kernel/drivers/staging/lustre/lustre/libcfs/hash.c

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 only,
   8  * as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License version 2 for more details (a copy is included
  14  * in the LICENSE file that accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * version 2 along with this program; If not, see
  18  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  19  *
  20  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  21  * CA 95054 USA or visit www.sun.com if you need additional information or
  22  * have any questions.
  23  *
  24  * GPL HEADER END
  25  */
  26 /*
  27  * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2011, 2012, Intel Corporation.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  *
  36  * libcfs/libcfs/hash.c
  37  *
  38  * Implement a hash class for hash process in lustre system.
  39  *
  40  * Author: YuZhangyong <yzy@clusterfs.com>
  41  *
  42  * 2008-08-15: Brian Behlendorf <behlendorf1@llnl.gov>
  43  * - Simplified API and improved documentation
  44  * - Added per-hash feature flags:
  45  *   * CFS_HASH_DEBUG additional validation
  46  *   * CFS_HASH_REHASH dynamic rehashing
  47  * - Added per-hash statistics
  48  * - General performance enhancements
  49  *
  50  * 2009-07-31: Liang Zhen <zhen.liang@sun.com>
  51  * - move all stuff to libcfs
  52  * - don't allow cur_bits != max_bits without setting of CFS_HASH_REHASH
  53  * - ignore hs_rwlock if without CFS_HASH_REHASH setting
  54  * - buckets are allocated one by one(instead of contiguous memory),
  55  *   to avoid unnecessary cacheline conflict
  56  *
  57  * 2010-03-01: Liang Zhen <zhen.liang@sun.com>
  58  * - "bucket" is a group of hlist_head now, user can specify bucket size
  59  *   by bkt_bits of cfs_hash_create(), all hlist_heads in a bucket share
  60  *   one lock for reducing memory overhead.
  61  *
  62  * - support lockless hash, caller will take care of locks:
  63  *   avoid lock overhead for hash tables that are already protected
  64  *   by locking in the caller for another reason
  65  *
  66  * - support both spin_lock/rwlock for bucket:
  67  *   overhead of spinlock contention is lower than read/write
  68  *   contention of rwlock, so using spinlock to serialize operations on
  69  *   bucket is more reasonable for those frequently changed hash tables
  70  *
  71  * - support one-single lock mode:
  72  *   one lock to protect all hash operations to avoid overhead of
  73  *   multiple locks if hash table is always small
  74  *
  75  * - removed a lot of unnecessary addref & decref on hash element:
  76  *   addref & decref are atomic operations in many use-cases which
  77  *   are expensive.
  78  *
  79  * - support non-blocking cfs_hash_add() and cfs_hash_findadd():
  80  *   some lustre use-cases require these functions to be strictly
  81  *   non-blocking, we need to schedule required rehash on a different
  82  *   thread on those cases.
  83  *
  84  * - safer rehash on large hash table
  85  *   In old implementation, rehash function will exclusively lock the
  86  *   hash table and finish rehash in one batch, it's dangerous on SMP
  87  *   system because rehash millions of elements could take long time.
  88  *   New implemented rehash can release lock and relax CPU in middle
  89  *   of rehash, it's safe for another thread to search/change on the
  90  *   hash table even it's in rehasing.
  91  *
  92  * - support two different refcount modes
  93  *   . hash table has refcount on element
  94  *   . hash table doesn't change refcount on adding/removing element
  95  *
  96  * - support long name hash table (for param-tree)
  97  *
  98  * - fix a bug for cfs_hash_rehash_key:
  99  *   in old implementation, cfs_hash_rehash_key could screw up the
 100  *   hash-table because @key is overwritten without any protection.
 101  *   Now we need user to define hs_keycpy for those rehash enabled
 102  *   hash tables, cfs_hash_rehash_key will overwrite hash-key
 103  *   inside lock by calling hs_keycpy.
 104  *
 105  * - better hash iteration:
 106  *   Now we support both locked iteration & lockless iteration of hash
 107  *   table. Also, user can break the iteration by return 1 in callback.
 108  */
 109
 110 #include "../../include/linux/libcfs/libcfs.h"
 111 #include <linux/seq_file.h>
 112
 113 #if CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1
 114 static unsigned int warn_on_depth = 8;
 115 module_param(warn_on_depth, uint, 0644);
 116 MODULE_PARM_DESC(warn_on_depth, "warning when hash depth is high.");
 117 #endif
 118
 119 struct cfs_wi_sched *cfs_sched_rehash;
 120
 121 static inline void
 122 cfs_hash_nl_lock(union cfs_hash_lock *lock, int exclusive) {}
 123
 124 static inline void
 125 cfs_hash_nl_unlock(union cfs_hash_lock *lock, int exclusive) {}
 126
 127 static inline void
 128 cfs_hash_spin_lock(union cfs_hash_lock *lock, int exclusive)
 129         __acquires(&lock->spin)
 130 {
 131         spin_lock(&lock->spin);
 132 }
 133
 134 static inline void
 135 cfs_hash_spin_unlock(union cfs_hash_lock *lock, int exclusive)
 136         __releases(&lock->spin)
 137 {
 138         spin_unlock(&lock->spin);
 139 }
 140
 141 static inline void
 142 cfs_hash_rw_lock(union cfs_hash_lock *lock, int exclusive)
 143         __acquires(&lock->rw)
 144 {
 145         if (!exclusive)
 146                 read_lock(&lock->rw);
 147         else
 148                 write_lock(&lock->rw);
 149 }
 150
 151 static inline void
 152 cfs_hash_rw_unlock(union cfs_hash_lock *lock, int exclusive)
 153         __releases(&lock->rw)
 154 {
 155         if (!exclusive)
 156                 read_unlock(&lock->rw);
 157         else
 158                 write_unlock(&lock->rw);
 159 }
 160
 161 /** No lock hash */
 162 static cfs_hash_lock_ops_t cfs_hash_nl_lops = {
 163         .hs_lock        = cfs_hash_nl_lock,
 164         .hs_unlock      = cfs_hash_nl_unlock,
 165         .hs_bkt_lock    = cfs_hash_nl_lock,
 166         .hs_bkt_unlock  = cfs_hash_nl_unlock,
 167 };
 168
 169 /** no bucket lock, one spinlock to protect everything */
 170 static cfs_hash_lock_ops_t cfs_hash_nbl_lops = {
 171         .hs_lock        = cfs_hash_spin_lock,
 172         .hs_unlock      = cfs_hash_spin_unlock,
 173         .hs_bkt_lock    = cfs_hash_nl_lock,
 174         .hs_bkt_unlock  = cfs_hash_nl_unlock,
 175 };
 176
 177 /** spin bucket lock, rehash is enabled */
 178 static cfs_hash_lock_ops_t cfs_hash_bkt_spin_lops = {
 179         .hs_lock        = cfs_hash_rw_lock,
 180         .hs_unlock      = cfs_hash_rw_unlock,
 181         .hs_bkt_lock    = cfs_hash_spin_lock,
 182         .hs_bkt_unlock  = cfs_hash_spin_unlock,
 183 };
 184
 185 /** rw bucket lock, rehash is enabled */
 186 static cfs_hash_lock_ops_t cfs_hash_bkt_rw_lops = {
 187         .hs_lock        = cfs_hash_rw_lock,
 188         .hs_unlock      = cfs_hash_rw_unlock,
 189         .hs_bkt_lock    = cfs_hash_rw_lock,
 190         .hs_bkt_unlock  = cfs_hash_rw_unlock,
 191 };
 192
 193 /** spin bucket lock, rehash is disabled */
 194 static cfs_hash_lock_ops_t cfs_hash_nr_bkt_spin_lops = {
 195         .hs_lock        = cfs_hash_nl_lock,
 196         .hs_unlock      = cfs_hash_nl_unlock,
 197         .hs_bkt_lock    = cfs_hash_spin_lock,
 198         .hs_bkt_unlock  = cfs_hash_spin_unlock,
 199 };
 200
 201 /** rw bucket lock, rehash is disabled */
 202 static cfs_hash_lock_ops_t cfs_hash_nr_bkt_rw_lops = {
 203         .hs_lock        = cfs_hash_nl_lock,
 204         .hs_unlock      = cfs_hash_nl_unlock,
 205         .hs_bkt_lock    = cfs_hash_rw_lock,
 206         .hs_bkt_unlock  = cfs_hash_rw_unlock,
 207 };
 208
 209 static void
 210 cfs_hash_lock_setup(struct cfs_hash *hs)
 211 {
 212         if (cfs_hash_with_no_lock(hs)) {
 213                 hs->hs_lops = &cfs_hash_nl_lops;
 214
 215         } else if (cfs_hash_with_no_bktlock(hs)) {
 216                 hs->hs_lops = &cfs_hash_nbl_lops;
 217                 spin_lock_init(&hs->hs_lock.spin);
 218
 219         } else if (cfs_hash_with_rehash(hs)) {
 220                 rwlock_init(&hs->hs_lock.rw);
 221
 222                 if (cfs_hash_with_rw_bktlock(hs))
 223                         hs->hs_lops = &cfs_hash_bkt_rw_lops;
 224                 else if (cfs_hash_with_spin_bktlock(hs))
 225                         hs->hs_lops = &cfs_hash_bkt_spin_lops;
 226                 else
 227                         LBUG();
 228         } else {
 229                 if (cfs_hash_with_rw_bktlock(hs))
 230                         hs->hs_lops = &cfs_hash_nr_bkt_rw_lops;
 231                 else if (cfs_hash_with_spin_bktlock(hs))
 232                         hs->hs_lops = &cfs_hash_nr_bkt_spin_lops;
 233                 else
 234                         LBUG();
 235         }
 236 }
 237
 238 /**
 239  * Simple hash head without depth tracking
 240  * new element is always added to head of hlist
 241  */
 242 typedef struct {
 243         struct hlist_head       hh_head;        /**< entries list */
 244 } cfs_hash_head_t;
 245
 246 static int
 247 cfs_hash_hh_hhead_size(struct cfs_hash *hs)
 248 {
 249         return sizeof(cfs_hash_head_t);
 250 }
 251
 252 static struct hlist_head *
 253 cfs_hash_hh_hhead(struct cfs_hash *hs, struct cfs_hash_bd *bd)
 254 {
 255         cfs_hash_head_t *head = (cfs_hash_head_t *)&bd->bd_bucket->hsb_head[0];
 256
 257         return &head[bd->bd_offset].hh_head;
 258 }
 259
 260 static int
 261 cfs_hash_hh_hnode_add(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 262                       struct hlist_node *hnode)
 263 {
 264         hlist_add_head(hnode, cfs_hash_hh_hhead(hs, bd));
 265         return -1; /* unknown depth */
 266 }
 267
 268 static int
 269 cfs_hash_hh_hnode_del(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 270                       struct hlist_node *hnode)
 271 {
 272         hlist_del_init(hnode);
 273         return -1; /* unknown depth */
 274 }
 275
 276 /**
 277  * Simple hash head with depth tracking
 278  * new element is always added to head of hlist
 279  */
 280 typedef struct {
 281         struct hlist_head       hd_head;        /**< entries list */
 282         unsigned int        hd_depth;       /**< list length */
 283 } cfs_hash_head_dep_t;
 284
 285 static int
 286 cfs_hash_hd_hhead_size(struct cfs_hash *hs)
 287 {
 288         return sizeof(cfs_hash_head_dep_t);
 289 }
 290
 291 static struct hlist_head *
 292 cfs_hash_hd_hhead(struct cfs_hash *hs, struct cfs_hash_bd *bd)
 293 {
 294         cfs_hash_head_dep_t   *head;
 295
 296         head = (cfs_hash_head_dep_t *)&bd->bd_bucket->hsb_head[0];
 297         return &head[bd->bd_offset].hd_head;
 298 }
 299
 300 static int
 301 cfs_hash_hd_hnode_add(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 302                       struct hlist_node *hnode)
 303 {
 304         cfs_hash_head_dep_t *hh = container_of(cfs_hash_hd_hhead(hs, bd),
 305                                                cfs_hash_head_dep_t, hd_head);
 306         hlist_add_head(hnode, &hh->hd_head);
 307         return ++hh->hd_depth;
 308 }
 309
 310 static int
 311 cfs_hash_hd_hnode_del(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 312                       struct hlist_node *hnode)
 313 {
 314         cfs_hash_head_dep_t *hh = container_of(cfs_hash_hd_hhead(hs, bd),
 315                                                cfs_hash_head_dep_t, hd_head);
 316         hlist_del_init(hnode);
 317         return --hh->hd_depth;
 318 }
 319
 320 /**
 321  * double links hash head without depth tracking
 322  * new element is always added to tail of hlist
 323  */
 324 typedef struct {
 325         struct hlist_head       dh_head;        /**< entries list */
 326         struct hlist_node       *dh_tail;       /**< the last entry */
 327 } cfs_hash_dhead_t;
 328
 329 static int
 330 cfs_hash_dh_hhead_size(struct cfs_hash *hs)
 331 {
 332         return sizeof(cfs_hash_dhead_t);
 333 }
 334
 335 static struct hlist_head *
 336 cfs_hash_dh_hhead(struct cfs_hash *hs, struct cfs_hash_bd *bd)
 337 {
 338         cfs_hash_dhead_t *head;
 339
 340         head = (cfs_hash_dhead_t *)&bd->bd_bucket->hsb_head[0];
 341         return &head[bd->bd_offset].dh_head;
 342 }
 343
 344 static int
 345 cfs_hash_dh_hnode_add(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 346                       struct hlist_node *hnode)
 347 {
 348         cfs_hash_dhead_t *dh = container_of(cfs_hash_dh_hhead(hs, bd),
 349                                             cfs_hash_dhead_t, dh_head);
 350
 351         if (dh->dh_tail != NULL) /* not empty */
 352                 hlist_add_behind(hnode, dh->dh_tail);
 353         else /* empty list */
 354                 hlist_add_head(hnode, &dh->dh_head);
 355         dh->dh_tail = hnode;
 356         return -1; /* unknown depth */
 357 }
 358
 359 static int
 360 cfs_hash_dh_hnode_del(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 361                       struct hlist_node *hnd)
 362 {
 363         cfs_hash_dhead_t *dh = container_of(cfs_hash_dh_hhead(hs, bd),
 364                                             cfs_hash_dhead_t, dh_head);
 365
 366         if (hnd->next == NULL) { /* it's the tail */
 367                 dh->dh_tail = (hnd->pprev == &dh->dh_head.first) ? NULL :
 368                               container_of(hnd->pprev, struct hlist_node, next);
 369         }
 370         hlist_del_init(hnd);
 371         return -1; /* unknown depth */
 372 }
 373
 374 /**
 375  * double links hash head with depth tracking
 376  * new element is always added to tail of hlist
 377  */
 378 typedef struct {
 379         struct hlist_head       dd_head;        /**< entries list */
 380         struct hlist_node       *dd_tail;       /**< the last entry */
 381         unsigned int        dd_depth;       /**< list length */
 382 } cfs_hash_dhead_dep_t;
 383
 384 static int
 385 cfs_hash_dd_hhead_size(struct cfs_hash *hs)
 386 {
 387         return sizeof(cfs_hash_dhead_dep_t);
 388 }
 389
 390 static struct hlist_head *
 391 cfs_hash_dd_hhead(struct cfs_hash *hs, struct cfs_hash_bd *bd)
 392 {
 393         cfs_hash_dhead_dep_t *head;
 394
 395         head = (cfs_hash_dhead_dep_t *)&bd->bd_bucket->hsb_head[0];
 396         return &head[bd->bd_offset].dd_head;
 397 }
 398
 399 static int
 400 cfs_hash_dd_hnode_add(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 401                       struct hlist_node *hnode)
 402 {
 403         cfs_hash_dhead_dep_t *dh = container_of(cfs_hash_dd_hhead(hs, bd),
 404                                                 cfs_hash_dhead_dep_t, dd_head);
 405
 406         if (dh->dd_tail != NULL) /* not empty */
 407                 hlist_add_behind(hnode, dh->dd_tail);
 408         else /* empty list */
 409                 hlist_add_head(hnode, &dh->dd_head);
 410         dh->dd_tail = hnode;
 411         return ++dh->dd_depth;
 412 }
 413
 414 static int
 415 cfs_hash_dd_hnode_del(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 416                       struct hlist_node *hnd)
 417 {
 418         cfs_hash_dhead_dep_t *dh = container_of(cfs_hash_dd_hhead(hs, bd),
 419                                                 cfs_hash_dhead_dep_t, dd_head);
 420
 421         if (hnd->next == NULL) { /* it's the tail */
 422                 dh->dd_tail = (hnd->pprev == &dh->dd_head.first) ? NULL :
 423                               container_of(hnd->pprev, struct hlist_node, next);
 424         }
 425         hlist_del_init(hnd);
 426         return --dh->dd_depth;
 427 }
 428
 429 static cfs_hash_hlist_ops_t cfs_hash_hh_hops = {
 430        .hop_hhead      = cfs_hash_hh_hhead,
 431        .hop_hhead_size = cfs_hash_hh_hhead_size,
 432        .hop_hnode_add  = cfs_hash_hh_hnode_add,
 433        .hop_hnode_del  = cfs_hash_hh_hnode_del,
 434 };
 435
 436 static cfs_hash_hlist_ops_t cfs_hash_hd_hops = {
 437        .hop_hhead      = cfs_hash_hd_hhead,
 438        .hop_hhead_size = cfs_hash_hd_hhead_size,
 439        .hop_hnode_add  = cfs_hash_hd_hnode_add,
 440        .hop_hnode_del  = cfs_hash_hd_hnode_del,
 441 };
 442
 443 static cfs_hash_hlist_ops_t cfs_hash_dh_hops = {
 444        .hop_hhead      = cfs_hash_dh_hhead,
 445        .hop_hhead_size = cfs_hash_dh_hhead_size,
 446        .hop_hnode_add  = cfs_hash_dh_hnode_add,
 447        .hop_hnode_del  = cfs_hash_dh_hnode_del,
 448 };
 449
 450 static cfs_hash_hlist_ops_t cfs_hash_dd_hops = {
 451        .hop_hhead      = cfs_hash_dd_hhead,
 452        .hop_hhead_size = cfs_hash_dd_hhead_size,
 453        .hop_hnode_add  = cfs_hash_dd_hnode_add,
 454        .hop_hnode_del  = cfs_hash_dd_hnode_del,
 455 };
 456
 457 static void
 458 cfs_hash_hlist_setup(struct cfs_hash *hs)
 459 {
 460         if (cfs_hash_with_add_tail(hs)) {
 461                 hs->hs_hops = cfs_hash_with_depth(hs) ?
 462                               &cfs_hash_dd_hops : &cfs_hash_dh_hops;
 463         } else {
 464                 hs->hs_hops = cfs_hash_with_depth(hs) ?
 465                               &cfs_hash_hd_hops : &cfs_hash_hh_hops;
 466         }
 467 }
 468
 469 static void
 470 cfs_hash_bd_from_key(struct cfs_hash *hs, struct cfs_hash_bucket **bkts,
 471                      unsigned int bits, const void *key, struct cfs_hash_bd *bd)
 472 {
 473         unsigned int index = cfs_hash_id(hs, key, (1U << bits) - 1);
 474
 475         LASSERT(bits == hs->hs_cur_bits || bits == hs->hs_rehash_bits);
 476
 477         bd->bd_bucket = bkts[index & ((1U << (bits - hs->hs_bkt_bits)) - 1)];
 478         bd->bd_offset = index >> (bits - hs->hs_bkt_bits);
 479 }
 480
 481 void
 482 cfs_hash_bd_get(struct cfs_hash *hs, const void *key, struct cfs_hash_bd *bd)
 483 {
 484         /* NB: caller should hold hs->hs_rwlock if REHASH is set */
 485         if (likely(hs->hs_rehash_buckets == NULL)) {
 486                 cfs_hash_bd_from_key(hs, hs->hs_buckets,
 487                                      hs->hs_cur_bits, key, bd);
 488         } else {
 489                 LASSERT(hs->hs_rehash_bits != 0);
 490                 cfs_hash_bd_from_key(hs, hs->hs_rehash_buckets,
 491                                      hs->hs_rehash_bits, key, bd);
 492         }
 493 }
 494 EXPORT_SYMBOL(cfs_hash_bd_get);
 495
 496 static inline void
 497 cfs_hash_bd_dep_record(struct cfs_hash *hs, struct cfs_hash_bd *bd, int dep_cur)
 498 {
 499         if (likely(dep_cur <= bd->bd_bucket->hsb_depmax))
 500                 return;
 501
 502         bd->bd_bucket->hsb_depmax = dep_cur;
 503 # if CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1
 504         if (likely(warn_on_depth == 0 ||
 505                    max(warn_on_depth, hs->hs_dep_max) >= dep_cur))
 506                 return;
 507
 508         spin_lock(&hs->hs_dep_lock);
 509         hs->hs_dep_max  = dep_cur;
 510         hs->hs_dep_bkt  = bd->bd_bucket->hsb_index;
 511         hs->hs_dep_off  = bd->bd_offset;
 512         hs->hs_dep_bits = hs->hs_cur_bits;
 513         spin_unlock(&hs->hs_dep_lock);
 514
 515         cfs_wi_schedule(cfs_sched_rehash, &hs->hs_dep_wi);
 516 # endif
 517 }
 518
 519 void
 520 cfs_hash_bd_add_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 521                        struct hlist_node *hnode)
 522 {
 523         int             rc;
 524
 525         rc = hs->hs_hops->hop_hnode_add(hs, bd, hnode);
 526         cfs_hash_bd_dep_record(hs, bd, rc);
 527         bd->bd_bucket->hsb_version++;
 528         if (unlikely(bd->bd_bucket->hsb_version == 0))
 529                 bd->bd_bucket->hsb_version++;
 530         bd->bd_bucket->hsb_count++;
 531
 532         if (cfs_hash_with_counter(hs))
 533                 atomic_inc(&hs->hs_count);
 534         if (!cfs_hash_with_no_itemref(hs))
 535                 cfs_hash_get(hs, hnode);
 536 }
 537 EXPORT_SYMBOL(cfs_hash_bd_add_locked);
 538
 539 void
 540 cfs_hash_bd_del_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 541                        struct hlist_node *hnode)
 542 {
 543         hs->hs_hops->hop_hnode_del(hs, bd, hnode);
 544
 545         LASSERT(bd->bd_bucket->hsb_count > 0);
 546         bd->bd_bucket->hsb_count--;
 547         bd->bd_bucket->hsb_version++;
 548         if (unlikely(bd->bd_bucket->hsb_version == 0))
 549                 bd->bd_bucket->hsb_version++;
 550
 551         if (cfs_hash_with_counter(hs)) {
 552                 LASSERT(atomic_read(&hs->hs_count) > 0);
 553                 atomic_dec(&hs->hs_count);
 554         }
 555         if (!cfs_hash_with_no_itemref(hs))
 556                 cfs_hash_put_locked(hs, hnode);
 557 }
 558 EXPORT_SYMBOL(cfs_hash_bd_del_locked);
 559
 560 void
 561 cfs_hash_bd_move_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd_old,
 562                         struct cfs_hash_bd *bd_new, struct hlist_node *hnode)
 563 {
 564         struct cfs_hash_bucket *obkt = bd_old->bd_bucket;
 565         struct cfs_hash_bucket *nbkt = bd_new->bd_bucket;
 566         int             rc;
 567
 568         if (cfs_hash_bd_compare(bd_old, bd_new) == 0)
 569                 return;
 570
 571         /* use cfs_hash_bd_hnode_add/del, to avoid atomic & refcount ops
 572          * in cfs_hash_bd_del/add_locked */
 573         hs->hs_hops->hop_hnode_del(hs, bd_old, hnode);
 574         rc = hs->hs_hops->hop_hnode_add(hs, bd_new, hnode);
 575         cfs_hash_bd_dep_record(hs, bd_new, rc);
 576
 577         LASSERT(obkt->hsb_count > 0);
 578         obkt->hsb_count--;
 579         obkt->hsb_version++;
 580         if (unlikely(obkt->hsb_version == 0))
 581                 obkt->hsb_version++;
 582         nbkt->hsb_count++;
 583         nbkt->hsb_version++;
 584         if (unlikely(nbkt->hsb_version == 0))
 585                 nbkt->hsb_version++;
 586 }
 587 EXPORT_SYMBOL(cfs_hash_bd_move_locked);
 588
 589 enum {
 590         /** always set, for sanity (avoid ZERO intent) */
 591         CFS_HS_LOOKUP_MASK_FIND     = 1 << 0,
 592         /** return entry with a ref */
 593         CFS_HS_LOOKUP_MASK_REF      = 1 << 1,
 594         /** add entry if not existing */
 595         CFS_HS_LOOKUP_MASK_ADD      = 1 << 2,
 596         /** delete entry, ignore other masks */
 597         CFS_HS_LOOKUP_MASK_DEL      = 1 << 3,
 598 };
 599
 600 typedef enum cfs_hash_lookup_intent {
 601         /** return item w/o refcount */
 602         CFS_HS_LOOKUP_IT_PEEK       = CFS_HS_LOOKUP_MASK_FIND,
 603         /** return item with refcount */
 604         CFS_HS_LOOKUP_IT_FIND       = (CFS_HS_LOOKUP_MASK_FIND |
 605                                        CFS_HS_LOOKUP_MASK_REF),
 606         /** return item w/o refcount if existed, otherwise add */
 607         CFS_HS_LOOKUP_IT_ADD    = (CFS_HS_LOOKUP_MASK_FIND |
 608                                        CFS_HS_LOOKUP_MASK_ADD),
 609         /** return item with refcount if existed, otherwise add */
 610         CFS_HS_LOOKUP_IT_FINDADD    = (CFS_HS_LOOKUP_IT_FIND |
 611                                        CFS_HS_LOOKUP_MASK_ADD),
 612         /** delete if existed */
 613         CFS_HS_LOOKUP_IT_FINDDEL    = (CFS_HS_LOOKUP_MASK_FIND |
 614                                        CFS_HS_LOOKUP_MASK_DEL)
 615 } cfs_hash_lookup_intent_t;
 616
 617 static struct hlist_node *
 618 cfs_hash_bd_lookup_intent(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 619                           const void *key, struct hlist_node *hnode,
 620                           cfs_hash_lookup_intent_t intent)
 621
 622 {
 623         struct hlist_head  *hhead = cfs_hash_bd_hhead(hs, bd);
 624         struct hlist_node  *ehnode;
 625         struct hlist_node  *match;
 626         int  intent_add = (intent & CFS_HS_LOOKUP_MASK_ADD) != 0;
 627
 628         /* with this function, we can avoid a lot of useless refcount ops,
 629          * which are expensive atomic operations most time. */
 630         match = intent_add ? NULL : hnode;
 631         hlist_for_each(ehnode, hhead) {
 632                 if (!cfs_hash_keycmp(hs, key, ehnode))
 633                         continue;
 634
 635                 if (match != NULL && match != ehnode) /* can't match */
 636                         continue;
 637
 638                 /* match and ... */
 639                 if ((intent & CFS_HS_LOOKUP_MASK_DEL) != 0) {
 640                         cfs_hash_bd_del_locked(hs, bd, ehnode);
 641                         return ehnode;
 642                 }
 643
 644                 /* caller wants refcount? */
 645                 if ((intent & CFS_HS_LOOKUP_MASK_REF) != 0)
 646                         cfs_hash_get(hs, ehnode);
 647                 return ehnode;
 648         }
 649         /* no match item */
 650         if (!intent_add)
 651                 return NULL;
 652
 653         LASSERT(hnode != NULL);
 654         cfs_hash_bd_add_locked(hs, bd, hnode);
 655         return hnode;
 656 }
 657
 658 struct hlist_node *
 659 cfs_hash_bd_lookup_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd, const void *key)
 660 {
 661         return cfs_hash_bd_lookup_intent(hs, bd, key, NULL,
 662                                          CFS_HS_LOOKUP_IT_FIND);
 663 }
 664 EXPORT_SYMBOL(cfs_hash_bd_lookup_locked);
 665
 666 struct hlist_node *
 667 cfs_hash_bd_peek_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd, const void *key)
 668 {
 669         return cfs_hash_bd_lookup_intent(hs, bd, key, NULL,
 670                                          CFS_HS_LOOKUP_IT_PEEK);
 671 }
 672 EXPORT_SYMBOL(cfs_hash_bd_peek_locked);
 673
 674 struct hlist_node *
 675 cfs_hash_bd_findadd_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 676                            const void *key, struct hlist_node *hnode,
 677                            int noref)
 678 {
 679         return cfs_hash_bd_lookup_intent(hs, bd, key, hnode,
 680                                          CFS_HS_LOOKUP_IT_ADD |
 681                                          (!noref * CFS_HS_LOOKUP_MASK_REF));
 682 }
 683 EXPORT_SYMBOL(cfs_hash_bd_findadd_locked);
 684
 685 struct hlist_node *
 686 cfs_hash_bd_finddel_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
 687                            const void *key, struct hlist_node *hnode)
 688 {
 689         /* hnode can be NULL, we find the first item with @key */
 690         return cfs_hash_bd_lookup_intent(hs, bd, key, hnode,
 691                                          CFS_HS_LOOKUP_IT_FINDDEL);
 692 }
 693 EXPORT_SYMBOL(cfs_hash_bd_finddel_locked);
 694
 695 static void
 696 cfs_hash_multi_bd_lock(struct cfs_hash *hs, struct cfs_hash_bd *bds,
 697                        unsigned n, int excl)
 698 {
 699         struct cfs_hash_bucket *prev = NULL;
 700         int             i;
 701
 702         /**
 703          * bds must be ascendantly ordered by bd->bd_bucket->hsb_index.
 704          * NB: it's possible that several bds point to the same bucket but
 705          * have different bd::bd_offset, so need take care of deadlock.
 706          */
 707         cfs_hash_for_each_bd(bds, n, i) {
 708                 if (prev == bds[i].bd_bucket)
 709                         continue;
 710
 711                 LASSERT(prev == NULL ||
 712                         prev->hsb_index < bds[i].bd_bucket->hsb_index);
 713                 cfs_hash_bd_lock(hs, &bds[i], excl);
 714                 prev = bds[i].bd_bucket;
 715         }
 716 }
 717
 718 static void
 719 cfs_hash_multi_bd_unlock(struct cfs_hash *hs, struct cfs_hash_bd *bds,
 720                          unsigned n, int excl)
 721 {
 722         struct cfs_hash_bucket *prev = NULL;
 723         int             i;
 724
 725         cfs_hash_for_each_bd(bds, n, i) {
 726                 if (prev != bds[i].bd_bucket) {
 727                         cfs_hash_bd_unlock(hs, &bds[i], excl);
 728                         prev = bds[i].bd_bucket;
 729                 }
 730         }
 731 }
 732
 733 static struct hlist_node *
 734 cfs_hash_multi_bd_lookup_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
 735                                 unsigned n, const void *key)
 736 {
 737         struct hlist_node  *ehnode;
 738         unsigned           i;
 739
 740         cfs_hash_for_each_bd(bds, n, i) {
 741                 ehnode = cfs_hash_bd_lookup_intent(hs, &bds[i], key, NULL,
 742                                                    CFS_HS_LOOKUP_IT_FIND);
 743                 if (ehnode != NULL)
 744                         return ehnode;
 745         }
 746         return NULL;
 747 }
 748
 749 static struct hlist_node *
 750 cfs_hash_multi_bd_findadd_locked(struct cfs_hash *hs,
 751                                  struct cfs_hash_bd *bds, unsigned n, const void *key,
 752                                  struct hlist_node *hnode, int noref)
 753 {
 754         struct hlist_node  *ehnode;
 755         int             intent;
 756         unsigned           i;
 757
 758         LASSERT(hnode != NULL);
 759         intent = CFS_HS_LOOKUP_IT_PEEK | (!noref * CFS_HS_LOOKUP_MASK_REF);
 760
 761         cfs_hash_for_each_bd(bds, n, i) {
 762                 ehnode = cfs_hash_bd_lookup_intent(hs, &bds[i], key,
 763                                                    NULL, intent);
 764                 if (ehnode != NULL)
 765                         return ehnode;
 766         }
 767
 768         if (i == 1) { /* only one bucket */
 769                 cfs_hash_bd_add_locked(hs, &bds[0], hnode);
 770         } else {
 771                 struct cfs_hash_bd      mybd;
 772
 773                 cfs_hash_bd_get(hs, key, &mybd);
 774                 cfs_hash_bd_add_locked(hs, &mybd, hnode);
 775         }
 776
 777         return hnode;
 778 }
 779
 780 static struct hlist_node *
 781 cfs_hash_multi_bd_finddel_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
 782                                  unsigned n, const void *key,
 783                                  struct hlist_node *hnode)
 784 {
 785         struct hlist_node  *ehnode;
 786         unsigned           i;
 787
 788         cfs_hash_for_each_bd(bds, n, i) {
 789                 ehnode = cfs_hash_bd_lookup_intent(hs, &bds[i], key, hnode,
 790                                                    CFS_HS_LOOKUP_IT_FINDDEL);
 791                 if (ehnode != NULL)
 792                         return ehnode;
 793         }
 794         return NULL;
 795 }
 796
 797 static void
 798 cfs_hash_bd_order(struct cfs_hash_bd *bd1, struct cfs_hash_bd *bd2)
 799 {
 800         int     rc;
 801
 802         if (bd2->bd_bucket == NULL)
 803                 return;
 804
 805         if (bd1->bd_bucket == NULL) {
 806                 *bd1 = *bd2;
 807                 bd2->bd_bucket = NULL;
 808                 return;
 809         }
 810
 811         rc = cfs_hash_bd_compare(bd1, bd2);
 812         if (rc == 0) {
 813                 bd2->bd_bucket = NULL;
 814
 815         } else if (rc > 0) { /* swab bd1 and bd2 */
 816                 struct cfs_hash_bd tmp;
 817
 818                 tmp = *bd2;
 819                 *bd2 = *bd1;
 820                 *bd1 = tmp;
 821         }
 822 }
 823
 824 void
 825 cfs_hash_dual_bd_get(struct cfs_hash *hs, const void *key, struct cfs_hash_bd *bds)
 826 {
 827         /* NB: caller should hold hs_lock.rw if REHASH is set */
 828         cfs_hash_bd_from_key(hs, hs->hs_buckets,
 829                              hs->hs_cur_bits, key, &bds[0]);
 830         if (likely(hs->hs_rehash_buckets == NULL)) {
 831                 /* no rehash or not rehashing */
 832                 bds[1].bd_bucket = NULL;
 833                 return;
 834         }
 835
 836         LASSERT(hs->hs_rehash_bits != 0);
 837         cfs_hash_bd_from_key(hs, hs->hs_rehash_buckets,
 838                              hs->hs_rehash_bits, key, &bds[1]);
 839
 840         cfs_hash_bd_order(&bds[0], &bds[1]);
 841 }
 842 EXPORT_SYMBOL(cfs_hash_dual_bd_get);
 843
 844 void
 845 cfs_hash_dual_bd_lock(struct cfs_hash *hs, struct cfs_hash_bd *bds, int excl)
 846 {
 847         cfs_hash_multi_bd_lock(hs, bds, 2, excl);
 848 }
 849 EXPORT_SYMBOL(cfs_hash_dual_bd_lock);
 850
 851 void
 852 cfs_hash_dual_bd_unlock(struct cfs_hash *hs, struct cfs_hash_bd *bds, int excl)
 853 {
 854         cfs_hash_multi_bd_unlock(hs, bds, 2, excl);
 855 }
 856 EXPORT_SYMBOL(cfs_hash_dual_bd_unlock);
 857
 858 struct hlist_node *
 859 cfs_hash_dual_bd_lookup_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
 860                                const void *key)
 861 {
 862         return cfs_hash_multi_bd_lookup_locked(hs, bds, 2, key);
 863 }
 864 EXPORT_SYMBOL(cfs_hash_dual_bd_lookup_locked);
 865
 866 struct hlist_node *
 867 cfs_hash_dual_bd_findadd_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
 868                                 const void *key, struct hlist_node *hnode,
 869                                 int noref)
 870 {
 871         return cfs_hash_multi_bd_findadd_locked(hs, bds, 2, key,
 872                                                 hnode, noref);
 873 }
 874 EXPORT_SYMBOL(cfs_hash_dual_bd_findadd_locked);
 875
 876 struct hlist_node *
 877 cfs_hash_dual_bd_finddel_locked(struct cfs_hash *hs, struct cfs_hash_bd *bds,
 878                                 const void *key, struct hlist_node *hnode)
 879 {
 880         return cfs_hash_multi_bd_finddel_locked(hs, bds, 2, key, hnode);
 881 }
 882 EXPORT_SYMBOL(cfs_hash_dual_bd_finddel_locked);
 883
 884 static void
 885 cfs_hash_buckets_free(struct cfs_hash_bucket **buckets,
 886                       int bkt_size, int prev_size, int size)
 887 {
 888         int     i;
 889
 890         for (i = prev_size; i < size; i++) {
 891                 if (buckets[i] != NULL)
 892                         LIBCFS_FREE(buckets[i], bkt_size);
 893         }
 894
 895         LIBCFS_FREE(buckets, sizeof(buckets[0]) * size);
 896 }
 897
 898 /*
 899  * Create or grow bucket memory. Return old_buckets if no allocation was
 900  * needed, the newly allocated buckets if allocation was needed and
 901  * successful, and NULL on error.
 902  */
 903 static struct cfs_hash_bucket **
 904 cfs_hash_buckets_realloc(struct cfs_hash *hs, struct cfs_hash_bucket **old_bkts,
 905                          unsigned int old_size, unsigned int new_size)
 906 {
 907         struct cfs_hash_bucket **new_bkts;
 908         int              i;
 909
 910         LASSERT(old_size == 0 || old_bkts != NULL);
 911
 912         if (old_bkts != NULL && old_size == new_size)
 913                 return old_bkts;
 914
 915         LIBCFS_ALLOC(new_bkts, sizeof(new_bkts[0]) * new_size);
 916         if (new_bkts == NULL)
 917                 return NULL;
 918
 919         if (old_bkts != NULL) {
 920                 memcpy(new_bkts, old_bkts,
 921                        min(old_size, new_size) * sizeof(*old_bkts));
 922         }
 923
 924         for (i = old_size; i < new_size; i++) {
 925                 struct hlist_head *hhead;
 926                 struct cfs_hash_bd     bd;
 927
 928                 LIBCFS_ALLOC(new_bkts[i], cfs_hash_bkt_size(hs));
 929                 if (new_bkts[i] == NULL) {
 930                         cfs_hash_buckets_free(new_bkts, cfs_hash_bkt_size(hs),
 931                                               old_size, new_size);
 932                         return NULL;
 933                 }
 934
 935                 new_bkts[i]->hsb_index   = i;
 936                 new_bkts[i]->hsb_version = 1;  /* shouldn't be zero */
 937                 new_bkts[i]->hsb_depmax  = -1; /* unknown */
 938                 bd.bd_bucket = new_bkts[i];
 939                 cfs_hash_bd_for_each_hlist(hs, &bd, hhead)
 940                         INIT_HLIST_HEAD(hhead);
 941
 942                 if (cfs_hash_with_no_lock(hs) ||
 943                     cfs_hash_with_no_bktlock(hs))
 944                         continue;
 945
 946                 if (cfs_hash_with_rw_bktlock(hs))
 947                         rwlock_init(&new_bkts[i]->hsb_lock.rw);
 948                 else if (cfs_hash_with_spin_bktlock(hs))
 949                         spin_lock_init(&new_bkts[i]->hsb_lock.spin);
 950                 else
 951                         LBUG(); /* invalid use-case */
 952         }
 953         return new_bkts;
 954 }
 955
 956 /**
 957  * Initialize new libcfs hash, where:
 958  * @name     - Descriptive hash name
 959  * @cur_bits - Initial hash table size, in bits
 960  * @max_bits - Maximum allowed hash table resize, in bits
 961  * @ops      - Registered hash table operations
 962  * @flags    - CFS_HASH_REHASH enable synamic hash resizing
 963  *         - CFS_HASH_SORT enable chained hash sort
 964  */
 965 static int cfs_hash_rehash_worker(cfs_workitem_t *wi);
 966
 967 #if CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1
 968 static int cfs_hash_dep_print(cfs_workitem_t *wi)
 969 {
 970         struct cfs_hash *hs = container_of(wi, struct cfs_hash, hs_dep_wi);
 971         int      dep;
 972         int      bkt;
 973         int      off;
 974         int      bits;
 975
 976         spin_lock(&hs->hs_dep_lock);
 977         dep  = hs->hs_dep_max;
 978         bkt  = hs->hs_dep_bkt;
 979         off  = hs->hs_dep_off;
 980         bits = hs->hs_dep_bits;
 981         spin_unlock(&hs->hs_dep_lock);
 982
 983         LCONSOLE_WARN("#### HASH %s (bits: %d): max depth %d at bucket %d/%d\n",
 984                       hs->hs_name, bits, dep, bkt, off);
 985         spin_lock(&hs->hs_dep_lock);
 986         hs->hs_dep_bits = 0; /* mark as workitem done */
 987         spin_unlock(&hs->hs_dep_lock);
 988         return 0;
 989 }
 990
 991 static void cfs_hash_depth_wi_init(struct cfs_hash *hs)
 992 {
 993         spin_lock_init(&hs->hs_dep_lock);
 994         cfs_wi_init(&hs->hs_dep_wi, hs, cfs_hash_dep_print);
 995 }
 996
 997 static void cfs_hash_depth_wi_cancel(struct cfs_hash *hs)
 998 {
 999         if (cfs_wi_deschedule(cfs_sched_rehash, &hs->hs_dep_wi))
1000                 return;
1001
1002         spin_lock(&hs->hs_dep_lock);
1003         while (hs->hs_dep_bits != 0) {
1004                 spin_unlock(&hs->hs_dep_lock);
1005                 cond_resched();
1006                 spin_lock(&hs->hs_dep_lock);
1007         }
1008         spin_unlock(&hs->hs_dep_lock);
1009 }
1010
1011 #else /* CFS_HASH_DEBUG_LEVEL < CFS_HASH_DEBUG_1 */
1012
1013 static inline void cfs_hash_depth_wi_init(struct cfs_hash *hs) {}
1014 static inline void cfs_hash_depth_wi_cancel(struct cfs_hash *hs) {}
1015
1016 #endif /* CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1 */
1017
1018 struct cfs_hash *
1019 cfs_hash_create(char *name, unsigned cur_bits, unsigned max_bits,
1020                 unsigned bkt_bits, unsigned extra_bytes,
1021                 unsigned min_theta, unsigned max_theta,
1022                 cfs_hash_ops_t *ops, unsigned flags)
1023 {
1024         struct cfs_hash *hs;
1025         int      len;
1026
1027         CLASSERT(CFS_HASH_THETA_BITS < 15);
1028
1029         LASSERT(name != NULL);
1030         LASSERT(ops != NULL);
1031         LASSERT(ops->hs_key);
1032         LASSERT(ops->hs_hash);
1033         LASSERT(ops->hs_object);
1034         LASSERT(ops->hs_keycmp);
1035         LASSERT(ops->hs_get != NULL);
1036         LASSERT(ops->hs_put_locked != NULL);
1037
1038         if ((flags & CFS_HASH_REHASH) != 0)
1039                 flags |= CFS_HASH_COUNTER; /* must have counter */
1040
1041         LASSERT(cur_bits > 0);
1042         LASSERT(cur_bits >= bkt_bits);
1043         LASSERT(max_bits >= cur_bits && max_bits < 31);
1044         LASSERT(ergo((flags & CFS_HASH_REHASH) == 0, cur_bits == max_bits));
1045         LASSERT(ergo((flags & CFS_HASH_REHASH) != 0,
1046                      (flags & CFS_HASH_NO_LOCK) == 0));
1047         LASSERT(ergo((flags & CFS_HASH_REHASH_KEY) != 0,
1048                       ops->hs_keycpy != NULL));
1049
1050         len = (flags & CFS_HASH_BIGNAME) == 0 ?
1051               CFS_HASH_NAME_LEN : CFS_HASH_BIGNAME_LEN;
1052         LIBCFS_ALLOC(hs, offsetof(struct cfs_hash, hs_name[len]));
1053         if (hs == NULL)
1054                 return NULL;
1055
1056         strncpy(hs->hs_name, name, len);
1057         hs->hs_name[len - 1] = '\0';
1058         hs->hs_flags = flags;
1059
1060         atomic_set(&hs->hs_refcount, 1);
1061         atomic_set(&hs->hs_count, 0);
1062
1063         cfs_hash_lock_setup(hs);
1064         cfs_hash_hlist_setup(hs);
1065
1066         hs->hs_cur_bits = (__u8)cur_bits;
1067         hs->hs_min_bits = (__u8)cur_bits;
1068         hs->hs_max_bits = (__u8)max_bits;
1069         hs->hs_bkt_bits = (__u8)bkt_bits;
1070
1071         hs->hs_ops       = ops;
1072         hs->hs_extra_bytes = extra_bytes;
1073         hs->hs_rehash_bits = 0;
1074         cfs_wi_init(&hs->hs_rehash_wi, hs, cfs_hash_rehash_worker);
1075         cfs_hash_depth_wi_init(hs);
1076
1077         if (cfs_hash_with_rehash(hs))
1078                 __cfs_hash_set_theta(hs, min_theta, max_theta);
1079
1080         hs->hs_buckets = cfs_hash_buckets_realloc(hs, NULL, 0,
1081                                                   CFS_HASH_NBKT(hs));
1082         if (hs->hs_buckets != NULL)
1083                 return hs;
1084
1085         LIBCFS_FREE(hs, offsetof(struct cfs_hash, hs_name[len]));
1086         return NULL;
1087 }
1088 EXPORT_SYMBOL(cfs_hash_create);
1089
1090 /**
1091  * Cleanup libcfs hash @hs.
1092  */
1093 static void
1094 cfs_hash_destroy(struct cfs_hash *hs)
1095 {
1096         struct hlist_node     *hnode;
1097         struct hlist_node     *pos;
1098         struct cfs_hash_bd       bd;
1099         int                i;
1100
1101         LASSERT(hs != NULL);
1102         LASSERT(!cfs_hash_is_exiting(hs) &&
1103                 !cfs_hash_is_iterating(hs));
1104
1105         /**
1106          * prohibit further rehashes, don't need any lock because
1107          * I'm the only (last) one can change it.
1108          */
1109         hs->hs_exiting = 1;
1110         if (cfs_hash_with_rehash(hs))
1111                 cfs_hash_rehash_cancel(hs);
1112
1113         cfs_hash_depth_wi_cancel(hs);
1114         /* rehash should be done/canceled */
1115         LASSERT(hs->hs_buckets != NULL &&
1116                 hs->hs_rehash_buckets == NULL);
1117
1118         cfs_hash_for_each_bucket(hs, &bd, i) {
1119                 struct hlist_head *hhead;
1120
1121                 LASSERT(bd.bd_bucket != NULL);
1122                 /* no need to take this lock, just for consistent code */
1123                 cfs_hash_bd_lock(hs, &bd, 1);
1124
1125                 cfs_hash_bd_for_each_hlist(hs, &bd, hhead) {
1126                         hlist_for_each_safe(hnode, pos, hhead) {
1127                                 LASSERTF(!cfs_hash_with_assert_empty(hs),
1128                                          "hash %s bucket %u(%u) is not empty: %u items left\n",
1129                                          hs->hs_name, bd.bd_bucket->hsb_index,
1130                                          bd.bd_offset, bd.bd_bucket->hsb_count);
1131                                 /* can't assert key valicate, because we
1132                                  * can interrupt rehash */
1133                                 cfs_hash_bd_del_locked(hs, &bd, hnode);
1134                                 cfs_hash_exit(hs, hnode);
1135                         }
1136                 }
1137                 LASSERT(bd.bd_bucket->hsb_count == 0);
1138                 cfs_hash_bd_unlock(hs, &bd, 1);
1139                 cond_resched();
1140         }
1141
1142         LASSERT(atomic_read(&hs->hs_count) == 0);
1143
1144         cfs_hash_buckets_free(hs->hs_buckets, cfs_hash_bkt_size(hs),
1145                               0, CFS_HASH_NBKT(hs));
1146         i = cfs_hash_with_bigname(hs) ?
1147             CFS_HASH_BIGNAME_LEN : CFS_HASH_NAME_LEN;
1148         LIBCFS_FREE(hs, offsetof(struct cfs_hash, hs_name[i]));
1149 }
1150
1151 struct cfs_hash *cfs_hash_getref(struct cfs_hash *hs)
1152 {
1153         if (atomic_inc_not_zero(&hs->hs_refcount))
1154                 return hs;
1155         return NULL;
1156 }
1157 EXPORT_SYMBOL(cfs_hash_getref);
1158
1159 void cfs_hash_putref(struct cfs_hash *hs)
1160 {
1161         if (atomic_dec_and_test(&hs->hs_refcount))
1162                 cfs_hash_destroy(hs);
1163 }
1164 EXPORT_SYMBOL(cfs_hash_putref);
1165
1166 static inline int
1167 cfs_hash_rehash_bits(struct cfs_hash *hs)
1168 {
1169         if (cfs_hash_with_no_lock(hs) ||
1170             !cfs_hash_with_rehash(hs))
1171                 return -EOPNOTSUPP;
1172
1173         if (unlikely(cfs_hash_is_exiting(hs)))
1174                 return -ESRCH;
1175
1176         if (unlikely(cfs_hash_is_rehashing(hs)))
1177                 return -EALREADY;
1178
1179         if (unlikely(cfs_hash_is_iterating(hs)))
1180                 return -EAGAIN;
1181
1182         /* XXX: need to handle case with max_theta != 2.0
1183          *      and the case with min_theta != 0.5 */
1184         if ((hs->hs_cur_bits < hs->hs_max_bits) &&
1185             (__cfs_hash_theta(hs) > hs->hs_max_theta))
1186                 return hs->hs_cur_bits + 1;
1187
1188         if (!cfs_hash_with_shrink(hs))
1189                 return 0;
1190
1191         if ((hs->hs_cur_bits > hs->hs_min_bits) &&
1192             (__cfs_hash_theta(hs) < hs->hs_min_theta))
1193                 return hs->hs_cur_bits - 1;
1194
1195         return 0;
1196 }
1197
1198 /**
1199  * don't allow inline rehash if:
1200  * - user wants non-blocking change (add/del) on hash table
1201  * - too many elements
1202  */
1203 static inline int
1204 cfs_hash_rehash_inline(struct cfs_hash *hs)
1205 {
1206         return !cfs_hash_with_nblk_change(hs) &&
1207                atomic_read(&hs->hs_count) < CFS_HASH_LOOP_HOG;
1208 }
1209
1210 /**
1211  * Add item @hnode to libcfs hash @hs using @key.  The registered
1212  * ops->hs_get function will be called when the item is added.
1213  */
1214 void
1215 cfs_hash_add(struct cfs_hash *hs, const void *key, struct hlist_node *hnode)
1216 {
1217         struct cfs_hash_bd   bd;
1218         int          bits;
1219
1220         LASSERT(hlist_unhashed(hnode));
1221
1222         cfs_hash_lock(hs, 0);
1223         cfs_hash_bd_get_and_lock(hs, key, &bd, 1);
1224
1225         cfs_hash_key_validate(hs, key, hnode);
1226         cfs_hash_bd_add_locked(hs, &bd, hnode);
1227
1228         cfs_hash_bd_unlock(hs, &bd, 1);
1229
1230         bits = cfs_hash_rehash_bits(hs);
1231         cfs_hash_unlock(hs, 0);
1232         if (bits > 0)
1233                 cfs_hash_rehash(hs, cfs_hash_rehash_inline(hs));
1234 }
1235 EXPORT_SYMBOL(cfs_hash_add);
1236
1237 static struct hlist_node *
1238 cfs_hash_find_or_add(struct cfs_hash *hs, const void *key,
1239                      struct hlist_node *hnode, int noref)
1240 {
1241         struct hlist_node *ehnode;
1242         struct cfs_hash_bd     bds[2];
1243         int            bits = 0;
1244
1245         LASSERT(hlist_unhashed(hnode));
1246
1247         cfs_hash_lock(hs, 0);
1248         cfs_hash_dual_bd_get_and_lock(hs, key, bds, 1);
1249
1250         cfs_hash_key_validate(hs, key, hnode);
1251         ehnode = cfs_hash_dual_bd_findadd_locked(hs, bds, key,
1252                                                  hnode, noref);
1253         cfs_hash_dual_bd_unlock(hs, bds, 1);
1254
1255         if (ehnode == hnode) /* new item added */
1256                 bits = cfs_hash_rehash_bits(hs);
1257         cfs_hash_unlock(hs, 0);
1258         if (bits > 0)
1259                 cfs_hash_rehash(hs, cfs_hash_rehash_inline(hs));
1260
1261         return ehnode;
1262 }
1263
1264 /**
1265  * Add item @hnode to libcfs hash @hs using @key.  The registered
1266  * ops->hs_get function will be called if the item was added.
1267  * Returns 0 on success or -EALREADY on key collisions.
1268  */
1269 int
1270 cfs_hash_add_unique(struct cfs_hash *hs, const void *key, struct hlist_node *hnode)
1271 {
1272         return cfs_hash_find_or_add(hs, key, hnode, 1) != hnode ?
1273                -EALREADY : 0;
1274 }
1275 EXPORT_SYMBOL(cfs_hash_add_unique);
1276
1277 /**
1278  * Add item @hnode to libcfs hash @hs using @key.  If this @key
1279  * already exists in the hash then ops->hs_get will be called on the
1280  * conflicting entry and that entry will be returned to the caller.
1281  * Otherwise ops->hs_get is called on the item which was added.
1282  */
1283 void *
1284 cfs_hash_findadd_unique(struct cfs_hash *hs, const void *key,
1285                         struct hlist_node *hnode)
1286 {
1287         hnode = cfs_hash_find_or_add(hs, key, hnode, 0);
1288
1289         return cfs_hash_object(hs, hnode);
1290 }
1291 EXPORT_SYMBOL(cfs_hash_findadd_unique);
1292
1293 /**
1294  * Delete item @hnode from the libcfs hash @hs using @key.  The @key
1295  * is required to ensure the correct hash bucket is locked since there
1296  * is no direct linkage from the item to the bucket.  The object
1297  * removed from the hash will be returned and obs->hs_put is called
1298  * on the removed object.
1299  */
1300 void *
1301 cfs_hash_del(struct cfs_hash *hs, const void *key, struct hlist_node *hnode)
1302 {
1303         void       *obj  = NULL;
1304         int          bits = 0;
1305         struct cfs_hash_bd   bds[2];
1306
1307         cfs_hash_lock(hs, 0);
1308         cfs_hash_dual_bd_get_and_lock(hs, key, bds, 1);
1309
1310         /* NB: do nothing if @hnode is not in hash table */
1311         if (hnode == NULL || !hlist_unhashed(hnode)) {
1312                 if (bds[1].bd_bucket == NULL && hnode != NULL) {
1313                         cfs_hash_bd_del_locked(hs, &bds[0], hnode);
1314                 } else {
1315                         hnode = cfs_hash_dual_bd_finddel_locked(hs, bds,
1316                                                                 key, hnode);
1317                 }
1318         }
1319
1320         if (hnode != NULL) {
1321                 obj  = cfs_hash_object(hs, hnode);
1322                 bits = cfs_hash_rehash_bits(hs);
1323         }
1324
1325         cfs_hash_dual_bd_unlock(hs, bds, 1);
1326         cfs_hash_unlock(hs, 0);
1327         if (bits > 0)
1328                 cfs_hash_rehash(hs, cfs_hash_rehash_inline(hs));
1329
1330         return obj;
1331 }
1332 EXPORT_SYMBOL(cfs_hash_del);
1333
1334 /**
1335  * Delete item given @key in libcfs hash @hs.  The first @key found in
1336  * the hash will be removed, if the key exists multiple times in the hash
1337  * @hs this function must be called once per key.  The removed object
1338  * will be returned and ops->hs_put is called on the removed object.
1339  */
1340 void *
1341 cfs_hash_del_key(struct cfs_hash *hs, const void *key)
1342 {
1343         return cfs_hash_del(hs, key, NULL);
1344 }
1345 EXPORT_SYMBOL(cfs_hash_del_key);
1346
1347 /**
1348  * Lookup an item using @key in the libcfs hash @hs and return it.
1349  * If the @key is found in the hash hs->hs_get() is called and the
1350  * matching objects is returned.  It is the callers responsibility
1351  * to call the counterpart ops->hs_put using the cfs_hash_put() macro
1352  * when when finished with the object.  If the @key was not found
1353  * in the hash @hs NULL is returned.
1354  */
1355 void *
1356 cfs_hash_lookup(struct cfs_hash *hs, const void *key)
1357 {
1358         void             *obj = NULL;
1359         struct hlist_node     *hnode;
1360         struct cfs_hash_bd       bds[2];
1361
1362         cfs_hash_lock(hs, 0);
1363         cfs_hash_dual_bd_get_and_lock(hs, key, bds, 0);
1364
1365         hnode = cfs_hash_dual_bd_lookup_locked(hs, bds, key);
1366         if (hnode != NULL)
1367                 obj = cfs_hash_object(hs, hnode);
1368
1369         cfs_hash_dual_bd_unlock(hs, bds, 0);
1370         cfs_hash_unlock(hs, 0);
1371
1372         return obj;
1373 }
1374 EXPORT_SYMBOL(cfs_hash_lookup);
1375
1376 static void
1377 cfs_hash_for_each_enter(struct cfs_hash *hs) {
1378         LASSERT(!cfs_hash_is_exiting(hs));
1379
1380         if (!cfs_hash_with_rehash(hs))
1381                 return;
1382         /*
1383          * NB: it's race on cfs_has_t::hs_iterating, but doesn't matter
1384          * because it's just an unreliable signal to rehash-thread,
1385          * rehash-thread will try to finish rehash ASAP when seeing this.
1386          */
1387         hs->hs_iterating = 1;
1388
1389         cfs_hash_lock(hs, 1);
1390         hs->hs_iterators++;
1391
1392         /* NB: iteration is mostly called by service thread,
1393          * we tend to cancel pending rehash-request, instead of
1394          * blocking service thread, we will relaunch rehash request
1395          * after iteration */
1396         if (cfs_hash_is_rehashing(hs))
1397                 cfs_hash_rehash_cancel_locked(hs);
1398         cfs_hash_unlock(hs, 1);
1399 }
1400
1401 static void
1402 cfs_hash_for_each_exit(struct cfs_hash *hs) {
1403         int remained;
1404         int bits;
1405
1406         if (!cfs_hash_with_rehash(hs))
1407                 return;
1408         cfs_hash_lock(hs, 1);
1409         remained = --hs->hs_iterators;
1410         bits = cfs_hash_rehash_bits(hs);
1411         cfs_hash_unlock(hs, 1);
1412         /* NB: it's race on cfs_has_t::hs_iterating, see above */
1413         if (remained == 0)
1414                 hs->hs_iterating = 0;
1415         if (bits > 0) {
1416                 cfs_hash_rehash(hs, atomic_read(&hs->hs_count) <
1417                                     CFS_HASH_LOOP_HOG);
1418         }
1419 }
1420
1421 /**
1422  * For each item in the libcfs hash @hs call the passed callback @func
1423  * and pass to it as an argument each hash item and the private @data.
1424  *
1425  * a) the function may sleep!
1426  * b) during the callback:
1427  *    . the bucket lock is held so the callback must never sleep.
1428  *    . if @removal_safe is true, use can remove current item by
1429  *      cfs_hash_bd_del_locked
1430  */
1431 static __u64
1432 cfs_hash_for_each_tight(struct cfs_hash *hs, cfs_hash_for_each_cb_t func,
1433                         void *data, int remove_safe) {
1434         struct hlist_node     *hnode;
1435         struct hlist_node     *pos;
1436         struct cfs_hash_bd       bd;
1437         __u64            count = 0;
1438         int                excl  = !!remove_safe;
1439         int                loop  = 0;
1440         int                i;
1441
1442         cfs_hash_for_each_enter(hs);
1443
1444         cfs_hash_lock(hs, 0);
1445         LASSERT(!cfs_hash_is_rehashing(hs));
1446
1447         cfs_hash_for_each_bucket(hs, &bd, i) {
1448                 struct hlist_head *hhead;
1449
1450                 cfs_hash_bd_lock(hs, &bd, excl);
1451                 if (func == NULL) { /* only glimpse size */
1452                         count += bd.bd_bucket->hsb_count;
1453                         cfs_hash_bd_unlock(hs, &bd, excl);
1454                         continue;
1455                 }
1456
1457                 cfs_hash_bd_for_each_hlist(hs, &bd, hhead) {
1458                         hlist_for_each_safe(hnode, pos, hhead) {
1459                                 cfs_hash_bucket_validate(hs, &bd, hnode);
1460                                 count++;
1461                                 loop++;
1462                                 if (func(hs, &bd, hnode, data)) {
1463                                         cfs_hash_bd_unlock(hs, &bd, excl);
1464                                         goto out;
1465                                 }
1466                         }
1467                 }
1468                 cfs_hash_bd_unlock(hs, &bd, excl);
1469                 if (loop < CFS_HASH_LOOP_HOG)
1470                         continue;
1471                 loop = 0;
1472                 cfs_hash_unlock(hs, 0);
1473                 cond_resched();
1474                 cfs_hash_lock(hs, 0);
1475         }
1476  out:
1477         cfs_hash_unlock(hs, 0);
1478
1479         cfs_hash_for_each_exit(hs);
1480         return count;
1481 }
1482
1483 typedef struct {
1484         cfs_hash_cond_opt_cb_t  func;
1485         void               *arg;
1486 } cfs_hash_cond_arg_t;
1487
1488 static int
1489 cfs_hash_cond_del_locked(struct cfs_hash *hs, struct cfs_hash_bd *bd,
1490                          struct hlist_node *hnode, void *data)
1491 {
1492         cfs_hash_cond_arg_t *cond = data;
1493
1494         if (cond->func(cfs_hash_object(hs, hnode), cond->arg))
1495                 cfs_hash_bd_del_locked(hs, bd, hnode);
1496         return 0;
1497 }
1498
1499 /**
1500  * Delete item from the libcfs hash @hs when @func return true.
1501  * The write lock being hold during loop for each bucket to avoid
1502  * any object be reference.
1503  */
1504 void
1505 cfs_hash_cond_del(struct cfs_hash *hs, cfs_hash_cond_opt_cb_t func, void *data)
1506 {
1507         cfs_hash_cond_arg_t arg = {
1508                 .func   = func,
1509                 .arg    = data,
1510         };
1511
1512         cfs_hash_for_each_tight(hs, cfs_hash_cond_del_locked, &arg, 1);
1513 }
1514 EXPORT_SYMBOL(cfs_hash_cond_del);
1515
1516 void
1517 cfs_hash_for_each(struct cfs_hash *hs,
1518                   cfs_hash_for_each_cb_t func, void *data)
1519 {
1520         cfs_hash_for_each_tight(hs, func, data, 0);
1521 }
1522 EXPORT_SYMBOL(cfs_hash_for_each);
1523
1524 void
1525 cfs_hash_for_each_safe(struct cfs_hash *hs,
1526                        cfs_hash_for_each_cb_t func, void *data) {
1527         cfs_hash_for_each_tight(hs, func, data, 1);
1528 }
1529 EXPORT_SYMBOL(cfs_hash_for_each_safe);
1530
1531 static int
1532 cfs_hash_peek(struct cfs_hash *hs, struct cfs_hash_bd *bd,
1533               struct hlist_node *hnode, void *data)
1534 {
1535         *(int *)data = 0;
1536         return 1; /* return 1 to break the loop */
1537 }
1538
1539 int
1540 cfs_hash_is_empty(struct cfs_hash *hs)
1541 {
1542         int empty = 1;
1543
1544         cfs_hash_for_each_tight(hs, cfs_hash_peek, &empty, 0);
1545         return empty;
1546 }
1547 EXPORT_SYMBOL(cfs_hash_is_empty);
1548
1549 __u64
1550 cfs_hash_size_get(struct cfs_hash *hs)
1551 {
1552         return cfs_hash_with_counter(hs) ?
1553                atomic_read(&hs->hs_count) :
1554                cfs_hash_for_each_tight(hs, NULL, NULL, 0);
1555 }
1556 EXPORT_SYMBOL(cfs_hash_size_get);
1557
1558 /*
1559  * cfs_hash_for_each_relax:
1560  * Iterate the hash table and call @func on each item without
1561  * any lock. This function can't guarantee to finish iteration
1562  * if these features are enabled:
1563  *
1564  *  a. if rehash_key is enabled, an item can be moved from
1565  *     one bucket to another bucket
1566  *  b. user can remove non-zero-ref item from hash-table,
1567  *     so the item can be removed from hash-table, even worse,
1568  *     it's possible that user changed key and insert to another
1569  *     hash bucket.
1570  * there's no way for us to finish iteration correctly on previous
1571  * two cases, so iteration has to be stopped on change.
1572  */
1573 static int
1574 cfs_hash_for_each_relax(struct cfs_hash *hs, cfs_hash_for_each_cb_t func,
1575                         void *data) {
1576         struct hlist_node *hnode;
1577         struct hlist_node *tmp;
1578         struct cfs_hash_bd     bd;
1579         __u32        version;
1580         int            count = 0;
1581         int            stop_on_change;
1582         int            rc;
1583         int            i;
1584
1585         stop_on_change = cfs_hash_with_rehash_key(hs) ||
1586                          !cfs_hash_with_no_itemref(hs) ||
1587                          hs->hs_ops->hs_put_locked == NULL;
1588         cfs_hash_lock(hs, 0);
1589         LASSERT(!cfs_hash_is_rehashing(hs));
1590
1591         cfs_hash_for_each_bucket(hs, &bd, i) {
1592                 struct hlist_head *hhead;
1593
1594                 cfs_hash_bd_lock(hs, &bd, 0);
1595                 version = cfs_hash_bd_version_get(&bd);
1596
1597                 cfs_hash_bd_for_each_hlist(hs, &bd, hhead) {
1598                         for (hnode = hhead->first; hnode != NULL;) {
1599                                 cfs_hash_bucket_validate(hs, &bd, hnode);
1600                                 cfs_hash_get(hs, hnode);
1601                                 cfs_hash_bd_unlock(hs, &bd, 0);
1602                                 cfs_hash_unlock(hs, 0);
1603
1604                                 rc = func(hs, &bd, hnode, data);
1605                                 if (stop_on_change)
1606                                         cfs_hash_put(hs, hnode);
1607                                 cond_resched();
1608                                 count++;
1609
1610                                 cfs_hash_lock(hs, 0);
1611                                 cfs_hash_bd_lock(hs, &bd, 0);
1612                                 if (!stop_on_change) {
1613                                         tmp = hnode->next;
1614                                         cfs_hash_put_locked(hs, hnode);
1615                                         hnode = tmp;
1616                                 } else { /* bucket changed? */
1617                                         if (version !=
1618                                             cfs_hash_bd_version_get(&bd))
1619                                                 break;
1620                                         /* safe to continue because no change */
1621                                         hnode = hnode->next;
1622                                 }
1623                                 if (rc) /* callback wants to break iteration */
1624                                         break;
1625                         }
1626                 }
1627                 cfs_hash_bd_unlock(hs, &bd, 0);
1628         }
1629         cfs_hash_unlock(hs, 0);
1630
1631         return count;
1632 }
1633
1634 int
1635 cfs_hash_for_each_nolock(struct cfs_hash *hs,
1636                          cfs_hash_for_each_cb_t func, void *data) {
1637         if (cfs_hash_with_no_lock(hs) ||
1638             cfs_hash_with_rehash_key(hs) ||
1639             !cfs_hash_with_no_itemref(hs))
1640                 return -EOPNOTSUPP;
1641
1642         if (hs->hs_ops->hs_get == NULL ||
1643             (hs->hs_ops->hs_put == NULL &&
1644              hs->hs_ops->hs_put_locked == NULL))
1645                 return -EOPNOTSUPP;
1646
1647         cfs_hash_for_each_enter(hs);
1648         cfs_hash_for_each_relax(hs, func, data);
1649         cfs_hash_for_each_exit(hs);
1650
1651         return 0;
1652 }
1653 EXPORT_SYMBOL(cfs_hash_for_each_nolock);
1654
1655 /**
1656  * For each hash bucket in the libcfs hash @hs call the passed callback
1657  * @func until all the hash buckets are empty.  The passed callback @func
1658  * or the previously registered callback hs->hs_put must remove the item
1659  * from the hash.  You may either use the cfs_hash_del() or hlist_del()
1660  * functions.  No rwlocks will be held during the callback @func it is
1661  * safe to sleep if needed.  This function will not terminate until the
1662  * hash is empty.  Note it is still possible to concurrently add new
1663  * items in to the hash.  It is the callers responsibility to ensure
1664  * the required locking is in place to prevent concurrent insertions.
1665  */
1666 int
1667 cfs_hash_for_each_empty(struct cfs_hash *hs,
1668                         cfs_hash_for_each_cb_t func, void *data) {
1669         unsigned  i = 0;
1670
1671         if (cfs_hash_with_no_lock(hs))
1672                 return -EOPNOTSUPP;
1673
1674         if (hs->hs_ops->hs_get == NULL ||
1675             (hs->hs_ops->hs_put == NULL &&
1676              hs->hs_ops->hs_put_locked == NULL))
1677                 return -EOPNOTSUPP;
1678
1679         cfs_hash_for_each_enter(hs);
1680         while (cfs_hash_for_each_relax(hs, func, data)) {
1681                 CDEBUG(D_INFO, "Try to empty hash: %s, loop: %u\n",
1682                        hs->hs_name, i++);
1683         }
1684         cfs_hash_for_each_exit(hs);
1685         return 0;
1686 }
1687 EXPORT_SYMBOL(cfs_hash_for_each_empty);
1688
1689 void
1690 cfs_hash_hlist_for_each(struct cfs_hash *hs, unsigned hindex,
1691                         cfs_hash_for_each_cb_t func, void *data)
1692 {
1693         struct hlist_head   *hhead;
1694         struct hlist_node   *hnode;
1695         struct cfs_hash_bd       bd;
1696
1697         cfs_hash_for_each_enter(hs);
1698         cfs_hash_lock(hs, 0);
1699         if (hindex >= CFS_HASH_NHLIST(hs))
1700                 goto out;
1701
1702         cfs_hash_bd_index_set(hs, hindex, &bd);
1703
1704         cfs_hash_bd_lock(hs, &bd, 0);
1705         hhead = cfs_hash_bd_hhead(hs, &bd);
1706         hlist_for_each(hnode, hhead) {
1707                 if (func(hs, &bd, hnode, data))
1708                         break;
1709         }
1710         cfs_hash_bd_unlock(hs, &bd, 0);
1711  out:
1712         cfs_hash_unlock(hs, 0);
1713         cfs_hash_for_each_exit(hs);
1714 }
1715
1716 EXPORT_SYMBOL(cfs_hash_hlist_for_each);
1717
1718 /*
1719  * For each item in the libcfs hash @hs which matches the @key call
1720  * the passed callback @func and pass to it as an argument each hash
1721  * item and the private @data. During the callback the bucket lock
1722  * is held so the callback must never sleep.
1723    */
1724 void
1725 cfs_hash_for_each_key(struct cfs_hash *hs, const void *key,
1726                       cfs_hash_for_each_cb_t func, void *data) {
1727         struct hlist_node   *hnode;
1728         struct cfs_hash_bd       bds[2];
1729         unsigned            i;
1730
1731         cfs_hash_lock(hs, 0);
1732
1733         cfs_hash_dual_bd_get_and_lock(hs, key, bds, 0);
1734
1735         cfs_hash_for_each_bd(bds, 2, i) {
1736                 struct hlist_head *hlist = cfs_hash_bd_hhead(hs, &bds[i]);
1737
1738                 hlist_for_each(hnode, hlist) {
1739                         cfs_hash_bucket_validate(hs, &bds[i], hnode);
1740
1741                         if (cfs_hash_keycmp(hs, key, hnode)) {
1742                                 if (func(hs, &bds[i], hnode, data))
1743                                         break;
1744                         }
1745                 }
1746         }
1747
1748         cfs_hash_dual_bd_unlock(hs, bds, 0);
1749         cfs_hash_unlock(hs, 0);
1750 }
1751 EXPORT_SYMBOL(cfs_hash_for_each_key);
1752
1753 /**
1754  * Rehash the libcfs hash @hs to the given @bits.  This can be used
1755  * to grow the hash size when excessive chaining is detected, or to
1756  * shrink the hash when it is larger than needed.  When the CFS_HASH_REHASH
1757  * flag is set in @hs the libcfs hash may be dynamically rehashed
1758  * during addition or removal if the hash's theta value exceeds
1759  * either the hs->hs_min_theta or hs->max_theta values.  By default
1760  * these values are tuned to keep the chained hash depth small, and
1761  * this approach assumes a reasonably uniform hashing function.  The
1762  * theta thresholds for @hs are tunable via cfs_hash_set_theta().
1763  */
1764 void
1765 cfs_hash_rehash_cancel_locked(struct cfs_hash *hs)
1766 {
1767         int     i;
1768
1769         /* need hold cfs_hash_lock(hs, 1) */
1770         LASSERT(cfs_hash_with_rehash(hs) &&
1771                 !cfs_hash_with_no_lock(hs));
1772
1773         if (!cfs_hash_is_rehashing(hs))
1774                 return;
1775
1776         if (cfs_wi_deschedule(cfs_sched_rehash, &hs->hs_rehash_wi)) {
1777                 hs->hs_rehash_bits = 0;
1778                 return;
1779         }
1780
1781         for (i = 2; cfs_hash_is_rehashing(hs); i++) {
1782                 cfs_hash_unlock(hs, 1);
1783                 /* raise console warning while waiting too long */
1784                 CDEBUG(IS_PO2(i >> 3) ? D_WARNING : D_INFO,
1785                        "hash %s is still rehashing, rescheded %d\n",
1786                        hs->hs_name, i - 1);
1787                 cond_resched();
1788                 cfs_hash_lock(hs, 1);
1789         }
1790 }
1791 EXPORT_SYMBOL(cfs_hash_rehash_cancel_locked);
1792
1793 void
1794 cfs_hash_rehash_cancel(struct cfs_hash *hs)
1795 {
1796         cfs_hash_lock(hs, 1);
1797         cfs_hash_rehash_cancel_locked(hs);
1798         cfs_hash_unlock(hs, 1);
1799 }
1800 EXPORT_SYMBOL(cfs_hash_rehash_cancel);
1801
1802 int
1803 cfs_hash_rehash(struct cfs_hash *hs, int do_rehash)
1804 {
1805         int     rc;
1806
1807         LASSERT(cfs_hash_with_rehash(hs) && !cfs_hash_with_no_lock(hs));
1808
1809         cfs_hash_lock(hs, 1);
1810
1811         rc = cfs_hash_rehash_bits(hs);
1812         if (rc <= 0) {
1813                 cfs_hash_unlock(hs, 1);
1814                 return rc;
1815         }
1816
1817         hs->hs_rehash_bits = rc;
1818         if (!do_rehash) {
1819                 /* launch and return */
1820                 cfs_wi_schedule(cfs_sched_rehash, &hs->hs_rehash_wi);
1821                 cfs_hash_unlock(hs, 1);
1822                 return 0;
1823         }
1824
1825         /* rehash right now */
1826         cfs_hash_unlock(hs, 1);
1827
1828         return cfs_hash_rehash_worker(&hs->hs_rehash_wi);
1829 }
1830 EXPORT_SYMBOL(cfs_hash_rehash);
1831
1832 static int
1833 cfs_hash_rehash_bd(struct cfs_hash *hs, struct cfs_hash_bd *old)
1834 {
1835         struct cfs_hash_bd      new;
1836         struct hlist_head  *hhead;
1837         struct hlist_node  *hnode;
1838         struct hlist_node  *pos;
1839         void          *key;
1840         int             c = 0;
1841
1842         /* hold cfs_hash_lock(hs, 1), so don't need any bucket lock */
1843         cfs_hash_bd_for_each_hlist(hs, old, hhead) {
1844                 hlist_for_each_safe(hnode, pos, hhead) {
1845                         key = cfs_hash_key(hs, hnode);
1846                         LASSERT(key != NULL);
1847                         /* Validate hnode is in the correct bucket. */
1848                         cfs_hash_bucket_validate(hs, old, hnode);
1849                         /*
1850                          * Delete from old hash bucket; move to new bucket.
1851                          * ops->hs_key must be defined.
1852                          */
1853                         cfs_hash_bd_from_key(hs, hs->hs_rehash_buckets,
1854                                              hs->hs_rehash_bits, key, &new);
1855                         cfs_hash_bd_move_locked(hs, old, &new, hnode);
1856                         c++;
1857                 }
1858         }
1859
1860         return c;
1861 }
1862
1863 static int
1864 cfs_hash_rehash_worker(cfs_workitem_t *wi)
1865 {
1866         struct cfs_hash  *hs = container_of(wi, struct cfs_hash, hs_rehash_wi);
1867         struct cfs_hash_bucket **bkts;
1868         struct cfs_hash_bd       bd;
1869         unsigned int    old_size;
1870         unsigned int    new_size;
1871         int              bsize;
1872         int              count = 0;
1873         int              rc = 0;
1874         int              i;
1875
1876         LASSERT (hs != NULL && cfs_hash_with_rehash(hs));
1877
1878         cfs_hash_lock(hs, 0);
1879         LASSERT(cfs_hash_is_rehashing(hs));
1880
1881         old_size = CFS_HASH_NBKT(hs);
1882         new_size = CFS_HASH_RH_NBKT(hs);
1883
1884         cfs_hash_unlock(hs, 0);
1885
1886         /*
1887          * don't need hs::hs_rwlock for hs::hs_buckets,
1888          * because nobody can change bkt-table except me.
1889          */
1890         bkts = cfs_hash_buckets_realloc(hs, hs->hs_buckets,
1891                                         old_size, new_size);
1892         cfs_hash_lock(hs, 1);
1893         if (bkts == NULL) {
1894                 rc = -ENOMEM;
1895                 goto out;
1896         }
1897
1898         if (bkts == hs->hs_buckets) {
1899                 bkts = NULL; /* do nothing */
1900                 goto out;
1901         }
1902
1903         rc = __cfs_hash_theta(hs);
1904         if ((rc >= hs->hs_min_theta) && (rc <= hs->hs_max_theta)) {
1905                 /* free the new allocated bkt-table */
1906                 old_size = new_size;
1907                 new_size = CFS_HASH_NBKT(hs);
1908                 rc = -EALREADY;
1909                 goto out;
1910         }
1911
1912         LASSERT(hs->hs_rehash_buckets == NULL);
1913         hs->hs_rehash_buckets = bkts;
1914
1915         rc = 0;
1916         cfs_hash_for_each_bucket(hs, &bd, i) {
1917                 if (cfs_hash_is_exiting(hs)) {
1918                         rc = -ESRCH;
1919                         /* someone wants to destroy the hash, abort now */
1920                         if (old_size < new_size) /* OK to free old bkt-table */
1921                                 break;
1922                         /* it's shrinking, need free new bkt-table */
1923                         hs->hs_rehash_buckets = NULL;
1924                         old_size = new_size;
1925                         new_size = CFS_HASH_NBKT(hs);
1926                         goto out;
1927                 }
1928
1929                 count += cfs_hash_rehash_bd(hs, &bd);
1930                 if (count < CFS_HASH_LOOP_HOG ||
1931                     cfs_hash_is_iterating(hs)) { /* need to finish ASAP */
1932                         continue;
1933                 }
1934
1935                 count = 0;
1936                 cfs_hash_unlock(hs, 1);
1937                 cond_resched();
1938                 cfs_hash_lock(hs, 1);
1939         }
1940
1941         hs->hs_rehash_count++;
1942
1943         bkts = hs->hs_buckets;
1944         hs->hs_buckets = hs->hs_rehash_buckets;
1945         hs->hs_rehash_buckets = NULL;
1946
1947         hs->hs_cur_bits = hs->hs_rehash_bits;
1948  out:
1949         hs->hs_rehash_bits = 0;
1950         if (rc == -ESRCH) /* never be scheduled again */
1951                 cfs_wi_exit(cfs_sched_rehash, wi);
1952         bsize = cfs_hash_bkt_size(hs);
1953         cfs_hash_unlock(hs, 1);
1954         /* can't refer to @hs anymore because it could be destroyed */
1955         if (bkts != NULL)
1956                 cfs_hash_buckets_free(bkts, bsize, new_size, old_size);
1957         if (rc != 0)
1958                 CDEBUG(D_INFO, "early quit of rehashing: %d\n", rc);
1959         /* return 1 only if cfs_wi_exit is called */
1960         return rc == -ESRCH;
1961 }
1962
1963 /**
1964  * Rehash the object referenced by @hnode in the libcfs hash @hs.  The
1965  * @old_key must be provided to locate the objects previous location
1966  * in the hash, and the @new_key will be used to reinsert the object.
1967  * Use this function instead of a cfs_hash_add() + cfs_hash_del()
1968  * combo when it is critical that there is no window in time where the
1969  * object is missing from the hash.  When an object is being rehashed
1970  * the registered cfs_hash_get() and cfs_hash_put() functions will
1971  * not be called.
1972  */
1973 void cfs_hash_rehash_key(struct cfs_hash *hs, const void *old_key,
1974                          void *new_key, struct hlist_node *hnode)
1975 {
1976         struct cfs_hash_bd      bds[3];
1977         struct cfs_hash_bd      old_bds[2];
1978         struct cfs_hash_bd      new_bd;
1979
1980         LASSERT(!hlist_unhashed(hnode));
1981
1982         cfs_hash_lock(hs, 0);
1983
1984         cfs_hash_dual_bd_get(hs, old_key, old_bds);
1985         cfs_hash_bd_get(hs, new_key, &new_bd);
1986
1987         bds[0] = old_bds[0];
1988         bds[1] = old_bds[1];
1989         bds[2] = new_bd;
1990
1991         /* NB: bds[0] and bds[1] are ordered already */
1992         cfs_hash_bd_order(&bds[1], &bds[2]);
1993         cfs_hash_bd_order(&bds[0], &bds[1]);
1994
1995         cfs_hash_multi_bd_lock(hs, bds, 3, 1);
1996         if (likely(old_bds[1].bd_bucket == NULL)) {
1997                 cfs_hash_bd_move_locked(hs, &old_bds[0], &new_bd, hnode);
1998         } else {
1999                 cfs_hash_dual_bd_finddel_locked(hs, old_bds, old_key, hnode);
2000                 cfs_hash_bd_add_locked(hs, &new_bd, hnode);
2001         }
2002         /* overwrite key inside locks, otherwise may screw up with
2003          * other operations, i.e: rehash */
2004         cfs_hash_keycpy(hs, new_key, hnode);
2005
2006         cfs_hash_multi_bd_unlock(hs, bds, 3, 1);
2007         cfs_hash_unlock(hs, 0);
2008 }
2009 EXPORT_SYMBOL(cfs_hash_rehash_key);
2010
2011 void cfs_hash_debug_header(struct seq_file *m)
2012 {
2013         seq_printf(m, "%-*s   cur   min   max theta t-min t-max flags rehash   count  maxdep maxdepb distribution\n",
2014                    CFS_HASH_BIGNAME_LEN, "name");
2015 }
2016 EXPORT_SYMBOL(cfs_hash_debug_header);
2017
2018 static struct cfs_hash_bucket **
2019 cfs_hash_full_bkts(struct cfs_hash *hs)
2020 {
2021         /* NB: caller should hold hs->hs_rwlock if REHASH is set */
2022         if (hs->hs_rehash_buckets == NULL)
2023                 return hs->hs_buckets;
2024
2025         LASSERT(hs->hs_rehash_bits != 0);
2026         return hs->hs_rehash_bits > hs->hs_cur_bits ?
2027                hs->hs_rehash_buckets : hs->hs_buckets;
2028 }
2029
2030 static unsigned int
2031 cfs_hash_full_nbkt(struct cfs_hash *hs)
2032 {
2033         /* NB: caller should hold hs->hs_rwlock if REHASH is set */
2034         if (hs->hs_rehash_buckets == NULL)
2035                 return CFS_HASH_NBKT(hs);
2036
2037         LASSERT(hs->hs_rehash_bits != 0);
2038         return hs->hs_rehash_bits > hs->hs_cur_bits ?
2039                CFS_HASH_RH_NBKT(hs) : CFS_HASH_NBKT(hs);
2040 }
2041
2042 void cfs_hash_debug_str(struct cfs_hash *hs, struct seq_file *m)
2043 {
2044         int                 dist[8] = { 0, };
2045         int                 maxdep  = -1;
2046         int                 maxdepb = -1;
2047         int                 total   = 0;
2048         int                 theta;
2049         int                 i;
2050
2051         cfs_hash_lock(hs, 0);
2052         theta = __cfs_hash_theta(hs);
2053
2054         seq_printf(m, "%-*s %5d %5d %5d %d.%03d %d.%03d %d.%03d  0x%02x %6d ",
2055                       CFS_HASH_BIGNAME_LEN, hs->hs_name,
2056                       1 << hs->hs_cur_bits, 1 << hs->hs_min_bits,
2057                       1 << hs->hs_max_bits,
2058                       __cfs_hash_theta_int(theta), __cfs_hash_theta_frac(theta),
2059                       __cfs_hash_theta_int(hs->hs_min_theta),
2060                       __cfs_hash_theta_frac(hs->hs_min_theta),
2061                       __cfs_hash_theta_int(hs->hs_max_theta),
2062                       __cfs_hash_theta_frac(hs->hs_max_theta),
2063                       hs->hs_flags, hs->hs_rehash_count);
2064
2065         /*
2066          * The distribution is a summary of the chained hash depth in
2067          * each of the libcfs hash buckets.  Each buckets hsb_count is
2068          * divided by the hash theta value and used to generate a
2069          * histogram of the hash distribution.  A uniform hash will
2070          * result in all hash buckets being close to the average thus
2071          * only the first few entries in the histogram will be non-zero.
2072          * If you hash function results in a non-uniform hash the will
2073          * be observable by outlier bucks in the distribution histogram.
2074          *
2075          * Uniform hash distribution:      128/128/0/0/0/0/0/0
2076          * Non-Uniform hash distribution:  128/125/0/0/0/0/2/1
2077          */
2078         for (i = 0; i < cfs_hash_full_nbkt(hs); i++) {
2079                 struct cfs_hash_bd  bd;
2080
2081                 bd.bd_bucket = cfs_hash_full_bkts(hs)[i];
2082                 cfs_hash_bd_lock(hs, &bd, 0);
2083                 if (maxdep < bd.bd_bucket->hsb_depmax) {
2084                         maxdep  = bd.bd_bucket->hsb_depmax;
2085                         maxdepb = ffz(~maxdep);
2086                 }
2087                 total += bd.bd_bucket->hsb_count;
2088                 dist[min(fls(bd.bd_bucket->hsb_count / max(theta, 1)), 7)]++;
2089                 cfs_hash_bd_unlock(hs, &bd, 0);
2090         }
2091
2092         seq_printf(m, "%7d %7d %7d ", total, maxdep, maxdepb);
2093         for (i = 0; i < 8; i++)
2094                 seq_printf(m, "%d%c",  dist[i], (i == 7) ? '\n' : '/');
2095
2096         cfs_hash_unlock(hs, 0);
2097 }
2098 EXPORT_SYMBOL(cfs_hash_debug_str);