net/ipv6/udp.c

   1 /*
   2  *      UDP over IPv6
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      Based on linux/ipv4/udp.c
   9  *
  10  *      $Id: udp.c,v 1.65 2002/02/01 22:01:04 davem Exp $
  11  *
  12  *      Fixes:
  13  *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  14  *      YOSHIFUJI Hideaki @USAGI and:   Support IPV6_V6ONLY socket option, which
  15  *      Alexey Kuznetsov                allow both IPv4 and IPv6 sockets to bind
  16  *                                      a single port at the same time.
  17  *      Kazunori MIYAZAWA @USAGI:       change process style to use ip6_append_data
  18  *      YOSHIFUJI Hideaki @USAGI:       convert /proc/net/udp6 to seq_file.
  19  *
  20  *      This program is free software; you can redistribute it and/or
  21  *      modify it under the terms of the GNU General Public License
  22  *      as published by the Free Software Foundation; either version
  23  *      2 of the License, or (at your option) any later version.
  24  */
  25
  26 #include <linux/config.h>
  27 #include <linux/errno.h>
  28 #include <linux/types.h>
  29 #include <linux/socket.h>
  30 #include <linux/sockios.h>
  31 #include <linux/sched.h>
  32 #include <linux/net.h>
  33 #include <linux/in6.h>
  34 #include <linux/netdevice.h>
  35 #include <linux/if_arp.h>
  36 #include <linux/ipv6.h>
  37 #include <linux/icmpv6.h>
  38 #include <linux/init.h>
  39 #include <asm/uaccess.h>
  40
  41 #include <net/sock.h>
  42 #include <net/snmp.h>
  43
  44 #include <net/ipv6.h>
  45 #include <net/ndisc.h>
  46 #include <net/protocol.h>
  47 #include <net/transp_v6.h>
  48 #include <net/ip6_route.h>
  49 #include <net/addrconf.h>
  50 #include <net/ip.h>
  51 #include <net/udp.h>
  52 #include <net/inet_common.h>
  53
  54 #include <net/checksum.h>
  55 #include <net/xfrm.h>
  56
  57 #include <linux/proc_fs.h>
  58 #include <linux/seq_file.h>
  59
  60 DEFINE_SNMP_STAT(struct udp_mib, udp_stats_in6);
  61
  62 /* Grrr, addr_type already calculated by caller, but I don't want
  63  * to add some silly "cookie" argument to this method just for that.
  64  */
  65 static int udp_v6_get_port(struct sock *sk, unsigned short snum)
  66 {
  67         struct sock *sk2;
  68         struct hlist_node *node;
  69
  70         write_lock_bh(&udp_hash_lock);
  71         if (snum == 0) {
  72                 int best_size_so_far, best, result, i;
  73
  74                 if (udp_port_rover > sysctl_local_port_range[1] ||
  75                     udp_port_rover < sysctl_local_port_range[0])
  76                         udp_port_rover = sysctl_local_port_range[0];
  77                 best_size_so_far = 32767;
  78                 best = result = udp_port_rover;
  79                 for (i = 0; i < UDP_HTABLE_SIZE; i++, result++) {
  80                         int size;
  81                         struct hlist_head *list;
  82
  83                         list = &udp_hash[result & (UDP_HTABLE_SIZE - 1)];
  84                         if (hlist_empty(list)) {
  85                                 if (result > sysctl_local_port_range[1])
  86                                         result = sysctl_local_port_range[0] +
  87                                                 ((result - sysctl_local_port_range[0]) &
  88                                                  (UDP_HTABLE_SIZE - 1));
  89                                 goto gotit;
  90                         }
  91                         size = 0;
  92                         sk_for_each(sk2, node, list)
  93                                 if (++size >= best_size_so_far)
  94                                         goto next;
  95                         best_size_so_far = size;
  96                         best = result;
  97                 next:;
  98                 }
  99                 result = best;
 100                 for(;; result += UDP_HTABLE_SIZE) {
 101                         if (result > sysctl_local_port_range[1])
 102                                 result = sysctl_local_port_range[0]
 103                                         + ((result - sysctl_local_port_range[0]) &
 104                                            (UDP_HTABLE_SIZE - 1));
 105                         if (!udp_lport_inuse(result))
 106                                 break;
 107                 }
 108 gotit:
 109                 udp_port_rover = snum = result;
 110         } else {
 111                 sk_for_each(sk2, node,
 112                             &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]) {
 113                         if (inet_sk(sk2)->num == snum &&
 114                             sk2 != sk &&
 115                             (!sk2->sk_bound_dev_if ||
 116                              !sk->sk_bound_dev_if ||
 117                              sk2->sk_bound_dev_if == sk->sk_bound_dev_if) &&
 118                             (!sk2->sk_reuse || !sk->sk_reuse) &&
 119                             ipv6_rcv_saddr_equal(sk, sk2))
 120                                 goto fail;
 121                 }
 122         }
 123
 124         inet_sk(sk)->num = snum;
 125         if (sk_unhashed(sk)) {
 126                 sk_add_node(sk, &udp_hash[snum & (UDP_HTABLE_SIZE - 1)]);
 127                 sock_prot_inc_use(sk->sk_prot);
 128         }
 129         write_unlock_bh(&udp_hash_lock);
 130         return 0;
 131
 132 fail:
 133         write_unlock_bh(&udp_hash_lock);
 134         return 1;
 135 }
 136
 137 static void udp_v6_hash(struct sock *sk)
 138 {
 139         BUG();
 140 }
 141
 142 static void udp_v6_unhash(struct sock *sk)
 143 {
 144         write_lock_bh(&udp_hash_lock);
 145         if (sk_del_node_init(sk)) {
 146                 inet_sk(sk)->num = 0;
 147                 sock_prot_dec_use(sk->sk_prot);
 148         }
 149         write_unlock_bh(&udp_hash_lock);
 150 }
 151
 152 static struct sock *udp_v6_lookup(struct in6_addr *saddr, u16 sport,
 153                                   struct in6_addr *daddr, u16 dport, int dif)
 154 {
 155         struct sock *sk, *result = NULL;
 156         struct hlist_node *node;
 157         unsigned short hnum = ntohs(dport);
 158         int badness = -1;
 159
 160         read_lock(&udp_hash_lock);
 161         sk_for_each(sk, node, &udp_hash[hnum & (UDP_HTABLE_SIZE - 1)]) {
 162                 struct inet_opt *inet = inet_sk(sk);
 163
 164                 if (inet->num == hnum && sk->sk_family == PF_INET6) {
 165                         struct ipv6_pinfo *np = inet6_sk(sk);
 166                         int score = 0;
 167                         if (inet->dport) {
 168                                 if (inet->dport != sport)
 169                                         continue;
 170                                 score++;
 171                         }
 172                         if (!ipv6_addr_any(&np->rcv_saddr)) {
 173                                 if (ipv6_addr_cmp(&np->rcv_saddr, daddr))
 174                                         continue;
 175                                 score++;
 176                         }
 177                         if (!ipv6_addr_any(&np->daddr)) {
 178                                 if (ipv6_addr_cmp(&np->daddr, saddr))
 179                                         continue;
 180                                 score++;
 181                         }
 182                         if (sk->sk_bound_dev_if) {
 183                                 if (sk->sk_bound_dev_if != dif)
 184                                         continue;
 185                                 score++;
 186                         }
 187                         if(score == 4) {
 188                                 result = sk;
 189                                 break;
 190                         } else if(score > badness) {
 191                                 result = sk;
 192                                 badness = score;
 193                         }
 194                 }
 195         }
 196         if (result)
 197                 sock_hold(result);
 198         read_unlock(&udp_hash_lock);
 199         return result;
 200 }
 201
 202 /*
 203  *
 204  */
 205
 206 int udpv6_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 207 {
 208         struct sockaddr_in6     *usin = (struct sockaddr_in6 *) uaddr;
 209         struct inet_opt         *inet = inet_sk(sk);
 210         struct ipv6_pinfo       *np = inet6_sk(sk);
 211         struct in6_addr         *daddr;
 212         struct dst_entry        *dst;
 213         struct flowi            fl;
 214         struct ip6_flowlabel    *flowlabel = NULL;
 215         int                     addr_type;
 216         int                     err;
 217
 218         if (usin->sin6_family == AF_INET) {
 219                 if (__ipv6_only_sock(sk))
 220                         return -EAFNOSUPPORT;
 221                 err = udp_connect(sk, uaddr, addr_len);
 222                 goto ipv4_connected;
 223         }
 224
 225         if (addr_len < SIN6_LEN_RFC2133)
 226                 return -EINVAL;
 227
 228         if (usin->sin6_family != AF_INET6)
 229                 return -EAFNOSUPPORT;
 230
 231         memset(&fl, 0, sizeof(fl));
 232         if (np->sndflow) {
 233                 fl.fl6_flowlabel = usin->sin6_flowinfo&IPV6_FLOWINFO_MASK;
 234                 if (fl.fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
 235                         flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel);
 236                         if (flowlabel == NULL)
 237                                 return -EINVAL;
 238                         ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst);
 239                 }
 240         }
 241
 242         addr_type = ipv6_addr_type(&usin->sin6_addr);
 243
 244         if (addr_type == IPV6_ADDR_ANY) {
 245                 /*
 246                  *      connect to self
 247                  */
 248                 usin->sin6_addr.s6_addr[15] = 0x01;
 249         }
 250
 251         daddr = &usin->sin6_addr;
 252
 253         if (addr_type == IPV6_ADDR_MAPPED) {
 254                 struct sockaddr_in sin;
 255
 256                 if (__ipv6_only_sock(sk)) {
 257                         err = -ENETUNREACH;
 258                         goto out;
 259                 }
 260                 sin.sin_family = AF_INET;
 261                 sin.sin_addr.s_addr = daddr->s6_addr32[3];
 262                 sin.sin_port = usin->sin6_port;
 263
 264                 err = udp_connect(sk, (struct sockaddr*) &sin, sizeof(sin));
 265
 266 ipv4_connected:
 267                 if (err)
 268                         goto out;
 269
 270                 ipv6_addr_set(&np->daddr, 0, 0, htonl(0x0000ffff), inet->daddr);
 271
 272                 if (ipv6_addr_any(&np->saddr)) {
 273                         ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000ffff),
 274                                       inet->saddr);
 275                 }
 276
 277                 if (ipv6_addr_any(&np->rcv_saddr)) {
 278                         ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000ffff),
 279                                       inet->rcv_saddr);
 280                 }
 281                 goto out;
 282         }
 283
 284         if (addr_type&IPV6_ADDR_LINKLOCAL) {
 285                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 286                     usin->sin6_scope_id) {
 287                         if (sk->sk_bound_dev_if &&
 288                             sk->sk_bound_dev_if != usin->sin6_scope_id) {
 289                                 err = -EINVAL;
 290                                 goto out;
 291                         }
 292                         sk->sk_bound_dev_if = usin->sin6_scope_id;
 293                         if (!sk->sk_bound_dev_if &&
 294                             (addr_type & IPV6_ADDR_MULTICAST))
 295                                 fl.oif = np->mcast_oif;
 296                 }
 297
 298                 /* Connect to link-local address requires an interface */
 299                 if (!sk->sk_bound_dev_if) {
 300                         err = -EINVAL;
 301                         goto out;
 302                 }
 303         }
 304
 305         ipv6_addr_copy(&np->daddr, daddr);
 306         np->flow_label = fl.fl6_flowlabel;
 307
 308         inet->dport = usin->sin6_port;
 309
 310         /*
 311          *      Check for a route to destination an obtain the
 312          *      destination cache for it.
 313          */
 314
 315         fl.proto = IPPROTO_UDP;
 316         ipv6_addr_copy(&fl.fl6_dst, &np->daddr);
 317         ipv6_addr_copy(&fl.fl6_src, &np->saddr);
 318         fl.oif = sk->sk_bound_dev_if;
 319         fl.fl_ip_dport = inet->dport;
 320         fl.fl_ip_sport = inet->sport;
 321
 322         if (!fl.oif && (addr_type&IPV6_ADDR_MULTICAST))
 323                 fl.oif = np->mcast_oif;
 324
 325         if (flowlabel) {
 326                 if (flowlabel->opt && flowlabel->opt->srcrt) {
 327                         struct rt0_hdr *rt0 = (struct rt0_hdr *) flowlabel->opt->srcrt;
 328                         ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 329                 }
 330         } else if (np->opt && np->opt->srcrt) {
 331                 struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt;
 332                 ipv6_addr_copy(&fl.fl6_dst, rt0->addr);
 333         }
 334
 335         err = ip6_dst_lookup(sk, &dst, &fl);
 336         if (err)
 337                 goto out;
 338
 339         /* source address lookup done in ip6_dst_lookup */
 340
 341         if (ipv6_addr_any(&np->saddr))
 342                 ipv6_addr_copy(&np->saddr, &fl.fl6_src);
 343
 344         if (ipv6_addr_any(&np->rcv_saddr)) {
 345                 ipv6_addr_copy(&np->rcv_saddr, &fl.fl6_src);
 346                 inet->rcv_saddr = LOOPBACK4_IPV6;
 347         }
 348
 349         ip6_dst_store(sk, dst,
 350                       !ipv6_addr_cmp(&fl.fl6_dst, &np->daddr) ?
 351                       &np->daddr : NULL);
 352
 353         sk->sk_state = TCP_ESTABLISHED;
 354 out:
 355         fl6_sock_release(flowlabel);
 356         return err;
 357 }
 358
 359 static void udpv6_close(struct sock *sk, long timeout)
 360 {
 361         inet_sock_release(sk);
 362 }
 363
 364 /*
 365  *      This should be easy, if there is something there we
 366  *      return it, otherwise we block.
 367  */
 368
 369 static int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
 370                   struct msghdr *msg, size_t len,
 371                   int noblock, int flags, int *addr_len)
 372 {
 373         struct ipv6_pinfo *np = inet6_sk(sk);
 374         struct sk_buff *skb;
 375         size_t copied;
 376         int err;
 377
 378         if (addr_len)
 379                 *addr_len=sizeof(struct sockaddr_in6);
 380
 381         if (flags & MSG_ERRQUEUE)
 382                 return ipv6_recv_error(sk, msg, len);
 383
 384 try_again:
 385         skb = skb_recv_datagram(sk, flags, noblock, &err);
 386         if (!skb)
 387                 goto out;
 388
 389         copied = skb->len - sizeof(struct udphdr);
 390         if (copied > len) {
 391                 copied = len;
 392                 msg->msg_flags |= MSG_TRUNC;
 393         }
 394
 395         if (skb->ip_summed==CHECKSUM_UNNECESSARY) {
 396                 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
 397                                               copied);
 398         } else if (msg->msg_flags&MSG_TRUNC) {
 399                 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
 400                         goto csum_copy_err;
 401                 err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov,
 402                                               copied);
 403         } else {
 404                 err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
 405                 if (err == -EINVAL)
 406                         goto csum_copy_err;
 407         }
 408         if (err)
 409                 goto out_free;
 410
 411         sock_recv_timestamp(msg, sk, skb);
 412
 413         /* Copy the address. */
 414         if (msg->msg_name) {
 415                 struct sockaddr_in6 *sin6;
 416
 417                 sin6 = (struct sockaddr_in6 *) msg->msg_name;
 418                 sin6->sin6_family = AF_INET6;
 419                 sin6->sin6_port = skb->h.uh->source;
 420                 sin6->sin6_flowinfo = 0;
 421                 sin6->sin6_scope_id = 0;
 422
 423                 if (skb->protocol == htons(ETH_P_IP)) {
 424                         struct inet_opt *inet = inet_sk(sk);
 425
 426                         ipv6_addr_set(&sin6->sin6_addr, 0, 0,
 427                                       htonl(0xffff), skb->nh.iph->saddr);
 428                         if (inet->cmsg_flags)
 429                                 ip_cmsg_recv(msg, skb);
 430                 } else {
 431                         ipv6_addr_copy(&sin6->sin6_addr, &skb->nh.ipv6h->saddr);
 432
 433                         if (np->rxopt.all)
 434                                 datagram_recv_ctl(sk, msg, skb);
 435                         if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL) {
 436                                 struct inet6_skb_parm *opt = (struct inet6_skb_parm *) skb->cb;
 437                                 sin6->sin6_scope_id = opt->iif;
 438                         }
 439                 }
 440         }
 441         err = copied;
 442
 443 out_free:
 444         skb_free_datagram(sk, skb);
 445 out:
 446         return err;
 447
 448 csum_copy_err:
 449         /* Clear queue. */
 450         if (flags&MSG_PEEK) {
 451                 int clear = 0;
 452                 spin_lock_irq(&sk->sk_receive_queue.lock);
 453                 if (skb == skb_peek(&sk->sk_receive_queue)) {
 454                         __skb_unlink(skb, &sk->sk_receive_queue);
 455                         clear = 1;
 456                 }
 457                 spin_unlock_irq(&sk->sk_receive_queue.lock);
 458                 if (clear)
 459                         kfree_skb(skb);
 460         }
 461
 462         skb_free_datagram(sk, skb);
 463
 464         if (flags & MSG_DONTWAIT) {
 465                 UDP6_INC_STATS_USER(UdpInErrors);
 466                 return -EAGAIN;
 467         }
 468         goto try_again;
 469 }
 470
 471 static void udpv6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 472                int type, int code, int offset, __u32 info)
 473 {
 474         struct ipv6_pinfo *np;
 475         struct ipv6hdr *hdr = (struct ipv6hdr*)skb->data;
 476         struct net_device *dev = skb->dev;
 477         struct in6_addr *saddr = &hdr->saddr;
 478         struct in6_addr *daddr = &hdr->daddr;
 479         struct udphdr *uh = (struct udphdr*)(skb->data+offset);
 480         struct sock *sk;
 481         int err;
 482
 483         sk = udp_v6_lookup(daddr, uh->dest, saddr, uh->source, dev->ifindex);
 484
 485         if (sk == NULL)
 486                 return;
 487
 488         np = inet6_sk(sk);
 489
 490         if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
 491                 goto out;
 492
 493         if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
 494                 goto out;
 495
 496         if (np->recverr)
 497                 ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
 498
 499         sk->sk_err = err;
 500         sk->sk_error_report(sk);
 501 out:
 502         sock_put(sk);
 503 }
 504
 505 static inline int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
 506 {
 507         if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) {
 508                 kfree_skb(skb);
 509                 return -1;
 510         }
 511
 512         if (sk->sk_filter && skb->ip_summed != CHECKSUM_UNNECESSARY) {
 513                 if ((unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum))) {
 514                         UDP6_INC_STATS_BH(UdpInErrors);
 515                         kfree_skb(skb);
 516                         return 0;
 517                 }
 518                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 519         }
 520
 521         if (sock_queue_rcv_skb(sk,skb)<0) {
 522                 UDP6_INC_STATS_BH(UdpInErrors);
 523                 kfree_skb(skb);
 524                 return 0;
 525         }
 526         UDP6_INC_STATS_BH(UdpInDatagrams);
 527         return 0;
 528 }
 529
 530 static struct sock *udp_v6_mcast_next(struct sock *sk,
 531                                       u16 loc_port, struct in6_addr *loc_addr,
 532                                       u16 rmt_port, struct in6_addr *rmt_addr,
 533                                       int dif)
 534 {
 535         struct hlist_node *node;
 536         struct sock *s = sk;
 537         unsigned short num = ntohs(loc_port);
 538
 539         sk_for_each_from(s, node) {
 540                 struct inet_opt *inet = inet_sk(s);
 541
 542                 if (inet->num == num && s->sk_family == PF_INET6) {
 543                         struct ipv6_pinfo *np = inet6_sk(s);
 544                         if (inet->dport) {
 545                                 if (inet->dport != rmt_port)
 546                                         continue;
 547                         }
 548                         if (!ipv6_addr_any(&np->daddr) &&
 549                             ipv6_addr_cmp(&np->daddr, rmt_addr))
 550                                 continue;
 551
 552                         if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
 553                                 continue;
 554
 555                         if (!ipv6_addr_any(&np->rcv_saddr)) {
 556                                 if (!ipv6_addr_cmp(&np->rcv_saddr, loc_addr))
 557                                         return s;
 558                                 continue;
 559                         }
 560                         if(!inet6_mc_check(s, loc_addr, rmt_addr))
 561                                 continue;
 562                         return s;
 563                 }
 564         }
 565         return NULL;
 566 }
 567
 568 /*
 569  * Note: called only from the BH handler context,
 570  * so we don't need to lock the hashes.
 571  */
 572 static void udpv6_mcast_deliver(struct udphdr *uh,
 573                                 struct in6_addr *saddr, struct in6_addr *daddr,
 574                                 struct sk_buff *skb)
 575 {
 576         struct sock *sk, *sk2;
 577         struct sk_buff *buff;
 578         int dif;
 579
 580         read_lock(&udp_hash_lock);
 581         sk = sk_head(&udp_hash[ntohs(uh->dest) & (UDP_HTABLE_SIZE - 1)]);
 582         dif = skb->dev->ifindex;
 583         sk = udp_v6_mcast_next(sk, uh->dest, daddr, uh->source, saddr, dif);
 584         if (!sk)
 585                 goto free_skb;
 586
 587         buff = NULL;
 588         sk2 = sk;
 589         while ((sk2 = udp_v6_mcast_next(sk_next(sk2), uh->dest, daddr,
 590                                         uh->source, saddr, dif))) {
 591                 if (!buff) {
 592                         buff = skb_clone(skb, GFP_ATOMIC);
 593                         if (!buff)
 594                                 continue;
 595                 }
 596                 if (udpv6_queue_rcv_skb(sk2, buff) >= 0)
 597                         buff = NULL;
 598         }
 599         if (buff)
 600                 kfree_skb(buff);
 601         if (udpv6_queue_rcv_skb(sk, skb) < 0) {
 602 free_skb:
 603                 kfree_skb(skb);
 604         }
 605         read_unlock(&udp_hash_lock);
 606 }
 607
 608 static int udpv6_rcv(struct sk_buff **pskb, unsigned int *nhoffp)
 609 {
 610         struct sk_buff *skb = *pskb;
 611         struct sock *sk;
 612         struct udphdr *uh;
 613         struct net_device *dev = skb->dev;
 614         struct in6_addr *saddr, *daddr;
 615         u32 ulen = 0;
 616
 617         if (!pskb_may_pull(skb, sizeof(struct udphdr)))
 618                 goto short_packet;
 619
 620         saddr = &skb->nh.ipv6h->saddr;
 621         daddr = &skb->nh.ipv6h->daddr;
 622         uh = skb->h.uh;
 623
 624         ulen = ntohs(uh->len);
 625
 626         /* Check for jumbo payload */
 627         if (ulen == 0)
 628                 ulen = skb->len;
 629
 630         if (ulen > skb->len || ulen < sizeof(*uh))
 631                 goto short_packet;
 632
 633         if (uh->check == 0) {
 634                 /* RFC 2460 section 8.1 says that we SHOULD log
 635                    this error. Well, it is reasonable.
 636                  */
 637                 LIMIT_NETDEBUG(
 638                         printk(KERN_INFO "IPv6: udp checksum is 0\n"));
 639                 goto discard;
 640         }
 641
 642         if (ulen < skb->len) {
 643                 if (__pskb_trim(skb, ulen))
 644                         goto discard;
 645                 saddr = &skb->nh.ipv6h->saddr;
 646                 daddr = &skb->nh.ipv6h->daddr;
 647                 uh = skb->h.uh;
 648         }
 649
 650         if (skb->ip_summed==CHECKSUM_HW) {
 651                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 652                 if (csum_ipv6_magic(saddr, daddr, ulen, IPPROTO_UDP, skb->csum)) {
 653                         LIMIT_NETDEBUG(printk(KERN_DEBUG "udp v6 hw csum failure.\n"));
 654                         skb->ip_summed = CHECKSUM_NONE;
 655                 }
 656         }
 657         if (skb->ip_summed != CHECKSUM_UNNECESSARY)
 658                 skb->csum = ~csum_ipv6_magic(saddr, daddr, ulen, IPPROTO_UDP, 0);
 659
 660         /*
 661          *      Multicast receive code
 662          */
 663         if (ipv6_addr_is_multicast(daddr)) {
 664                 udpv6_mcast_deliver(uh, saddr, daddr, skb);
 665                 return 0;
 666         }
 667
 668         /* Unicast */
 669
 670         /*
 671          * check socket cache ... must talk to Alan about his plans
 672          * for sock caches... i'll skip this for now.
 673          */
 674         sk = udp_v6_lookup(saddr, uh->source, daddr, uh->dest, dev->ifindex);
 675
 676         if (sk == NULL) {
 677                 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 678                         goto discard;
 679
 680                 if (skb->ip_summed != CHECKSUM_UNNECESSARY &&
 681                     (unsigned short)csum_fold(skb_checksum(skb, 0, skb->len, skb->csum)))
 682                         goto discard;
 683                 UDP6_INC_STATS_BH(UdpNoPorts);
 684
 685                 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0, dev);
 686
 687                 kfree_skb(skb);
 688                 return(0);
 689         }
 690
 691         /* deliver */
 692
 693         udpv6_queue_rcv_skb(sk, skb);
 694         sock_put(sk);
 695         return(0);
 696
 697 short_packet:
 698         if (net_ratelimit())
 699                 printk(KERN_DEBUG "UDP: short packet: %d/%u\n", ulen, skb->len);
 700
 701 discard:
 702         UDP6_INC_STATS_BH(UdpInErrors);
 703         kfree_skb(skb);
 704         return(0);
 705 }
 706 /*
 707  * Throw away all pending data and cancel the corking. Socket is locked.
 708  */
 709 static void udp_v6_flush_pending_frames(struct sock *sk)
 710 {
 711         struct udp_opt *up = udp_sk(sk);
 712
 713         if (up->pending) {
 714                 up->len = 0;
 715                 up->pending = 0;
 716                 ip6_flush_pending_frames(sk);
 717         }
 718 }
 719
 720 /*
 721  *      Sending
 722  */
 723
 724 static int udp_v6_push_pending_frames(struct sock *sk, struct udp_opt *up)
 725 {
 726         struct sk_buff *skb;
 727         struct udphdr *uh;
 728         struct inet_opt *inet = inet_sk(sk);
 729         struct flowi *fl = &inet->cork.fl;
 730         int err = 0;
 731
 732         /* Grab the skbuff where UDP header space exists. */
 733         if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
 734                 goto out;
 735
 736         /*
 737          * Create a UDP header
 738          */
 739         uh = skb->h.uh;
 740         uh->source = fl->fl_ip_sport;
 741         uh->dest = fl->fl_ip_dport;
 742         uh->len = htons(up->len);
 743         uh->check = 0;
 744
 745         if (sk->sk_no_check == UDP_CSUM_NOXMIT) {
 746                 skb->ip_summed = CHECKSUM_NONE;
 747                 goto send;
 748         }
 749
 750         if (skb_queue_len(&sk->sk_write_queue) == 1) {
 751                 skb->csum = csum_partial((char *)uh,
 752                                 sizeof(struct udphdr), skb->csum);
 753                 uh->check = csum_ipv6_magic(&fl->fl6_src,
 754                                             &fl->fl6_dst,
 755                                             up->len, fl->proto, skb->csum);
 756         } else {
 757                 u32 tmp_csum = 0;
 758
 759                 skb_queue_walk(&sk->sk_write_queue, skb) {
 760                         tmp_csum = csum_add(tmp_csum, skb->csum);
 761                 }
 762                 tmp_csum = csum_partial((char *)uh,
 763                                 sizeof(struct udphdr), tmp_csum);
 764                 tmp_csum = csum_ipv6_magic(&fl->fl6_src,
 765                                            &fl->fl6_dst,
 766                                            up->len, fl->proto, tmp_csum);
 767                 uh->check = tmp_csum;
 768
 769         }
 770         if (uh->check == 0)
 771                 uh->check = -1;
 772
 773 send:
 774         err = ip6_push_pending_frames(sk);
 775 out:
 776         up->len = 0;
 777         up->pending = 0;
 778         return err;
 779 }
 780
 781 static int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
 782                   struct msghdr *msg, size_t len)
 783 {
 784         struct ipv6_txoptions opt_space;
 785         struct udp_opt *up = udp_sk(sk);
 786         struct inet_opt *inet = inet_sk(sk);
 787         struct ipv6_pinfo *np = inet6_sk(sk);
 788         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
 789         struct in6_addr *daddr;
 790         struct ipv6_txoptions *opt = NULL;
 791         struct ip6_flowlabel *flowlabel = NULL;
 792         struct flowi *fl = &inet->cork.fl;
 793         struct dst_entry *dst;
 794         int addr_len = msg->msg_namelen;
 795         int ulen = len;
 796         int hlimit = -1;
 797         int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
 798         int err;
 799
 800         /* destination address check */
 801         if (sin6) {
 802                 if (addr_len < offsetof(struct sockaddr, sa_data))
 803                         return -EINVAL;
 804
 805                 switch (sin6->sin6_family) {
 806                 case AF_INET6:
 807                         if (addr_len < SIN6_LEN_RFC2133)
 808                                 return -EINVAL;
 809                         daddr = &sin6->sin6_addr;
 810                         break;
 811                 case AF_INET:
 812                         goto do_udp_sendmsg;
 813                 case AF_UNSPEC:
 814                         msg->msg_name = sin6 = NULL;
 815                         msg->msg_namelen = addr_len = 0;
 816                         daddr = NULL;
 817                         break;
 818                 default:
 819                         return -EINVAL;
 820                 }
 821         } else if (!up->pending) {
 822                 if (sk->sk_state != TCP_ESTABLISHED)
 823                         return -EDESTADDRREQ;
 824                 daddr = &np->daddr;
 825         } else
 826                 daddr = NULL;
 827
 828         if (daddr) {
 829                 if (ipv6_addr_type(daddr) == IPV6_ADDR_MAPPED) {
 830                         struct sockaddr_in sin;
 831                         sin.sin_family = AF_INET;
 832                         sin.sin_port = sin6 ? sin6->sin6_port : inet->dport;
 833                         sin.sin_addr.s_addr = daddr->s6_addr32[3];
 834                         msg->msg_name = &sin;
 835                         msg->msg_namelen = sizeof(sin);
 836 do_udp_sendmsg:
 837                         if (__ipv6_only_sock(sk))
 838                                 return -ENETUNREACH;
 839                         return udp_sendmsg(iocb, sk, msg, len);
 840                 }
 841         }
 842
 843         if (up->pending == AF_INET)
 844                 return udp_sendmsg(iocb, sk, msg, len);
 845
 846         /* Rough check on arithmetic overflow,
 847            better check is made in ip6_build_xmit
 848            */
 849         if (len > INT_MAX - sizeof(struct udphdr))
 850                 return -EMSGSIZE;
 851
 852         if (up->pending) {
 853                 /*
 854                  * There are pending frames.
 855                  * The socket lock must be held while it's corked.
 856                  */
 857                 lock_sock(sk);
 858                 if (likely(up->pending)) {
 859                         if (unlikely(up->pending != AF_INET6)) {
 860                                 release_sock(sk);
 861                                 return -EINVAL;
 862                         }
 863                         dst = NULL;
 864                         goto do_append_data;
 865                 }
 866                 release_sock(sk);
 867         }
 868         ulen += sizeof(struct udphdr);
 869
 870         memset(fl, 0, sizeof(*fl));
 871
 872         if (sin6) {
 873                 if (sin6->sin6_port == 0)
 874                         return -EINVAL;
 875
 876                 fl->fl_ip_dport = sin6->sin6_port;
 877                 daddr = &sin6->sin6_addr;
 878
 879                 if (np->sndflow) {
 880                         fl->fl6_flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
 881                         if (fl->fl6_flowlabel&IPV6_FLOWLABEL_MASK) {
 882                                 flowlabel = fl6_sock_lookup(sk, fl->fl6_flowlabel);
 883                                 if (flowlabel == NULL)
 884                                         return -EINVAL;
 885                                 daddr = &flowlabel->dst;
 886                         }
 887                 }
 888
 889                 /*
 890                  * Otherwise it will be difficult to maintain
 891                  * sk->sk_dst_cache.
 892                  */
 893                 if (sk->sk_state == TCP_ESTABLISHED &&
 894                     !ipv6_addr_cmp(daddr, &np->daddr))
 895                         daddr = &np->daddr;
 896
 897                 if (addr_len >= sizeof(struct sockaddr_in6) &&
 898                     sin6->sin6_scope_id &&
 899                     ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
 900                         fl->oif = sin6->sin6_scope_id;
 901         } else {
 902                 if (sk->sk_state != TCP_ESTABLISHED)
 903                         return -EDESTADDRREQ;
 904
 905                 fl->fl_ip_dport = inet->dport;
 906                 daddr = &np->daddr;
 907                 fl->fl6_flowlabel = np->flow_label;
 908         }
 909
 910         if (!fl->oif)
 911                 fl->oif = sk->sk_bound_dev_if;
 912
 913         if (msg->msg_controllen) {
 914                 opt = &opt_space;
 915                 memset(opt, 0, sizeof(struct ipv6_txoptions));
 916                 opt->tot_len = sizeof(*opt);
 917
 918                 err = datagram_send_ctl(msg, fl, opt, &hlimit);
 919                 if (err < 0) {
 920                         fl6_sock_release(flowlabel);
 921                         return err;
 922                 }
 923                 if ((fl->fl6_flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
 924                         flowlabel = fl6_sock_lookup(sk, fl->fl6_flowlabel);
 925                         if (flowlabel == NULL)
 926                                 return -EINVAL;
 927                 }
 928                 if (!(opt->opt_nflen|opt->opt_flen))
 929                         opt = NULL;
 930         }
 931         if (opt == NULL)
 932                 opt = np->opt;
 933         if (flowlabel)
 934                 opt = fl6_merge_options(&opt_space, flowlabel, opt);
 935
 936         fl->proto = IPPROTO_UDP;
 937         ipv6_addr_copy(&fl->fl6_dst, daddr);
 938         if (ipv6_addr_any(&fl->fl6_src) && !ipv6_addr_any(&np->saddr))
 939                 ipv6_addr_copy(&fl->fl6_src, &np->saddr);
 940         fl->fl_ip_sport = inet->sport;
 941
 942         /* merge ip6_build_xmit from ip6_output */
 943         if (opt && opt->srcrt) {
 944                 struct rt0_hdr *rt0 = (struct rt0_hdr *) opt->srcrt;
 945                 ipv6_addr_copy(&fl->fl6_dst, rt0->addr);
 946         }
 947
 948         if (!fl->oif && ipv6_addr_is_multicast(&fl->fl6_dst))
 949                 fl->oif = np->mcast_oif;
 950
 951         err = ip6_dst_lookup(sk, &dst, fl);
 952         if (err)
 953                 goto out;
 954
 955         if (hlimit < 0) {
 956                 if (ipv6_addr_is_multicast(&fl->fl6_dst))
 957                         hlimit = np->mcast_hops;
 958                 else
 959                         hlimit = np->hop_limit;
 960                 if (hlimit < 0)
 961                         hlimit = dst_metric(dst, RTAX_HOPLIMIT);
 962         }
 963
 964         if (msg->msg_flags&MSG_CONFIRM)
 965                 goto do_confirm;
 966 back_from_confirm:
 967
 968         lock_sock(sk);
 969         if (unlikely(up->pending)) {
 970                 /* The socket is already corked while preparing it. */
 971                 /* ... which is an evident application bug. --ANK */
 972                 release_sock(sk);
 973
 974                 LIMIT_NETDEBUG(printk(KERN_DEBUG "udp cork app bug 2\n"));
 975                 err = -EINVAL;
 976                 goto out;
 977         }
 978
 979         up->pending = AF_INET6;
 980
 981 do_append_data:
 982         up->len += ulen;
 983         err = ip6_append_data(sk, ip_generic_getfrag, msg->msg_iov, ulen, sizeof(struct udphdr),
 984                               hlimit, opt, fl, (struct rt6_info*)dst,
 985                               corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags);
 986         if (err)
 987                 udp_v6_flush_pending_frames(sk);
 988         else if (!corkreq)
 989                 err = udp_v6_push_pending_frames(sk, up);
 990
 991         if (dst)
 992                 ip6_dst_store(sk, dst,
 993                               !ipv6_addr_cmp(&fl->fl6_dst, &np->daddr) ?
 994                               &np->daddr : NULL);
 995         if (err > 0)
 996                 err = np->recverr ? net_xmit_errno(err) : 0;
 997         release_sock(sk);
 998 out:
 999         fl6_sock_release(flowlabel);
1000         if (!err) {
1001                 UDP6_INC_STATS_USER(UdpOutDatagrams);
1002                 return len;
1003         }
1004         return err;
1005
1006 do_confirm:
1007         dst_confirm(dst);
1008         if (!(msg->msg_flags&MSG_PROBE) || len)
1009                 goto back_from_confirm;
1010         err = 0;
1011         goto out;
1012 }
1013
1014 static int udpv6_destroy_sock(struct sock *sk)
1015 {
1016         lock_sock(sk);
1017         udp_v6_flush_pending_frames(sk);
1018         release_sock(sk);
1019
1020         inet6_destroy_sock(sk);
1021
1022         return 0;
1023 }
1024
1025 /*
1026  *      Socket option code for UDP
1027  */
1028 static int udpv6_setsockopt(struct sock *sk, int level, int optname,
1029                           char *optval, int optlen)
1030 {
1031         struct udp_opt *up = udp_sk(sk);
1032         int val;
1033         int err = 0;
1034
1035         if (level != SOL_UDP)
1036                 return ipv6_setsockopt(sk, level, optname, optval, optlen);
1037
1038         if(optlen<sizeof(int))
1039                 return -EINVAL;
1040
1041         if (get_user(val, (int *)optval))
1042                 return -EFAULT;
1043
1044         switch(optname) {
1045         case UDP_CORK:
1046                 if (val != 0) {
1047                         up->corkflag = 1;
1048                 } else {
1049                         up->corkflag = 0;
1050                         lock_sock(sk);
1051                         udp_v6_push_pending_frames(sk, up);
1052                         release_sock(sk);
1053                 }
1054                 break;
1055
1056         case UDP_ENCAP:
1057                 up->encap_type = val;
1058                 break;
1059
1060         default:
1061                 err = -ENOPROTOOPT;
1062                 break;
1063         };
1064
1065         return err;
1066 }
1067
1068 static int udpv6_getsockopt(struct sock *sk, int level, int optname,
1069                           char *optval, int *optlen)
1070 {
1071         struct udp_opt *up = udp_sk(sk);
1072         int val, len;
1073
1074         if (level != SOL_UDP)
1075                 return ipv6_getsockopt(sk, level, optname, optval, optlen);
1076
1077         if(get_user(len,optlen))
1078                 return -EFAULT;
1079
1080         len = min_t(unsigned int, len, sizeof(int));
1081
1082         if(len < 0)
1083                 return -EINVAL;
1084
1085         switch(optname) {
1086         case UDP_CORK:
1087                 val = up->corkflag;
1088                 break;
1089
1090         case UDP_ENCAP:
1091                 val = up->encap_type;
1092                 break;
1093
1094         default:
1095                 return -ENOPROTOOPT;
1096         };
1097
1098         if(put_user(len, optlen))
1099                 return -EFAULT;
1100         if(copy_to_user(optval, &val,len))
1101                 return -EFAULT;
1102         return 0;
1103 }
1104
1105 static struct inet6_protocol udpv6_protocol = {
1106         .handler        =       udpv6_rcv,
1107         .err_handler    =       udpv6_err,
1108         .flags          =       INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1109 };
1110
1111 /* ------------------------------------------------------------------------ */
1112 #ifdef CONFIG_PROC_FS
1113
1114 static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
1115 {
1116         struct inet_opt *inet = inet_sk(sp);
1117         struct ipv6_pinfo *np = inet6_sk(sp);
1118         struct in6_addr *dest, *src;
1119         __u16 destp, srcp;
1120
1121         dest  = &np->daddr;
1122         src   = &np->rcv_saddr;
1123         destp = ntohs(inet->dport);
1124         srcp  = ntohs(inet->sport);
1125         seq_printf(seq,
1126                    "%4d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1127                    "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p\n",
1128                    bucket,
1129                    src->s6_addr32[0], src->s6_addr32[1],
1130                    src->s6_addr32[2], src->s6_addr32[3], srcp,
1131                    dest->s6_addr32[0], dest->s6_addr32[1],
1132                    dest->s6_addr32[2], dest->s6_addr32[3], destp,
1133                    sp->sk_state,
1134                    atomic_read(&sp->sk_wmem_alloc),
1135                    atomic_read(&sp->sk_rmem_alloc),
1136                    0, 0L, 0,
1137                    sock_i_uid(sp), 0,
1138                    sock_i_ino(sp),
1139                    atomic_read(&sp->sk_refcnt), sp);
1140 }
1141
1142 static int udp6_seq_show(struct seq_file *seq, void *v)
1143 {
1144         if (v == SEQ_START_TOKEN)
1145                 seq_printf(seq,
1146                            "  sl  "
1147                            "local_address                         "
1148                            "remote_address                        "
1149                            "st tx_queue rx_queue tr tm->when retrnsmt"
1150                            "   uid  timeout inode\n");
1151         else
1152                 udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1153         return 0;
1154 }
1155
1156 static struct file_operations udp6_seq_fops;
1157 static struct udp_seq_afinfo udp6_seq_afinfo = {
1158         .owner          = THIS_MODULE,
1159         .name           = "udp6",
1160         .family         = AF_INET6,
1161         .seq_show       = udp6_seq_show,
1162         .seq_fops       = &udp6_seq_fops,
1163 };
1164
1165 int __init udp6_proc_init(void)
1166 {
1167         return udp_proc_register(&udp6_seq_afinfo);
1168 }
1169
1170 void udp6_proc_exit(void) {
1171         udp_proc_unregister(&udp6_seq_afinfo);
1172 }
1173 #endif /* CONFIG_PROC_FS */
1174
1175 /* ------------------------------------------------------------------------ */
1176
1177 struct proto udpv6_prot = {
1178         .name =         "UDP",
1179         .close =        udpv6_close,
1180         .connect =      udpv6_connect,
1181         .disconnect =   udp_disconnect,
1182         .ioctl =        udp_ioctl,
1183         .destroy =      udpv6_destroy_sock,
1184         .setsockopt =   udpv6_setsockopt,
1185         .getsockopt =   udpv6_getsockopt,
1186         .sendmsg =      udpv6_sendmsg,
1187         .recvmsg =      udpv6_recvmsg,
1188         .backlog_rcv =  udpv6_queue_rcv_skb,
1189         .hash =         udp_v6_hash,
1190         .unhash =       udp_v6_unhash,
1191         .get_port =     udp_v6_get_port,
1192 };
1193
1194 extern struct proto_ops inet6_dgram_ops;
1195
1196 static struct inet_protosw udpv6_protosw = {
1197         .type =      SOCK_DGRAM,
1198         .protocol =  IPPROTO_UDP,
1199         .prot =      &udpv6_prot,
1200         .ops =       &inet6_dgram_ops,
1201         .capability =-1,
1202         .no_check =  UDP_CSUM_DEFAULT,
1203         .flags =     INET_PROTOSW_PERMANENT,
1204 };
1205
1206
1207 void __init udpv6_init(void)
1208 {
1209         if (inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP) < 0)
1210                 printk(KERN_ERR "udpv6_init: Could not register protocol\n");
1211         inet6_register_protosw(&udpv6_protosw);
1212 }