net/ipv4/inet_hashtables.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              Generic INET transport hashtables
   7  *
   8  * Authors:     Lotsa people, from code originally in tcp
   9  *
  10  *      This program is free software; you can redistribute it and/or
  11  *      modify it under the terms of the GNU General Public License
  12  *      as published by the Free Software Foundation; either version
  13  *      2 of the License, or (at your option) any later version.
  14  */
  15
  16 #include <linux/module.h>
  17 #include <linux/random.h>
  18 #include <linux/sched.h>
  19 #include <linux/slab.h>
  20 #include <linux/wait.h>
  21
  22 #include <net/inet_connection_sock.h>
  23 #include <net/inet_hashtables.h>
  24 #include <net/ip.h>
  25
  26 /*
  27  * Allocate and initialize a new local port bind bucket.
  28  * The bindhash mutex for snum's hash chain must be held here.
  29  */
  30 struct inet_bind_bucket *inet_bind_bucket_create(struct kmem_cache *cachep,
  31                                                  struct net *net,
  32                                                  struct inet_bind_hashbucket *head,
  33                                                  const unsigned short snum)
  34 {
  35         struct inet_bind_bucket *tb = kmem_cache_alloc(cachep, GFP_ATOMIC);
  36
  37         if (tb != NULL) {
  38                 tb->ib_net       = hold_net(net);
  39                 tb->port      = snum;
  40                 tb->fastreuse = 0;
  41                 INIT_HLIST_HEAD(&tb->owners);
  42                 hlist_add_head(&tb->node, &head->chain);
  43         }
  44         return tb;
  45 }
  46
  47 /*
  48  * Caller must hold hashbucket lock for this tb with local BH disabled
  49  */
  50 void inet_bind_bucket_destroy(struct kmem_cache *cachep, struct inet_bind_bucket *tb)
  51 {
  52         if (hlist_empty(&tb->owners)) {
  53                 __hlist_del(&tb->node);
  54                 release_net(tb->ib_net);
  55                 kmem_cache_free(cachep, tb);
  56         }
  57 }
  58
  59 void inet_bind_hash(struct sock *sk, struct inet_bind_bucket *tb,
  60                     const unsigned short snum)
  61 {
  62         inet_sk(sk)->num = snum;
  63         sk_add_bind_node(sk, &tb->owners);
  64         inet_csk(sk)->icsk_bind_hash = tb;
  65 }
  66
  67 /*
  68  * Get rid of any references to a local port held by the given sock.
  69  */
  70 static void __inet_put_port(struct sock *sk)
  71 {
  72         struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
  73         const int bhash = inet_bhashfn(sock_net(sk), inet_sk(sk)->num,
  74                         hashinfo->bhash_size);
  75         struct inet_bind_hashbucket *head = &hashinfo->bhash[bhash];
  76         struct inet_bind_bucket *tb;
  77
  78         spin_lock(&head->lock);
  79         tb = inet_csk(sk)->icsk_bind_hash;
  80         __sk_del_bind_node(sk);
  81         inet_csk(sk)->icsk_bind_hash = NULL;
  82         inet_sk(sk)->num = 0;
  83         inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
  84         spin_unlock(&head->lock);
  85 }
  86
  87 void inet_put_port(struct sock *sk)
  88 {
  89         local_bh_disable();
  90         __inet_put_port(sk);
  91         local_bh_enable();
  92 }
  93
  94 EXPORT_SYMBOL(inet_put_port);
  95
  96 void __inet_inherit_port(struct sock *sk, struct sock *child)
  97 {
  98         struct inet_hashinfo *table = sk->sk_prot->h.hashinfo;
  99         const int bhash = inet_bhashfn(sock_net(sk), inet_sk(child)->num,
 100                         table->bhash_size);
 101         struct inet_bind_hashbucket *head = &table->bhash[bhash];
 102         struct inet_bind_bucket *tb;
 103
 104         spin_lock(&head->lock);
 105         tb = inet_csk(sk)->icsk_bind_hash;
 106         sk_add_bind_node(child, &tb->owners);
 107         inet_csk(child)->icsk_bind_hash = tb;
 108         spin_unlock(&head->lock);
 109 }
 110
 111 EXPORT_SYMBOL_GPL(__inet_inherit_port);
 112
 113 /*
 114  * This lock without WQ_FLAG_EXCLUSIVE is good on UP and it can be very bad on SMP.
 115  * Look, when several writers sleep and reader wakes them up, all but one
 116  * immediately hit write lock and grab all the cpus. Exclusive sleep solves
 117  * this, _but_ remember, it adds useless work on UP machines (wake up each
 118  * exclusive lock release). It should be ifdefed really.
 119  */
 120 void inet_listen_wlock(struct inet_hashinfo *hashinfo)
 121         __acquires(hashinfo->lhash_lock)
 122 {
 123         write_lock(&hashinfo->lhash_lock);
 124
 125         if (atomic_read(&hashinfo->lhash_users)) {
 126                 DEFINE_WAIT(wait);
 127
 128                 for (;;) {
 129                         prepare_to_wait_exclusive(&hashinfo->lhash_wait,
 130                                                   &wait, TASK_UNINTERRUPTIBLE);
 131                         if (!atomic_read(&hashinfo->lhash_users))
 132                                 break;
 133                         write_unlock_bh(&hashinfo->lhash_lock);
 134                         schedule();
 135                         write_lock_bh(&hashinfo->lhash_lock);
 136                 }
 137
 138                 finish_wait(&hashinfo->lhash_wait, &wait);
 139         }
 140 }
 141
 142 /*
 143  * Don't inline this cruft. Here are some nice properties to exploit here. The
 144  * BSD API does not allow a listening sock to specify the remote port nor the
 145  * remote address for the connection. So always assume those are both
 146  * wildcarded during the search since they can never be otherwise.
 147  */
 148 static struct sock *inet_lookup_listener_slow(struct net *net,
 149                                               const struct hlist_head *head,
 150                                               const __be32 daddr,
 151                                               const unsigned short hnum,
 152                                               const int dif)
 153 {
 154         struct sock *result = NULL, *sk;
 155         const struct hlist_node *node;
 156         int hiscore = -1;
 157
 158         sk_for_each(sk, node, head) {
 159                 const struct inet_sock *inet = inet_sk(sk);
 160
 161                 if (net_eq(sock_net(sk), net) && inet->num == hnum &&
 162                                 !ipv6_only_sock(sk)) {
 163                         const __be32 rcv_saddr = inet->rcv_saddr;
 164                         int score = sk->sk_family == PF_INET ? 1 : 0;
 165
 166                         if (rcv_saddr) {
 167                                 if (rcv_saddr != daddr)
 168                                         continue;
 169                                 score += 2;
 170                         }
 171                         if (sk->sk_bound_dev_if) {
 172                                 if (sk->sk_bound_dev_if != dif)
 173                                         continue;
 174                                 score += 2;
 175                         }
 176                         if (score == 5)
 177                                 return sk;
 178                         if (score > hiscore) {
 179                                 hiscore = score;
 180                                 result  = sk;
 181                         }
 182                 }
 183         }
 184         return result;
 185 }
 186
 187 /* Optimize the common listener case. */
 188 struct sock *__inet_lookup_listener(struct net *net,
 189                                     struct inet_hashinfo *hashinfo,
 190                                     const __be32 daddr, const unsigned short hnum,
 191                                     const int dif)
 192 {
 193         struct sock *sk = NULL;
 194         const struct hlist_head *head;
 195
 196         read_lock(&hashinfo->lhash_lock);
 197         head = &hashinfo->listening_hash[inet_lhashfn(net, hnum)];
 198         if (!hlist_empty(head)) {
 199                 const struct inet_sock *inet = inet_sk((sk = __sk_head(head)));
 200
 201                 if (inet->num == hnum && !sk->sk_node.next &&
 202                     (!inet->rcv_saddr || inet->rcv_saddr == daddr) &&
 203                     (sk->sk_family == PF_INET || !ipv6_only_sock(sk)) &&
 204                     !sk->sk_bound_dev_if && net_eq(sock_net(sk), net))
 205                         goto sherry_cache;
 206                 sk = inet_lookup_listener_slow(net, head, daddr, hnum, dif);
 207         }
 208         if (sk) {
 209 sherry_cache:
 210                 sock_hold(sk);
 211         }
 212         read_unlock(&hashinfo->lhash_lock);
 213         return sk;
 214 }
 215 EXPORT_SYMBOL_GPL(__inet_lookup_listener);
 216
 217 struct sock * __inet_lookup_established(struct net *net,
 218                                   struct inet_hashinfo *hashinfo,
 219                                   const __be32 saddr, const __be16 sport,
 220                                   const __be32 daddr, const u16 hnum,
 221                                   const int dif)
 222 {
 223         INET_ADDR_COOKIE(acookie, saddr, daddr)
 224         const __portpair ports = INET_COMBINED_PORTS(sport, hnum);
 225         struct sock *sk;
 226         const struct hlist_node *node;
 227         /* Optimize here for direct hit, only listening connections can
 228          * have wildcards anyways.
 229          */
 230         unsigned int hash = inet_ehashfn(net, daddr, hnum, saddr, sport);
 231         struct inet_ehash_bucket *head = inet_ehash_bucket(hashinfo, hash);
 232         rwlock_t *lock = inet_ehash_lockp(hashinfo, hash);
 233
 234         prefetch(head->chain.first);
 235         read_lock(lock);
 236         sk_for_each(sk, node, &head->chain) {
 237                 if (INET_MATCH(sk, net, hash, acookie,
 238                                         saddr, daddr, ports, dif))
 239                         goto hit; /* You sunk my battleship! */
 240         }
 241
 242         /* Must check for a TIME_WAIT'er before going to listener hash. */
 243         sk_for_each(sk, node, &head->twchain) {
 244                 if (INET_TW_MATCH(sk, net, hash, acookie,
 245                                         saddr, daddr, ports, dif))
 246                         goto hit;
 247         }
 248         sk = NULL;
 249 out:
 250         read_unlock(lock);
 251         return sk;
 252 hit:
 253         sock_hold(sk);
 254         goto out;
 255 }
 256 EXPORT_SYMBOL_GPL(__inet_lookup_established);
 257
 258 /* called with local bh disabled */
 259 static int __inet_check_established(struct inet_timewait_death_row *death_row,
 260                                     struct sock *sk, __u16 lport,
 261                                     struct inet_timewait_sock **twp)
 262 {
 263         struct inet_hashinfo *hinfo = death_row->hashinfo;
 264         struct inet_sock *inet = inet_sk(sk);
 265         __be32 daddr = inet->rcv_saddr;
 266         __be32 saddr = inet->daddr;
 267         int dif = sk->sk_bound_dev_if;
 268         INET_ADDR_COOKIE(acookie, saddr, daddr)
 269         const __portpair ports = INET_COMBINED_PORTS(inet->dport, lport);
 270         struct net *net = sock_net(sk);
 271         unsigned int hash = inet_ehashfn(net, daddr, lport, saddr, inet->dport);
 272         struct inet_ehash_bucket *head = inet_ehash_bucket(hinfo, hash);
 273         rwlock_t *lock = inet_ehash_lockp(hinfo, hash);
 274         struct sock *sk2;
 275         const struct hlist_node *node;
 276         struct inet_timewait_sock *tw;
 277
 278         prefetch(head->chain.first);
 279         write_lock(lock);
 280
 281         /* Check TIME-WAIT sockets first. */
 282         sk_for_each(sk2, node, &head->twchain) {
 283                 tw = inet_twsk(sk2);
 284
 285                 if (INET_TW_MATCH(sk2, net, hash, acookie,
 286                                         saddr, daddr, ports, dif)) {
 287                         if (twsk_unique(sk, sk2, twp))
 288                                 goto unique;
 289                         else
 290                                 goto not_unique;
 291                 }
 292         }
 293         tw = NULL;
 294
 295         /* And established part... */
 296         sk_for_each(sk2, node, &head->chain) {
 297                 if (INET_MATCH(sk2, net, hash, acookie,
 298                                         saddr, daddr, ports, dif))
 299                         goto not_unique;
 300         }
 301
 302 unique:
 303         /* Must record num and sport now. Otherwise we will see
 304          * in hash table socket with a funny identity. */
 305         inet->num = lport;
 306         inet->sport = htons(lport);
 307         sk->sk_hash = hash;
 308         WARN_ON(!sk_unhashed(sk));
 309         __sk_add_node(sk, &head->chain);
 310         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
 311         write_unlock(lock);
 312
 313         if (twp) {
 314                 *twp = tw;
 315                 NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
 316         } else if (tw) {
 317                 /* Silly. Should hash-dance instead... */
 318                 inet_twsk_deschedule(tw, death_row);
 319                 NET_INC_STATS_BH(net, LINUX_MIB_TIMEWAITRECYCLED);
 320
 321                 inet_twsk_put(tw);
 322         }
 323
 324         return 0;
 325
 326 not_unique:
 327         write_unlock(lock);
 328         return -EADDRNOTAVAIL;
 329 }
 330
 331 static inline u32 inet_sk_port_offset(const struct sock *sk)
 332 {
 333         const struct inet_sock *inet = inet_sk(sk);
 334         return secure_ipv4_port_ephemeral(inet->rcv_saddr, inet->daddr,
 335                                           inet->dport);
 336 }
 337
 338 void __inet_hash_nolisten(struct sock *sk)
 339 {
 340         struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
 341         struct hlist_head *list;
 342         rwlock_t *lock;
 343         struct inet_ehash_bucket *head;
 344
 345         WARN_ON(!sk_unhashed(sk));
 346
 347         sk->sk_hash = inet_sk_ehashfn(sk);
 348         head = inet_ehash_bucket(hashinfo, sk->sk_hash);
 349         list = &head->chain;
 350         lock = inet_ehash_lockp(hashinfo, sk->sk_hash);
 351
 352         write_lock(lock);
 353         __sk_add_node(sk, list);
 354         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
 355         write_unlock(lock);
 356 }
 357 EXPORT_SYMBOL_GPL(__inet_hash_nolisten);
 358
 359 static void __inet_hash(struct sock *sk)
 360 {
 361         struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
 362         struct hlist_head *list;
 363         rwlock_t *lock;
 364
 365         if (sk->sk_state != TCP_LISTEN) {
 366                 __inet_hash_nolisten(sk);
 367                 return;
 368         }
 369
 370         WARN_ON(!sk_unhashed(sk));
 371         list = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
 372         lock = &hashinfo->lhash_lock;
 373
 374         inet_listen_wlock(hashinfo);
 375         __sk_add_node(sk, list);
 376         sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
 377         write_unlock(lock);
 378         wake_up(&hashinfo->lhash_wait);
 379 }
 380
 381 void inet_hash(struct sock *sk)
 382 {
 383         if (sk->sk_state != TCP_CLOSE) {
 384                 local_bh_disable();
 385                 __inet_hash(sk);
 386                 local_bh_enable();
 387         }
 388 }
 389 EXPORT_SYMBOL_GPL(inet_hash);
 390
 391 void inet_unhash(struct sock *sk)
 392 {
 393         rwlock_t *lock;
 394         struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
 395
 396         if (sk_unhashed(sk))
 397                 goto out;
 398
 399         if (sk->sk_state == TCP_LISTEN) {
 400                 local_bh_disable();
 401                 inet_listen_wlock(hashinfo);
 402                 lock = &hashinfo->lhash_lock;
 403         } else {
 404                 lock = inet_ehash_lockp(hashinfo, sk->sk_hash);
 405                 write_lock_bh(lock);
 406         }
 407
 408         if (__sk_del_node_init(sk))
 409                 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 410         write_unlock_bh(lock);
 411 out:
 412         if (sk->sk_state == TCP_LISTEN)
 413                 wake_up(&hashinfo->lhash_wait);
 414 }
 415 EXPORT_SYMBOL_GPL(inet_unhash);
 416
 417 int __inet_hash_connect(struct inet_timewait_death_row *death_row,
 418                 struct sock *sk, u32 port_offset,
 419                 int (*check_established)(struct inet_timewait_death_row *,
 420                         struct sock *, __u16, struct inet_timewait_sock **),
 421                 void (*hash)(struct sock *sk))
 422 {
 423         struct inet_hashinfo *hinfo = death_row->hashinfo;
 424         const unsigned short snum = inet_sk(sk)->num;
 425         struct inet_bind_hashbucket *head;
 426         struct inet_bind_bucket *tb;
 427         int ret;
 428         struct net *net = sock_net(sk);
 429
 430         if (!snum) {
 431                 int i, remaining, low, high, port;
 432                 static u32 hint;
 433                 u32 offset = hint + port_offset;
 434                 struct hlist_node *node;
 435                 struct inet_timewait_sock *tw = NULL;
 436
 437                 inet_get_local_port_range(&low, &high);
 438                 remaining = (high - low) + 1;
 439
 440                 local_bh_disable();
 441                 for (i = 1; i <= remaining; i++) {
 442                         port = low + (i + offset) % remaining;
 443                         head = &hinfo->bhash[inet_bhashfn(net, port,
 444                                         hinfo->bhash_size)];
 445                         spin_lock(&head->lock);
 446
 447                         /* Does not bother with rcv_saddr checks,
 448                          * because the established check is already
 449                          * unique enough.
 450                          */
 451                         inet_bind_bucket_for_each(tb, node, &head->chain) {
 452                                 if (tb->ib_net == net && tb->port == port) {
 453                                         WARN_ON(hlist_empty(&tb->owners));
 454                                         if (tb->fastreuse >= 0)
 455                                                 goto next_port;
 456                                         if (!check_established(death_row, sk,
 457                                                                 port, &tw))
 458                                                 goto ok;
 459                                         goto next_port;
 460                                 }
 461                         }
 462
 463                         tb = inet_bind_bucket_create(hinfo->bind_bucket_cachep,
 464                                         net, head, port);
 465                         if (!tb) {
 466                                 spin_unlock(&head->lock);
 467                                 break;
 468                         }
 469                         tb->fastreuse = -1;
 470                         goto ok;
 471
 472                 next_port:
 473                         spin_unlock(&head->lock);
 474                 }
 475                 local_bh_enable();
 476
 477                 return -EADDRNOTAVAIL;
 478
 479 ok:
 480                 hint += i;
 481
 482                 /* Head lock still held and bh's disabled */
 483                 inet_bind_hash(sk, tb, port);
 484                 if (sk_unhashed(sk)) {
 485                         inet_sk(sk)->sport = htons(port);
 486                         hash(sk);
 487                 }
 488                 spin_unlock(&head->lock);
 489
 490                 if (tw) {
 491                         inet_twsk_deschedule(tw, death_row);
 492                         inet_twsk_put(tw);
 493                 }
 494
 495                 ret = 0;
 496                 goto out;
 497         }
 498
 499         head = &hinfo->bhash[inet_bhashfn(net, snum, hinfo->bhash_size)];
 500         tb  = inet_csk(sk)->icsk_bind_hash;
 501         spin_lock_bh(&head->lock);
 502         if (sk_head(&tb->owners) == sk && !sk->sk_bind_node.next) {
 503                 hash(sk);
 504                 spin_unlock_bh(&head->lock);
 505                 return 0;
 506         } else {
 507                 spin_unlock(&head->lock);
 508                 /* No definite answer... Walk to established hash table */
 509                 ret = check_established(death_row, sk, snum, NULL);
 510 out:
 511                 local_bh_enable();
 512                 return ret;
 513         }
 514 }
 515
 516 /*
 517  * Bind a port for a connect operation and hash it.
 518  */
 519 int inet_hash_connect(struct inet_timewait_death_row *death_row,
 520                       struct sock *sk)
 521 {
 522         return __inet_hash_connect(death_row, sk, inet_sk_port_offset(sk),
 523                         __inet_check_established, __inet_hash_nolisten);
 524 }
 525
 526 EXPORT_SYMBOL_GPL(inet_hash_connect);