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/addrconf.h>
  41#include <net/ndisc.h>
  42#include <net/protocol.h>
  43#include <net/transp_v6.h>
  44#include <net/ip6_route.h>
  45#include <net/raw.h>
  46#include <net/tcp_states.h>
  47#include <net/ip6_checksum.h>
  48#include <net/xfrm.h>
  49#include <net/inet6_hashtables.h>
  50#include <net/busy_poll.h>
  51#include <net/sock_reuseport.h>
  52
  53#include <linux/proc_fs.h>
  54#include <linux/seq_file.h>
  55#include <trace/events/skb.h>
  56#include "udp_impl.h"
  57
  58static u32 udp6_ehashfn(const struct net *net,
  59                        const struct in6_addr *laddr,
  60                        const u16 lport,
  61                        const struct in6_addr *faddr,
  62                        const __be16 fport)
  63{
  64        static u32 udp6_ehash_secret __read_mostly;
  65        static u32 udp_ipv6_hash_secret __read_mostly;
  66
  67        u32 lhash, fhash;
  68
  69        net_get_random_once(&udp6_ehash_secret,
  70                            sizeof(udp6_ehash_secret));
  71        net_get_random_once(&udp_ipv6_hash_secret,
  72                            sizeof(udp_ipv6_hash_secret));
  73
  74        lhash = (__force u32)laddr->s6_addr32[3];
  75        fhash = __ipv6_addr_jhash(faddr, udp_ipv6_hash_secret);
  76
  77        return __inet6_ehashfn(lhash, lport, fhash, fport,
  78                               udp_ipv6_hash_secret + net_hash_mix(net));
  79}
  80
  81static u32 udp6_portaddr_hash(const struct net *net,
  82                              const struct in6_addr *addr6,
  83                              unsigned int port)
  84{
  85        unsigned int hash, mix = net_hash_mix(net);
  86
  87        if (ipv6_addr_any(addr6))
  88                hash = jhash_1word(0, mix);
  89        else if (ipv6_addr_v4mapped(addr6))
  90                hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
  91        else
  92                hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
  93
  94        return hash ^ port;
  95}
  96
  97int udp_v6_get_port(struct sock *sk, unsigned short snum)
  98{
  99        unsigned int hash2_nulladdr =
 100                udp6_portaddr_hash(sock_net(sk), &in6addr_any, snum);
 101        unsigned int hash2_partial =
 102                udp6_portaddr_hash(sock_net(sk), &sk->sk_v6_rcv_saddr, 0);
 103
 104        /* precompute partial secondary hash */
 105        udp_sk(sk)->udp_portaddr_hash = hash2_partial;
 106        return udp_lib_get_port(sk, snum, ipv6_rcv_saddr_equal, hash2_nulladdr);
 107}
 108
 109static void udp_v6_rehash(struct sock *sk)
 110{
 111        u16 new_hash = udp6_portaddr_hash(sock_net(sk),
 112                                          &sk->sk_v6_rcv_saddr,
 113                                          inet_sk(sk)->inet_num);
 114
 115        udp_lib_rehash(sk, new_hash);
 116}
 117
 118static int compute_score(struct sock *sk, struct net *net,
 119                         const struct in6_addr *saddr, __be16 sport,
 120                         const struct in6_addr *daddr, unsigned short hnum,
 121                         int dif)
 122{
 123        int score;
 124        struct inet_sock *inet;
 125
 126        if (!net_eq(sock_net(sk), net) ||
 127            udp_sk(sk)->udp_port_hash != hnum ||
 128            sk->sk_family != PF_INET6)
 129                return -1;
 130
 131        score = 0;
 132        inet = inet_sk(sk);
 133
 134        if (inet->inet_dport) {
 135                if (inet->inet_dport != sport)
 136                        return -1;
 137                score++;
 138        }
 139
 140        if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) {
 141                if (!ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
 142                        return -1;
 143                score++;
 144        }
 145
 146        if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
 147                if (!ipv6_addr_equal(&sk->sk_v6_daddr, saddr))
 148                        return -1;
 149                score++;
 150        }
 151
 152        if (sk->sk_bound_dev_if) {
 153                if (sk->sk_bound_dev_if != dif)
 154                        return -1;
 155                score++;
 156        }
 157
 158        if (sk->sk_incoming_cpu == raw_smp_processor_id())
 159                score++;
 160
 161        return score;
 162}
 163
 164/* called with rcu_read_lock() */
 165static struct sock *udp6_lib_lookup2(struct net *net,
 166                const struct in6_addr *saddr, __be16 sport,
 167                const struct in6_addr *daddr, unsigned int hnum, int dif,
 168                struct udp_hslot *hslot2,
 169                struct sk_buff *skb)
 170{
 171        struct sock *sk, *result;
 172        int score, badness, matches = 0, reuseport = 0;
 173        u32 hash = 0;
 174
 175        result = NULL;
 176        badness = -1;
 177        udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
 178                score = compute_score(sk, net, saddr, sport,
 179                                      daddr, hnum, dif);
 180                if (score > badness) {
 181                        reuseport = sk->sk_reuseport;
 182                        if (reuseport) {
 183                                hash = udp6_ehashfn(net, daddr, hnum,
 184                                                    saddr, sport);
 185
 186                                result = reuseport_select_sock(sk, hash, skb,
 187                                                        sizeof(struct udphdr));
 188                                if (result)
 189                                        return result;
 190                                matches = 1;
 191                        }
 192                        result = sk;
 193                        badness = score;
 194                } else if (score == badness && reuseport) {
 195                        matches++;
 196                        if (reciprocal_scale(hash, matches) == 0)
 197                                result = sk;
 198                        hash = next_pseudo_random32(hash);
 199                }
 200        }
 201        return result;
 202}
 203
 204/* rcu_read_lock() must be held */
 205struct sock *__udp6_lib_lookup(struct net *net,
 206                                      const struct in6_addr *saddr, __be16 sport,
 207                                      const struct in6_addr *daddr, __be16 dport,
 208                                      int dif, struct udp_table *udptable,
 209                                      struct sk_buff *skb)
 210{
 211        struct sock *sk, *result;
 212        unsigned short hnum = ntohs(dport);
 213        unsigned int hash2, slot2, slot = udp_hashfn(net, hnum, udptable->mask);
 214        struct udp_hslot *hslot2, *hslot = &udptable->hash[slot];
 215        int score, badness, matches = 0, reuseport = 0;
 216        u32 hash = 0;
 217
 218        if (hslot->count > 10) {
 219                hash2 = udp6_portaddr_hash(net, daddr, hnum);
 220                slot2 = hash2 & udptable->mask;
 221                hslot2 = &udptable->hash2[slot2];
 222                if (hslot->count < hslot2->count)
 223                        goto begin;
 224
 225                result = udp6_lib_lookup2(net, saddr, sport,
 226                                          daddr, hnum, dif,
 227                                          hslot2, skb);
 228                if (!result) {
 229                        unsigned int old_slot2 = slot2;
 230                        hash2 = udp6_portaddr_hash(net, &in6addr_any, hnum);
 231                        slot2 = hash2 & udptable->mask;
 232                        /* avoid searching the same slot again. */
 233                        if (unlikely(slot2 == old_slot2))
 234                                return result;
 235
 236                        hslot2 = &udptable->hash2[slot2];
 237                        if (hslot->count < hslot2->count)
 238                                goto begin;
 239
 240                        result = udp6_lib_lookup2(net, saddr, sport,
 241                                                  daddr, hnum, dif,
 242                                                  hslot2, skb);
 243                }
 244                return result;
 245        }
 246begin:
 247        result = NULL;
 248        badness = -1;
 249        sk_for_each_rcu(sk, &hslot->head) {
 250                score = compute_score(sk, net, saddr, sport, daddr, hnum, dif);
 251                if (score > badness) {
 252                        reuseport = sk->sk_reuseport;
 253                        if (reuseport) {
 254                                hash = udp6_ehashfn(net, daddr, hnum,
 255                                                    saddr, sport);
 256                                result = reuseport_select_sock(sk, hash, skb,
 257                                                        sizeof(struct udphdr));
 258                                if (result)
 259                                        return result;
 260                                matches = 1;
 261                        }
 262                        result = sk;
 263                        badness = score;
 264                } else if (score == badness && reuseport) {
 265                        matches++;
 266                        if (reciprocal_scale(hash, matches) == 0)
 267                                result = sk;
 268                        hash = next_pseudo_random32(hash);
 269                }
 270        }
 271        return result;
 272}
 273EXPORT_SYMBOL_GPL(__udp6_lib_lookup);
 274
 275static struct sock *__udp6_lib_lookup_skb(struct sk_buff *skb,
 276                                          __be16 sport, __be16 dport,
 277                                          struct udp_table *udptable)
 278{
 279        const struct ipv6hdr *iph = ipv6_hdr(skb);
 280        struct sock *sk;
 281
 282        sk = skb_steal_sock(skb);
 283        if (unlikely(sk))
 284                return sk;
 285        return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
 286                                 &iph->daddr, dport, inet6_iif(skb),
 287                                 udptable, skb);
 288}
 289
 290struct sock *udp6_lib_lookup_skb(struct sk_buff *skb,
 291                                 __be16 sport, __be16 dport)
 292{
 293        const struct ipv6hdr *iph = ipv6_hdr(skb);
 294
 295        return __udp6_lib_lookup(dev_net(skb->dev), &iph->saddr, sport,
 296                                 &iph->daddr, dport, inet6_iif(skb),
 297                                 &udp_table, skb);
 298}
 299EXPORT_SYMBOL_GPL(udp6_lib_lookup_skb);
 300
 301/* Must be called under rcu_read_lock().
 302 * Does increment socket refcount.
 303 */
 304#if IS_ENABLED(CONFIG_NETFILTER_XT_MATCH_SOCKET) || \
 305    IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TPROXY)
 306struct sock *udp6_lib_lookup(struct net *net, const struct in6_addr *saddr, __be16 sport,
 307                             const struct in6_addr *daddr, __be16 dport, int dif)
 308{
 309        struct sock *sk;
 310
 311        sk =  __udp6_lib_lookup(net, saddr, sport, daddr, dport,
 312                                dif, &udp_table, NULL);
 313        if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
 314                sk = NULL;
 315        return sk;
 316}
 317EXPORT_SYMBOL_GPL(udp6_lib_lookup);
 318#endif
 319
 320/*
 321 *      This should be easy, if there is something there we
 322 *      return it, otherwise we block.
 323 */
 324
 325int udpv6_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
 326                  int noblock, int flags, int *addr_len)
 327{
 328        struct ipv6_pinfo *np = inet6_sk(sk);
 329        struct inet_sock *inet = inet_sk(sk);
 330        struct sk_buff *skb;
 331        unsigned int ulen, copied;
 332        int peeked, peeking, off;
 333        int err;
 334        int is_udplite = IS_UDPLITE(sk);
 335        bool checksum_valid = false;
 336        int is_udp4;
 337        bool slow;
 338
 339        if (flags & MSG_ERRQUEUE)
 340                return ipv6_recv_error(sk, msg, len, addr_len);
 341
 342        if (np->rxpmtu && np->rxopt.bits.rxpmtu)
 343                return ipv6_recv_rxpmtu(sk, msg, len, addr_len);
 344
 345try_again:
 346        peeking = off = sk_peek_offset(sk, flags);
 347        skb = __skb_recv_datagram(sk, flags | (noblock ? MSG_DONTWAIT : 0),
 348                                  &peeked, &off, &err);
 349        if (!skb)
 350                return err;
 351
 352        ulen = skb->len;
 353        copied = len;
 354        if (copied > ulen - off)
 355                copied = ulen - off;
 356        else if (copied < ulen)
 357                msg->msg_flags |= MSG_TRUNC;
 358
 359        is_udp4 = (skb->protocol == htons(ETH_P_IP));
 360
 361        /*
 362         * If checksum is needed at all, try to do it while copying the
 363         * data.  If the data is truncated, or if we only want a partial
 364         * coverage checksum (UDP-Lite), do it before the copy.
 365         */
 366
 367        if (copied < ulen || UDP_SKB_CB(skb)->partial_cov || peeking) {
 368                checksum_valid = !udp_lib_checksum_complete(skb);
 369                if (!checksum_valid)
 370                        goto csum_copy_err;
 371        }
 372
 373        if (checksum_valid || skb_csum_unnecessary(skb))
 374                err = skb_copy_datagram_msg(skb, off, msg, copied);
 375        else {
 376                err = skb_copy_and_csum_datagram_msg(skb, off, msg);
 377                if (err == -EINVAL)
 378                        goto csum_copy_err;
 379        }
 380        if (unlikely(err)) {
 381                trace_kfree_skb(skb, udpv6_recvmsg);
 382                if (!peeked) {
 383                        atomic_inc(&sk->sk_drops);
 384                        if (is_udp4)
 385                                UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
 386                                              is_udplite);
 387                        else
 388                                UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS,
 389                                               is_udplite);
 390                }
 391                skb_free_datagram_locked(sk, skb);
 392                return err;
 393        }
 394        if (!peeked) {
 395                if (is_udp4)
 396                        UDP_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
 397                                      is_udplite);
 398                else
 399                        UDP6_INC_STATS(sock_net(sk), UDP_MIB_INDATAGRAMS,
 400                                       is_udplite);
 401        }
 402
 403        sock_recv_ts_and_drops(msg, sk, skb);
 404
 405        /* Copy the address. */
 406        if (msg->msg_name) {
 407                DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 408                sin6->sin6_family = AF_INET6;
 409                sin6->sin6_port = udp_hdr(skb)->source;
 410                sin6->sin6_flowinfo = 0;
 411
 412                if (is_udp4) {
 413                        ipv6_addr_set_v4mapped(ip_hdr(skb)->saddr,
 414                                               &sin6->sin6_addr);
 415                        sin6->sin6_scope_id = 0;
 416                } else {
 417                        sin6->sin6_addr = ipv6_hdr(skb)->saddr;
 418                        sin6->sin6_scope_id =
 419                                ipv6_iface_scope_id(&sin6->sin6_addr,
 420                                                    inet6_iif(skb));
 421                }
 422                *addr_len = sizeof(*sin6);
 423        }
 424
 425        if (np->rxopt.all)
 426                ip6_datagram_recv_common_ctl(sk, msg, skb);
 427
 428        if (is_udp4) {
 429                if (inet->cmsg_flags)
 430                        ip_cmsg_recv(msg, skb);
 431        } else {
 432                if (np->rxopt.all)
 433                        ip6_datagram_recv_specific_ctl(sk, msg, skb);
 434        }
 435
 436        err = copied;
 437        if (flags & MSG_TRUNC)
 438                err = ulen;
 439
 440        __skb_free_datagram_locked(sk, skb, peeking ? -err : err);
 441        return err;
 442
 443csum_copy_err:
 444        slow = lock_sock_fast(sk);
 445        if (!skb_kill_datagram(sk, skb, flags)) {
 446                if (is_udp4) {
 447                        UDP_INC_STATS(sock_net(sk),
 448                                      UDP_MIB_CSUMERRORS, is_udplite);
 449                        UDP_INC_STATS(sock_net(sk),
 450                                      UDP_MIB_INERRORS, is_udplite);
 451                } else {
 452                        UDP6_INC_STATS(sock_net(sk),
 453                                       UDP_MIB_CSUMERRORS, is_udplite);
 454                        UDP6_INC_STATS(sock_net(sk),
 455                                       UDP_MIB_INERRORS, is_udplite);
 456                }
 457        }
 458        unlock_sock_fast(sk, slow);
 459
 460        /* starting over for a new packet, but check if we need to yield */
 461        cond_resched();
 462        msg->msg_flags &= ~MSG_TRUNC;
 463        goto try_again;
 464}
 465
 466void __udp6_lib_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 467                    u8 type, u8 code, int offset, __be32 info,
 468                    struct udp_table *udptable)
 469{
 470        struct ipv6_pinfo *np;
 471        const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
 472        const struct in6_addr *saddr = &hdr->saddr;
 473        const struct in6_addr *daddr = &hdr->daddr;
 474        struct udphdr *uh = (struct udphdr *)(skb->data+offset);
 475        struct sock *sk;
 476        int harderr;
 477        int err;
 478        struct net *net = dev_net(skb->dev);
 479
 480        sk = __udp6_lib_lookup(net, daddr, uh->dest, saddr, uh->source,
 481                               inet6_iif(skb), udptable, skb);
 482        if (!sk) {
 483                __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
 484                                  ICMP6_MIB_INERRORS);
 485                return;
 486        }
 487
 488        harderr = icmpv6_err_convert(type, code, &err);
 489        np = inet6_sk(sk);
 490
 491        if (type == ICMPV6_PKT_TOOBIG) {
 492                if (!ip6_sk_accept_pmtu(sk))
 493                        goto out;
 494                ip6_sk_update_pmtu(skb, sk, info);
 495                if (np->pmtudisc != IPV6_PMTUDISC_DONT)
 496                        harderr = 1;
 497        }
 498        if (type == NDISC_REDIRECT) {
 499                ip6_sk_redirect(skb, sk);
 500                goto out;
 501        }
 502
 503        if (!np->recverr) {
 504                if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 505                        goto out;
 506        } else {
 507                ipv6_icmp_error(sk, skb, err, uh->dest, ntohl(info), (u8 *)(uh+1));
 508        }
 509
 510        sk->sk_err = err;
 511        sk->sk_error_report(sk);
 512out:
 513        return;
 514}
 515
 516static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 517{
 518        int rc;
 519
 520        if (!ipv6_addr_any(&sk->sk_v6_daddr)) {
 521                sock_rps_save_rxhash(sk, skb);
 522                sk_mark_napi_id(sk, skb);
 523                sk_incoming_cpu_update(sk);
 524        }
 525
 526        rc = __sock_queue_rcv_skb(sk, skb);
 527        if (rc < 0) {
 528                int is_udplite = IS_UDPLITE(sk);
 529
 530                /* Note that an ENOMEM error is charged twice */
 531                if (rc == -ENOMEM)
 532                        UDP6_INC_STATS(sock_net(sk),
 533                                         UDP_MIB_RCVBUFERRORS, is_udplite);
 534                UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 535                kfree_skb(skb);
 536                return -1;
 537        }
 538        return 0;
 539}
 540
 541static __inline__ void udpv6_err(struct sk_buff *skb,
 542                                 struct inet6_skb_parm *opt, u8 type,
 543                                 u8 code, int offset, __be32 info)
 544{
 545        __udp6_lib_err(skb, opt, type, code, offset, info, &udp_table);
 546}
 547
 548static struct static_key udpv6_encap_needed __read_mostly;
 549void udpv6_encap_enable(void)
 550{
 551        if (!static_key_enabled(&udpv6_encap_needed))
 552                static_key_slow_inc(&udpv6_encap_needed);
 553}
 554EXPORT_SYMBOL(udpv6_encap_enable);
 555
 556int udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 557{
 558        struct udp_sock *up = udp_sk(sk);
 559        int rc;
 560        int is_udplite = IS_UDPLITE(sk);
 561
 562        if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
 563                goto drop;
 564
 565        if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
 566                int (*encap_rcv)(struct sock *sk, struct sk_buff *skb);
 567
 568                /*
 569                 * This is an encapsulation socket so pass the skb to
 570                 * the socket's udp_encap_rcv() hook. Otherwise, just
 571                 * fall through and pass this up the UDP socket.
 572                 * up->encap_rcv() returns the following value:
 573                 * =0 if skb was successfully passed to the encap
 574                 *    handler or was discarded by it.
 575                 * >0 if skb should be passed on to UDP.
 576                 * <0 if skb should be resubmitted as proto -N
 577                 */
 578
 579                /* if we're overly short, let UDP handle it */
 580                encap_rcv = ACCESS_ONCE(up->encap_rcv);
 581                if (encap_rcv) {
 582                        int ret;
 583
 584                        /* Verify checksum before giving to encap */
 585                        if (udp_lib_checksum_complete(skb))
 586                                goto csum_error;
 587
 588                        ret = encap_rcv(sk, skb);
 589                        if (ret <= 0) {
 590                                __UDP_INC_STATS(sock_net(sk),
 591                                                UDP_MIB_INDATAGRAMS,
 592                                                is_udplite);
 593                                return -ret;
 594                        }
 595                }
 596
 597                /* FALLTHROUGH -- it's a UDP Packet */
 598        }
 599
 600        /*
 601         * UDP-Lite specific tests, ignored on UDP sockets (see net/ipv4/udp.c).
 602         */
 603        if ((is_udplite & UDPLITE_RECV_CC)  &&  UDP_SKB_CB(skb)->partial_cov) {
 604
 605                if (up->pcrlen == 0) {          /* full coverage was set  */
 606                        net_dbg_ratelimited("UDPLITE6: partial coverage %d while full coverage %d requested\n",
 607                                            UDP_SKB_CB(skb)->cscov, skb->len);
 608                        goto drop;
 609                }
 610                if (UDP_SKB_CB(skb)->cscov  <  up->pcrlen) {
 611                        net_dbg_ratelimited("UDPLITE6: coverage %d too small, need min %d\n",
 612                                            UDP_SKB_CB(skb)->cscov, up->pcrlen);
 613                        goto drop;
 614                }
 615        }
 616
 617        if (rcu_access_pointer(sk->sk_filter) &&
 618            udp_lib_checksum_complete(skb))
 619                goto csum_error;
 620
 621        if (sk_filter_trim_cap(sk, skb, sizeof(struct udphdr)))
 622                goto drop;
 623
 624        udp_csum_pull_header(skb);
 625        if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
 626                __UDP6_INC_STATS(sock_net(sk),
 627                                 UDP_MIB_RCVBUFERRORS, is_udplite);
 628                goto drop;
 629        }
 630
 631        skb_dst_drop(skb);
 632
 633        bh_lock_sock(sk);
 634        rc = 0;
 635        if (!sock_owned_by_user(sk))
 636                rc = __udpv6_queue_rcv_skb(sk, skb);
 637        else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
 638                bh_unlock_sock(sk);
 639                goto drop;
 640        }
 641        bh_unlock_sock(sk);
 642
 643        return rc;
 644
 645csum_error:
 646        __UDP6_INC_STATS(sock_net(sk), UDP_MIB_CSUMERRORS, is_udplite);
 647drop:
 648        __UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 649        atomic_inc(&sk->sk_drops);
 650        kfree_skb(skb);
 651        return -1;
 652}
 653
 654static bool __udp_v6_is_mcast_sock(struct net *net, struct sock *sk,
 655                                   __be16 loc_port, const struct in6_addr *loc_addr,
 656                                   __be16 rmt_port, const struct in6_addr *rmt_addr,
 657                                   int dif, unsigned short hnum)
 658{
 659        struct inet_sock *inet = inet_sk(sk);
 660
 661        if (!net_eq(sock_net(sk), net))
 662                return false;
 663
 664        if (udp_sk(sk)->udp_port_hash != hnum ||
 665            sk->sk_family != PF_INET6 ||
 666            (inet->inet_dport && inet->inet_dport != rmt_port) ||
 667            (!ipv6_addr_any(&sk->sk_v6_daddr) &&
 668                    !ipv6_addr_equal(&sk->sk_v6_daddr, rmt_addr)) ||
 669            (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif) ||
 670            (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 671                    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, loc_addr)))
 672                return false;
 673        if (!inet6_mc_check(sk, loc_addr, rmt_addr))
 674                return false;
 675        return true;
 676}
 677
 678static void udp6_csum_zero_error(struct sk_buff *skb)
 679{
 680        /* RFC 2460 section 8.1 says that we SHOULD log
 681         * this error. Well, it is reasonable.
 682         */
 683        net_dbg_ratelimited("IPv6: udp checksum is 0 for [%pI6c]:%u->[%pI6c]:%u\n",
 684                            &ipv6_hdr(skb)->saddr, ntohs(udp_hdr(skb)->source),
 685                            &ipv6_hdr(skb)->daddr, ntohs(udp_hdr(skb)->dest));
 686}
 687
 688/*
 689 * Note: called only from the BH handler context,
 690 * so we don't need to lock the hashes.
 691 */
 692static int __udp6_lib_mcast_deliver(struct net *net, struct sk_buff *skb,
 693                const struct in6_addr *saddr, const struct in6_addr *daddr,
 694                struct udp_table *udptable, int proto)
 695{
 696        struct sock *sk, *first = NULL;
 697        const struct udphdr *uh = udp_hdr(skb);
 698        unsigned short hnum = ntohs(uh->dest);
 699        struct udp_hslot *hslot = udp_hashslot(udptable, net, hnum);
 700        unsigned int offset = offsetof(typeof(*sk), sk_node);
 701        unsigned int hash2 = 0, hash2_any = 0, use_hash2 = (hslot->count > 10);
 702        int dif = inet6_iif(skb);
 703        struct hlist_node *node;
 704        struct sk_buff *nskb;
 705
 706        if (use_hash2) {
 707                hash2_any = udp6_portaddr_hash(net, &in6addr_any, hnum) &
 708                            udp_table.mask;
 709                hash2 = udp6_portaddr_hash(net, daddr, hnum) & udp_table.mask;
 710start_lookup:
 711                hslot = &udp_table.hash2[hash2];
 712                offset = offsetof(typeof(*sk), __sk_common.skc_portaddr_node);
 713        }
 714
 715        sk_for_each_entry_offset_rcu(sk, node, &hslot->head, offset) {
 716                if (!__udp_v6_is_mcast_sock(net, sk, uh->dest, daddr,
 717                                            uh->source, saddr, dif, hnum))
 718                        continue;
 719                /* If zero checksum and no_check is not on for
 720                 * the socket then skip it.
 721                 */
 722                if (!uh->check && !udp_sk(sk)->no_check6_rx)
 723                        continue;
 724                if (!first) {
 725                        first = sk;
 726                        continue;
 727                }
 728                nskb = skb_clone(skb, GFP_ATOMIC);
 729                if (unlikely(!nskb)) {
 730                        atomic_inc(&sk->sk_drops);
 731                        __UDP6_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
 732                                         IS_UDPLITE(sk));
 733                        __UDP6_INC_STATS(net, UDP_MIB_INERRORS,
 734                                         IS_UDPLITE(sk));
 735                        continue;
 736                }
 737
 738                if (udpv6_queue_rcv_skb(sk, nskb) > 0)
 739                        consume_skb(nskb);
 740        }
 741
 742        /* Also lookup *:port if we are using hash2 and haven't done so yet. */
 743        if (use_hash2 && hash2 != hash2_any) {
 744                hash2 = hash2_any;
 745                goto start_lookup;
 746        }
 747
 748        if (first) {
 749                if (udpv6_queue_rcv_skb(first, skb) > 0)
 750                        consume_skb(skb);
 751        } else {
 752                kfree_skb(skb);
 753                __UDP6_INC_STATS(net, UDP_MIB_IGNOREDMULTI,
 754                                 proto == IPPROTO_UDPLITE);
 755        }
 756        return 0;
 757}
 758
 759int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
 760                   int proto)
 761{
 762        const struct in6_addr *saddr, *daddr;
 763        struct net *net = dev_net(skb->dev);
 764        struct udphdr *uh;
 765        struct sock *sk;
 766        u32 ulen = 0;
 767
 768        if (!pskb_may_pull(skb, sizeof(struct udphdr)))
 769                goto discard;
 770
 771        saddr = &ipv6_hdr(skb)->saddr;
 772        daddr = &ipv6_hdr(skb)->daddr;
 773        uh = udp_hdr(skb);
 774
 775        ulen = ntohs(uh->len);
 776        if (ulen > skb->len)
 777                goto short_packet;
 778
 779        if (proto == IPPROTO_UDP) {
 780                /* UDP validates ulen. */
 781
 782                /* Check for jumbo payload */
 783                if (ulen == 0)
 784                        ulen = skb->len;
 785
 786                if (ulen < sizeof(*uh))
 787                        goto short_packet;
 788
 789                if (ulen < skb->len) {
 790                        if (pskb_trim_rcsum(skb, ulen))
 791                                goto short_packet;
 792                        saddr = &ipv6_hdr(skb)->saddr;
 793                        daddr = &ipv6_hdr(skb)->daddr;
 794                        uh = udp_hdr(skb);
 795                }
 796        }
 797
 798        if (udp6_csum_init(skb, uh, proto))
 799                goto csum_error;
 800
 801        /*
 802         *      Multicast receive code
 803         */
 804        if (ipv6_addr_is_multicast(daddr))
 805                return __udp6_lib_mcast_deliver(net, skb,
 806                                saddr, daddr, udptable, proto);
 807
 808        /* Unicast */
 809
 810        /*
 811         * check socket cache ... must talk to Alan about his plans
 812         * for sock caches... i'll skip this for now.
 813         */
 814        sk = __udp6_lib_lookup_skb(skb, uh->source, uh->dest, udptable);
 815        if (sk) {
 816                int ret;
 817
 818                if (!uh->check && !udp_sk(sk)->no_check6_rx) {
 819                        udp6_csum_zero_error(skb);
 820                        goto csum_error;
 821                }
 822
 823                if (inet_get_convert_csum(sk) && uh->check && !IS_UDPLITE(sk))
 824                        skb_checksum_try_convert(skb, IPPROTO_UDP, uh->check,
 825                                                 ip6_compute_pseudo);
 826
 827                ret = udpv6_queue_rcv_skb(sk, skb);
 828
 829                /* a return value > 0 means to resubmit the input */
 830                if (ret > 0)
 831                        return ret;
 832
 833                return 0;
 834        }
 835
 836        if (!uh->check) {
 837                udp6_csum_zero_error(skb);
 838                goto csum_error;
 839        }
 840
 841        if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 842                goto discard;
 843
 844        if (udp_lib_checksum_complete(skb))
 845                goto csum_error;
 846
 847        __UDP6_INC_STATS(net, UDP_MIB_NOPORTS, proto == IPPROTO_UDPLITE);
 848        icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_PORT_UNREACH, 0);
 849
 850        kfree_skb(skb);
 851        return 0;
 852
 853short_packet:
 854        net_dbg_ratelimited("UDP%sv6: short packet: From [%pI6c]:%u %d/%d to [%pI6c]:%u\n",
 855                            proto == IPPROTO_UDPLITE ? "-Lite" : "",
 856                            saddr, ntohs(uh->source),
 857                            ulen, skb->len,
 858                            daddr, ntohs(uh->dest));
 859        goto discard;
 860csum_error:
 861        __UDP6_INC_STATS(net, UDP_MIB_CSUMERRORS, proto == IPPROTO_UDPLITE);
 862discard:
 863        __UDP6_INC_STATS(net, UDP_MIB_INERRORS, proto == IPPROTO_UDPLITE);
 864        kfree_skb(skb);
 865        return 0;
 866}
 867
 868static __inline__ int udpv6_rcv(struct sk_buff *skb)
 869{
 870        return __udp6_lib_rcv(skb, &udp_table, IPPROTO_UDP);
 871}
 872
 873/*
 874 * Throw away all pending data and cancel the corking. Socket is locked.
 875 */
 876static void udp_v6_flush_pending_frames(struct sock *sk)
 877{
 878        struct udp_sock *up = udp_sk(sk);
 879
 880        if (up->pending == AF_INET)
 881                udp_flush_pending_frames(sk);
 882        else if (up->pending) {
 883                up->len = 0;
 884                up->pending = 0;
 885                ip6_flush_pending_frames(sk);
 886        }
 887}
 888
 889/**
 890 *      udp6_hwcsum_outgoing  -  handle outgoing HW checksumming
 891 *      @sk:    socket we are sending on
 892 *      @skb:   sk_buff containing the filled-in UDP header
 893 *              (checksum field must be zeroed out)
 894 */
 895static void udp6_hwcsum_outgoing(struct sock *sk, struct sk_buff *skb,
 896                                 const struct in6_addr *saddr,
 897                                 const struct in6_addr *daddr, int len)
 898{
 899        unsigned int offset;
 900        struct udphdr *uh = udp_hdr(skb);
 901        struct sk_buff *frags = skb_shinfo(skb)->frag_list;
 902        __wsum csum = 0;
 903
 904        if (!frags) {
 905                /* Only one fragment on the socket.  */
 906                skb->csum_start = skb_transport_header(skb) - skb->head;
 907                skb->csum_offset = offsetof(struct udphdr, check);
 908                uh->check = ~csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP, 0);
 909        } else {
 910                /*
 911                 * HW-checksum won't work as there are two or more
 912                 * fragments on the socket so that all csums of sk_buffs
 913                 * should be together
 914                 */
 915                offset = skb_transport_offset(skb);
 916                skb->csum = skb_checksum(skb, offset, skb->len - offset, 0);
 917
 918                skb->ip_summed = CHECKSUM_NONE;
 919
 920                do {
 921                        csum = csum_add(csum, frags->csum);
 922                } while ((frags = frags->next));
 923
 924                uh->check = csum_ipv6_magic(saddr, daddr, len, IPPROTO_UDP,
 925                                            csum);
 926                if (uh->check == 0)
 927                        uh->check = CSUM_MANGLED_0;
 928        }
 929}
 930
 931/*
 932 *      Sending
 933 */
 934
 935static int udp_v6_send_skb(struct sk_buff *skb, struct flowi6 *fl6)
 936{
 937        struct sock *sk = skb->sk;
 938        struct udphdr *uh;
 939        int err = 0;
 940        int is_udplite = IS_UDPLITE(sk);
 941        __wsum csum = 0;
 942        int offset = skb_transport_offset(skb);
 943        int len = skb->len - offset;
 944
 945        /*
 946         * Create a UDP header
 947         */
 948        uh = udp_hdr(skb);
 949        uh->source = fl6->fl6_sport;
 950        uh->dest = fl6->fl6_dport;
 951        uh->len = htons(len);
 952        uh->check = 0;
 953
 954        if (is_udplite)
 955                csum = udplite_csum(skb);
 956        else if (udp_sk(sk)->no_check6_tx) {   /* UDP csum disabled */
 957                skb->ip_summed = CHECKSUM_NONE;
 958                goto send;
 959        } else if (skb->ip_summed == CHECKSUM_PARTIAL) { /* UDP hardware csum */
 960                udp6_hwcsum_outgoing(sk, skb, &fl6->saddr, &fl6->daddr, len);
 961                goto send;
 962        } else
 963                csum = udp_csum(skb);
 964
 965        /* add protocol-dependent pseudo-header */
 966        uh->check = csum_ipv6_magic(&fl6->saddr, &fl6->daddr,
 967                                    len, fl6->flowi6_proto, csum);
 968        if (uh->check == 0)
 969                uh->check = CSUM_MANGLED_0;
 970
 971send:
 972        err = ip6_send_skb(skb);
 973        if (err) {
 974                if (err == -ENOBUFS && !inet6_sk(sk)->recverr) {
 975                        UDP6_INC_STATS(sock_net(sk),
 976                                       UDP_MIB_SNDBUFERRORS, is_udplite);
 977                        err = 0;
 978                }
 979        } else {
 980                UDP6_INC_STATS(sock_net(sk),
 981                               UDP_MIB_OUTDATAGRAMS, is_udplite);
 982        }
 983        return err;
 984}
 985
 986static int udp_v6_push_pending_frames(struct sock *sk)
 987{
 988        struct sk_buff *skb;
 989        struct udp_sock  *up = udp_sk(sk);
 990        struct flowi6 fl6;
 991        int err = 0;
 992
 993        if (up->pending == AF_INET)
 994                return udp_push_pending_frames(sk);
 995
 996        /* ip6_finish_skb will release the cork, so make a copy of
 997         * fl6 here.
 998         */
 999        fl6 = inet_sk(sk)->cork.fl.u.ip6;
1000
1001        skb = ip6_finish_skb(sk);
1002        if (!skb)
1003                goto out;
1004
1005        err = udp_v6_send_skb(skb, &fl6);
1006
1007out:
1008        up->len = 0;
1009        up->pending = 0;
1010        return err;
1011}
1012
1013int udpv6_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
1014{
1015        struct ipv6_txoptions opt_space;
1016        struct udp_sock *up = udp_sk(sk);
1017        struct inet_sock *inet = inet_sk(sk);
1018        struct ipv6_pinfo *np = inet6_sk(sk);
1019        DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1020        struct in6_addr *daddr, *final_p, final;
1021        struct ipv6_txoptions *opt = NULL;
1022        struct ipv6_txoptions *opt_to_free = NULL;
1023        struct ip6_flowlabel *flowlabel = NULL;
1024        struct flowi6 fl6;
1025        struct dst_entry *dst;
1026        struct ipcm6_cookie ipc6;
1027        int addr_len = msg->msg_namelen;
1028        int ulen = len;
1029        int corkreq = up->corkflag || msg->msg_flags&MSG_MORE;
1030        int err;
1031        int connected = 0;
1032        int is_udplite = IS_UDPLITE(sk);
1033        int (*getfrag)(void *, char *, int, int, int, struct sk_buff *);
1034        struct sockcm_cookie sockc;
1035
1036        ipc6.hlimit = -1;
1037        ipc6.tclass = -1;
1038        ipc6.dontfrag = -1;
1039
1040        /* destination address check */
1041        if (sin6) {
1042                if (addr_len < offsetof(struct sockaddr, sa_data))
1043                        return -EINVAL;
1044
1045                switch (sin6->sin6_family) {
1046                case AF_INET6:
1047                        if (addr_len < SIN6_LEN_RFC2133)
1048                                return -EINVAL;
1049                        daddr = &sin6->sin6_addr;
1050                        break;
1051                case AF_INET:
1052                        goto do_udp_sendmsg;
1053                case AF_UNSPEC:
1054                        msg->msg_name = sin6 = NULL;
1055                        msg->msg_namelen = addr_len = 0;
1056                        daddr = NULL;
1057                        break;
1058                default:
1059                        return -EINVAL;
1060                }
1061        } else if (!up->pending) {
1062                if (sk->sk_state != TCP_ESTABLISHED)
1063                        return -EDESTADDRREQ;
1064                daddr = &sk->sk_v6_daddr;
1065        } else
1066                daddr = NULL;
1067
1068        if (daddr) {
1069                if (ipv6_addr_v4mapped(daddr)) {
1070                        struct sockaddr_in sin;
1071                        sin.sin_family = AF_INET;
1072                        sin.sin_port = sin6 ? sin6->sin6_port : inet->inet_dport;
1073                        sin.sin_addr.s_addr = daddr->s6_addr32[3];
1074                        msg->msg_name = &sin;
1075                        msg->msg_namelen = sizeof(sin);
1076do_udp_sendmsg:
1077                        if (__ipv6_only_sock(sk))
1078                                return -ENETUNREACH;
1079                        return udp_sendmsg(sk, msg, len);
1080                }
1081        }
1082
1083        if (up->pending == AF_INET)
1084                return udp_sendmsg(sk, msg, len);
1085
1086        /* Rough check on arithmetic overflow,
1087           better check is made in ip6_append_data().
1088           */
1089        if (len > INT_MAX - sizeof(struct udphdr))
1090                return -EMSGSIZE;
1091
1092        getfrag  =  is_udplite ?  udplite_getfrag : ip_generic_getfrag;
1093        if (up->pending) {
1094                /*
1095                 * There are pending frames.
1096                 * The socket lock must be held while it's corked.
1097                 */
1098                lock_sock(sk);
1099                if (likely(up->pending)) {
1100                        if (unlikely(up->pending != AF_INET6)) {
1101                                release_sock(sk);
1102                                return -EAFNOSUPPORT;
1103                        }
1104                        dst = NULL;
1105                        goto do_append_data;
1106                }
1107                release_sock(sk);
1108        }
1109        ulen += sizeof(struct udphdr);
1110
1111        memset(&fl6, 0, sizeof(fl6));
1112
1113        if (sin6) {
1114                if (sin6->sin6_port == 0)
1115                        return -EINVAL;
1116
1117                fl6.fl6_dport = sin6->sin6_port;
1118                daddr = &sin6->sin6_addr;
1119
1120                if (np->sndflow) {
1121                        fl6.flowlabel = sin6->sin6_flowinfo&IPV6_FLOWINFO_MASK;
1122                        if (fl6.flowlabel&IPV6_FLOWLABEL_MASK) {
1123                                flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1124                                if (!flowlabel)
1125                                        return -EINVAL;
1126                        }
1127                }
1128
1129                /*
1130                 * Otherwise it will be difficult to maintain
1131                 * sk->sk_dst_cache.
1132                 */
1133                if (sk->sk_state == TCP_ESTABLISHED &&
1134                    ipv6_addr_equal(daddr, &sk->sk_v6_daddr))
1135                        daddr = &sk->sk_v6_daddr;
1136
1137                if (addr_len >= sizeof(struct sockaddr_in6) &&
1138                    sin6->sin6_scope_id &&
1139                    __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
1140                        fl6.flowi6_oif = sin6->sin6_scope_id;
1141        } else {
1142                if (sk->sk_state != TCP_ESTABLISHED)
1143                        return -EDESTADDRREQ;
1144
1145                fl6.fl6_dport = inet->inet_dport;
1146                daddr = &sk->sk_v6_daddr;
1147                fl6.flowlabel = np->flow_label;
1148                connected = 1;
1149        }
1150
1151        if (!fl6.flowi6_oif)
1152                fl6.flowi6_oif = sk->sk_bound_dev_if;
1153
1154        if (!fl6.flowi6_oif)
1155                fl6.flowi6_oif = np->sticky_pktinfo.ipi6_ifindex;
1156
1157        fl6.flowi6_mark = sk->sk_mark;
1158        sockc.tsflags = sk->sk_tsflags;
1159
1160        if (msg->msg_controllen) {
1161                opt = &opt_space;
1162                memset(opt, 0, sizeof(struct ipv6_txoptions));
1163                opt->tot_len = sizeof(*opt);
1164                ipc6.opt = opt;
1165
1166                err = ip6_datagram_send_ctl(sock_net(sk), sk, msg, &fl6, &ipc6, &sockc);
1167                if (err < 0) {
1168                        fl6_sock_release(flowlabel);
1169                        return err;
1170                }
1171                if ((fl6.flowlabel&IPV6_FLOWLABEL_MASK) && !flowlabel) {
1172                        flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
1173                        if (!flowlabel)
1174                                return -EINVAL;
1175                }
1176                if (!(opt->opt_nflen|opt->opt_flen))
1177                        opt = NULL;
1178                connected = 0;
1179        }
1180        if (!opt) {
1181                opt = txopt_get(np);
1182                opt_to_free = opt;
1183        }
1184        if (flowlabel)
1185                opt = fl6_merge_options(&opt_space, flowlabel, opt);
1186        opt = ipv6_fixup_options(&opt_space, opt);
1187        ipc6.opt = opt;
1188
1189        fl6.flowi6_proto = sk->sk_protocol;
1190        if (!ipv6_addr_any(daddr))
1191                fl6.daddr = *daddr;
1192        else
1193                fl6.daddr.s6_addr[15] = 0x1; /* :: means loopback (BSD'ism) */
1194        if (ipv6_addr_any(&fl6.saddr) && !ipv6_addr_any(&np->saddr))
1195                fl6.saddr = np->saddr;
1196        fl6.fl6_sport = inet->inet_sport;
1197
1198        final_p = fl6_update_dst(&fl6, opt, &final);
1199        if (final_p)
1200                connected = 0;
1201
1202        if (!fl6.flowi6_oif && ipv6_addr_is_multicast(&fl6.daddr)) {
1203                fl6.flowi6_oif = np->mcast_oif;
1204                connected = 0;
1205        } else if (!fl6.flowi6_oif)
1206                fl6.flowi6_oif = np->ucast_oif;
1207
1208        security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
1209
1210        if (ipc6.tclass < 0)
1211                ipc6.tclass = np->tclass;
1212
1213        fl6.flowlabel = ip6_make_flowinfo(ipc6.tclass, fl6.flowlabel);
1214
1215        dst = ip6_sk_dst_lookup_flow(sk, &fl6, final_p);
1216        if (IS_ERR(dst)) {
1217                err = PTR_ERR(dst);
1218                dst = NULL;
1219                goto out;
1220        }
1221
1222        if (ipc6.hlimit < 0)
1223                ipc6.hlimit = ip6_sk_dst_hoplimit(np, &fl6, dst);
1224
1225        if (msg->msg_flags&MSG_CONFIRM)
1226                goto do_confirm;
1227back_from_confirm:
1228
1229        /* Lockless fast path for the non-corking case */
1230        if (!corkreq) {
1231                struct sk_buff *skb;
1232
1233                skb = ip6_make_skb(sk, getfrag, msg, ulen,
1234                                   sizeof(struct udphdr), &ipc6,
1235                                   &fl6, (struct rt6_info *)dst,
1236                                   msg->msg_flags, &sockc);
1237                err = PTR_ERR(skb);
1238                if (!IS_ERR_OR_NULL(skb))
1239                        err = udp_v6_send_skb(skb, &fl6);
1240                goto release_dst;
1241        }
1242
1243        lock_sock(sk);
1244        if (unlikely(up->pending)) {
1245                /* The socket is already corked while preparing it. */
1246                /* ... which is an evident application bug. --ANK */
1247                release_sock(sk);
1248
1249                net_dbg_ratelimited("udp cork app bug 2\n");
1250                err = -EINVAL;
1251                goto out;
1252        }
1253
1254        up->pending = AF_INET6;
1255
1256do_append_data:
1257        if (ipc6.dontfrag < 0)
1258                ipc6.dontfrag = np->dontfrag;
1259        up->len += ulen;
1260        err = ip6_append_data(sk, getfrag, msg, ulen, sizeof(struct udphdr),
1261                              &ipc6, &fl6, (struct rt6_info *)dst,
1262                              corkreq ? msg->msg_flags|MSG_MORE : msg->msg_flags, &sockc);
1263        if (err)
1264                udp_v6_flush_pending_frames(sk);
1265        else if (!corkreq)
1266                err = udp_v6_push_pending_frames(sk);
1267        else if (unlikely(skb_queue_empty(&sk->sk_write_queue)))
1268                up->pending = 0;
1269
1270        if (err > 0)
1271                err = np->recverr ? net_xmit_errno(err) : 0;
1272        release_sock(sk);
1273
1274release_dst:
1275        if (dst) {
1276                if (connected) {
1277                        ip6_dst_store(sk, dst,
1278                                      ipv6_addr_equal(&fl6.daddr, &sk->sk_v6_daddr) ?
1279                                      &sk->sk_v6_daddr : NULL,
1280#ifdef CONFIG_IPV6_SUBTREES
1281                                      ipv6_addr_equal(&fl6.saddr, &np->saddr) ?
1282                                      &np->saddr :
1283#endif
1284                                      NULL);
1285                } else {
1286                        dst_release(dst);
1287                }
1288                dst = NULL;
1289        }
1290
1291out:
1292        dst_release(dst);
1293        fl6_sock_release(flowlabel);
1294        txopt_put(opt_to_free);
1295        if (!err)
1296                return len;
1297        /*
1298         * ENOBUFS = no kernel mem, SOCK_NOSPACE = no sndbuf space.  Reporting
1299         * ENOBUFS might not be good (it's not tunable per se), but otherwise
1300         * we don't have a good statistic (IpOutDiscards but it can be too many
1301         * things).  We could add another new stat but at least for now that
1302         * seems like overkill.
1303         */
1304        if (err == -ENOBUFS || test_bit(SOCK_NOSPACE, &sk->sk_socket->flags)) {
1305                UDP6_INC_STATS(sock_net(sk),
1306                               UDP_MIB_SNDBUFERRORS, is_udplite);
1307        }
1308        return err;
1309
1310do_confirm:
1311        dst_confirm(dst);
1312        if (!(msg->msg_flags&MSG_PROBE) || len)
1313                goto back_from_confirm;
1314        err = 0;
1315        goto out;
1316}
1317
1318void udpv6_destroy_sock(struct sock *sk)
1319{
1320        struct udp_sock *up = udp_sk(sk);
1321        lock_sock(sk);
1322        udp_v6_flush_pending_frames(sk);
1323        release_sock(sk);
1324
1325        if (static_key_false(&udpv6_encap_needed) && up->encap_type) {
1326                void (*encap_destroy)(struct sock *sk);
1327                encap_destroy = ACCESS_ONCE(up->encap_destroy);
1328                if (encap_destroy)
1329                        encap_destroy(sk);
1330        }
1331
1332        inet6_destroy_sock(sk);
1333}
1334
1335/*
1336 *      Socket option code for UDP
1337 */
1338int udpv6_setsockopt(struct sock *sk, int level, int optname,
1339                     char __user *optval, unsigned int optlen)
1340{
1341        if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1342                return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1343                                          udp_v6_push_pending_frames);
1344        return ipv6_setsockopt(sk, level, optname, optval, optlen);
1345}
1346
1347#ifdef CONFIG_COMPAT
1348int compat_udpv6_setsockopt(struct sock *sk, int level, int optname,
1349                            char __user *optval, unsigned int optlen)
1350{
1351        if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1352                return udp_lib_setsockopt(sk, level, optname, optval, optlen,
1353                                          udp_v6_push_pending_frames);
1354        return compat_ipv6_setsockopt(sk, level, optname, optval, optlen);
1355}
1356#endif
1357
1358int udpv6_getsockopt(struct sock *sk, int level, int optname,
1359                     char __user *optval, int __user *optlen)
1360{
1361        if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1362                return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1363        return ipv6_getsockopt(sk, level, optname, optval, optlen);
1364}
1365
1366#ifdef CONFIG_COMPAT
1367int compat_udpv6_getsockopt(struct sock *sk, int level, int optname,
1368                            char __user *optval, int __user *optlen)
1369{
1370        if (level == SOL_UDP  ||  level == SOL_UDPLITE)
1371                return udp_lib_getsockopt(sk, level, optname, optval, optlen);
1372        return compat_ipv6_getsockopt(sk, level, optname, optval, optlen);
1373}
1374#endif
1375
1376static const struct inet6_protocol udpv6_protocol = {
1377        .handler        =       udpv6_rcv,
1378        .err_handler    =       udpv6_err,
1379        .flags          =       INET6_PROTO_NOPOLICY|INET6_PROTO_FINAL,
1380};
1381
1382/* ------------------------------------------------------------------------ */
1383#ifdef CONFIG_PROC_FS
1384int udp6_seq_show(struct seq_file *seq, void *v)
1385{
1386        if (v == SEQ_START_TOKEN) {
1387                seq_puts(seq, IPV6_SEQ_DGRAM_HEADER);
1388        } else {
1389                int bucket = ((struct udp_iter_state *)seq->private)->bucket;
1390                struct inet_sock *inet = inet_sk(v);
1391                __u16 srcp = ntohs(inet->inet_sport);
1392                __u16 destp = ntohs(inet->inet_dport);
1393                ip6_dgram_sock_seq_show(seq, v, srcp, destp, bucket);
1394        }
1395        return 0;
1396}
1397
1398static const struct file_operations udp6_afinfo_seq_fops = {
1399        .owner    = THIS_MODULE,
1400        .open     = udp_seq_open,
1401        .read     = seq_read,
1402        .llseek   = seq_lseek,
1403        .release  = seq_release_net
1404};
1405
1406static struct udp_seq_afinfo udp6_seq_afinfo = {
1407        .name           = "udp6",
1408        .family         = AF_INET6,
1409        .udp_table      = &udp_table,
1410        .seq_fops       = &udp6_afinfo_seq_fops,
1411        .seq_ops        = {
1412                .show           = udp6_seq_show,
1413        },
1414};
1415
1416int __net_init udp6_proc_init(struct net *net)
1417{
1418        return udp_proc_register(net, &udp6_seq_afinfo);
1419}
1420
1421void udp6_proc_exit(struct net *net)
1422{
1423        udp_proc_unregister(net, &udp6_seq_afinfo);
1424}
1425#endif /* CONFIG_PROC_FS */
1426
1427void udp_v6_clear_sk(struct sock *sk, int size)
1428{
1429        struct inet_sock *inet = inet_sk(sk);
1430
1431        /* we do not want to clear pinet6 field, because of RCU lookups */
1432        sk_prot_clear_portaddr_nulls(sk, offsetof(struct inet_sock, pinet6));
1433
1434        size -= offsetof(struct inet_sock, pinet6) + sizeof(inet->pinet6);
1435        memset(&inet->pinet6 + 1, 0, size);
1436}
1437
1438/* ------------------------------------------------------------------------ */
1439
1440struct proto udpv6_prot = {
1441        .name              = "UDPv6",
1442        .owner             = THIS_MODULE,
1443        .close             = udp_lib_close,
1444        .connect           = ip6_datagram_connect,
1445        .disconnect        = udp_disconnect,
1446        .ioctl             = udp_ioctl,
1447        .destroy           = udpv6_destroy_sock,
1448        .setsockopt        = udpv6_setsockopt,
1449        .getsockopt        = udpv6_getsockopt,
1450        .sendmsg           = udpv6_sendmsg,
1451        .recvmsg           = udpv6_recvmsg,
1452        .backlog_rcv       = __udpv6_queue_rcv_skb,
1453        .release_cb        = ip6_datagram_release_cb,
1454        .hash              = udp_lib_hash,
1455        .unhash            = udp_lib_unhash,
1456        .rehash            = udp_v6_rehash,
1457        .get_port          = udp_v6_get_port,
1458        .memory_allocated  = &udp_memory_allocated,
1459        .sysctl_mem        = sysctl_udp_mem,
1460        .sysctl_wmem       = &sysctl_udp_wmem_min,
1461        .sysctl_rmem       = &sysctl_udp_rmem_min,
1462        .obj_size          = sizeof(struct udp6_sock),
1463        .h.udp_table       = &udp_table,
1464#ifdef CONFIG_COMPAT
1465        .compat_setsockopt = compat_udpv6_setsockopt,
1466        .compat_getsockopt = compat_udpv6_getsockopt,
1467#endif
1468        .clear_sk          = udp_v6_clear_sk,
1469};
1470
1471static struct inet_protosw udpv6_protosw = {
1472        .type =      SOCK_DGRAM,
1473        .protocol =  IPPROTO_UDP,
1474        .prot =      &udpv6_prot,
1475        .ops =       &inet6_dgram_ops,
1476        .flags =     INET_PROTOSW_PERMANENT,
1477};
1478
1479int __init udpv6_init(void)
1480{
1481        int ret;
1482
1483        ret = inet6_add_protocol(&udpv6_protocol, IPPROTO_UDP);
1484        if (ret)
1485                goto out;
1486
1487        ret = inet6_register_protosw(&udpv6_protosw);
1488        if (ret)
1489                goto out_udpv6_protocol;
1490out:
1491        return ret;
1492
1493out_udpv6_protocol:
1494        inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1495        goto out;
1496}
1497
1498void udpv6_exit(void)
1499{
1500        inet6_unregister_protosw(&udpv6_protosw);
1501        inet6_del_protocol(&udpv6_protocol, IPPROTO_UDP);
1502}
1503