linux/net/ipv6/udp.c
<<
>>
Prefs
   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 *      Fixes:
  11 *      Hideaki YOSHIFUJI       :       sin6_scope_id support
  12 *      YOSHIFUJI Hideaki @USAGI and:   Support IPV6_V6ONLY socket option, which
  13 *      Alexey Kuznetsov                allow both IPv4 and IPv6 sockets to bind
  14 *                                      a single port at the same time.
  15 *      Kazunori MIYAZAWA @USAGI:       change process style to use ip6_append_data
  16 *      YOSHIFUJI Hideaki @USAGI:       convert /proc/net/udp6 to seq_file.
  17 *
  18 *      This program is free software; you can redistribute it and/or
  19 *      modify it under the terms of the GNU General Public License
  20 *      as published by the Free Software Foundation; either version
  21 *      2 of the License, or (at your option) any later version.
  22 */
  23
  24#include <linux/errno.h>
  25#include <linux/types.h>
  26#include <linux/socket.h>
  27#include <linux/sockios.h>
  28#include <linux/net.h>
  29#include <linux/in6.h>
  30#include <linux/netdevice.h>
  31#include <linux/if_arp.h>
  32#include <linux/ipv6.h>
  33#include <linux/icmpv6.h>
  34#include <linux/init.h>
  35#include <linux/module.h>
  36#include <linux/skbuff.h>
  37#include <linux/slab.h>
  38#include <asm/uaccess.h>
  39
  40#include <net/ndisc.h>
  41#include <net/protocol.h>
  42#include <net/transp_v6.h>
  43#include <net/ip6_route.h>
  44#include <net/raw.h>
  45#include <net/tcp_states.h>
  46#include <net/ip6_checksum.h>
  47#include <net/xfrm.h>
  48
  49#include <linux/proc_fs.h>
  50#include <linux/seq_file.h>
  51#include "udp_impl.h"
  52
  53int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
  54{
  55        const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
  56        const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
  57        __be32 sk1_rcv_saddr = sk_rcv_saddr(sk);
  58        __be32 sk2_rcv_saddr = sk_rcv_saddr(sk2);
  59        int sk_ipv6only = ipv6_only_sock(sk);
  60        int sk2_ipv6only = inet_v6_ipv6only(sk2);
  61        int addr_type = ipv6_addr_type(sk_rcv_saddr6);
  62        int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
  63
  64        /* if both are mapped, treat as IPv4 */
  65        if (addr_type == IPV6_ADDR_MAPPED && addr_type2 == IPV6_ADDR_MAPPED)
  66                return (!sk2_ipv6only &&
  67                        (!sk1_rcv_saddr || !sk2_rcv_saddr ||
  68                          sk1_rcv_saddr == sk2_rcv_saddr));
  69
  70        if (addr_type2 == IPV6_ADDR_ANY &&
  71            !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
  72                return 1;
  73
  74        if (addr_type == IPV6_ADDR_ANY &&
  75            !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
  76                return 1;
  77
  78        if (sk2_rcv_saddr6 &&
  79            ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
  80                return 1;
  81
  82        return 0;
  83}
  84
  85static unsigned int udp6_portaddr_hash(struct net *net,
  86                                       const struct in6_addr *addr6,
  87                                       unsigned int port)
  88{
  89        unsigned int hash, mix = net_hash_mix(net);
  90
  91        if (ipv6_addr_any(addr6))
  92                hash = jhash_1word(0, mix);
  93        else if (ipv6_addr_v4mapped(addr6))
  94                hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
  95        else
  96                hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
  97
  98        return hash ^ port;
  99}
 100
 101
 102int udp_v6_get_port(struct sock *sk, unsigned short snum)
 103{
 104        unsigned int hash2_nulladdr =
 105                udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
 106        unsigned int hash2_partial = 
 107                udp6_portaddr_hash(sock_net(sk), &inet6_sk(sk)->rcv_saddr, 0);
 108
 109        /* precompute partial secondary hash */
 110        udp_sk(sk)->udp_portaddr_hash = hash2_partial;
 111        return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
 112}
 113
 114static void udp_v6_rehash(struct sock *sk)
 115{
 116        u16 new_hash = udp6_portaddr_hash(sock_net(sk),
 117                                          &inet6_sk(sk)->rcv_saddr,
 118                                          inet_sk(sk)->inet_num);
 119
 120        udp_lib_rehash(sk, new_hash);
 121}
 122
 123static inline int compute_score(struct sock *sk, struct net *net,
 124                                unsigned short hnum,
 125                                const struct in6_addr *saddr, __be16 sport,
 126                                const struct in6_addr *daddr, __be16 dport,
 127                                int dif)
 128{
 129        int score = -1;
 130
 131        if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
 132                        sk->sk_family == PF_INET6) {
 133                struct ipv6_pinfo *np = inet6_sk(sk);
 134                struct inet_sock *inet = inet_sk(sk);
 135
 136                score = 0;
 137                if (inet->inet_dport) {
 138                        if (inet->inet_dport != sport)
 139                                return -1;
 140                        score++;
 141                }
 142                if (!ipv6_addr_any(&np->rcv_saddr)) {
 143                        if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
 144                                return -1;
 145                        score++;
 146                }
 147                if (!ipv6_addr_any(&np->daddr)) {
 148                        if (!ipv6_addr_equal(&np->daddr, saddr))
 149                                return -1;
 150                        score++;
 151                }
 152                if (sk->sk_bound_dev_if) {
 153                        if (sk->sk_bound_dev_if != dif)
 154                                return -1;
 155                        score++;
 156                }
 157        }
 158        return score;
 159}
 160
 161#define SCORE2_MAX (1 + 1 + 1)
 162static inline int compute_score2(struct sock *sk, struct net *net,
 163                                const struct in6_addr *saddr, __be16 sport,
 164                                const struct in6_addr *daddr, unsigned short hnum,
 165                                int dif)
 166{
 167        int score = -1;
 168
 169        if (net_eq(sock_net(sk), net) && udp_sk(sk)->udp_port_hash == hnum &&
 170                        sk->sk_family == PF_INET6) {
 171                struct ipv6_pinfo *np = inet6_sk(sk);
 172                struct inet_sock *inet = inet_sk(sk);
 173
 174                if (!ipv6_addr_equal(&np->rcv_saddr, daddr))
 175                        return -1;
 176                score = 0;
 177                if (inet->inet_dport) {
 178                        if (inet->inet_dport != sport)
 179                                return -1;
 180                        score++;
 181                }
 182                if (!ipv6_addr_any(&np->daddr)) {
 183                        if (!ipv6_addr_equal(&np->daddr, saddr))
 184                                return -1;
 185                        score++;
 186                }
 187                if (sk->sk_bound_dev_if) {
 188                        if (sk->sk_bound_dev_if != dif)
 189                                return -1;
 190                        score++;
 191                }
 192        }
 193        return score;
 194}
 195
 196
 197/* called with read_rcu_lock() */
 198static struct sock *udp6_lib_lookup2(struct net *net,
 199                const struct in6_addr *saddr, __be16 sport,
 200                const struct in6_addr *daddr, unsigned int hnum, int dif,
 201                struct udp_hslot *hslot2, unsigned int slot2)
 202{
 203        struct sock *sk, *result;
 204        struct hlist_nulls_node *node;
 205        int score, badness;
 206
 207begin:
 208        result = NULL;
 209        badness = -1;
 210        udp_portaddr_for_each_entry_rcu(sk, node, &hslot2->head) {
 211                score = compute_score2(sk, net, saddr, sport,
 212                                      daddr, hnum, dif);
 213                if (score > badness) {
 214                        result = sk;
 215                        badness = score;
 216                        if (score == SCORE2_MAX)
 217                                goto exact_match;
 218                }
 219        }
 220        /*
 221         * if the nulls value we got at the end of this lookup is
 222         * not the expected one, we must restart lookup.
 223         * We probably met an item that was moved to another chain.
 224         */
 225        if (get_nulls_value(node) != slot2)
 226                goto begin;
 227
 228        if (result) {
 229exact_match:
 230                if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
 231                        result = NULL;
 232                else if (unlikely(compute_score2(result, net, saddr, sport,
 233                                  daddr, hnum, dif) < badness)) {
 234                        sock_put(result);
 235                        goto begin;
 236                }
 237        }
 238        return result;
 239}
 240
 241struct sock *__udp6_lib_lookup(struct net *net,
 242                                      const struct in6_addr *saddr, __be16 sport,
 243                                      const struct in6_addr *daddr, __be16 dport,
 244                                      int dif, struct udp_table *udptable)
 245{
 246        struct sock *sk, *result;
 247        struct hlist_nulls_node *node;
 248        unsigned short hnum = ntohs(dport);
 249        unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
 250        struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
 251        int score, badness;
 252
 253        rcu_read_lock();
 254        if (hslot->count > 10) {
 255                hash2 = udp6_portaddr_hash(net, daddr, hnum);
 256                slot2 = hash2 & udptable->mask;
 257                hslot2 = &udptable->hash2[slot2];
 258                if (hslot->count < hslot2->count)
 259                        goto begin;
 260
 261                result = udp6_lib_lookup2(net, saddr, sport,
 262                                          daddr, hnum, dif,
 263                                          hslot2, slot2);
 264                if (!result) {
 265                        hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
 266                        slot2 = hash2 & udptable->mask;
 267                        hslot2 = &udptable->hash2[slot2];
 268                        if (hslot->count < hslot2->count)
 269                                goto begin;
 270
 271                        result = udp6_lib_lookup2(net, saddr, sport,
 272                                                  &in6addr_any, hnum, dif,
 273                                                  hslot2, slot2);
 274                }
 275                rcu_read_unlock();
 276                return result;
 277        }
 278begin:
 279        result = NULL;
 280        badness = -1;
 281        sk_nulls_for_each_rcu(sk, node, &hslot->head) {
 282                score = compute_score(sk, net, hnum, saddr, sport, daddr, dport, dif);
 283                if (score > badness) {
 284                        result = sk;
 285                        badness = score;
 286                }
 287        }
 288        /*
 289         * if the nulls value we got at the end of this lookup is
 290         * not the expected one, we must restart lookup.
 291         * We probably met an item that was moved to another chain.
 292         */
 293        if (get_nulls_value(node) != slot)
 294                goto begin;
 295
 296        if (result) {
 297                if (unlikely(!atomic_inc_not_zero_hint(&result->sk_refcnt, 2)))
 298                        result = NULL;
 299                else if (unlikely(compute_score(result, net, hnum, saddr, sport,
 300                                        daddr, dport, dif) < badness)) {
 301                        sock_put(result);
 302                        goto begin;
 303                }
 304        }
 305        rcu_read_unlock();
 306        return result;
 307}
 308EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
 309
 310static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
 311                                          __be16 sport, __be16 dport,
 312                                          struct udp_table *udptable)
 313{
 314        struct sock *sk;
 315        const struct ipv6hdr *iph = ipv6_hdr(skb);
 316
 317        if (unlikely(sk = skb_steal_sock(skb)))
 318                return sk;
 319        return __udp6_lib_lookup(dev_net(skb_dst(skb)->dev), &iph->saddr, sport,
 320                                 &iph->daddr, dport, inet6_iif(skb),
 321                                 udptable);
 322}
 323
 324struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
 325                             const struct in6_addr *daddr, __be16 dport, int dif)
 326{
 327        return __udp6_lib_lookup(net, saddr, sport, daddr, dport, dif, &udp_table);
 328}
 329EXPORT_SYMBOL_GPL(udp6_lib_lookup);
 330
 331
 332/*
 333 *      This should be easy, if there is something there we
 334 *      return it, otherwise we block.
 335 */
 336
 337int udpv6_recvmsg(struct kiocb *iocb, struct sock *sk,
 338                  struct msghdr *msg, size_t len,
 339                  int noblock, int flags, int *addr_len)
 340{
 341        struct ipv6_pinfo *np = inet6_sk(sk);
 342        struct inet_sock *inet = inet_sk(sk);
 343        struct sk_buff *skb;
 344        unsigned int ulen, copied;
 345        int peeked, off = 0;
 346        int err;
 347        int is_udplite = IS_UDPLITE(sk);
 348        int is_udp4;
 349        bool slow;
 350
 351        if (addr_len)
 352                *addr_len=sizeof(struct sockaddr_in6);
 353
 354        if (flags & MSG_ERRQUEUE)
 355                return ipv6_recv_error(sk, msg, len);
 356
 357        if (np->rxpmtu && np->rxopt.bits.rxpmtu)
 358                return ipv6_recv_rxpmtu(sk, msg, len);
 359
 360try_again:
 361        skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
 362                                  &peeked, &off, &err);
 363        if (!skb)
 364                goto out;
 365
 366        ulen = skb->len - sizeof(struct udphdr);
 367        copied = len;
 368        if (copied > ulen)
 369                copied = ulen;
 370        else if (copied < ulen)
 371                msg->msg_flags |= MSG_TRUNC;
 372
 373        is_udp4 = (skb->protocol == htons(ETH_P_IP));
 374
 375        /*
 376         * If checksum is needed at all, try to do it while copying the
 377         * data.  If the data is truncated, or if we only want a partial
 378         * coverage checksum (UDP-Lite), do it before the copy.
 379         */
 380
 381        if (copied < ulen || UDP_SKB_CB(skb)->partial_cov) {
 382                if (udp_lib_checksum_complete(skb))
 383                        goto csum_copy_err;
 384        }
 385
 386        if (skb_csum_unnecessary(skb))
 387                err = skb_copy_datagram_iovec(skb, sizeof(struct udphdr),
 388                                              msg->msg_iov, copied       );
 389        else {
 390                err = skb_copy_and_csum_datagram_iovec(skb, sizeof(struct udphdr), msg->msg_iov);
 391                if (err == -EINVAL)
 392                        goto csum_copy_err;
 393        }
 394        if (err)
 395                goto out_free;
 396
 397        if (!peeked) {
 398                if (is_udp4)
 399                        UDP_INC_STATS_USER(sock_net(sk),
 400                                        UDP_MIB_INDATAGRAMS, is_udplite);
 401                else
 402                        UDP6_INC_STATS_USER(sock_net(sk),
 403                                        UDP_MIB_INDATAGRAMS, is_udplite);
 404        }
 405
 406        sock_recv_ts_and_drops(msg, sk, skb);
 407
 408        /* Copy the address. */
 409        if (msg->msg_name) {
 410                struct sockaddr_in6 *sin6;
 411
 412                sin6 = (struct sockaddr_in6 *) msg->msg_name;
 413                sin6->sin6_family = AF_INET6;
 414                sin6->sin6_port = udp_hdr(skb)->source;
 415                sin6->sin6_flowinfo = 0;
 416                sin6->sin6_scope_id = 0;
 417
 418                if (is_udp4)
 419                        ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
 420                                               &sin6->sin6_addr);
 421                else {
 422                        sin6->sin6_addr = ipv6_hdr(skb)->saddr;
 423                        if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
 424                                sin6->sin6_scope_id = IP6CB(skb)->iif;
 425                }
 426
 427        }
 428        if (is_udp4) {
 429                if (inet->cmsg_flags)
 430                        ip_cmsg_recv(msg, skb);
 431        } else {
 432                if (np->rxopt.all)
 433                        datagram_recv_ctl(sk, msg, skb);
 434        }
 435
 436        err = copied;
 437        if (flags & MSG_TRUNC)
 438                err = ulen;
 439
 440out_free:
 441        skb_free_datagram_locked(sk, skb);
 442out:
 443        return err;
 444
 445csum_copy_err:
 446        slow = lock_sock_fast(sk);
 447        if (!skb_kill_datagram(sk, skb, flags)) {
 448                if (is_udp4)
 449                        UDP_INC_STATS_USER(sock_net(sk),
 450                                        UDP_MIB_INERRORS, is_udplite);
 451                else
 452                        UDP6_INC_STATS_USER(sock_net(sk),
 453                                        UDP_MIB_INERRORS, is_udplite);
 454        }
 455        unlock_sock_fast(sk, slow);
 456
 457        if (noblock)
 458                return -EAGAIN;
 459
 460        /* starting over for a new packet */
 461        msg->msg_flags &= ~MSG_TRUNC;
 462        goto try_again;
 463}
 464
 465void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 466                    u8 type, u8 code, int offset, __be32 info,
 467                    struct udp_table *udptable)
 468{
 469        struct ipv6_pinfo *np;
 470        const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
 471        const struct in6_addr *saddr = &hdr->saddr;
 472        const struct in6_addr *daddr = &hdr->daddr;
 473        struct udphdr *uh = (struct udphdr*)(skb->data+offset);
 474        struct sock *sk;
 475        int err;
 476
 477        sk = __udp6_lib_lookup(dev_net(skb->dev), daddr, uh->dest,
 478                               saddr, uh->source, inet6_iif(skb), udptable);
 479        if (sk == NULL)
 480                return;
 481
 482        np = inet6_sk(sk);
 483
 484        if (!icmpv6_err_convert(type, code, &err) && !np->recverr)
 485                goto out;
 486
 487        if (sk->sk_state != TCP_ESTABLISHED && !np->recverr)
 488                goto out;
 489
 490        if (np->recverr)
 491                ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
 492
 493        sk->sk_err = err;
 494        sk->sk_error_report(sk);
 495out:
 496        sock_put(sk);
 497}
 498
 499static __inline__ void udpv6_err(struct sk_buff *skb,
 500                                 struct inet6_skb_parm *opt, u8 type,
 501                                 u8 code, int offset, __be32 info     )
 502{
 503        __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
 504}
 505
 506int udpv6_queue_rcv_skb(struct sock * sk, struct sk_buff *skb)
 507{
 508        struct udp_sock *up = udp_sk(sk);
 509        int rc;
 510        int is_udplite = IS_UDPLITE(sk);
 511
 512        if (!ipv6_addr_any(&inet6_sk(sk)->daddr))
 513                sock_rps_save_rxhash(sk, skb);
 514
 515        if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
 516                goto drop;
 517
 518        /*
 519         * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
 520         */
 521        if ((is_udplite & UDPLITE_RECV_CC)  &&  UDP_SKB_CB(skb)->partial_cov) {
 522
 523                if (up->pcrlen == 0) {          /* full coverage was set  */
 524                        LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: partial coverage"
 525                                " %d while full coverage %d requested\n",
 526                                UDP_SKB_CB(skb)->cscov, skb->len);
 527                        goto drop;
 528                }
 529                if (UDP_SKB_CB(skb)->cscov  <  up->pcrlen) {
 530                        LIMIT_NETDEBUG(KERN_WARNING "UDPLITE6: coverage %d "
 531                                                    "too small, need min %d\n",
 532                                       UDP_SKB_CB(skb)->cscov, up->pcrlen);
 533                        goto drop;
 534                }
 535        }
 536
 537        if (rcu_access_pointer(sk->sk_filter)) {
 538                if (udp_lib_checksum_complete(skb))
 539                        goto drop;
 540        }
 541
 542        skb_dst_drop(skb);
 543        rc = sock_queue_rcv_skb(sk, skb);
 544        if (rc < 0) {
 545                /* Note that an ENOMEM error is charged twice */
 546                if (rc == -ENOMEM)
 547                        UDP6_INC_STATS_BH(sock_net(sk),
 548                                        UDP_MIB_RCVBUFERRORS, is_udplite);
 549                goto drop_no_sk_drops_inc;
 550        }
 551
 552        return 0;
 553drop:
 554        atomic_inc(&sk->sk_drops);
 555drop_no_sk_drops_inc:
 556        UDP6_INC_STATS_BH(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 557        kfree_skb(skb);
 558        return -1;
 559}
 560
 561static struct sock *udp_v6_mcast_next(struct net *net, struct sock *sk,
 562                                      __be16 loc_port, const struct in6_addr *loc_addr,
 563                                      __be16 rmt_port, const struct in6_addr *rmt_addr,
 564                                      int dif)
 565{
 566        struct hlist_nulls_node *node;
 567        struct sock *s = sk;
 568        unsigned short num = ntohs(loc_port);
 569
 570        sk_nulls_for_each_from(s, node) {
 571                struct inet_sock *inet = inet_sk(s);
 572
 573                if (!net_eq(sock_net(s), net))
 574                        continue;
 575
 576                if (udp_sk(s)->udp_port_hash == num &&
 577                    s->sk_family == PF_INET6) {
 578                        struct ipv6_pinfo *np = inet6_sk(s);
 579                        if (inet->inet_dport) {
 580                                if (inet->inet_dport != rmt_port)
 581                                        continue;
 582                        }
 583                        if (!ipv6_addr_any(&np->daddr) &&
 584                            !ipv6_addr_equal(&np->daddr, rmt_addr))
 585                                continue;
 586
 587                        if (s->sk_bound_dev_if && s->sk_bound_dev_if != dif)
 588                                continue;
 589
 590                        if (!ipv6_addr_any(&np->rcv_saddr)) {
 591                                if (!ipv6_addr_equal(&np->rcv_saddr, loc_addr))
 592                                        continue;
 593                        }
 594                        if (!inet6_mc_check(s, loc_addr, rmt_addr))
 595                                continue;
 596                        return s;
 597                }
 598        }
 599        return NULL;
 600}
 601
 602static void flush_stack(struct sock **stack, unsigned int count,
 603                        struct sk_buff *skb, unsigned int final)
 604{
 605        unsigned int i;
 606        struct sock *sk;
 607        struct sk_buff *skb1;
 608
 609        for (i = 0; i < count; i++) {
 610                skb1 = (i == final) ? skb : skb_clone(skb, GFP_ATOMIC);
 611
 612                sk = stack[i];
 613                if (skb1) {
 614                        if (sk_rcvqueues_full(sk, skb1)) {
 615                                kfree_skb(skb1);
 616                                goto drop;
 617                        }
 618                        bh_lock_sock(sk);
 619                        if (!sock_owned_by_user(sk))
 620                                udpv6_queue_rcv_skb(sk, skb1);
 621                        else if (sk_add_backlog(sk, skb1)) {
 622                                kfree_skb(skb1);
 623                                bh_unlock_sock(sk);
 624                                goto drop;
 625                        }
 626                        bh_unlock_sock(sk);
 627                        continue;
 628                }
 629drop:
 630                atomic_inc(&sk->sk_drops);
 631                UDP6_INC_STATS_BH(sock_net(sk),
 632                                UDP_MIB_RCVBUFERRORS, IS_UDPLITE(sk));
 633                UDP6_INC_STATS_BH(sock_net(sk),
 634                                UDP_MIB_INERRORS, IS_UDPLITE(sk));
 635        }
 636}
 637/*
 638 * Note: called only from the BH handler context,
 639 * so we don't need to lock the hashes.
 640 */
 641static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
 642                const struct in6_addr *saddr, const struct in6_addr *daddr,
 643                struct udp_table *udptable)
 644{
 645        struct sock *sk, *stack[256 / sizeof(struct sock *)];
 646        const struct udphdr *uh = udp_hdr(skb);
 647        struct udp_hslot *hslot = udp_hashslot(udptable, net, ntohs(uh->dest));
 648        int dif;
 649        unsigned int i, count = 0;
 650
 651        spin_lock(&hslot->lock);
 652        sk = sk_nulls_head(&hslot->head);
 653        dif = inet6_iif(skb);
 654        sk = udp_v6_mcast_next(net, sk, uh->dest, daddr, uh->source, saddr, dif);
 655        while (sk) {
 656                stack[count++] = sk;
 657                sk = udp_v6_mcast_next(net, sk_nulls_next(sk), uh->dest, daddr,
 658                                       uh->source, saddr, dif);
 659                if (unlikely(count == ARRAY_SIZE(stack))) {
 660                        if (!sk)
 661                                break;
 662                        flush_stack(stack, count, skb, ~0);
 663                        count = 0;
 664                }
 665        }
 666        /*
 667         * before releasing the lock, we must take reference on sockets
 668         */
 669        for (i = 0; i < count; i++)
 670                sock_hold(stack[i]);
 671
 672        spin_unlock(&hslot->lock);
 673
 674        if (count) {
 675                flush_stack(stack, count, skb, count - 1);
 676
 677                for (i = 0; i < count; i++)
 678                        sock_put(stack[i]);
 679        } else {
 680                kfree_skb(skb);
 681        }
 682        return 0;
 683}
 684
 685static inline int udp6_csum_init(struct sk_buff *skb, struct udphdr *uh,
 686                                 int proto)
 687{
 688        int err;
 689
 690        UDP_SKB_CB(skb)->partial_cov = 0;
 691        UDP_SKB_CB(skb)->cscov = skb->len;
 692
 693        if (proto == IPPROTO_UDPLITE) {
 694                err = udplite_checksum_init(skb, uh);
 695                if (err)
 696                        return err;
 697        }
 698
 699        if (uh->check == 0) {
 700                /* RFC 2460 section 8.1 says that we SHOULD log
 701                   this error. Well, it is reasonable.
 702                 */
 703                LIMIT_NETDEBUG(KERN_INFO "IPv6: udp checksum is 0\n");
 704                return 1;
 705        }
 706        if (skb->ip_summed == CHECKSUM_COMPLETE &&
 707            !csum_ipv6_magic(&ipv6_hdr(skb)->saddr, &ipv6_hdr(skb)->daddr,
 708                             skb->len, proto, skb->csum))
 709                skb->ip_summed = CHECKSUM_UNNECESSARY;
 710
 711        if (!skb_csum_unnecessary(skb))
 712                skb->csum = ~csum_unfold(csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
 713                                                         &ipv6_hdr(skb)->daddr,
 714                                                         skb->len, proto, 0));
 715
 716        return 0;
 717}
 718
 719int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
 720                   int proto)
 721{
 722        struct net *net = dev_net(skb->dev);
 723        struct sock *sk;
 724        struct udphdr *uh;
 725        const struct in6_addr *saddr, *daddr;
 726        u32 ulen = 0;
 727
 728        if (!pskb_may_pull(skb, sizeof(struct udphdr)))
 729                goto discard;
 730
 731        saddr = &ipv6_hdr(skb)->saddr;
 732        daddr = &ipv6_hdr(skb)->daddr;
 733        uh = udp_hdr(skb);
 734
 735        ulen = ntohs(uh->len);
 736        if (ulen > skb->len)
 737                goto short_packet;
 738
 739        if (proto == IPPROTO_UDP) {
 740                /* UDP validates ulen. */
 741
 742                /* Check for jumbo payload */
 743                if (ulen == 0)
 744                        ulen = skb->len;
 745
 746                if (ulen < sizeof(*uh))
 747                        goto short_packet;
 748
 749                if (ulen < skb->len) {
 750                        if (pskb_trim_rcsum(skb, ulen))
 751                                goto short_packet;
 752                        saddr = &ipv6_hdr(skb)->saddr;
 753                        daddr = &ipv6_hdr(skb)->daddr;
 754                        uh = udp_hdr(skb);
 755                }
 756        }
 757
 758        if (udp6_csum_init(skb, uh, proto))
 759                goto discard;
 760
 761        /*
 762         *      Multicast receive code
 763         */
 764        if (ipv6_addr_is_multicast(daddr))
 765                return __udp6_lib_mcast_deliver(net, skb,
 766                                saddr, daddr, udptable);
 767
 768        /* Unicast */
 769
 770        /*
 771         * check socket cache ... must talk to Alan about his plans
 772         * for sock caches... i'll skip this for now.
 773         */
 774        sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
 775
 776        if (sk == NULL) {
 777                if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 778                        goto discard;
 779
 780                if (udp_lib_checksum_complete(skb))
 781                        goto discard;
 782                UDP6_INC_STATS_BH(net, UDP_MIB_NOPORTS,
 783                                proto == IPPROTO_UDPLITE);
 784
 785                icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 786
 787                kfree_skb(skb);
 788                return 0;
 789        }
 790
 791        /* deliver */
 792
 793        if (sk_rcvqueues_full(sk, skb)) {
 794                sock_put(sk);
 795                goto discard;
 796        }
 797        bh_lock_sock(sk);
 798        if (!sock_owned_by_user(sk))
 799                udpv6_queue_rcv_skb(sk, skb);
 800        else if (sk_add_backlog(sk, skb)) {
 801                atomic_inc(&sk->sk_drops);
 802                bh_unlock_sock(sk);
 803                sock_put(sk);
 804                goto discard;
 805        }
 806        bh_unlock_sock(sk);
 807        sock_put(sk);
 808        return 0;
 809
 810short_packet:
 811        LIMIT_NETDEBUG(KERN_DEBUG "UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
 812                       proto == IPPROTO_UDPLITE ? "-Lite" : "",
 813                       saddr,
 814                       ntohs(uh->source),
 815                       ulen,
 816                       skb->len,
 817                       daddr,
 818                       ntohs(uh->dest));
 819
 820discard:
 821        UDP6_INC_STATS_BH(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
 822        kfree_skb(skb);
 823        return 0;
 824}
 825
 826static __inline__ int udpv6_rcv(struct sk_buff *skb)
 827{
 828        return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
 829}
 830
 831/*
 832 * Throw away all pending data and cancel the corking. Socket is locked.
 833 */
 834static void udp_v6_flush_pending_frames(struct sock *sk)
 835{
 836        struct udp_sock *up = udp_sk(sk);
 837
 838        if (up->pending == AF_INET)
 839                udp_flush_pending_frames(sk);
 840        else if (up->pending) {
 841                up->len = 0;
 842                up->pending = 0;
 843                ip6_flush_pending_frames(sk);
 844        }
 845}
 846
 847/**
 848 *      udp6_hwcsum_outgoing  -  handle outgoing HW checksumming
 849 *      @sk:    socket we are sending on
 850 *      @skb:   sk_buff containing the filled-in UDP header
 851 *              (checksum field must be zeroed out)
 852 */
 853static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
 854                                 const struct in6_addr *saddr,
 855                                 const struct in6_addr *daddr, int len)
 856{
 857        unsigned int offset;
 858        struct udphdr *uh = udp_hdr(skb);
 859        __wsum csum = 0;
 860
 861        if (skb_queue_len(&sk->sk_write_queue) == 1) {
 862                /* Only one fragment on the socket.  */
 863                skb->csum_start = skb_transport_header(skb) - skb->head;
 864                skb->csum_offset = offsetof(struct udphdr, check);
 865                uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
 866        } else {
 867                /*
 868                 * HW-checksum won't work as there are two or more
 869                 * fragments on the socket so that all csums of sk_buffs
 870                 * should be together
 871                 */
 872                offset = skb_transport_offset(skb);
 873                skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
 874
 875                skb->ip_summed = CHECKSUM_NONE;
 876
 877                skb_queue_walk(&sk->sk_write_queue, skb) {
 878                        csum = csum_add(csum, skb->csum);
 879                }
 880
 881                uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
 882                                            csum);
 883                if (uh->check == 0)
 884                        uh->check = CSUM_MANGLED_0;
 885        }
 886}
 887
 888/*
 889 *      Sending
 890 */
 891
 892static int udp_v6_push_pending_frames(struct sock *sk)
 893{
 894        struct sk_buff *skb;
 895        struct udphdr *uh;
 896        struct udp_sock  *up = udp_sk(sk);
 897        struct inet_sock *inet = inet_sk(sk);
 898        struct flowi6 *fl6 = &inet->cork.fl.u.ip6;
 899        int err = 0;
 900        int is_udplite = IS_UDPLITE(sk);
 901        __wsum csum = 0;
 902
 903        /* Grab the skbuff where UDP header space exists. */
 904        if ((skb = skb_peek(&sk->sk_write_queue)) == NULL)
 905                goto out;
 906
 907        /*
 908         * Create a UDP header
 909         */
 910        uh = udp_hdr(skb);
 911        uh->source = fl6->fl6_sport;
 912        uh->dest = fl6->fl6_dport;
 913        uh->len = htons(up->len);
 914        uh->check = 0;
 915
 916        if (is_udplite)
 917                csum = udplite_csum_outgoing(sk, skb);
 918        else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
 919                udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr,
 920                                     up->len);
 921                goto send;
 922        } else
 923                csum = udp_csum_outgoing(sk, skb);
 924
 925        /* add protocol-dependent pseudo-header */
 926        uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
 927                                    up->len, fl6->flowi6_proto, csum);
 928        if (uh->check == 0)
 929                uh->check = CSUM_MANGLED_0;
 930
 931send:
 932        err = ip6_push_pending_frames(sk);
 933        if (err) {
 934                if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
 935                        UDP6_INC_STATS_USER(sock_net(sk),
 936                                            UDP_MIB_SNDBUFERRORS, is_udplite);
 937                        err = 0;
 938                }
 939        } else
 940                UDP6_INC_STATS_USER(sock_net(sk),
 941                                    UDP_MIB_OUTDATAGRAMS, is_udplite);
 942out:
 943        up->len = 0;
 944        up->pending = 0;
 945        return err;
 946}
 947
 948int udpv6_sendmsg(struct kiocb *iocb, struct sock *sk,
 949                  struct msghdr *msg, size_t len)
 950{
 951        struct ipv6_txoptions opt_space;
 952        struct udp_sock *up = udp_sk(sk);
 953        struct inet_sock *inet = inet_sk(sk);
 954        struct ipv6_pinfo *np = inet6_sk(sk);
 955        struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) msg->msg_name;
 956        struct in6_addr *daddr, *final_p, final;
 957        struct ipv6_txoptions *opt = NULL;
 958        struct ip6_flowlabel *flowlabel = NULL;
 959        struct flowi6 fl6;
 960        struct dst_entry *dst;
 961        int addr_len = msg->msg_namelen;
 962        int ulen = len;
 963        int hlimit = -1;
 964        int tclass = -1;
 965        int dontfrag = -1;
 966        int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
 967        int err;
 968        int connected = 0;
 969        int is_udplite = IS_UDPLITE(sk);
 970        int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
 971
 972        /* destination address check */
 973        if (sin6) {
 974                if (addr_len < offsetof(struct sockaddr, sa_data))
 975                        return -EINVAL;
 976
 977                switch (sin6->sin6_family) {
 978                case AF_INET6:
 979                        if (addr_len < SIN6_LEN_RFC2133)
 980                                return -EINVAL;
 981                        daddr = &sin6->sin6_addr;
 982                        break;
 983                case AF_INET:
 984                        goto do_udp_sendmsg;
 985                case AF_UNSPEC:
 986                        msg->msg_name = sin6 = NULL;
 987                        msg->msg_namelen = addr_len = 0;
 988                        daddr = NULL;
 989                        break;
 990                default:
 991                        return -EINVAL;
 992                }
 993        } else if (!up->pending) {
 994                if (sk->sk_state != TCP_ESTABLISHED)
 995                        return -EDESTADDRREQ;
 996                daddr = &np->daddr;
 997        } else
 998                daddr = NULL;
 999
1000        if (daddr) {
1001                if (ipv6_addr_v4mapped(daddr)) {
1002                        struct sockaddr_in sin;
1003                        sin.sin_family = AF_INET;
1004                        sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1005                        sin.sin_addr.s_addr = daddr->s6_addr32[3];
1006                        msg->msg_name = &sin;
1007                        msg->msg_namelen = sizeof(sin);
1008do_udp_sendmsg:
1009                        if (__ipv6_only_sock(sk))
1010                                return -ENETUNREACH;
1011                        return udp_sendmsg(iocb, sk, msg, len);
1012                }
1013        }
1014
1015        if (up->pending == AF_INET)
1016                return udp_sendmsg(iocb, sk, msg, len);
1017
1018        /* Rough check on arithmetic overflow,
1019           better check is made in ip6_append_data().
1020           */
1021        if (len > INT_MAX - sizeof(struct udphdr))
1022                return -EMSGSIZE;
1023
1024        if (up->pending) {
1025                /*
1026                 * There are pending frames.
1027                 * The socket lock must be held while it's corked.
1028                 */
1029                lock_sock(sk);
1030                if (likely(up->pending)) {
1031                        if (unlikely(up->pending != AF_INET6)) {
1032                                release_sock(sk);
1033                                return -EAFNOSUPPORT;
1034                        }
1035                        dst = NULL;
1036                        goto do_append_data;
1037                }
1038                release_sock(sk);
1039        }
1040        ulen += sizeof(struct udphdr);
1041
1042        memset(&fl6, 0, sizeof(fl6));
1043
1044        if (sin6) {
1045                if (sin6->sin6_port == 0)
1046                        return -EINVAL;
1047
1048                fl6.fl6_dport = sin6->sin6_port;
1049                daddr = &sin6->sin6_addr;
1050
1051                if (np->sndflow) {
1052                        fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1053                        if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1054                                flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1055                                if (flowlabel == NULL)
1056                                        return -EINVAL;
1057                                daddr = &flowlabel->dst;
1058                        }
1059                }
1060
1061                /*
1062                 * Otherwise it will be difficult to maintain
1063                 * sk->sk_dst_cache.
1064                 */
1065                if (sk->sk_state == TCP_ESTABLISHED &&
1066                    ipv6_addr_equal(daddr, &np->daddr))
1067                        daddr = &np->daddr;
1068
1069                if (addr_len >= sizeof(struct sockaddr_in6) &&
1070                    sin6->sin6_scope_id &&
1071                    ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
1072                        fl6.flowi6_oif = sin6->sin6_scope_id;
1073        } else {
1074                if (sk->sk_state != TCP_ESTABLISHED)
1075                        return -EDESTADDRREQ;
1076
1077                fl6.fl6_dport = inet->inet_dport;
1078                daddr = &np->daddr;
1079                fl6.flowlabel = np->flow_label;
1080                connected = 1;
1081        }
1082
1083        if (!fl6.flowi6_oif)
1084                fl6.flowi6_oif = sk->sk_bound_dev_if;
1085
1086        if (!fl6.flowi6_oif)
1087                fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1088
1089        fl6.flowi6_mark = sk->sk_mark;
1090
1091        if (msg->msg_controllen) {
1092                opt = &opt_space;
1093                memset(opt, 0, sizeof(struct ipv6_txoptions));
1094                opt->tot_len = sizeof(*opt);
1095
1096                err = datagram_send_ctl(sock_net(sk), sk, msg, &fl6, opt,
1097                                        &hlimit, &tclass, &dontfrag);
1098                if (err < 0) {
1099                        fl6_sock_release(flowlabel);
1100                        return err;
1101                }
1102                if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1103                        flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1104                        if (flowlabel == NULL)
1105                                return -EINVAL;
1106                }
1107                if (!(opt->opt_nflen|opt->opt_flen))
1108                        opt = NULL;
1109                connected = 0;
1110        }
1111        if (opt == NULL)
1112                opt = np->opt;
1113        if (flowlabel)
1114                opt = fl6_merge_options(&opt_space, flowlabel, opt);
1115        opt = ipv6_fixup_options(&opt_space, opt);
1116
1117        fl6.flowi6_proto = sk->sk_protocol;
1118        if (!ipv6_addr_any(daddr))
1119                fl6.daddr = *daddr;
1120        else
1121                fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1122        if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1123                fl6.saddr = np->saddr;
1124        fl6.fl6_sport = inet->inet_sport;
1125
1126        final_p = fl6_update_dst(&fl6, opt, &final);
1127        if (final_p)
1128                connected = 0;
1129
1130        if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1131                fl6.flowi6_oif = np->mcast_oif;
1132                connected = 0;
1133        } else if (!fl6.flowi6_oif)
1134                fl6.flowi6_oif = np->ucast_oif;
1135
1136        security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1137
1138        dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p, true);
1139        if (IS_ERR(dst)) {
1140                err = PTR_ERR(dst);
1141                dst = NULL;
1142                goto out;
1143        }
1144
1145        if (hlimit < 0) {
1146                if (ipv6_addr_is_multicast(&fl6.daddr))
1147                        hlimit = np->mcast_hops;
1148                else
1149                        hlimit = np->hop_limit;
1150                if (hlimit < 0)
1151                        hlimit = ip6_dst_hoplimit(dst);
1152        }
1153
1154        if (tclass < 0)
1155                tclass = np->tclass;
1156
1157        if (dontfrag < 0)
1158                dontfrag = np->dontfrag;
1159
1160        if (msg->msg_flags&MSG_CONFIRM)
1161                goto do_confirm;
1162back_from_confirm:
1163
1164        lock_sock(sk);
1165        if (unlikely(up->pending)) {
1166                /* The socket is already corked while preparing it. */
1167                /* ... which is an evident application bug. --ANK */
1168                release_sock(sk);
1169
1170                LIMIT_NETDEBUG(KERN_DEBUG "udp cork app bug 2\n");
1171                err = -EINVAL;
1172                goto out;
1173        }
1174
1175        up->pending = AF_INET6;
1176
1177do_append_data:
1178        up->len += ulen;
1179        getfrag  =  is_udplite ?  udplite_getfrag : ip_generic_getfrag;
1180        err = ip6_append_data(sk, getfrag, msg->msg_iov, ulen,
1181                sizeof(struct udphdr), hlimit, tclass, opt, &fl6,
1182                (struct rt6_info*)dst,
1183                corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, dontfrag);
1184        if (err)
1185                udp_v6_flush_pending_frames(sk);
1186        else if (!corkreq)
1187                err = udp_v6_push_pending_frames(sk);
1188        else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1189                up->pending = 0;
1190
1191        if (dst) {
1192                if (connected) {
1193                        ip6_dst_store(sk, dst,
1194                                      ipv6_addr_equal(&fl6.daddr, &np->daddr) ?
1195                                      &np->daddr : NULL,
1196#ifdef CONFIG_IPV6_SUBTREES
1197                                      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1198                                      &np->saddr :
1199#endif
1200                                      NULL);
1201                } else {
1202                        dst_release(dst);
1203                }
1204                dst = NULL;
1205        }
1206
1207        if (err > 0)
1208                err = np->recverr ? net_xmit_errno(err) : 0;
1209        release_sock(sk);
1210out:
1211        dst_release(dst);
1212        fl6_sock_release(flowlabel);
1213        if (!err)
1214                return len;
1215        /*
1216         * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1217         * ENOBUFS might not be good (it's not tunable per se), but otherwise
1218         * we don't have a good statistic (IpOutDiscards but it can be too many
1219         * things).  We could add another new stat but at least for now that
1220         * seems like overkill.
1221         */
1222        if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1223                UDP6_INC_STATS_USER(sock_net(sk),
1224                                UDP_MIB_SNDBUFERRORS, is_udplite);
1225        }
1226        return err;
1227
1228do_confirm:
1229        dst_confirm(dst);
1230        if (!(msg->msg_flags&MSG_PROBE) || len)
1231                goto back_from_confirm;
1232        err = 0;
1233        goto out;
1234}
1235
1236void udpv6_destroy_sock(struct sock *sk)
1237{
1238        lock_sock(sk);
1239        udp_v6_flush_pending_frames(sk);
1240        release_sock(sk);
1241
1242        inet6_destroy_sock(sk);
1243}
1244
1245/*
1246 *      Socket option code for UDP
1247 */
1248int udpv6_setsockopt(struct sock *sk, int level, int optname,
1249                     char __user *optval, unsigned int optlen)
1250{
1251        if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1252                return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1253                                          udp_v6_push_pending_frames);
1254        return ipv6_setsockopt(sk, level, optname, optval, optlen);
1255}
1256
1257#ifdef CONFIG_COMPAT
1258int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1259                            char __user *optval, unsigned int optlen)
1260{
1261        if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1262                return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1263                                          udp_v6_push_pending_frames);
1264        return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1265}
1266#endif
1267
1268int udpv6_getsockopt(struct sock *sk, int level, int optname,
1269                     char __user *optval, int __user *optlen)
1270{
1271        if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1272                return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1273        return ipv6_getsockopt(sk, level, optname, optval, optlen);
1274}
1275
1276#ifdef CONFIG_COMPAT
1277int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1278                            char __user *optval, int __user *optlen)
1279{
1280        if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1281                return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1282        return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1283}
1284#endif
1285
1286static int udp6_ufo_send_check(struct sk_buff *skb)
1287{
1288        const struct ipv6hdr *ipv6h;
1289        struct udphdr *uh;
1290
1291        if (!pskb_may_pull(skb, sizeof(*uh)))
1292                return -EINVAL;
1293
1294        ipv6h = ipv6_hdr(skb);
1295        uh = udp_hdr(skb);
1296
1297        uh->check = ~csum_ipv6_magic(&ipv6h->saddr, &ipv6h->daddr, skb->len,
1298                                     IPPROTO_UDP, 0);
1299        skb->csum_start = skb_transport_header(skb) - skb->head;
1300        skb->csum_offset = offsetof(struct udphdr, check);
1301        skb->ip_summed = CHECKSUM_PARTIAL;
1302        return 0;
1303}
1304
1305static struct sk_buff *udp6_ufo_fragment(struct sk_buff *skb,
1306        netdev_features_t features)
1307{
1308        struct sk_buff *segs = ERR_PTR(-EINVAL);
1309        unsigned int mss;
1310        unsigned int unfrag_ip6hlen, unfrag_len;
1311        struct frag_hdr *fptr;
1312        u8 *mac_start, *prevhdr;
1313        u8 nexthdr;
1314        u8 frag_hdr_sz = sizeof(struct frag_hdr);
1315        int offset;
1316        __wsum csum;
1317
1318        mss = skb_shinfo(skb)->gso_size;
1319        if (unlikely(skb->len <= mss))
1320                goto out;
1321
1322        if (skb_gso_ok(skb, features | NETIF_F_GSO_ROBUST)) {
1323                /* Packet is from an untrusted source, reset gso_segs. */
1324                int type = skb_shinfo(skb)->gso_type;
1325
1326                if (unlikely(type & ~(SKB_GSO_UDP | SKB_GSO_DODGY) ||
1327                             !(type & (SKB_GSO_UDP))))
1328                        goto out;
1329
1330                skb_shinfo(skb)->gso_segs = DIV_ROUND_UP(skb->len, mss);
1331
1332                segs = NULL;
1333                goto out;
1334        }
1335
1336        /* Do software UFO. Complete and fill in the UDP checksum as HW cannot
1337         * do checksum of UDP packets sent as multiple IP fragments.
1338         */
1339        offset = skb_checksum_start_offset(skb);
1340        csum = skb_checksum(skb, offset, skb->len- offset, 0);
1341        offset += skb->csum_offset;
1342        *(__sum16 *)(skb->data + offset) = csum_fold(csum);
1343        skb->ip_summed = CHECKSUM_NONE;
1344
1345        /* Check if there is enough headroom to insert fragment header. */
1346        if ((skb_mac_header(skb) < skb->head + frag_hdr_sz) &&
1347            pskb_expand_head(skb, frag_hdr_sz, 0, GFP_ATOMIC))
1348                goto out;
1349
1350        /* Find the unfragmentable header and shift it left by frag_hdr_sz
1351         * bytes to insert fragment header.
1352         */
1353        unfrag_ip6hlen = ip6_find_1stfragopt(skb, &prevhdr);
1354        nexthdr = *prevhdr;
1355        *prevhdr = NEXTHDR_FRAGMENT;
1356        unfrag_len = skb_network_header(skb) - skb_mac_header(skb) +
1357                     unfrag_ip6hlen;
1358        mac_start = skb_mac_header(skb);
1359        memmove(mac_start-frag_hdr_sz, mac_start, unfrag_len);
1360
1361        skb->mac_header -= frag_hdr_sz;
1362        skb->network_header -= frag_hdr_sz;
1363
1364        fptr = (struct frag_hdr *)(skb_network_header(skb) + unfrag_ip6hlen);
1365        fptr->nexthdr = nexthdr;
1366        fptr->reserved = 0;
1367        ipv6_select_ident(fptr, (struct rt6_info *)skb_dst(skb));
1368
1369        /* Fragment the skb. ipv6 header and the remaining fields of the
1370         * fragment header are updated in ipv6_gso_segment()
1371         */
1372        segs = skb_segment(skb, features);
1373
1374out:
1375        return segs;
1376}
1377
1378static const struct inet6_protocol udpv6_protocol = {
1379        .handler        =       udpv6_rcv,
1380        .err_handler    =       udpv6_err,
1381        .gso_send_check =       udp6_ufo_send_check,
1382        .gso_segment    =       udp6_ufo_fragment,
1383        .flags          =       INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1384};
1385
1386/* ------------------------------------------------------------------------ */
1387#ifdef CONFIG_PROC_FS
1388
1389static void udp6_sock_seq_show(struct seq_file *seq, struct sock *sp, int bucket)
1390{
1391        struct inet_sock *inet = inet_sk(sp);
1392        struct ipv6_pinfo *np = inet6_sk(sp);
1393        const struct in6_addr *dest, *src;
1394        __u16 destp, srcp;
1395
1396        dest  = &np->daddr;
1397        src   = &np->rcv_saddr;
1398        destp = ntohs(inet->inet_dport);
1399        srcp  = ntohs(inet->inet_sport);
1400        seq_printf(seq,
1401                   "%5d: %08X%08X%08X%08X:%04X %08X%08X%08X%08X:%04X "
1402                   "%02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %pK %d\n",
1403                   bucket,
1404                   src->s6_addr32[0], src->s6_addr32[1],
1405                   src->s6_addr32[2], src->s6_addr32[3], srcp,
1406                   dest->s6_addr32[0], dest->s6_addr32[1],
1407                   dest->s6_addr32[2], dest->s6_addr32[3], destp,
1408                   sp->sk_state,
1409                   sk_wmem_alloc_get(sp),
1410                   sk_rmem_alloc_get(sp),
1411                   0, 0L, 0,
1412                   sock_i_uid(sp), 0,
1413                   sock_i_ino(sp),
1414                   atomic_read(&sp->sk_refcnt), sp,
1415                   atomic_read(&sp->sk_drops));
1416}
1417
1418int udp6_seq_show(struct seq_file *seq, void *v)
1419{
1420        if (v == SEQ_START_TOKEN)
1421                seq_printf(seq,
1422                           "  sl  "
1423                           "local_address                         "
1424                           "remote_address                        "
1425                           "st tx_queue rx_queue tr tm->when retrnsmt"
1426                           "   uid  timeout inode ref pointer drops\n");
1427        else
1428                udp6_sock_seq_show(seq, v, ((struct udp_iter_state *)seq->private)->bucket);
1429        return 0;
1430}
1431
1432static const struct file_operations udp6_afinfo_seq_fops = {
1433        .owner    = THIS_MODULE,
1434        .open     = udp_seq_open,
1435        .read     = seq_read,
1436        .llseek   = seq_lseek,
1437        .release  = seq_release_net
1438};
1439
1440static struct udp_seq_afinfo udp6_seq_afinfo = {
1441        .name           = "udp6",
1442        .family         = AF_INET6,
1443        .udp_table      = &udp_table,
1444        .seq_fops       = &udp6_afinfo_seq_fops,
1445        .seq_ops        = {
1446                .show           = udp6_seq_show,
1447        },
1448};
1449
1450int __net_init udp6_proc_init(struct net *net)
1451{
1452        return udp_proc_register(net, &udp6_seq_afinfo);
1453}
1454
1455void udp6_proc_exit(struct net *net) {
1456        udp_proc_unregister(net, &udp6_seq_afinfo);
1457}
1458#endif /* CONFIG_PROC_FS */
1459
1460/* ------------------------------------------------------------------------ */
1461
1462struct proto udpv6_prot = {
1463        .name              = "UDPv6",
1464        .owner             = THIS_MODULE,
1465        .close             = udp_lib_close,
1466        .connect           = ip6_datagram_connect,
1467        .disconnect        = udp_disconnect,
1468        .ioctl             = udp_ioctl,
1469        .destroy           = udpv6_destroy_sock,
1470        .setsockopt        = udpv6_setsockopt,
1471        .getsockopt        = udpv6_getsockopt,
1472        .sendmsg           = udpv6_sendmsg,
1473        .recvmsg           = udpv6_recvmsg,
1474        .backlog_rcv       = udpv6_queue_rcv_skb,
1475        .hash              = udp_lib_hash,
1476        .unhash            = udp_lib_unhash,
1477        .rehash            = udp_v6_rehash,
1478        .get_port          = udp_v6_get_port,
1479        .memory_allocated  = &udp_memory_allocated,
1480        .sysctl_mem        = sysctl_udp_mem,
1481        .sysctl_wmem       = &sysctl_udp_wmem_min,
1482        .sysctl_rmem       = &sysctl_udp_rmem_min,
1483        .obj_size          = sizeof(struct udp6_sock),
1484        .slab_flags        = SLAB_DESTROY_BY_RCU,
1485        .h.udp_table       = &udp_table,
1486#ifdef CONFIG_COMPAT
1487        .compat_setsockopt = compat_udpv6_setsockopt,
1488        .compat_getsockopt = compat_udpv6_getsockopt,
1489#endif
1490        .clear_sk          = sk_prot_clear_portaddr_nulls,
1491};
1492
1493static struct inet_protosw udpv6_protosw = {
1494        .type =      SOCK_DGRAM,
1495        .protocol =  IPPROTO_UDP,
1496        .prot =      &udpv6_prot,
1497        .ops =       &inet6_dgram_ops,
1498        .no_check =  UDP_CSUM_DEFAULT,
1499        .flags =     INET_PROTOSW_PERMANENT,
1500};
1501
1502
1503int __init udpv6_init(void)
1504{
1505        int ret;
1506
1507        ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1508        if (ret)
1509                goto out;
1510
1511        ret = inet6_register_protosw(&udpv6_protosw);
1512        if (ret)
1513                goto out_udpv6_protocol;
1514out:
1515        return ret;
1516
1517out_udpv6_protocol:
1518        inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1519        goto out;
1520}
1521
1522void udpv6_exit(void)
1523{
1524        inet6_unregister_protosw(&udpv6_protosw);
1525        inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1526}
1527