1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/obdclass/class_hash.c
38 * Implement a hash class for hash process in lustre system.
40 * Author: YuZhangyong <yzy@clusterfs.com>
42 * 2008-08-15: Brian Behlendorf <behlendorf1@llnl.gov>
43 * - Simplified API and improved documentation
44 * - Added per-hash feature flags:
45 * * LH_DEBUG additional validation
46 * * LH_REHASH dynamic rehashing
47 * - Added per-hash statistics
48 * - General performance enhancements
52 #include <liblustre.h>
56 #include <class_hash.h>
59 lh_read_lock(lustre_hash_t *lh)
61 if ((lh->lh_flags & LH_REHASH) != 0)
62 read_lock(&lh->lh_rwlock);
66 lh_read_unlock(lustre_hash_t *lh)
68 if ((lh->lh_flags & LH_REHASH) != 0)
69 read_unlock(&lh->lh_rwlock);
73 lh_write_lock(lustre_hash_t *lh)
75 if ((lh->lh_flags & LH_REHASH) != 0)
76 write_lock(&lh->lh_rwlock);
80 lh_write_unlock(lustre_hash_t *lh)
82 if ((lh->lh_flags & LH_REHASH) != 0)
83 write_unlock(&lh->lh_rwlock);
87 * Initialize new lustre hash, where:
88 * @name - Descriptive hash name
89 * @cur_bits - Initial hash table size, in bits
90 * @max_bits - Maximum allowed hash table resize, in bits
91 * @ops - Registered hash table operations
92 * @flags - LH_REHASH enable synamic hash resizing
93 * - LH_SORT enable chained hash sort
96 lustre_hash_init(char *name, unsigned int cur_bits, unsigned int max_bits,
97 lustre_hash_ops_t *ops, int flags)
103 LASSERT(name != NULL);
104 LASSERT(ops != NULL);
106 LASSERT(cur_bits > 0);
107 LASSERT(max_bits >= cur_bits);
108 LASSERT(max_bits < 31);
110 LIBCFS_ALLOC_PTR(lh);
114 strncpy(lh->lh_name, name, sizeof(lh->lh_name));
115 lh->lh_name[sizeof(lh->lh_name) - 1] = '\0';
116 atomic_set(&lh->lh_rehash_count, 0);
117 atomic_set(&lh->lh_count, 0);
118 rwlock_init(&lh->lh_rwlock);
119 lh->lh_cur_bits = cur_bits;
120 lh->lh_cur_mask = (1 << cur_bits) - 1;
121 lh->lh_min_bits = cur_bits;
122 lh->lh_max_bits = max_bits;
123 /* XXX: need to fixup lustre_hash_rehash_bits() before this can be
124 * anything other than 0.5 and 2.0 */
125 lh->lh_min_theta = 1 << (LH_THETA_BITS - 1);
126 lh->lh_max_theta = 1 << (LH_THETA_BITS + 1);
128 lh->lh_flags = flags;
129 if (cur_bits != max_bits && (lh->lh_flags & LH_REHASH) == 0)
130 CERROR("Rehash is disabled for %s, ignore max_bits %d\n",
134 __lustre_hash_set_theta(lh, 500, 2000);
136 LIBCFS_ALLOC(lh->lh_buckets, sizeof(*lh->lh_buckets) << lh->lh_cur_bits);
137 if (!lh->lh_buckets) {
142 for (i = 0; i <= lh->lh_cur_mask; i++) {
143 LIBCFS_ALLOC(lh->lh_buckets[i], sizeof(lustre_hash_bucket_t));
144 if (lh->lh_buckets[i] == NULL) {
145 lustre_hash_exit(lh);
149 INIT_HLIST_HEAD(&lh->lh_buckets[i]->lhb_head);
150 rwlock_init(&lh->lh_buckets[i]->lhb_rwlock);
151 atomic_set(&lh->lh_buckets[i]->lhb_count, 0);
156 EXPORT_SYMBOL(lustre_hash_init);
159 * Cleanup lustre hash @lh.
162 lustre_hash_exit(lustre_hash_t *lh)
164 lustre_hash_bucket_t *lhb;
165 struct hlist_node *hnode;
166 struct hlist_node *pos;
174 lh_for_each_bucket(lh, lhb, i) {
178 write_lock(&lhb->lhb_rwlock);
179 hlist_for_each_safe(hnode, pos, &(lhb->lhb_head)) {
180 __lustre_hash_bucket_validate(lh, lhb, hnode);
181 __lustre_hash_bucket_del(lh, lhb, hnode);
185 LASSERT(hlist_empty(&(lhb->lhb_head)));
186 LASSERT(atomic_read(&lhb->lhb_count) == 0);
187 write_unlock(&lhb->lhb_rwlock);
188 LIBCFS_FREE_PTR(lhb);
191 LASSERT(atomic_read(&lh->lh_count) == 0);
194 LIBCFS_FREE(lh->lh_buckets, sizeof(*lh->lh_buckets) << lh->lh_cur_bits);
198 EXPORT_SYMBOL(lustre_hash_exit);
200 static inline unsigned int lustre_hash_rehash_bits(lustre_hash_t *lh)
202 if (!(lh->lh_flags & LH_REHASH))
205 /* XXX: need to handle case with max_theta != 2.0
206 * and the case with min_theta != 0.5 */
207 if ((lh->lh_cur_bits < lh->lh_max_bits) &&
208 (__lustre_hash_theta(lh) > lh->lh_max_theta))
209 return lh->lh_cur_bits + 1;
211 if ((lh->lh_cur_bits > lh->lh_min_bits) &&
212 (__lustre_hash_theta(lh) < lh->lh_min_theta))
213 return lh->lh_cur_bits - 1;
219 * Add item @hnode to lustre hash @lh using @key. The registered
220 * ops->lh_get function will be called when the item is added.
223 lustre_hash_add(lustre_hash_t *lh, void *key, struct hlist_node *hnode)
225 lustre_hash_bucket_t *lhb;
230 __lustre_hash_key_validate(lh, key, hnode);
233 i = lh_hash(lh, key, lh->lh_cur_mask);
234 lhb = lh->lh_buckets[i];
235 LASSERT(i <= lh->lh_cur_mask);
236 LASSERT(hlist_unhashed(hnode));
238 write_lock(&lhb->lhb_rwlock);
239 __lustre_hash_bucket_add(lh, lhb, hnode);
240 write_unlock(&lhb->lhb_rwlock);
242 bits = lustre_hash_rehash_bits(lh);
245 lustre_hash_rehash(lh, bits);
249 EXPORT_SYMBOL(lustre_hash_add);
251 static struct hlist_node *
252 lustre_hash_findadd_unique_hnode(lustre_hash_t *lh, void *key,
253 struct hlist_node *hnode)
256 struct hlist_node *ehnode;
257 lustre_hash_bucket_t *lhb;
261 __lustre_hash_key_validate(lh, key, hnode);
264 i = lh_hash(lh, key, lh->lh_cur_mask);
265 lhb = lh->lh_buckets[i];
266 LASSERT(i <= lh->lh_cur_mask);
267 LASSERT(hlist_unhashed(hnode));
269 write_lock(&lhb->lhb_rwlock);
270 ehnode = __lustre_hash_bucket_lookup(lh, lhb, key);
274 __lustre_hash_bucket_add(lh, lhb, hnode);
276 bits = lustre_hash_rehash_bits(lh);
278 write_unlock(&lhb->lhb_rwlock);
281 lustre_hash_rehash(lh, bits);
287 * Add item @hnode to lustre hash @lh using @key. The registered
288 * ops->lh_get function will be called if the item was added.
289 * Returns 0 on success or -EALREADY on key collisions.
292 lustre_hash_add_unique(lustre_hash_t *lh, void *key, struct hlist_node *hnode)
294 struct hlist_node *ehnode;
297 ehnode = lustre_hash_findadd_unique_hnode(lh, key, hnode);
298 if (ehnode != hnode) {
304 EXPORT_SYMBOL(lustre_hash_add_unique);
307 * Add item @hnode to lustre hash @lh using @key. If this @key
308 * already exists in the hash then ops->lh_get will be called on the
309 * conflicting entry and that entry will be returned to the caller.
310 * Otherwise ops->lh_get is called on the item which was added.
313 lustre_hash_findadd_unique(lustre_hash_t *lh, void *key,
314 struct hlist_node *hnode)
316 struct hlist_node *ehnode;
320 ehnode = lustre_hash_findadd_unique_hnode(lh, key, hnode);
321 obj = lh_get(lh, ehnode);
325 EXPORT_SYMBOL(lustre_hash_findadd_unique);
328 * Delete item @hnode from the lustre hash @lh using @key. The @key
329 * is required to ensure the correct hash bucket is locked since there
330 * is no direct linkage from the item to the bucket. The object
331 * removed from the hash will be returned and obs->lh_put is called
332 * on the removed object.
335 lustre_hash_del(lustre_hash_t *lh, void *key, struct hlist_node *hnode)
337 lustre_hash_bucket_t *lhb;
342 __lustre_hash_key_validate(lh, key, hnode);
345 i = lh_hash(lh, key, lh->lh_cur_mask);
346 lhb = lh->lh_buckets[i];
347 LASSERT(i <= lh->lh_cur_mask);
348 LASSERT(!hlist_unhashed(hnode));
350 write_lock(&lhb->lhb_rwlock);
351 obj = __lustre_hash_bucket_del(lh, lhb, hnode);
352 write_unlock(&lhb->lhb_rwlock);
357 EXPORT_SYMBOL(lustre_hash_del);
360 * Delete item given @key in lustre hash @lh. The first @key found in
361 * the hash will be removed, if the key exists multiple times in the hash
362 * @lh this function must be called once per key. The removed object
363 * will be returned and ops->lh_put is called on the removed object.
366 lustre_hash_del_key(lustre_hash_t *lh, void *key)
368 struct hlist_node *hnode;
369 lustre_hash_bucket_t *lhb;
375 i = lh_hash(lh, key, lh->lh_cur_mask);
376 lhb = lh->lh_buckets[i];
377 LASSERT(i <= lh->lh_cur_mask);
379 write_lock(&lhb->lhb_rwlock);
380 hnode = __lustre_hash_bucket_lookup(lh, lhb, key);
382 obj = __lustre_hash_bucket_del(lh, lhb, hnode);
384 write_unlock(&lhb->lhb_rwlock);
389 EXPORT_SYMBOL(lustre_hash_del_key);
392 * Lookup an item using @key in the lustre hash @lh and return it.
393 * If the @key is found in the hash lh->lh_get() is called and the
394 * matching objects is returned. It is the callers responsibility
395 * to call the counterpart ops->lh_put using the lh_put() macro
396 * when when finished with the object. If the @key was not found
397 * in the hash @lh NULL is returned.
400 lustre_hash_lookup(lustre_hash_t *lh, void *key)
402 struct hlist_node *hnode;
403 lustre_hash_bucket_t *lhb;
409 i = lh_hash(lh, key, lh->lh_cur_mask);
410 lhb = lh->lh_buckets[i];
411 LASSERT(i <= lh->lh_cur_mask);
413 read_lock(&lhb->lhb_rwlock);
414 hnode = __lustre_hash_bucket_lookup(lh, lhb, key);
416 obj = lh_get(lh, hnode);
418 read_unlock(&lhb->lhb_rwlock);
423 EXPORT_SYMBOL(lustre_hash_lookup);
426 * For each item in the lustre hash @lh call the passed callback @func
427 * and pass to it as an argument each hash item and the private @data.
428 * Before each callback ops->lh_get will be called, and after each
429 * callback ops->lh_put will be called. Finally, during the callback
430 * the bucket lock is held so the callback must never sleep.
433 lustre_hash_for_each(lustre_hash_t *lh, lh_for_each_cb func, void *data)
435 struct hlist_node *hnode;
436 lustre_hash_bucket_t *lhb;
442 lh_for_each_bucket(lh, lhb, i) {
443 read_lock(&lhb->lhb_rwlock);
444 hlist_for_each(hnode, &(lhb->lhb_head)) {
445 __lustre_hash_bucket_validate(lh, lhb, hnode);
446 obj = lh_get(lh, hnode);
448 (void)lh_put(lh, hnode);
450 read_unlock(&lhb->lhb_rwlock);
456 EXPORT_SYMBOL(lustre_hash_for_each);
459 * For each item in the lustre hash @lh call the passed callback @func
460 * and pass to it as an argument each hash item and the private @data.
461 * Before each callback ops->lh_get will be called, and after each
462 * callback ops->lh_put will be called. During the callback the
463 * bucket lock will not be held will allows for the current item
464 * to be removed from the hash during the callback. However, care
465 * should be taken to prevent other callers from operating on the
466 * hash concurrently or list corruption may occur.
469 lustre_hash_for_each_safe(lustre_hash_t *lh, lh_for_each_cb func, void *data)
471 struct hlist_node *hnode;
472 struct hlist_node *pos;
473 lustre_hash_bucket_t *lhb;
479 lh_for_each_bucket(lh, lhb, i) {
480 read_lock(&lhb->lhb_rwlock);
481 hlist_for_each_safe(hnode, pos, &(lhb->lhb_head)) {
482 __lustre_hash_bucket_validate(lh, lhb, hnode);
483 obj = lh_get(lh, hnode);
484 read_unlock(&lhb->lhb_rwlock);
486 read_lock(&lhb->lhb_rwlock);
487 (void)lh_put(lh, hnode);
489 read_unlock(&lhb->lhb_rwlock);
494 EXPORT_SYMBOL(lustre_hash_for_each_safe);
497 * For each hash bucket in the lustre hash @lh call the passed callback
498 * @func until all the hash buckets are empty. The passed callback @func
499 * or the previously registered callback lh->lh_put must remove the item
500 * from the hash. You may either use the lustre_hash_del() or hlist_del()
501 * functions. No rwlocks will be held during the callback @func it is
502 * safe to sleep if needed. This function will not terminate until the
503 * hash is empty. Note it is still possible to concurrently add new
504 * items in to the hash. It is the callers responsibility to ensure
505 * the required locking is in place to prevent concurrent insertions.
508 lustre_hash_for_each_empty(lustre_hash_t *lh, lh_for_each_cb func, void *data)
510 struct hlist_node *hnode;
511 lustre_hash_bucket_t *lhb;
512 lustre_hash_bucket_t **lhb_last = NULL;
519 /* If the hash table has changed since we last held lh_rwlock,
520 * we need to start traversing the list from the start. */
521 if (lh->lh_buckets != lhb_last) {
523 lhb_last = lh->lh_buckets;
525 lh_for_each_bucket_restart(lh, lhb, i) {
526 write_lock(&lhb->lhb_rwlock);
527 while (!hlist_empty(&lhb->lhb_head)) {
528 hnode = lhb->lhb_head.first;
529 __lustre_hash_bucket_validate(lh, lhb, hnode);
530 obj = lh_get(lh, hnode);
531 write_unlock(&lhb->lhb_rwlock);
534 (void)lh_put(lh, hnode);
538 write_unlock(&lhb->lhb_rwlock);
543 EXPORT_SYMBOL(lustre_hash_for_each_empty);
546 * For each item in the lustre hash @lh which matches the @key call
547 * the passed callback @func and pass to it as an argument each hash
548 * item and the private @data. Before each callback ops->lh_get will
549 * be called, and after each callback ops->lh_put will be called.
550 * Finally, during the callback the bucket lock is held so the
551 * callback must never sleep.
554 lustre_hash_for_each_key(lustre_hash_t *lh, void *key,
555 lh_for_each_cb func, void *data)
557 struct hlist_node *hnode;
558 lustre_hash_bucket_t *lhb;
563 i = lh_hash(lh, key, lh->lh_cur_mask);
564 lhb = lh->lh_buckets[i];
565 LASSERT(i <= lh->lh_cur_mask);
567 read_lock(&lhb->lhb_rwlock);
568 hlist_for_each(hnode, &(lhb->lhb_head)) {
569 __lustre_hash_bucket_validate(lh, lhb, hnode);
571 if (!lh_compare(lh, key, hnode))
574 func(lh_get(lh, hnode), data);
575 (void)lh_put(lh, hnode);
578 read_unlock(&lhb->lhb_rwlock);
583 EXPORT_SYMBOL(lustre_hash_for_each_key);
586 * Rehash the lustre hash @lh to the given @bits. This can be used
587 * to grow the hash size when excessive chaining is detected, or to
588 * shrink the hash when it is larger than needed. When the LH_REHASH
589 * flag is set in @lh the lustre hash may be dynamically rehashed
590 * during addition or removal if the hash's theta value exceeds
591 * either the lh->lh_min_theta or lh->max_theta values. By default
592 * these values are tuned to keep the chained hash depth small, and
593 * this approach assumes a reasonably uniform hashing function. The
594 * theta thresholds for @lh are tunable via lustre_hash_set_theta().
597 lustre_hash_rehash(lustre_hash_t *lh, int bits)
599 struct hlist_node *hnode;
600 struct hlist_node *pos;
601 lustre_hash_bucket_t **lh_buckets;
602 lustre_hash_bucket_t **rehash_buckets;
603 lustre_hash_bucket_t *lh_lhb;
604 lustre_hash_bucket_t *rehash_lhb;
609 int mask = (1 << bits) - 1;
614 LASSERT(!in_interrupt());
616 LASSERT((lh->lh_flags & LH_REHASH) != 0);
618 LIBCFS_ALLOC(rehash_buckets, sizeof(*rehash_buckets) << bits);
622 for (i = 0; i <= mask; i++) {
623 LIBCFS_ALLOC(rehash_buckets[i], sizeof(*rehash_buckets[i]));
624 if (rehash_buckets[i] == NULL)
625 GOTO(free, rc = -ENOMEM);
627 INIT_HLIST_HEAD(&rehash_buckets[i]->lhb_head);
628 rwlock_init(&rehash_buckets[i]->lhb_rwlock);
629 atomic_set(&rehash_buckets[i]->lhb_count, 0);
635 * Early return for multiple concurrent racing callers,
636 * ensure we only trigger the rehash if it is still needed.
638 theta = __lustre_hash_theta(lh);
639 if ((theta >= lh->lh_min_theta) && (theta <= lh->lh_max_theta)) {
641 GOTO(free, rc = -EALREADY);
644 lh_bits = lh->lh_cur_bits;
645 lh_buckets = lh->lh_buckets;
646 lh_mask = (1 << lh_bits) - 1;
648 lh->lh_cur_bits = bits;
649 lh->lh_cur_mask = (1 << bits) - 1;
650 lh->lh_buckets = rehash_buckets;
651 atomic_inc(&lh->lh_rehash_count);
653 for (i = 0; i <= lh_mask; i++) {
654 lh_lhb = lh_buckets[i];
656 write_lock(&lh_lhb->lhb_rwlock);
657 hlist_for_each_safe(hnode, pos, &(lh_lhb->lhb_head)) {
658 key = lh_key(lh, hnode);
662 * Validate hnode is in the correct bucket.
664 if (unlikely(lh->lh_flags & LH_DEBUG))
665 LASSERT(lh_hash(lh, key, lh_mask) == i);
668 * Delete from old hash bucket.
671 LASSERT(atomic_read(&lh_lhb->lhb_count) > 0);
672 atomic_dec(&lh_lhb->lhb_count);
675 * Add to rehash bucket, ops->lh_key must be defined.
677 rehash_lhb = rehash_buckets[lh_hash(lh, key, mask)];
678 hlist_add_head(hnode, &(rehash_lhb->lhb_head));
679 atomic_inc(&rehash_lhb->lhb_count);
682 LASSERT(hlist_empty(&(lh_lhb->lhb_head)));
683 LASSERT(atomic_read(&lh_lhb->lhb_count) == 0);
684 write_unlock(&lh_lhb->lhb_rwlock);
688 rehash_buckets = lh_buckets;
693 LIBCFS_FREE(rehash_buckets[i], sizeof(*rehash_buckets[i]));
694 LIBCFS_FREE(rehash_buckets, sizeof(*rehash_buckets) << bits);
698 EXPORT_SYMBOL(lustre_hash_rehash);
701 * Rehash the object referenced by @hnode in the lustre hash @lh. The
702 * @old_key must be provided to locate the objects previous location
703 * in the hash, and the @new_key will be used to reinsert the object.
704 * Use this function instead of a lustre_hash_add() + lustre_hash_del()
705 * combo when it is critical that there is no window in time where the
706 * object is missing from the hash. When an object is being rehashed
707 * the registered lh_get() and lh_put() functions will not be called.
709 void lustre_hash_rehash_key(lustre_hash_t *lh, void *old_key, void *new_key,
710 struct hlist_node *hnode)
712 lustre_hash_bucket_t *old_lhb;
713 lustre_hash_bucket_t *new_lhb;
718 __lustre_hash_key_validate(lh, new_key, hnode);
719 LASSERT(!hlist_unhashed(hnode));
723 i = lh_hash(lh, old_key, lh->lh_cur_mask);
724 old_lhb = lh->lh_buckets[i];
725 LASSERT(i <= lh->lh_cur_mask);
727 j = lh_hash(lh, new_key, lh->lh_cur_mask);
728 new_lhb = lh->lh_buckets[j];
729 LASSERT(j <= lh->lh_cur_mask);
731 if (i < j) { /* write_lock ordering */
732 write_lock(&old_lhb->lhb_rwlock);
733 write_lock(&new_lhb->lhb_rwlock);
735 write_lock(&new_lhb->lhb_rwlock);
736 write_lock(&old_lhb->lhb_rwlock);
737 } else { /* do nothing */
738 read_unlock(&lh->lh_rwlock);
744 * Migrate item between hash buckets without calling
745 * the lh_get() and lh_put() callback functions.
748 LASSERT(atomic_read(&old_lhb->lhb_count) > 0);
749 atomic_dec(&old_lhb->lhb_count);
750 hlist_add_head(hnode, &(new_lhb->lhb_head));
751 atomic_inc(&new_lhb->lhb_count);
753 write_unlock(&new_lhb->lhb_rwlock);
754 write_unlock(&old_lhb->lhb_rwlock);
759 EXPORT_SYMBOL(lustre_hash_rehash_key);
761 int lustre_hash_debug_header(char *str, int size)
763 return snprintf(str, size,
764 "%-*s%6s%6s%6s%6s%6s%6s%6s%7s%6s%s\n", LUSTRE_MAX_HASH_NAME,
765 "name", "cur", "min", "max", "theta", "t-min", "t-max",
766 "flags", "rehash", "count", " distribution");
768 EXPORT_SYMBOL(lustre_hash_debug_header);
770 int lustre_hash_debug_str(lustre_hash_t *lh, char *str, int size)
772 lustre_hash_bucket_t *lhb;
776 int dist[8] = { 0, };
778 if (str == NULL || size == 0)
782 theta = __lustre_hash_theta(lh);
784 c += snprintf(str + c, size - c, "%-*s ",
785 LUSTRE_MAX_HASH_NAME, lh->lh_name);
786 c += snprintf(str + c, size - c, "%5d ", 1 << lh->lh_cur_bits);
787 c += snprintf(str + c, size - c, "%5d ", 1 << lh->lh_min_bits);
788 c += snprintf(str + c, size - c, "%5d ", 1 << lh->lh_max_bits);
789 c += snprintf(str + c, size - c, "%d.%03d ",
790 __lustre_hash_theta_int(theta),
791 __lustre_hash_theta_frac(theta));
792 c += snprintf(str + c, size - c, "%d.%03d ",
793 __lustre_hash_theta_int(lh->lh_min_theta),
794 __lustre_hash_theta_frac(lh->lh_min_theta));
795 c += snprintf(str + c, size - c, "%d.%03d ",
796 __lustre_hash_theta_int(lh->lh_max_theta),
797 __lustre_hash_theta_frac(lh->lh_max_theta));
798 c += snprintf(str + c, size - c, " 0x%02x ", lh->lh_flags);
799 c += snprintf(str + c, size - c, "%6d ",
800 atomic_read(&lh->lh_rehash_count));
801 c += snprintf(str + c, size - c, "%5d ",
802 atomic_read(&lh->lh_count));
805 * The distribution is a summary of the chained hash depth in
806 * each of the lustre hash buckets. Each buckets lhb_count is
807 * divided by the hash theta value and used to generate a
808 * histogram of the hash distribution. A uniform hash will
809 * result in all hash buckets being close to the average thus
810 * only the first few entries in the histogram will be non-zero.
811 * If you hash function results in a non-uniform hash the will
812 * be observable by outlier bucks in the distribution histogram.
814 * Uniform hash distribution: 128/128/0/0/0/0/0/0
815 * Non-Uniform hash distribution: 128/125/0/0/0/0/2/1
817 lh_for_each_bucket(lh, lhb, i)
818 dist[min(__fls(atomic_read(&lhb->lhb_count)/max(theta,1)),7)]++;
820 for (i = 0; i < 8; i++)
821 c += snprintf(str + c, size - c, "%d%c", dist[i],
822 (i == 7) ? '\n' : '/');
828 EXPORT_SYMBOL(lustre_hash_debug_str);