_v6.h>
  41 #endif
  42
  43 #include <linux/errqueue.h>
  44 #include <asm/uaccess.h>
  45
  46 #define IP_CMSG_PKTINFO         1
  47 #define IP_CMSG_TTL             2
  48 #define IP_CMSG_TOS             4
  49 #define IP_CMSG_RECVOPTS        8
  50 #define IP_CMSG_RETOPTS         16
  51 #define IP_CMSG_PASSSEC         32
  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
 131 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
 132 {
 133         struct inet_sock *inet = inet_sk(skb->sk);
 134         unsigned flags = inet->cmsg_flags;
 135
 136         /* Ordered by supposed usage frequency */
 137         if (flags & 1)
 138                 ip_cmsg_recv_pktinfo(msg, skb);
 139         if ((flags>>=1) == 0)
 140                 return;
 141
 142         if (flags & 1)
 143                 ip_cmsg_recv_ttl(msg, skb);
 144         if ((flags>>=1) == 0)
 145                 return;
 146
 147         if (flags & 1)
 148                 ip_cmsg_recv_tos(msg, skb);
 149         if ((flags>>=1) == 0)
 150                 return;
 151
 152         if (flags & 1)
 153                 ip_cmsg_recv_opts(msg, skb);
 154         if ((flags>>=1) == 0)
 155                 return;
 156
 157         if (flags & 1)
 158                 ip_cmsg_recv_retopts(msg, skb);
 159         if ((flags>>=1) == 0)
 160                 return;
 161
 162         if (flags & 1)
 163                 ip_cmsg_recv_security(msg, skb);
 164 }
 165
 166 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc)
 167 {
 168         int err;
 169         struct cmsghdr *cmsg;
 170
 171         for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) {
 172                 if (!CMSG_OK(msg, cmsg))
 173                         return -EINVAL;
 174                 if (cmsg->cmsg_level != SOL_IP)
 175                         continue;
 176                 switch (cmsg->cmsg_type) {
 177                 case IP_RETOPTS:
 178                         err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr));
 179                         err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg), err < 40 ? err : 40);
 180                         if (err)
 181                                 return err;
 182                         break;
 183                 case IP_PKTINFO:
 184                 {
 185                         struct in_pktinfo *info;
 186                         if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo)))
 187                                 return -EINVAL;
 188                         info = (struct in_pktinfo *)CMSG_DATA(cmsg);
 189                         ipc->oif = info->ipi_ifindex;
 190                         ipc->addr = info->ipi_spec_dst.s_addr;
 191                         break;
 192                 }
 193                 default:
 194                         return -EINVAL;
 195                 }
 196         }
 197         return 0;
 198 }
 199
 200
 201 /* Special input handler for packets caught by router alert option.
 202    They are selected only by protocol field, and then processed likely
 203    local ones; but only if someone wants them! Otherwise, router
 204    not running rsvpd will kill RSVP.
 205
 206    It is user level problem, what it will make with them.
 207    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
 208    but receiver should be enough clever f.e. to forward mtrace requests,
 209    sent to multicast group to reach destination designated router.
 210  */
 211 struct ip_ra_chain *ip_ra_chain;
 212 DEFINE_RWLOCK(ip_ra_lock);
 213
 214 int ip_ra_control(struct sock *sk, unsigned char on, void (*destructor)(struct sock *))
 215 {
 216         struct ip_ra_chain *ra, *new_ra, **rap;
 217
 218         if (sk->sk_type != SOCK_RAW || inet_sk(sk)->num == IPPROTO_RAW)
 219                 return -EINVAL;
 220
 221         new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
 222
 223         write_lock_bh(&ip_ra_lock);
 224         for (rap = &ip_ra_chain; (ra=*rap) != NULL; rap = &ra->next) {
 225                 if (ra->sk == sk) {
 226                         if (on) {
 227                                 write_unlock_bh(&ip_ra_lock);
 228                                 kfree(new_ra);
 229                                 return -EADDRINUSE;
 230                         }
 231                         *rap = ra->next;
 232                         write_unlock_bh(&ip_ra_lock);
 233
 234                         if (ra->destructor)
 235                                 ra->destructor(sk);
 236                         sock_put(sk);
 237                         kfree(ra);
 238                         return 0;
 239                 }
 240         }
 241         if (new_ra == NULL) {
 242                 write_unlock_bh(&ip_ra_lock);
 243                 return -ENOBUFS;
 244         }
 245         new_ra->sk = sk;
 246         new_ra->destructor = destructor;
 247
 248         new_ra->next = ra;
 249         *rap = new_ra;
 250         sock_hold(sk);
 251         write_unlock_bh(&ip_ra_lock);
 252
 253         return 0;
 254 }
 255
 256 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err,
 257                    __be16 port, u32 info, u8 *payload)
 258 {
 259         struct inet_sock *inet = inet_sk(sk);
 260         struct sock_exterr_skb *serr;
 261
 262         if (!inet->recverr)
 263                 return;
 264
 265         skb = skb_clone(skb, GFP_ATOMIC);
 266         if (!skb)
 267                 return;
 268
 269         serr = SKB_EXT_ERR(skb);
 270         serr->ee.ee_errno = err;
 271         serr->ee.ee_origin = SO_EE_ORIGIN_ICMP;
 272         serr->ee.ee_type = icmp_hdr(skb)->type;
 273         serr->ee.ee_code = icmp_hdr(skb)->code;
 274         serr->ee.ee_pad = 0;
 275         serr->ee.ee_info = info;
 276         serr->ee.ee_data = 0;
 277         serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) -
 278                                    skb_network_header(skb);
 279         serr->port = port;
 280
 281         if (skb_pull(skb, payload - skb->data) != NULL) {
 282                 skb_reset_transport_header(skb);
 283                 if (sock_queue_err_skb(sk, skb) == 0)
 284                         return;
 285         }
 286         kfree_skb(skb);
 287 }
 288
 289 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info)
 290 {
 291         struct inet_sock *inet = inet_sk(sk);
 292         struct sock_exterr_skb *serr;
 293         struct iphdr *iph;
 294         struct sk_buff *skb;
 295
 296         if (!inet->recverr)
 297                 return;
 298
 299         skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC);
 300         if (!skb)
 301                 return;
 302
 303         skb_put(skb, sizeof(struct iphdr));
 304         skb_reset_network_header(skb);
 305         iph = ip_hdr(skb);
 306         iph->daddr = daddr;
 307
 308         serr = SKB_EXT_ERR(skb);
 309         serr->ee.ee_errno = err;
 310         serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL;
 311         serr->ee.ee_type = 0;
 312         serr->ee.ee_code = 0;
 313         serr->ee.ee_pad = 0;
 314         serr->ee.ee_info = info;
 315         serr->ee.ee_data = 0;
 316         serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb);
 317         serr->port = port;
 318
 319         __skb_pull(skb, skb_tail_pointer(skb) - skb->data);
 320         skb_reset_transport_header(skb);
 321
 322         if (sock_queue_err_skb(sk, skb))
 323                 kfree_skb(skb);
 324 }
 325
 326 /*
 327  *      Handle MSG_ERRQUEUE
 328  */
 329 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len)
 330 {
 331         struct sock_exterr_skb *serr;
 332         struct sk_buff *skb, *skb2;
 333         struct sockaddr_in *sin;
 334         struct {
 335                 struct sock_extended_err ee;
 336                 struct sockaddr_in       offender;
 337         } errhdr;
 338         int err;
 339         int copied;
 340
 341         err = -EAGAIN;
 342         skb = skb_dequeue(&sk->sk_error_queue);
 343         if (skb == NULL)
 344                 goto out;
 345
 346         copied = skb->len;
 347         if (copied > len) {
 348                 msg->msg_flags |= MSG_TRUNC;
 349                 copied = len;
 350         }
 351         err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 352         if (err)
 353                 goto out_free_skb;
 354
 355         sock_recv_timestamp(msg, sk, skb);
 356
 357         serr = SKB_EXT_ERR(skb);
 358
 359         sin = (struct sockaddr_in *)msg->msg_name;
 360         if (sin) {
 361                 sin->sin_family = AF_INET;
 362                 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) +
 363                                                    serr->addr_offset);
 364                 sin->sin_port = serr->port;
 365                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 366         }
 367
 368         memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err));
 369         sin = &errhdr.offender;
 370         sin->sin_family = AF_UNSPEC;
 371         if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) {
 372                 struct inet_sock *inet = inet_sk(sk);
 373
 374                 sin->sin_family = AF_INET;
 375                 sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 376                 sin->sin_port = 0;
 377                 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
 378                 if (inet->cmsg_flags)
 379                         ip_cmsg_recv(msg, skb);
 380         }
 381
 382         put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr);
 383
 384         /* Now we could try to dump offended packet options */
 385
 386         msg->msg_flags |= MSG_ERRQUEUE;
 387         err = copied;
 388
 389         /* Reset and regenerate socket error */
 390         spin_lock_bh(&sk->sk_error_queue.lock);
 391         sk->sk_err = 0;
 392         if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) {
 393                 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
 394                 spin_unlock_bh(&sk->sk_error_queue.lock);
 395                 sk->sk_error_report(sk);
 396         } else
 397                 spin_unlock_bh(&sk->sk_error_queue.lock);
 398
 399 out_free_skb:
 400         kfree_skb(skb);
 401 out:
 402         return err;
 403 }
 404
 405
 406 /*
 407  *      Socket option code for IP. This is the end of the line after any TCP,UDP etc options on
 408  *      an IP socket.
 409  */
 410
 411 static int do_ip_setsockopt(struct sock *sk, int level,
 412                             int optname, char __user *optval, int optlen)
 413 {
 414         struct inet_sock *inet = inet_sk(sk);
 415         int val=0,err;
 416
 417         if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) |
 418                              (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) |
 419                              (1<<IP_RETOPTS) | (1<<IP_TOS) |
 420                              (1<<IP_TTL) | (1<<IP_HDRINCL) |
 421                              (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) |
 422                              (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) |
 423                              (1<<IP_PASSSEC))) ||
 424             optname == IP_MULTICAST_TTL ||
 425             optname == IP_MULTICAST_LOOP) {
 426                 if (optlen >= sizeof(int)) {
 427                         if (get_user(val, (int __user *) optval))
 428                                 return -EFAULT;
 429                 } else if (optlen >= sizeof(char)) {
 430                         unsigned char ucval;
 431
 432                         if (get_user(ucval, (unsigned char __user *) optval))
 433                                 return -EFAULT;
 434                         val = (int) ucval;
 435                 }
 436         }
 437
 438         /* If optlen==0, it is equivalent to val == 0 */
 439
 440         if (ip_mroute_opt(optname))
 441                 return ip_mroute_setsockopt(sk,optname,optval,optlen);
 442
 443         err = 0;
 444         lock_sock(sk);
 445
 446         switch (optname) {
 447         case IP_OPTIONS:
 448         {
 449                 struct ip_options * opt = NULL;
 450                 if (optlen > 40 || optlen < 0)
 451                         goto e_inval;
 452                 err = ip_options_get_from_user(sock_net(sk), &opt,
 453                                                optval, optlen);
 454                 if (err)
 455                         break;
 456                 if (inet->is_icsk) {
 457                         struct inet_connection_sock *icsk = inet_csk(sk);
 458 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 459                         if (sk->sk_family == PF_INET ||
 460                             (!((1 << sk->sk_state) &
 461                                (TCPF_LISTEN | TCPF_CLOSE)) &&
 462                              inet->daddr != LOOPBACK4_IPV6)) {
 463 #endif
 464                                 if (inet->opt)
 465                                         icsk->icsk_ext_hdr_len -= inet->opt->optlen;
 466                                 if (opt)
 467                                         icsk->icsk_ext_hdr_len += opt->optlen;
 468                                 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
 469 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 470                         }
 471 #endif
 472                 }
 473                 opt = xchg(&inet->opt, opt);
 474                 kfree(opt);
 475                 break;
 476         }
 477         case IP_PKTINFO:
 478                 if (val)
 479                         inet->cmsg_flags |= IP_CMSG_PKTINFO;
 480                 else
 481                         inet->cmsg_flags &= ~IP_CMSG_PKTINFO;
 482                 break;
 483         case IP_RECVTTL:
 484                 if (val)
 485                         inet->cmsg_flags |=  IP_CMSG_TTL;
 486                 else
 487                         inet->cmsg_flags &= ~IP_CMSG_TTL;
 488                 break;
 489         case IP_RECVTOS:
 490                 if (val)
 491                         inet->cmsg_flags |=  IP_CMSG_TOS;
 492                 else
 493                         inet->cmsg_flags &= ~IP_CMSG_TOS;
 494                 break;
 495         case IP_RECVOPTS:
 496                 if (val)
 497                         inet->cmsg_flags |=  IP_CMSG_RECVOPTS;
 498                 else
 499                         inet->cmsg_flags &= ~IP_CMSG_RECVOPTS;
 500                 break;
 501         case IP_RETOPTS:
 502                 if (val)
 503                         inet->cmsg_flags |= IP_CMSG_RETOPTS;
 504                 else
 505                         inet->cmsg_flags &= ~IP_CMSG_RETOPTS;
 506                 break;
 507         case IP_PASSSEC:
 508                 if (val)
 509                         inet->cmsg_flags |= IP_CMSG_PASSSEC;
 510                 else
 511                         inet->cmsg_flags &= ~IP_CMSG_PASSSEC;
 512                 break;
 513         case IP_TOS:    /* This sets both TOS and Precedence */
 514                 if (sk->sk_type == SOCK_STREAM) {
 515                         val &= ~3;
 516                         val |= inet->tos & 3;
 517                 }
 518                 if (inet->tos != val) {
 519                         inet->tos = val;
 520                         sk->sk_priority = rt_tos2priority(val);
 521                         sk_dst_reset(sk);
 522                 }
 523                 break;
 524         case IP_TTL:
 525                 if (optlen<1)
 526                         goto e_inval;
 527                 if (val != -1 && (val < 1 || val>255))
 528                         goto e_inval;
 529                 inet->uc_ttl = val;
 530                 break;
 531         case IP_HDRINCL:
 532                 if (sk->sk_type != SOCK_RAW) {
 533                         err = -ENOPROTOOPT;
 534                         break;
 535                 }
 536                 inet->hdrincl = val ? 1 : 0;
 537                 break;
 538         case IP_MTU_DISCOVER:
 539                 if (val<0 || val>3)
 540                         goto e_inval;
 541                 inet->pmtudisc = val;
 542                 break;
 543         case IP_RECVERR:
 544                 inet->recverr = !!val;
 545                 if (!val)
 546                         skb_queue_purge(&sk->sk_error_queue);
 547                 break;
 548         case IP_MULTICAST_TTL:
 549                 if (sk->sk_type == SOCK_STREAM)
 550                         goto e_inval;
 551                 if (optlen<1)
 552                         goto e_inval;
 553                 if (val==-1)
 554                         val = 1;
 555                 if (val < 0 || val > 255)
 556                         goto e_inval;
 557                 inet->mc_ttl = val;
 558                 break;
 559         case IP_MULTICAST_LOOP:
 560                 if (optlen<1)
 561                         goto e_inval;
 562                 inet->mc_loop = !!val;
 563                 break;
 564         case IP_MULTICAST_IF:
 565         {
 566                 struct ip_mreqn mreq;
 567                 struct net_device *dev = NULL;
 568
 569                 if (sk->sk_type == SOCK_STREAM)
 570                         goto e_inval;
 571                 /*
 572                  *      Check the arguments are allowable
 573                  */
 574
 575                 err = -EFAULT;
 576                 if (optlen >= sizeof(struct ip_mreqn)) {
 577                         if (copy_from_user(&mreq,optval,sizeof(mreq)))
 578                                 break;
 579                 } else {
 580                         memset(&mreq, 0, sizeof(mreq));
 581                         if (optlen >= sizeof(struct in_addr) &&
 582                             copy_from_user(&mreq.imr_address,optval,sizeof(struct in_addr)))
 583                                 break;
 584                 }
 585
 586                 if (!mreq.imr_ifindex) {
 587                         if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) {
 588                                 inet->mc_index = 0;
 589                                 inet->mc_addr  = 0;
 590                                 err = 0;
 591                                 break;
 592                         }
 593                         dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr);
 594                         if (dev) {
 595                                 mreq.imr_ifindex = dev->ifindex;
 596                                 dev_put(dev);
 597                         }
 598                 } else
 599                         dev = __dev_get_by_index(sock_net(sk), mreq.imr_ifindex);
 600
 601
 602                 err = -EADDRNOTAVAIL;
 603                 if (!dev)
 604                         break;
 605
 606                 err = -EINVAL;
 607                 if (sk->sk_bound_dev_if &&
 608                     mreq.imr_ifindex != sk->sk_bound_dev_if)
 609                         break;
 610
 611                 inet->mc_index = mreq.imr_ifindex;
 612                 inet->mc_addr  = mreq.imr_address.s_addr;
 613                 err = 0;
 614                 break;
 615         }
 616
 617         case IP_ADD_MEMBERSHIP:
 618         case IP_DROP_MEMBERSHIP:
 619         {
 620                 struct ip_mreqn mreq;
 621
 622                 err = -EPROTO;
 623                 if (inet_sk(sk)->is_icsk)
 624                         break;
 625
 626                 if (optlen < sizeof(struct ip_mreq))
 627                         goto e_inval;
 628                 err = -EFAULT;
 629                 if (optlen >= sizeof(struct ip_mreqn)) {
 630                         if (copy_from_user(&mreq,optval,sizeof(mreq)))
 631                                 break;
 632                 } else {
 633                         memset(&mreq, 0, sizeof(mreq));
 634                         if (copy_from_user(&mreq,optval,sizeof(struct ip_mreq)))
 635                                 break;
 636                 }
 637
 638                 if (optname == IP_ADD_MEMBERSHIP)
 639                         err = ip_mc_join_group(sk, &mreq);
 640                 else
 641                         err = ip_mc_leave_group(sk, &mreq);
 642                 break;
 643         }
 644         case IP_MSFILTER:
 645         {
 646                 extern int sysctl_igmp_max_msf;
 647                 struct ip_msfilter *msf;
 648
 649                 if (optlen < IP_MSFILTER_SIZE(0))
 650                         goto e_inval;
 651                 if (optlen > sysctl_optmem_max) {
 652                         err = -ENOBUFS;
 653                         break;
 654                 }
 655                 msf = kmalloc(optlen, GFP_KERNEL);
 656                 if (!msf) {
 657                         err = -ENOBUFS;
 658                         break;
 659                 }
 660                 err = -EFAULT;
 661                 if (copy_from_user(msf, optval, optlen)) {
 662                         kfree(msf);
 663                         break;
 664                 }
 665                 /* numsrc >= (1G-4) overflow in 32 bits */
 666                 if (msf->imsf_numsrc >= 0x3ffffffcU ||
 667                     msf->imsf_numsrc > sysctl_igmp_max_msf) {
 668                         kfree(msf);
 669                         err = -ENOBUFS;
 670                         break;
 671                 }
 672                 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) {
 673                         kfree(msf);
 674                         err = -EINVAL;
 675                         break;
 676                 }
 677                 err = ip_mc_msfilter(sk, msf, 0);
 678                 kfree(msf);
 679                 break;
 680         }
 681         case IP_BLOCK_SOURCE:
 682         case IP_UNBLOCK_SOURCE:
 683         case IP_ADD_SOURCE_MEMBERSHIP:
 684         case IP_DROP_SOURCE_MEMBERSHIP:
 685         {
 686                 struct ip_mreq_source mreqs;
 687                 int omode, add;
 688
 689                 if (optlen != sizeof(struct ip_mreq_source))
 690                         goto e_inval;
 691                 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) {
 692                         err = -EFAULT;
 693                         break;
 694                 }
 695                 if (optname == IP_BLOCK_SOURCE) {
 696                         omode = MCAST_EXCLUDE;
 697                         add = 1;
 698                 } else if (optname == IP_UNBLOCK_SOURCE) {
 699                         omode = MCAST_EXCLUDE;
 700                         add = 0;
 701                 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) {
 702                         struct ip_mreqn mreq;
 703
 704                         mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
 705                         mreq.imr_address.s_addr = mreqs.imr_interface;
 706                         mreq.imr_ifindex = 0;
 707                         err = ip_mc_join_group(sk, &mreq);
 708                         if (err && err != -EADDRINUSE)
 709                                 break;
 710                         omode = MCAST_INCLUDE;
 711                         add = 1;
 712                 } else /* IP_DROP_SOURCE_MEMBERSHIP */ {
 713                         omode = MCAST_INCLUDE;
 714                         add = 0;
 715                 }
 716                 err = ip_mc_source(add, omode, sk, &mreqs, 0);
 717                 break;
 718         }
 719         case MCAST_JOIN_GROUP:
 720         case MCAST_LEAVE_GROUP:
 721         {
 722                 struct group_req greq;
 723                 struct sockaddr_in *psin;
 724                 struct ip_mreqn mreq;
 725
 726                 if (optlen < sizeof(struct group_req))
 727                         goto e_inval;
 728                 err = -EFAULT;
 729                 if (copy_from_user(&greq, optval, sizeof(greq)))
 730                         break;
 731                 psin = (struct sockaddr_in *)&greq.gr_group;
 732                 if (psin->sin_family != AF_INET)
 733                         goto e_inval;
 734                 memset(&mreq, 0, sizeof(mreq));
 735                 mreq.imr_multiaddr = psin->sin_addr;
 736                 mreq.imr_ifindex = greq.gr_interface;
 737
 738                 if (optname == MCAST_JOIN_GROUP)
 739                         err = ip_mc_join_group(sk, &mreq);
 740                 else
 741                         err = ip_mc_leave_group(sk, &mreq);
 742                 break;
 743         }
 744         case MCAST_JOIN_SOURCE_GROUP:
 745         case MCAST_LEAVE_SOURCE_GROUP:
 746         case MCAST_BLOCK_SOURCE:
 747         case MCAST_UNBLOCK_SOURCE:
 748         {
 749                 struct group_source_req greqs;
 750                 struct ip_mreq_source mreqs;
 751                 struct sockaddr_in *psin;
 752                 int omode, add;
 753
 754                 if (optlen != sizeof(struct group_source_req))
 755                         goto e_inval;
 756                 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
 757                         err = -EFAULT;
 758                         break;
 759                 }
 760                 if (greqs.gsr_group.ss_family != AF_INET ||
 761                     greqs.gsr_source.ss_family != AF_INET) {
 762                         err = -EADDRNOTAVAIL;
 763                         break;
 764                 }
 765                 psin = (struct sockaddr_in *)&greqs.gsr_group;
 766                 mreqs.imr_multiaddr = psin->sin_addr.s_addr;
 767                 psin = (struct sockaddr_in *)&greqs.gsr_source;
 768                 mreqs.imr_sourceaddr = psin->sin_addr.s_addr;
 769                 mreqs.imr_interface = 0; /* use index for mc_source */
 770
 771                 if (optname == MCAST_BLOCK_SOURCE) {
 772                         omode = MCAST_EXCLUDE;
 773                         add = 1;
 774                 } else if (optname == MCAST_UNBLOCK_SOURCE) {
 775                         omode = MCAST_EXCLUDE;
 776                         add = 0;
 777                 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
 778                         struct ip_mreqn mreq;
 779
 780                         psin = (struct sockaddr_in *)&greqs.gsr_group;
 781                         mreq.imr_multiaddr = psin->sin_addr;
 782                         mreq.imr_address.s_addr = 0;
 783                         mreq.imr_ifindex = greqs.gsr_interface;
 784                         err = ip_mc_join_group(sk, &mreq);
 785                         if (err && err != -EADDRINUSE)
 786                                 break;
 787                         greqs.gsr_interface = mreq.imr_ifindex;
 788                         omode = MCAST_INCLUDE;
 789                         add = 1;
 790                 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
 791                         omode = MCAST_INCLUDE;
 792                         add = 0;
 793                 }
 794                 err = ip_mc_source(add, omode, sk, &mreqs,
 795                                    greqs.gsr_interface);
 796                 break;
 797         }
 798         case MCAST_MSFILTER:
 799         {
 800                 extern int sysctl_igmp_max_msf;
 801                 struct sockaddr_in *psin;
 802                 struct ip_msfilter *msf = NULL;
 803                 struct group_filter *gsf = NULL;
 804                 int msize, i, ifindex;
 805
 806                 if (optlen < GROUP_FILTER_SIZE(0))
 807                         goto e_inval;
 808                 if (optlen > sysctl_optmem_max) {
 809                         err = -ENOBUFS;
 810                         break;
 811                 }
 812                 gsf = kmalloc(optlen,GFP_KERNEL);
 813                 if (!gsf) {
 814                         err = -ENOBUFS;
 815                         break;
 816                 }
 817                 err = -EFAULT;
 818                 if (copy_from_user(gsf, optval, optlen)) {
 819                         goto mc_msf_out;
 820                 }
 821                 /* numsrc >= (4G-140)/128 overflow in 32 bits */
 822                 if (gsf->gf_numsrc >= 0x1ffffff ||
 823                     gsf->gf_numsrc > sysctl_igmp_max_msf) {
 824                         err = -ENOBUFS;
 825                         goto mc_msf_out;
 826                 }
 827                 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
 828                         err = -EINVAL;
 829                         goto mc_msf_out;
 830                 }
 831                 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc);
 832                 msf = kmalloc(msize,GFP_KERNEL);
 833                 if (!msf) {
 834                         err = -ENOBUFS;
 835                         goto mc_msf_out;
 836                 }
 837                 ifindex = gsf->gf_interface;
 838                 psin = (struct sockaddr_in *)&gsf->gf_group;
 839                 if (psin->sin_family != AF_INET) {
 840                         err = -EADDRNOTAVAIL;
 841                         goto mc_msf_out;
 842                 }
 843                 msf->imsf_multiaddr = psin->sin_addr.s_addr;
 844                 msf->imsf_interface = 0;
 845                 msf->imsf_fmode = gsf->gf_fmode;
 846                 msf->imsf_numsrc = gsf->gf_numsrc;
 847                 err = -EADDRNOTAVAIL;
 848                 for (i=0; i<gsf->gf_numsrc; ++i) {
 849                         psin = (struct sockaddr_in *)&gsf->gf_slist[i];
 850
 851                         if (psin->sin_family != AF_INET)
 852                                 goto mc_msf_out;
 853                         msf->imsf_slist[i] = psin->sin_addr.s_addr;
 854                 }
 855                 kfree(gsf);
 856                 gsf = NULL;
 857
 858                 err = ip_mc_msfilter(sk, msf, ifindex);
 859         mc_msf_out:
 860                 kfree(msf);
 861                 kfree(gsf);
 862                 break;
 863         }
 864         case IP_ROUTER_ALERT:
 865                 err = ip_ra_control(sk, val ? 1 : 0, NULL);
 866                 break;
 867
 868         case IP_FREEBIND:
 869                 if (optlen<1)
 870                         goto e_inval;
 871                 inet->freebind = !!val;
 872                 break;
 873
 874         case IP_IPSEC_POLICY:
 875         case IP_XFRM_POLICY:
 876                 err = -EPERM;
 877                 if (!capable(CAP_NET_ADMIN))
 878                         break;
 879                 err = xfrm_user_policy(sk, optname, optval, optlen);
 880                 break;
 881
 882         default:
 883                 err = -ENOPROTOOPT;
 884                 break;
 885         }
 886         release_sock(sk);
 887         return err;
 888
 889 e_inval:
 890         release_sock(sk);
 891         return -EINVAL;
 892 }
 893
 894 int ip_setsockopt(struct sock *sk, int level,
 895                 int optname, char __user *optval, int optlen)
 896 {
 897         int err;
 898
 899         if (level != SOL_IP)
 900                 return -ENOPROTOOPT;
 901
 902         err = do_ip_setsockopt(sk, level, optname, optval, optlen);
 903 #ifdef CONFIG_NETFILTER
 904         /* we need to exclude all possible ENOPROTOOPTs except default case */
 905         if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
 906                         optname != IP_IPSEC_POLICY &&
 907                         optname != IP_XFRM_POLICY &&
 908                         !ip_mroute_opt(optname)) {
 909                 lock_sock(sk);
 910                 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen);
 911                 release_sock(sk);
 912         }
 913 #endif
 914         return err;
 915 }
 916
 917 #ifdef CONFIG_COMPAT
 918 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
 919                          char __user *optval, int optlen)
 920 {
 921         int err;
 922
 923         if (level != SOL_IP)
 924                 return -ENOPROTOOPT;
 925
 926         err = do_ip_setsockopt(sk, level, optname, optval, optlen);
 927 #ifdef CONFIG_NETFILTER
 928         /* we need to exclude all possible ENOPROTOOPTs except default case */
 929         if (err == -ENOPROTOOPT && optname != IP_HDRINCL &&
 930                         optname != IP_IPSEC_POLICY &&
 931                         optname != IP_XFRM_POLICY &&
 932                         !ip_mroute_opt(optname)) {
 933                 lock_sock(sk);
 934                 err = compat_nf_setsockopt(sk, PF_INET, optname,
 935                                            optval, optlen);
 936                 release_sock(sk);
 937         }
 938 #endif
 939         return err;
 940 }
 941
 942 EXPORT_SYMBOL(compat_ip_setsockopt);
 943 #endif
 944
 945 /*
 946  *      Get the options. Note for future reference. The GET of IP options gets the
 947  *      _received_ ones. The set sets the _sent_ ones.
 948  */
 949
 950 static int do_ip_getsockopt(struct sock *sk, int level, int optname,
 951                             char __user *optval, int __user *optlen)
 952 {
 953         struct inet_sock *inet = inet_sk(sk);
 954         int val;
 955         int len;
 956
 957         if (level != SOL_IP)
 958                 return -EOPNOTSUPP;
 959
 960         if (ip_mroute_opt(optname))
 961                 return ip_mroute_getsockopt(sk,optname,optval,optlen);
 962
 963         if (get_user(len,optlen))
 964                 return -EFAULT;
 965         if (len < 0)
 966                 return -EINVAL;
 967
 968         lock_sock(sk);
 969
 970         switch (optname) {
 971         case IP_OPTIONS:
 972         {
 973                 unsigned char optbuf[sizeof(struct ip_options)+40];
 974                 struct ip_options * opt = (struct ip_options*)optbuf;
 975                 opt->optlen = 0;
 976                 if (inet->opt)
 977                         memcpy(optbuf, inet->opt,
 978                                sizeof(struct ip_options)+
 979                                inet->opt->optlen);
 980                 release_sock(sk);
 981
 982                 if (opt->optlen == 0)
 983                         return put_user(0, optlen);
 984
 985                 ip_options_undo(opt);
 986
 987                 len = min_t(unsigned int, len, opt->optlen);
 988                 if (put_user(len, optlen))
 989                         return -EFAULT;
 990                 if (copy_to_user(optval, opt->__data, len))
 991                         return -EFAULT;
 992                 return 0;
 993         }
 994         case IP_PKTINFO:
 995                 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0;
 996                 break;
 997         case IP_RECVTTL:
 998                 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0;
 999                 break;
1000         case IP_RECVTOS:
1001                 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0;
1002                 break;
1003         case IP_RECVOPTS:
1004                 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0;
1005                 break;
1006         case IP_RETOPTS:
1007                 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0;
1008                 break;
1009         case IP_PASSSEC:
1010                 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0;
1011                 break;
1012         case IP_TOS:
1013                 val = inet->tos;
1014                 break;
1015         case IP_TTL:
1016                 val = (inet->uc_ttl == -1 ?
1017                        sysctl_ip_default_ttl :
1018                        inet->uc_ttl);
1019                 break;
1020         case IP_HDRINCL:
1021                 val = inet->hdrincl;
1022                 break;
1023         case IP_MTU_DISCOVER:
1024                 val = inet->pmtudisc;
1025                 break;
1026         case IP_MTU:
1027         {
1028                 struct dst_entry *dst;
1029                 val = 0;
1030                 dst = sk_dst_get(sk);
1031                 if (dst) {
1032                         val = dst_mtu(dst);
1033                         dst_release(dst);
1034                 }
1035                 if (!val) {
1036                         release_sock(sk);
1037                         return -ENOTCONN;
1038                 }
1039                 break;
1040         }
1041         case IP_RECVERR:
1042                 val = inet->recverr;
1043                 break;
1044         case IP_MULTICAST_TTL:
1045                 val = inet->mc_ttl;
1046                 break;
1047         case IP_MULTICAST_LOOP:
1048                 val = inet->mc_loop;
1049                 break;
1050         case IP_MULTICAST_IF:
1051         {
1052                 struct in_addr addr;
1053                 len = min_t(unsigned int, len, sizeof(struct in_addr));
1054                 addr.s_addr = inet->mc_addr;
1055                 release_sock(sk);
1056
1057                 if (put_user(len, optlen))
1058                         return -EFAULT;
1059                 if (copy_to_user(optval, &addr, len))
1060                         return -EFAULT;
1061                 return 0;
1062         }
1063         case IP_MSFILTER:
1064         {
1065                 struct ip_msfilter msf;
1066                 int err;
1067
1068                 if (len < IP_MSFILTER_SIZE(0)) {
1069                         release_sock(sk);
1070                         return -EINVAL;
1071                 }
1072                 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) {
1073                         release_sock(sk);
1074                         return -EFAULT;
1075                 }
1076                 err = ip_mc_msfget(sk, &msf,
1077                                    (struct ip_msfilter __user *)optval, optlen);
1078                 release_sock(sk);
1079                 return err;
1080         }
1081         case MCAST_MSFILTER:
1082         {
1083                 struct group_filter gsf;
1084                 int err;
1085
1086                 if (len < GROUP_FILTER_SIZE(0)) {
1087                         release_sock(sk);
1088                         return -EINVAL;
1089                 }
1090                 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) {
1091                         release_sock(sk);
1092                         return -EFAULT;
1093                 }
1094                 err = ip_mc_gsfget(sk, &gsf,
1095                                    (struct group_filter __user *)optval, optlen);
1096                 release_sock(sk);
1097                 return err;
1098         }
1099         case IP_PKTOPTIONS:
1100         {
1101                 struct msghdr msg;
1102
1103                 release_sock(sk);
1104
1105                 if (sk->sk_type != SOCK_STREAM)
1106                         return -ENOPROTOOPT;
1107
1108                 msg.msg_control = optval;
1109                 msg.msg_controllen = len;
1110                 msg.msg_flags = 0;
1111
1112                 if (inet->cmsg_flags & IP_CMSG_PKTINFO) {
1113                         struct in_pktinfo info;
1114
1115                         info.ipi_addr.s_addr = inet->rcv_saddr;
1116                         info.ipi_spec_dst.s_addr = inet->rcv_saddr;
1117                         info.ipi_ifindex = inet->mc_index;
1118                         put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info);
1119                 }
1120                 if (inet->cmsg_flags & IP_CMSG_TTL) {
1121                         int hlim = inet->mc_ttl;
1122                         put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim);
1123                 }
1124                 len -= msg.msg_controllen;
1125                 return put_user(len, optlen);
1126         }
1127         case IP_FREEBIND:
1128                 val = inet->freebind;
1129                 break;
1130         default:
1131                 release_sock(sk);
1132                 return -ENOPROTOOPT;
1133         }
1134         release_sock(sk);
1135
1136         if (len < sizeof(int) && len > 0 && val>=0 && val<=255) {
1137                 unsigned char ucval = (unsigned char)val;
1138                 len = 1;
1139                 if (put_user(len, optlen))
1140                         return -EFAULT;
1141                 if (copy_to_user(optval,&ucval,1))
1142                         return -EFAULT;
1143         } else {
1144                 len = min_t(unsigned int, sizeof(int), len);
1145                 if (put_user(len, optlen))
1146                         return -EFAULT;
1147                 if (copy_to_user(optval,&val,len))
1148                         return -EFAULT;
1149         }
1150         return 0;
1151 }
1152
1153 int ip_getsockopt(struct sock *sk, int level,
1154                   int optname, char __user *optval, int __user *optlen)
1155 {
1156         int err;
1157
1158         err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1159 #ifdef CONFIG_NETFILTER
1160         /* we need to exclude all possible ENOPROTOOPTs except default case */
1161         if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1162                         !ip_mroute_opt(optname)) {
1163                 int len;
1164
1165                 if (get_user(len,optlen))
1166                         return -EFAULT;
1167
1168                 lock_sock(sk);
1169                 err = nf_getsockopt(sk, PF_INET, optname, optval,
1170                                 &len);
1171                 release_sock(sk);
1172                 if (err >= 0)
1173                         err = put_user(len, optlen);
1174                 return err;
1175         }
1176 #endif
1177         return err;
1178 }
1179
1180 #ifdef CONFIG_COMPAT
1181 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
1182                          char __user *optval, int __user *optlen)
1183 {
1184         int err = do_ip_getsockopt(sk, level, optname, optval, optlen);
1185 #ifdef CONFIG_NETFILTER
1186         /* we need to exclude all possible ENOPROTOOPTs except default case */
1187         if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS &&
1188                         !ip_mroute_opt(optname)) {
1189                 int len;
1190
1191                 if (get_user(len, optlen))
1192                         return -EFAULT;
1193
1194                 lock_sock(sk);
1195                 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len);
1196                 release_sock(sk);
1197                 if (err >= 0)
1198                         err = put_user(len, optlen);
1199                 return err;
1200         }
1201 #endif
1202         return err;
1203 }
1204
1205 EXPORT_SYMBOL(compat_ip_getsockopt);
1206 #endif
1207
1208 EXPORT_SYMBOL(ip_cmsg_recv);
1209
1210 EXPORT_SYMBOL(ip_getsockopt);
1211 EXPORT_SYMBOL(ip_setsockopt);