net/ipv6/raw.c

   1 /*
   2  *      RAW sockets for IPv6
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <pedro_m@yahoo.com>
   7  *
   8  *      Adapted from linux/net/ipv4/raw.c
   9  *
  10  *      $Id: raw.c,v 1.50.2.1 2002/03/05 12:47:34 davem Exp $
  11  *
  12  *      Fixes:
  13  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  14  *      YOSHIFUJI,H.@USAGI      :       raw checksum (RFC2292(bis) compliance)
  15  *
  16  *      This program is free software; you can redistribute it and/or
  17  *      modify it under the terms of the GNU General Public License
  18  *      as published by the Free Software Foundation; either version
  19  *      2 of the License, or (at your option) any later version.
  20  */
  21
  22 #include <linux/errno.h>
  23 #include <linux/types.h>
  24 #include <linux/socket.h>
  25 #include <linux/sockios.h>
  26 #include <linux/sched.h>
  27 #include <linux/net.h>
  28 #include <linux/in6.h>
  29 #include <linux/netdevice.h>
  30 #include <linux/if_arp.h>
  31 #include <linux/icmpv6.h>
  32 #include <asm/uaccess.h>
  33 #include <asm/ioctls.h>
  34
  35 #include <net/sock.h>
  36 #include <net/snmp.h>
  37
  38 #include <net/ipv6.h>
  39 #include <net/ndisc.h>
  40 #include <net/protocol.h>
  41 #include <net/ip6_route.h>
  42 #include <net/addrconf.h>
  43 #include <net/transp_v6.h>
  44 #include <net/udp.h>
  45 #include <net/inet_common.h>
  46
  47 #include <net/rawv6.h>
  48
  49 struct sock *raw_v6_htable[RAWV6_HTABLE_SIZE];
  50 rwlock_t raw_v6_lock = RW_LOCK_UNLOCKED;
  51
  52 static void raw_v6_hash(struct sock *sk)
  53 {
  54         struct sock **skp = &raw_v6_htable[sk->num & (RAWV6_HTABLE_SIZE - 1)];
  55
  56         write_lock_bh(&raw_v6_lock);
  57         if ((sk->next = *skp) != NULL)
  58                 (*skp)->pprev = &sk->next;
  59         *skp = sk;
  60         sk->pprev = skp;
  61         sock_prot_inc_use(sk->prot);
  62         sock_hold(sk);
  63         write_unlock_bh(&raw_v6_lock);
  64 }
  65
  66 static void raw_v6_unhash(struct sock *sk)
  67 {
  68         write_lock_bh(&raw_v6_lock);
  69         if (sk->pprev) {
  70                 if (sk->next)
  71                         sk->next->pprev = sk->pprev;
  72                 *sk->pprev = sk->next;
  73                 sk->pprev = NULL;
  74                 sock_prot_dec_use(sk->prot);
  75                 __sock_put(sk);
  76         }
  77         write_unlock_bh(&raw_v6_lock);
  78 }
  79
  80
  81 /* Grumble... icmp and ip_input want to get at this... */
  82 struct sock *__raw_v6_lookup(struct sock *sk, unsigned short num,
  83                              struct in6_addr *loc_addr, struct in6_addr *rmt_addr)
  84 {
  85         struct sock *s = sk;
  86         int addr_type = ipv6_addr_type(loc_addr);
  87
  88         for(s = sk; s; s = s->next) {
  89                 if(s->num == num) {
  90                         struct ipv6_pinfo *np = &s->net_pinfo.af_inet6;
  91
  92                         if (!ipv6_addr_any(&np->daddr) &&
  93                             ipv6_addr_cmp(&np->daddr, rmt_addr))
  94                                 continue;
  95
  96                         if (!ipv6_addr_any(&np->rcv_saddr)) {
  97                                 if (ipv6_addr_cmp(&np->rcv_saddr, loc_addr) == 0)
  98                                         break;
  99                                 if ((addr_type & IPV6_ADDR_MULTICAST) &&
 100                                     inet6_mc_check(s, loc_addr, rmt_addr))
 101                                         break;
 102                                 continue;
 103                         }
 104                         break;
 105                 }
 106         }
 107         return s;
 108 }
 109
 110 /*
 111  *      0 - deliver
 112  *      1 - block
 113  */
 114 static __inline__ int icmpv6_filter(struct sock *sk, struct sk_buff *skb)
 115 {
 116         struct icmp6hdr *icmph;
 117         struct raw6_opt *opt;
 118
 119         opt = &sk->tp_pinfo.tp_raw;
 120         if (pskb_may_pull(skb, sizeof(struct icmp6hdr))) {
 121                 __u32 *data = &opt->filter.data[0];
 122                 int bit_nr;
 123
 124                 icmph = (struct icmp6hdr *) skb->data;
 125                 bit_nr = icmph->icmp6_type;
 126
 127                 return (data[bit_nr >> 5] & (1 << (bit_nr & 31))) != 0;
 128         }
 129         return 0;
 130 }
 131
 132 /*
 133  *      demultiplex raw sockets.
 134  *      (should consider queueing the skb in the sock receive_queue
 135  *      without calling rawv6.c)
 136  */
 137 struct sock * ipv6_raw_deliver(struct sk_buff *skb, int nexthdr)
 138 {
 139         struct in6_addr *saddr;
 140         struct in6_addr *daddr;
 141         struct sock *sk, *sk2;
 142         __u8 hash;
 143
 144         saddr = &skb->nh.ipv6h->saddr;
 145         daddr = saddr + 1;
 146
 147         hash = nexthdr & (MAX_INET_PROTOS - 1);
 148
 149         read_lock(&raw_v6_lock);
 150         sk = raw_v6_htable[hash];
 151
 152         /*
 153          *      The first socket found will be delivered after
 154          *      delivery to transport protocols.
 155          */
 156
 157         if (sk == NULL)
 158                 goto out;
 159
 160         sk = __raw_v6_lookup(sk, nexthdr, daddr, saddr);
 161
 162         if (sk) {
 163                 sk2 = sk;
 164
 165                 while ((sk2 = __raw_v6_lookup(sk2->next, nexthdr, daddr, saddr))) {
 166                         struct sk_buff *buff;
 167
 168                         if (nexthdr == IPPROTO_ICMPV6 &&
 169                             icmpv6_filter(sk2, skb))
 170                                 continue;
 171
 172                         buff = skb_clone(skb, GFP_ATOMIC);
 173                         if (buff)
 174                                 rawv6_rcv(sk2, buff);
 175                 }
 176         }
 177
 178         if (sk && nexthdr == IPPROTO_ICMPV6 && icmpv6_filter(sk, skb))
 179                 sk = NULL;
 180
 181 out:
 182         if (sk)
 183                 sock_hold(sk);
 184         read_unlock(&raw_v6_lock);
 185         return sk;
 186 }
 187
 188 /* This cleans up af_inet6 a bit. -DaveM */
 189 static int rawv6_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 190 {
 191         struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 192         __u32 v4addr = 0;
 193         int addr_type;
 194         int err;
 195
 196         if (addr_len < SIN6_LEN_RFC2133)
 197                 return -EINVAL;
 198         addr_type = ipv6_addr_type(&addr->sin6_addr);
 199
 200         /* Raw sockets are IPv6 only */
 201         if (addr_type == IPV6_ADDR_MAPPED)
 202                 return(-EADDRNOTAVAIL);
 203
 204         lock_sock(sk);
 205
 206         err = -EINVAL;
 207         if (sk->state != TCP_CLOSE)
 208                 goto out;
 209
 210         if (addr_type & IPV6_ADDR_LINKLOCAL) {
 211                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 212                     addr->sin6_scope_id) {
 213                         /* Override any existing binding, if another one
 214                          * is supplied by user.
 215                          */
 216                         sk->bound_dev_if = addr->sin6_scope_id;
 217                 }
 218
 219                 /* Binding to link-local address requires an interface */
 220                 if (sk->bound_dev_if == 0)
 221                         goto out;
 222         }
 223
 224         /* Check if the address belongs to the host. */
 225         if (addr_type != IPV6_ADDR_ANY) {
 226                 /* ipv4 addr of the socket is invalid.  Only the
 227                  * unpecified and mapped address have a v4 equivalent.
 228                  */
 229                 v4addr = LOOPBACK4_IPV6;
 230                 if (!(addr_type & IPV6_ADDR_MULTICAST)) {
 231                         err = -EADDRNOTAVAIL;
 232                         if (!ipv6_chk_addr(&addr->sin6_addr, NULL))
 233                                 goto out;
 234                 }
 235         }
 236
 237         sk->rcv_saddr = v4addr;
 238         sk->saddr = v4addr;
 239         ipv6_addr_copy(&sk->net_pinfo.af_inet6.rcv_saddr, &addr->sin6_addr);
 240         if (!(addr_type & IPV6_ADDR_MULTICAST))
 241                 ipv6_addr_copy(&sk->net_pinfo.af_inet6.saddr, &addr->sin6_addr);
 242         err = 0;
 243 out:
 244         release_sock(sk);
 245         return err;
 246 }
 247
 248 void rawv6_err(struct sock *sk, struct sk_buff *skb,
 249                struct inet6_skb_parm *opt,
 250                int type, int code, int offset, u32 info)
 251 {
 252         int err;
 253         int harderr;
 254
 255         /* Report error on raw socket, if:
 256            1. User requested recverr.
 257            2. Socket is connected (otherwise the error indication
 258               is useless without recverr and error is hard.
 259          */
 260         if (!sk->net_pinfo.af_inet6.recverr && sk->state != TCP_ESTABLISHED)
 261                 return;
 262
 263         harderr = icmpv6_err_convert(type, code, &err);
 264         if (type == ICMPV6_PKT_TOOBIG)
 265                 harderr = (sk->net_pinfo.af_inet6.pmtudisc == IPV6_PMTUDISC_DO);
 266
 267         if (sk->net_pinfo.af_inet6.recverr) {
 268                 u8 *payload = skb->data;
 269                 if (!sk->protinfo.af_inet.hdrincl)
 270                         payload += offset;
 271                 ipv6_icmp_error(sk, skb, err, 0, ntohl(info), payload);
 272         }
 273
 274         if (sk->net_pinfo.af_inet6.recverr || harderr) {
 275                 sk->err = err;
 276                 sk->error_report(sk);
 277         }
 278 }
 279
 280 static inline int rawv6_rcv_skb(struct sock * sk, struct sk_buff * skb)
 281 {
 282         if ((sk->tp_pinfo.tp_raw.checksum
 283 #if defined(CONFIG_FILTER)
 284             || sk->filter
 285 #endif
 286             ) && skb->ip_summed != CHECKSUM_UNNECESSARY) {
 287                 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
 288                         IP6_INC_STATS_BH(Ip6InDiscards);
 289                         kfree_skb(skb);
 290                         return 0;
 291                 }
 292                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 293         }
 294         /* Charge it to the socket. */
 295         if (sock_queue_rcv_skb(sk,skb)<0) {
 296                 IP6_INC_STATS_BH(Ip6InDiscards);
 297                 kfree_skb(skb);
 298                 return 0;
 299         }
 300
 301         IP6_INC_STATS_BH(Ip6InDelivers);
 302         return 0;
 303 }
 304
 305 /*
 306  *      This is next to useless...
 307  *      if we demultiplex in network layer we don't need the extra call
 308  *      just to queue the skb...
 309  *      maybe we could have the network decide uppon a hint if it
 310  *      should call raw_rcv for demultiplexing
 311  */
 312 int rawv6_rcv(struct sock *sk, struct sk_buff *skb)
 313 {
 314         if (!sk->tp_pinfo.tp_raw.checksum)
 315                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 316
 317         if (skb->ip_summed != CHECKSUM_UNNECESSARY) {
 318                 if (skb->ip_summed == CHECKSUM_HW) {
 319                         skb->ip_summed = CHECKSUM_UNNECESSARY;
 320                         if (csum_ipv6_magic(&skb->nh.ipv6h->saddr,
 321                                             &skb->nh.ipv6h->daddr,
 322                                             skb->len, sk->num, skb->csum)) {
 323                                 NETDEBUG(if (net_ratelimit()) printk(KERN_DEBUG "raw v6 hw csum failure.\n"));
 324                                 skb->ip_summed = CHECKSUM_NONE;
 325                         }
 326                 }
 327                 if (skb->ip_summed == CHECKSUM_NONE)
 328                         skb->csum = ~csum_ipv6_magic(&skb->nh.ipv6h->saddr,
 329                                                      &skb->nh.ipv6h->daddr,
 330                                                      skb->len, sk->num, 0);
 331         }
 332
 333         if (sk->protinfo.af_inet.hdrincl) {
 334                 if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
 335                     (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
 336                         IP6_INC_STATS_BH(Ip6InDiscards);
 337                         kfree_skb(skb);
 338                         return 0;
 339                 }
 340                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 341         }
 342
 343         rawv6_rcv_skb(sk, skb);
 344         return 0;
 345 }
 346
 347
 348 /*
 349  *      This should be easy, if there is something there
 350  *      we return it, otherwise we block.
 351  */
 352
 353 int rawv6_recvmsg(struct sock *sk, struct msghdr *msg, int len,
 354                   int noblock, int flags, int *addr_len)
 355 {
 356         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)msg->msg_name;
 357         struct sk_buff *skb;
 358         int copied, err;
 359
 360         if (flags & MSG_OOB)
 361                 return -EOPNOTSUPP;
 362
 363         if (addr_len)
 364                 *addr_len=sizeof(*sin6);
 365
 366         if (flags & MSG_ERRQUEUE)
 367                 return ipv6_recv_error(sk, msg, len);
 368
 369         skb = skb_recv_datagram(sk, flags, noblock, &err);
 370         if (!skb)
 371                 goto out;
 372
 373         copied = skb->len;
 374         if (copied > len) {
 375                 copied = len;
 376                 msg->msg_flags |= MSG_TRUNC;
 377         }
 378
 379         if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
 380                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 381         } else if (msg->msg_flags&MSG_TRUNC) {
 382                 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
 383                         goto csum_copy_err;
 384                 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
 385         } else {
 386                 err = skb_copy_and_csum_datagram_iovec(skb, 0, msg->msg_iov);
 387                 if (err == -EINVAL)
 388                         goto csum_copy_err;
 389         }
 390         if (err)
 391                 goto out_free;
 392
 393         /* Copy the address. */
 394         if (sin6) {
 395                 sin6->sin6_family = AF_INET6;
 396                 memcpy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr,
 397                        sizeof(struct in6_addr));
 398                 sin6->sin6_flowinfo = 0;
 399                 sin6->sin6_scope_id = 0;
 400                 if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL) {
 401                         struct inet6_skb_parm *opt = (struct inet6_skb_parm *) skb->cb;
 402                         sin6->sin6_scope_id = opt->iif;
 403                 }
 404         }
 405
 406         sock_recv_timestamp(msg, sk, skb);
 407
 408         if (sk->net_pinfo.af_inet6.rxopt.all)
 409                 datagram_recv_ctl(sk, msg, skb);
 410
 411         err = copied;
 412         if (flags & MSG_TRUNC)
 413                 err = skb->len;
 414
 415 out_free:
 416         skb_free_datagram(sk, skb);
 417 out:
 418         return err;
 419
 420 csum_copy_err:
 421         /* Clear queue. */
 422         if (flags&MSG_PEEK) {
 423                 int clear = 0;
 424                 spin_lock_irq(&sk->receive_queue.lock);
 425                 if (skb == skb_peek(&sk->receive_queue)) {
 426                         __skb_unlink(skb, &sk->receive_queue);
 427                         clear = 1;
 428                 }
 429                 spin_unlock_irq(&sk->receive_queue.lock);
 430                 if (clear)
 431                         kfree_skb(skb);
 432         }
 433
 434         /* Error for blocking case is chosen to masquerade
 435            as some normal condition.
 436          */
 437         err = (flags&MSG_DONTWAIT) ? -EAGAIN : -EHOSTUNREACH;
 438         IP6_INC_STATS_USER(Ip6InDiscards);
 439         goto out_free;
 440 }
 441
 442 /*
 443  *      Sending...
 444  */
 445
 446 struct rawv6_fakehdr {
 447         struct iovec    *iov;
 448         struct sock     *sk;
 449         __u32           len;
 450         __u32           cksum;
 451         __u32           proto;
 452         struct in6_addr *daddr;
 453 };
 454
 455 static int rawv6_getfrag(const void *data, struct in6_addr *saddr,
 456                           char *buff, unsigned int offset, unsigned int len)
 457 {
 458         struct iovec *iov = (struct iovec *) data;
 459
 460         return memcpy_fromiovecend(buff, iov, offset, len);
 461 }
 462
 463 static int rawv6_frag_cksum(const void *data, struct in6_addr *addr,
 464                              char *buff, unsigned int offset,
 465                              unsigned int len)
 466 {
 467         struct rawv6_fakehdr *hdr = (struct rawv6_fakehdr *) data;
 468
 469         if (csum_partial_copy_fromiovecend(buff, hdr->iov, offset,
 470                                                     len, &hdr->cksum))
 471                 return -EFAULT;
 472
 473         if (offset == 0) {
 474                 struct sock *sk;
 475                 struct raw6_opt *opt;
 476                 struct in6_addr *daddr;
 477
 478                 sk = hdr->sk;
 479                 opt = &sk->tp_pinfo.tp_raw;
 480
 481                 if (hdr->daddr)
 482                         daddr = hdr->daddr;
 483                 else
 484                         daddr = addr + 1;
 485
 486                 hdr->cksum = csum_ipv6_magic(addr, daddr, hdr->len,
 487                                              hdr->proto, hdr->cksum);
 488
 489                 if (opt->offset + 1 < len) {
 490                         __u16 *csum;
 491
 492                         csum = (__u16 *) (buff + opt->offset);
 493                         if (*csum) {
 494                                 /* in case cksum was not initialized */
 495                                 __u32 sum = hdr->cksum;
 496                                 sum += *csum;
 497                                 *csum = hdr->cksum = (sum + (sum>>16));
 498                         } else {
 499                                 *csum = hdr->cksum;
 500                         }
 501                 } else {
 502                         if (net_ratelimit())
 503                                 printk(KERN_DEBUG "icmp: cksum offset too big\n");
 504                         return -EINVAL;
 505                 }
 506         }
 507         return 0;
 508 }
 509
 510
 511 static int rawv6_sendmsg(struct sock *sk, struct msghdr *msg, int len)
 512 {
 513         struct ipv6_txoptions opt_space;
 514         struct sockaddr_in6 * sin6 = (struct sockaddr_in6 *) msg->msg_name;
 515         struct ipv6_pinfo *np = &sk->net_pinfo.af_inet6;
 516         struct ipv6_txoptions *opt = NULL;
 517         struct ip6_flowlabel *flowlabel = NULL;
 518         struct flowi fl;
 519         int addr_len = msg->msg_namelen;
 520         struct in6_addr *daddr;
 521         struct raw6_opt *raw_opt;
 522         int hlimit = -1;
 523         u16 proto;
 524         int err;
 525
 526         /* Rough check on arithmetic overflow,
 527            better check is made in ip6_build_xmit
 528          */
 529         if (len < 0)
 530                 return -EMSGSIZE;
 531
 532         /* Mirror BSD error message compatibility */
 533         if (msg->msg_flags & MSG_OOB)
 534                 return -EOPNOTSUPP;
 535
 536         /*
 537          *      Get and verify the address.
 538          */
 539
 540         fl.fl6_flowlabel = 0;
 541         fl.oif = 0;
 542
 543         if (sin6) {
 544                 if (addr_len < SIN6_LEN_RFC2133)
 545                         return -EINVAL;
 546
 547                 if (sin6->sin6_family && sin6->sin6_family != AF_INET6)
 548                         return(-EINVAL);
 549
 550                 /* port is the proto value [0..255] carried in nexthdr */
 551                 proto = ntohs(sin6->sin6_port);
 552
 553                 if (!proto)
 554                         proto = sk->num;
 555
 556                 if (proto > 255)
 557                         return(-EINVAL);
 558
 559                 daddr = &sin6->sin6_addr;
 560                 if (np->sndflow) {
 561                         fl.fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
 562                         if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
 563                                 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
 564                                 if (flowlabel == NULL)
 565                                         return -EINVAL;
 566                                 daddr = &flowlabel->dst;
 567                         }
 568                 }
 569
 570                 /* Otherwise it will be difficult to maintain sk->dst_cache. */
 571                 if (sk->state == TCP_ESTABLISHED &&
 572                     !ipv6_addr_cmp(daddr, &sk->net_pinfo.af_inet6.daddr))
 573                         daddr = &sk->net_pinfo.af_inet6.daddr;
 574
 575                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 576                     sin6->sin6_scope_id &&
 577                     ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
 578                         fl.oif = sin6->sin6_scope_id;
 579         } else {
 580                 if (sk->state != TCP_ESTABLISHED)
 581                         return -EDESTADDRREQ;
 582
 583                 proto = sk->num;
 584                 daddr = &(sk->net_pinfo.af_inet6.daddr);
 585                 fl.fl6_flowlabel = np->flow_label;
 586         }
 587
 588         if (ipv6_addr_any(daddr)) {
 589                 /*
 590                  * unspecfied destination address
 591                  * treated as error... is this correct ?
 592                  */
 593                 return(-EINVAL);
 594         }
 595
 596         if (fl.oif == 0)
 597                 fl.oif = sk->bound_dev_if;
 598         fl.fl6_src = NULL;
 599
 600         if (msg->msg_controllen) {
 601                 opt = &opt_space;
 602                 memset(opt, 0, sizeof(struct ipv6_txoptions));
 603
 604                 err = datagram_send_ctl(msg, &fl, opt, &hlimit);
 605                 if (err < 0) {
 606                         fl6_sock_release(flowlabel);
 607                         return err;
 608                 }
 609                 if ((fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
 610                         flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
 611                         if (flowlabel == NULL)
 612                                 return -EINVAL;
 613                 }
 614                 if (!(opt->opt_nflen|opt->opt_flen))
 615                         opt = NULL;
 616         }
 617         if (opt == NULL)
 618                 opt = np->opt;
 619         if (flowlabel)
 620                 opt = fl6_merge_options(&opt_space, flowlabel, opt);
 621
 622         raw_opt = &sk->tp_pinfo.tp_raw;
 623
 624         fl.proto = proto;
 625         fl.fl6_dst = daddr;
 626         if (fl.fl6_src == NULL && !ipv6_addr_any(&np->saddr))
 627                 fl.fl6_src = &np->saddr;
 628         fl.uli_u.icmpt.type = 0;
 629         fl.uli_u.icmpt.code = 0;
 630
 631         if (raw_opt->checksum) {
 632                 struct rawv6_fakehdr hdr;
 633
 634                 hdr.iov = msg->msg_iov;
 635                 hdr.sk  = sk;
 636                 hdr.len = len;
 637                 hdr.cksum = 0;
 638                 hdr.proto = proto;
 639
 640                 if (opt && opt->srcrt)
 641                         hdr.daddr = daddr;
 642                 else
 643                         hdr.daddr = NULL;
 644
 645                 err = ip6_build_xmit(sk, rawv6_frag_cksum, &hdr, &fl, len,
 646                                      opt, hlimit, msg->msg_flags);
 647         } else {
 648                 err = ip6_build_xmit(sk, rawv6_getfrag, msg->msg_iov, &fl, len,
 649                                      opt, hlimit, msg->msg_flags);
 650         }
 651
 652         fl6_sock_release(flowlabel);
 653
 654         return err<0?err:len;
 655 }
 656
 657 static int rawv6_seticmpfilter(struct sock *sk, int level, int optname,
 658                                char *optval, int optlen)
 659 {
 660         switch (optname) {
 661         case ICMPV6_FILTER:
 662                 if (optlen > sizeof(struct icmp6_filter))
 663                         optlen = sizeof(struct icmp6_filter);
 664                 if (copy_from_user(&sk->tp_pinfo.tp_raw.filter, optval, optlen))
 665                         return -EFAULT;
 666                 return 0;
 667         default:
 668                 return -ENOPROTOOPT;
 669         };
 670
 671         return 0;
 672 }
 673
 674 static int rawv6_geticmpfilter(struct sock *sk, int level, int optname,
 675                                char *optval, int *optlen)
 676 {
 677         int len;
 678
 679         switch (optname) {
 680         case ICMPV6_FILTER:
 681                 if (get_user(len, optlen))
 682                         return -EFAULT;
 683                 if (len < 0)
 684                         return -EINVAL;
 685                 if (len > sizeof(struct icmp6_filter))
 686                         len = sizeof(struct icmp6_filter);
 687                 if (put_user(len, optlen))
 688                         return -EFAULT;
 689                 if (copy_to_user(optval, &sk->tp_pinfo.tp_raw.filter, len))
 690                         return -EFAULT;
 691                 return 0;
 692         default:
 693                 return -ENOPROTOOPT;
 694         };
 695
 696         return 0;
 697 }
 698
 699
 700 static int rawv6_setsockopt(struct sock *sk, int level, int optname,
 701                             char *optval, int optlen)
 702 {
 703         struct raw6_opt *opt = &sk->tp_pinfo.tp_raw;
 704         int val;
 705
 706         switch(level) {
 707                 case SOL_RAW:
 708                         break;
 709
 710                 case SOL_ICMPV6:
 711                         if (sk->num != IPPROTO_ICMPV6)
 712                                 return -EOPNOTSUPP;
 713                         return rawv6_seticmpfilter(sk, level, optname, optval,
 714                                                    optlen);
 715                 case SOL_IPV6:
 716                         if (optname == IPV6_CHECKSUM)
 717                                 break;
 718                 default:
 719                         return ipv6_setsockopt(sk, level, optname, optval,
 720                                                optlen);
 721         };
 722
 723         if (get_user(val, (int *)optval))
 724                 return -EFAULT;
 725
 726         switch (optname) {
 727                 case IPV6_CHECKSUM:
 728                         /* You may get strange result with a positive odd offset;
 729                            RFC2292bis agrees with me. */
 730                         if (val > 0 && (val&1))
 731                                 return(-EINVAL);
 732                         if (val < 0) {
 733                                 opt->checksum = 0;
 734                         } else {
 735                                 opt->checksum = 1;
 736                                 opt->offset = val;
 737                         }
 738
 739                         return 0;
 740                         break;
 741
 742                 default:
 743                         return(-ENOPROTOOPT);
 744         }
 745 }
 746
 747 static int rawv6_getsockopt(struct sock *sk, int level, int optname,
 748                             char *optval, int *optlen)
 749 {
 750         struct raw6_opt *opt = &sk->tp_pinfo.tp_raw;
 751         int val, len;
 752
 753         switch(level) {
 754                 case SOL_RAW:
 755                         break;
 756
 757                 case SOL_ICMPV6:
 758                         if (sk->num != IPPROTO_ICMPV6)
 759                                 return -EOPNOTSUPP;
 760                         return rawv6_geticmpfilter(sk, level, optname, optval,
 761                                                    optlen);
 762                 case SOL_IPV6:
 763                         if (optname == IPV6_CHECKSUM)
 764                                 break;
 765                 default:
 766                         return ipv6_getsockopt(sk, level, optname, optval,
 767                                                optlen);
 768         };
 769
 770         if (get_user(len,optlen))
 771                 return -EFAULT;
 772
 773         switch (optname) {
 774         case IPV6_CHECKSUM:
 775                 if (opt->checksum == 0)
 776                         val = -1;
 777                 else
 778                         val = opt->offset;
 779                 break;
 780
 781         default:
 782                 return -ENOPROTOOPT;
 783         }
 784
 785         len = min_t(unsigned int, sizeof(int), len);
 786
 787         if (put_user(len, optlen))
 788                 return -EFAULT;
 789         if (copy_to_user(optval,&val,len))
 790                 return -EFAULT;
 791         return 0;
 792 }
 793
 794 static int rawv6_ioctl(struct sock *sk, int cmd, unsigned long arg)
 795 {
 796         switch(cmd) {
 797                 case SIOCOUTQ:
 798                 {
 799                         int amount = atomic_read(&sk->wmem_alloc);
 800                         return put_user(amount, (int *)arg);
 801                 }
 802                 case SIOCINQ:
 803                 {
 804                         struct sk_buff *skb;
 805                         int amount = 0;
 806
 807                         spin_lock_irq(&sk->receive_queue.lock);
 808                         skb = skb_peek(&sk->receive_queue);
 809                         if (skb != NULL)
 810                                 amount = skb->tail - skb->h.raw;
 811                         spin_unlock_irq(&sk->receive_queue.lock);
 812                         return put_user(amount, (int *)arg);
 813                 }
 814
 815                 default:
 816                         return -ENOIOCTLCMD;
 817         }
 818 }
 819
 820 static void rawv6_close(struct sock *sk, long timeout)
 821 {
 822         if (sk->num == IPPROTO_RAW)
 823                 ip6_ra_control(sk, -1, NULL);
 824
 825         inet_sock_release(sk);
 826 }
 827
 828 static int rawv6_init_sk(struct sock *sk)
 829 {
 830         if (sk->num == IPPROTO_ICMPV6){
 831                 struct raw6_opt *opt = &sk->tp_pinfo.tp_raw;
 832                 opt->checksum = 1;
 833                 opt->offset = 2;
 834         }
 835         return(0);
 836 }
 837
 838 #define LINE_LEN 190
 839 #define LINE_FMT "%-190s\n"
 840
 841 static void get_raw6_sock(struct sock *sp, char *tmpbuf, int i)
 842 {
 843         struct in6_addr *dest, *src;
 844         __u16 destp, srcp;
 845
 846         dest  = &sp->net_pinfo.af_inet6.daddr;
 847         src   = &sp->net_pinfo.af_inet6.rcv_saddr;
 848         destp = 0;
 849         srcp  = sp->num;
 850         sprintf(tmpbuf,
 851                 "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
 852                 "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
 853                 i,
 854                 src->s6_addr32[0], src->s6_addr32[1],
 855                 src->s6_addr32[2], src->s6_addr32[3], srcp,
 856                 dest->s6_addr32[0], dest->s6_addr32[1],
 857                 dest->s6_addr32[2], dest->s6_addr32[3], destp,
 858                 sp->state,
 859                 atomic_read(&sp->wmem_alloc), atomic_read(&sp->rmem_alloc),
 860                 0, 0L, 0,
 861                 sock_i_uid(sp), 0,
 862                 sock_i_ino(sp),
 863                 atomic_read(&sp->refcnt), sp);
 864 }
 865
 866 int raw6_get_info(char *buffer, char **start, off_t offset, int length)
 867 {
 868         int len = 0, num = 0, i;
 869         off_t pos = 0;
 870         off_t begin;
 871         char tmpbuf[LINE_LEN+2];
 872
 873         if (offset < LINE_LEN+1)
 874                 len += sprintf(buffer, LINE_FMT,
 875                                "  sl  "                                         /* 6 */
 876                                "local_address                         "         /* 38 */
 877                                "remote_address                        "         /* 38 */
 878                                "st tx_queue rx_queue tr tm->when retrnsmt"      /* 41 */
 879                                "   uid  timeout inode");                        /* 21 */
 880                                                                                 /*----*/
 881                                                                                 /*144 */
 882         pos = LINE_LEN+1;
 883         read_lock(&raw_v6_lock);
 884         for (i = 0; i < RAWV6_HTABLE_SIZE; i++) {
 885                 struct sock *sk;
 886
 887                 for (sk = raw_v6_htable[i]; sk; sk = sk->next, num++) {
 888                         if (sk->family != PF_INET6)
 889                                 continue;
 890                         pos += LINE_LEN+1;
 891                         if (pos <= offset)
 892                                 continue;
 893                         get_raw6_sock(sk, tmpbuf, i);
 894                         len += sprintf(buffer+len, LINE_FMT, tmpbuf);
 895                         if(len >= length)
 896                                 goto out;
 897                 }
 898         }
 899 out:
 900         read_unlock(&raw_v6_lock);
 901         begin = len - (pos - offset);
 902         *start = buffer + begin;
 903         len -= begin;
 904         if(len > length)
 905                 len = length;
 906         if (len < 0)
 907                 len = 0;
 908         return len;
 909 }
 910
 911 struct proto rawv6_prot = {
 912         name:           "RAW",
 913         close:          rawv6_close,
 914         connect:        udpv6_connect,
 915         disconnect:     udp_disconnect,
 916         ioctl:          rawv6_ioctl,
 917         init:           rawv6_init_sk,
 918         destroy:        inet6_destroy_sock,
 919         setsockopt:     rawv6_setsockopt,
 920         getsockopt:     rawv6_getsockopt,
 921         sendmsg:        rawv6_sendmsg,
 922         recvmsg:        rawv6_recvmsg,
 923         bind:           rawv6_bind,
 924         backlog_rcv:    rawv6_rcv_skb,
 925         hash:           raw_v6_hash,
 926         unhash:         raw_v6_unhash,
 927 };