xref: /illumos-gate/usr/src/uts/common/os/modhash.c (revision bd335c6465ddbafe543900df4b03247bfa288eff)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License, Version 1.0 only
6  * (the "License").  You may not use this file except in compliance
7  * with the License.
8  *
9  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10  * or http://www.opensolaris.org/os/licensing.
11  * See the License for the specific language governing permissions
12  * and limitations under the License.
13  *
14  * When distributing Covered Code, include this CDDL HEADER in each
15  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16  * If applicable, add the following below this CDDL HEADER, with the
17  * fields enclosed by brackets "[]" replaced with your own identifying
18  * information: Portions Copyright [yyyy] [name of copyright owner]
19  *
20  * CDDL HEADER END
21  */
22 /*
23  * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 #pragma ident	"%Z%%M%	%I%	%E% SMI"
28 
29 /*
30  * mod_hash: flexible hash table implementation.
31  *
32  * This is a reasonably fast, reasonably flexible hash table implementation
33  * which features pluggable hash algorithms to support storing arbitrary keys
34  * and values.  It is designed to handle small (< 100,000 items) amounts of
35  * data.  The hash uses chaining to resolve collisions, and does not feature a
36  * mechanism to grow the hash.  Care must be taken to pick nchains to be large
37  * enough for the application at hand, or lots of time will be wasted searching
38  * hash chains.
39  *
40  * The client of the hash is required to supply a number of items to support
41  * the various hash functions:
42  *
43  * 	- Destructor functions for the key and value being hashed.
44  *	  A destructor is responsible for freeing an object when the hash
45  *	  table is no longer storing it.  Since keys and values can be of
46  *	  arbitrary type, separate destructors for keys & values are used.
47  *	  These may be mod_hash_null_keydtor and mod_hash_null_valdtor if no
48  *	  destructor is needed for either a key or value.
49  *
50  *	- A hashing algorithm which returns a uint_t representing a hash index
51  *	  The number returned need _not_ be between 0 and nchains.  The mod_hash
52  *	  code will take care of doing that.  The second argument (after the
53  *	  key) to the hashing function is a void * that represents
54  *	  hash_alg_data-- this is provided so that the hashing algrorithm can
55  *	  maintain some state across calls, or keep algorithm-specific
56  *	  constants associated with the hash table.
57  *
58  *	  A pointer-hashing and a string-hashing algorithm are supplied in
59  *	  this file.
60  *
61  *	- A key comparator (a la qsort).
62  *	  This is used when searching the hash chain.  The key comparator
63  *	  determines if two keys match.  It should follow the return value
64  *	  semantics of strcmp.
65  *
66  *	  string and pointer comparators are supplied in this file.
67  *
68  * mod_hash_create_strhash() and mod_hash_create_ptrhash() provide good
69  * examples of how to create a customized hash table.
70  *
71  * Basic hash operations:
72  *
73  *   mod_hash_create_strhash(name, nchains, dtor),
74  *	create a hash using strings as keys.
75  *	NOTE: This create a hash which automatically cleans up the string
76  *	      values it is given for keys.
77  *
78  *   mod_hash_create_ptrhash(name, nchains, dtor, key_elem_size):
79  *	create a hash using pointers as keys.
80  *
81  *   mod_hash_create_extended(name, nchains, kdtor, vdtor,
82  *			      hash_alg, hash_alg_data,
83  *			      keycmp, sleep)
84  *	create a customized hash table.
85  *
86  *   mod_hash_destroy_hash(hash):
87  *	destroy the given hash table, calling the key and value destructors
88  *	on each key-value pair stored in the hash.
89  *
90  *   mod_hash_insert(hash, key, val):
91  *	place a key, value pair into the given hash.
92  *	duplicate keys are rejected.
93  *
94  *   mod_hash_insert_reserve(hash, key, val, handle):
95  *	place a key, value pair into the given hash, using handle to indicate
96  *	the reserved storage for the pair.  (no memory allocation is needed
97  *	during a mod_hash_insert_reserve.)  duplicate keys are rejected.
98  *
99  *   mod_hash_reserve(hash, *handle):
100  *      reserve storage for a key-value pair using the memory allocation
101  *      policy of 'hash', returning the storage handle in 'handle'.
102  *
103  *   mod_hash_reserve_nosleep(hash, *handle): reserve storage for a key-value
104  *	pair ignoring the memory allocation policy of 'hash' and always without
105  *	sleep, returning the storage handle in 'handle'.
106  *
107  *   mod_hash_remove(hash, key, *val):
108  *	remove a key-value pair with key 'key' from 'hash', destroying the
109  *	stored key, and returning the value in val.
110  *
111  *   mod_hash_replace(hash, key, val)
112  * 	atomically remove an existing key-value pair from a hash, and replace
113  * 	the key and value with the ones supplied.  The removed key and value
114  * 	(if any) are destroyed.
115  *
116  *   mod_hash_destroy(hash, key):
117  *	remove a key-value pair with key 'key' from 'hash', destroying both
118  *	stored key and stored value.
119  *
120  *   mod_hash_find(hash, key, val):
121  *	find a value in the hash table corresponding to the given key.
122  *
123  *   mod_hash_find_cb(hash, key, val, found_callback)
124  *	find a value in the hash table corresponding to the given key.
125  *	If a value is found, call specified callback passing key and val to it.
126  *      The callback is called with the hash lock held.
127  *	It is intended to be used in situations where the act of locating the
128  *	data must also modify it - such as in reference counting schemes.
129  *
130  *   mod_hash_walk(hash, callback(key, elem, arg), arg)
131  * 	walks all the elements in the hashtable and invokes the callback
132  * 	function with the key/value pair for each element.  the hashtable
133  * 	is locked for readers so the callback function should not attempt
134  * 	to do any updates to the hashable.  the callback function should
135  * 	return MH_WALK_CONTINUE to continue walking the hashtable or
136  * 	MH_WALK_TERMINATE to abort the walk of the hashtable.
137  *
138  *   mod_hash_clear(hash):
139  *	clears the given hash table of entries, calling the key and value
140  *	destructors for every element in the hash.
141  */
142 
143 #include <sys/bitmap.h>
144 #include <sys/debug.h>
145 #include <sys/kmem.h>
146 #include <sys/sunddi.h>
147 
148 #include <sys/modhash_impl.h>
149 
150 /*
151  * MH_KEY_DESTROY()
152  * 	Invoke the key destructor.
153  */
154 #define	MH_KEY_DESTROY(hash, key) ((hash->mh_kdtor)(key))
155 
156 /*
157  * MH_VAL_DESTROY()
158  * 	Invoke the value destructor.
159  */
160 #define	MH_VAL_DESTROY(hash, val) ((hash->mh_vdtor)(val))
161 
162 /*
163  * MH_KEYCMP()
164  * 	Call the key comparator for the given hash keys.
165  */
166 #define	MH_KEYCMP(hash, key1, key2) ((hash->mh_keycmp)(key1, key2))
167 
168 static void i_mod_hash_clear_nosync(mod_hash_t *);
169 static int i_mod_hash_find_nosync(mod_hash_t *, mod_hash_key_t,
170     mod_hash_val_t *);
171 static int i_mod_hash_insert_nosync(mod_hash_t *, mod_hash_key_t,
172     mod_hash_val_t, mod_hash_hndl_t);
173 static int i_mod_hash_remove_nosync(mod_hash_t *, mod_hash_key_t,
174     mod_hash_val_t *);
175 static uint_t i_mod_hash(mod_hash_t *, mod_hash_key_t);
176 
177 /*
178  * Cache for struct mod_hash_entry
179  */
180 kmem_cache_t *mh_e_cache = NULL;
181 mod_hash_t *mh_head = NULL;
182 kmutex_t mh_head_lock;
183 
184 /*
185  * mod_hash_null_keydtor()
186  * mod_hash_null_valdtor()
187  * 	no-op key and value destructors.
188  */
189 /*ARGSUSED*/
190 void
191 mod_hash_null_keydtor(mod_hash_key_t key)
192 {
193 }
194 
195 /*ARGSUSED*/
196 void
197 mod_hash_null_valdtor(mod_hash_val_t val)
198 {
199 }
200 
201 /*
202  * mod_hash_bystr()
203  * mod_hash_strkey_cmp()
204  * mod_hash_strkey_dtor()
205  * mod_hash_strval_dtor()
206  *	Hash and key comparison routines for hashes with string keys.
207  *
208  * mod_hash_create_strhash()
209  * 	Create a hash using strings as keys
210  *
211  *	The string hashing algorithm is from the "Dragon Book" --
212  *	"Compilers: Principles, Tools & Techniques", by Aho, Sethi, Ullman
213  */
214 
215 /*ARGSUSED*/
216 uint_t
217 mod_hash_bystr(void *hash_data, mod_hash_key_t key)
218 {
219 	uint_t hash = 0;
220 	uint_t g;
221 	char *p, *k = (char *)key;
222 
223 	ASSERT(k);
224 	for (p = k; *p != '\0'; p++) {
225 		hash = (hash << 4) + *p;
226 		if ((g = (hash & 0xf0000000)) != 0) {
227 			hash ^= (g >> 24);
228 			hash ^= g;
229 		}
230 	}
231 	return (hash);
232 }
233 
234 int
235 mod_hash_strkey_cmp(mod_hash_key_t key1, mod_hash_key_t key2)
236 {
237 	return (strcmp((char *)key1, (char *)key2));
238 }
239 
240 void
241 mod_hash_strkey_dtor(mod_hash_key_t key)
242 {
243 	char *c = (char *)key;
244 	kmem_free(c, strlen(c) + 1);
245 }
246 
247 void
248 mod_hash_strval_dtor(mod_hash_val_t val)
249 {
250 	char *c = (char *)val;
251 	kmem_free(c, strlen(c) + 1);
252 }
253 
254 mod_hash_t *
255 mod_hash_create_strhash(char *name, size_t nchains,
256     void (*val_dtor)(mod_hash_val_t))
257 {
258 	return mod_hash_create_extended(name, nchains, mod_hash_strkey_dtor,
259 	    val_dtor, mod_hash_bystr, NULL, mod_hash_strkey_cmp, KM_SLEEP);
260 }
261 
262 void
263 mod_hash_destroy_strhash(mod_hash_t *strhash)
264 {
265 	ASSERT(strhash);
266 	mod_hash_destroy_hash(strhash);
267 }
268 
269 
270 /*
271  * mod_hash_byptr()
272  * mod_hash_ptrkey_cmp()
273  *	Hash and key comparison routines for hashes with pointer keys.
274  *
275  * mod_hash_create_ptrhash()
276  * mod_hash_destroy_ptrhash()
277  * 	Create a hash that uses pointers as keys.  This hash algorithm
278  * 	picks an appropriate set of middle bits in the address to hash on
279  * 	based on the size of the hash table and a hint about the size of
280  * 	the items pointed at.
281  */
282 uint_t
283 mod_hash_byptr(void *hash_data, mod_hash_key_t key)
284 {
285 	uintptr_t k = (uintptr_t)key;
286 	k >>= (int)(uintptr_t)hash_data;
287 
288 	return ((uint_t)k);
289 }
290 
291 int
292 mod_hash_ptrkey_cmp(mod_hash_key_t key1, mod_hash_key_t key2)
293 {
294 	uintptr_t k1 = (uintptr_t)key1;
295 	uintptr_t k2 = (uintptr_t)key2;
296 	if (k1 > k2)
297 		return (-1);
298 	else if (k1 < k2)
299 		return (1);
300 	else
301 		return (0);
302 }
303 
304 mod_hash_t *
305 mod_hash_create_ptrhash(char *name, size_t nchains,
306     void (*val_dtor)(mod_hash_val_t), size_t key_elem_size)
307 {
308 	size_t rshift;
309 
310 	/*
311 	 * We want to hash on the bits in the middle of the address word
312 	 * Bits far to the right in the word have little significance, and
313 	 * are likely to all look the same (for example, an array of
314 	 * 256-byte structures will have the bottom 8 bits of address
315 	 * words the same).  So we want to right-shift each address to
316 	 * ignore the bottom bits.
317 	 *
318 	 * The high bits, which are also unused, will get taken out when
319 	 * mod_hash takes hashkey % nchains.
320 	 */
321 	rshift = highbit(key_elem_size);
322 
323 	return mod_hash_create_extended(name, nchains, mod_hash_null_keydtor,
324 	    val_dtor, mod_hash_byptr, (void *)rshift, mod_hash_ptrkey_cmp,
325 	    KM_SLEEP);
326 }
327 
328 void
329 mod_hash_destroy_ptrhash(mod_hash_t *hash)
330 {
331 	ASSERT(hash);
332 	mod_hash_destroy_hash(hash);
333 }
334 
335 /*
336  * mod_hash_byid()
337  * mod_hash_idkey_cmp()
338  *	Hash and key comparison routines for hashes with 32-bit unsigned keys.
339  *
340  * mod_hash_create_idhash()
341  * mod_hash_destroy_idhash()
342  * mod_hash_iddata_gen()
343  * 	Create a hash that uses numeric keys.
344  *
345  *	The hash algorithm is documented in "Introduction to Algorithms"
346  *	(Cormen, Leiserson, Rivest);  when the hash table is created, it
347  *	attempts to find the next largest prime above the number of hash
348  *	slots.  The hash index is then this number times the key modulo
349  *	the hash size, or (key * prime) % nchains.
350  */
351 uint_t
352 mod_hash_byid(void *hash_data, mod_hash_key_t key)
353 {
354 	uint_t kval = (uint_t)(uintptr_t)hash_data;
355 	return ((uint_t)(uintptr_t)key * (uint_t)kval);
356 }
357 
358 int
359 mod_hash_idkey_cmp(mod_hash_key_t key1, mod_hash_key_t key2)
360 {
361 	return ((uint_t)(uintptr_t)key1 - (uint_t)(uintptr_t)key2);
362 }
363 
364 /*
365  * Generate the next largest prime number greater than nchains; this value
366  * is intended to be later passed in to mod_hash_create_extended() as the
367  * hash_data.
368  */
369 uint_t
370 mod_hash_iddata_gen(size_t nchains)
371 {
372 	uint_t kval, i, prime;
373 
374 	/*
375 	 * Pick the first (odd) prime greater than nchains.  Make sure kval is
376 	 * odd (so start with nchains +1 or +2 as appropriate).
377 	 */
378 	kval = (nchains % 2 == 0) ? nchains + 1 : nchains + 2;
379 
380 	for (;;) {
381 		prime = 1;
382 		for (i = 3; i * i <= kval; i += 2) {
383 			if (kval % i == 0)
384 				prime = 0;
385 		}
386 		if (prime == 1)
387 			break;
388 		kval += 2;
389 	}
390 	return (kval);
391 }
392 
393 mod_hash_t *
394 mod_hash_create_idhash(char *name, size_t nchains,
395     void (*val_dtor)(mod_hash_val_t))
396 {
397 	uint_t kval = mod_hash_iddata_gen(nchains);
398 
399 	return (mod_hash_create_extended(name, nchains, mod_hash_null_keydtor,
400 	    val_dtor, mod_hash_byid, (void *)(uintptr_t)kval,
401 	    mod_hash_idkey_cmp, KM_SLEEP));
402 }
403 
404 void
405 mod_hash_destroy_idhash(mod_hash_t *hash)
406 {
407 	ASSERT(hash);
408 	mod_hash_destroy_hash(hash);
409 }
410 
411 /*
412  * mod_hash_init()
413  * 	sets up globals, etc for mod_hash_*
414  */
415 void
416 mod_hash_init(void)
417 {
418 	ASSERT(mh_e_cache == NULL);
419 	mh_e_cache = kmem_cache_create("mod_hash_entries",
420 	    sizeof (struct mod_hash_entry), 0, NULL, NULL, NULL, NULL,
421 	    NULL, 0);
422 }
423 
424 /*
425  * mod_hash_create_extended()
426  * 	The full-blown hash creation function.
427  *
428  * notes:
429  * 	nchains		- how many hash slots to create.  More hash slots will
430  *			  result in shorter hash chains, but will consume
431  *			  slightly more memory up front.
432  *	sleep		- should be KM_SLEEP or KM_NOSLEEP, to indicate whether
433  *			  to sleep for memory, or fail in low-memory conditions.
434  *
435  * 	Fails only if KM_NOSLEEP was specified, and no memory was available.
436  */
437 mod_hash_t *
438 mod_hash_create_extended(
439     char *hname,			/* descriptive name for hash */
440     size_t nchains,			/* number of hash slots */
441     void (*kdtor)(mod_hash_key_t),	/* key destructor */
442     void (*vdtor)(mod_hash_val_t),	/* value destructor */
443     uint_t (*hash_alg)(void *, mod_hash_key_t), /* hash algorithm */
444     void *hash_alg_data,		/* pass-thru arg for hash_alg */
445     int (*keycmp)(mod_hash_key_t, mod_hash_key_t), /* key comparator */
446     int sleep)				/* whether to sleep for mem */
447 {
448 	mod_hash_t *mod_hash;
449 	ASSERT(hname && keycmp && hash_alg && vdtor && kdtor);
450 
451 	if ((mod_hash = kmem_zalloc(MH_SIZE(nchains), sleep)) == NULL)
452 		return (NULL);
453 
454 	mod_hash->mh_name = kmem_alloc(strlen(hname) + 1, sleep);
455 	if (mod_hash->mh_name == NULL) {
456 		kmem_free(mod_hash, MH_SIZE(nchains));
457 		return (NULL);
458 	}
459 	(void) strcpy(mod_hash->mh_name, hname);
460 
461 	mod_hash->mh_sleep = sleep;
462 	mod_hash->mh_nchains = nchains;
463 	mod_hash->mh_kdtor = kdtor;
464 	mod_hash->mh_vdtor = vdtor;
465 	mod_hash->mh_hashalg = hash_alg;
466 	mod_hash->mh_hashalg_data = hash_alg_data;
467 	mod_hash->mh_keycmp = keycmp;
468 
469 	/*
470 	 * Link the hash up on the list of hashes
471 	 */
472 	mutex_enter(&mh_head_lock);
473 	mod_hash->mh_next = mh_head;
474 	mh_head = mod_hash;
475 	mutex_exit(&mh_head_lock);
476 
477 	return (mod_hash);
478 }
479 
480 /*
481  * mod_hash_destroy_hash()
482  * 	destroy a hash table, destroying all of its stored keys and values
483  * 	as well.
484  */
485 void
486 mod_hash_destroy_hash(mod_hash_t *hash)
487 {
488 	mod_hash_t *mhp, *mhpp;
489 
490 	mutex_enter(&mh_head_lock);
491 	/*
492 	 * Remove the hash from the hash list
493 	 */
494 	if (hash == mh_head) {		/* removing 1st list elem */
495 		mh_head = mh_head->mh_next;
496 	} else {
497 		/*
498 		 * mhpp can start out NULL since we know the 1st elem isn't the
499 		 * droid we're looking for.
500 		 */
501 		mhpp = NULL;
502 		for (mhp = mh_head; mhp != NULL; mhp = mhp->mh_next) {
503 			if (mhp == hash) {
504 				mhpp->mh_next = mhp->mh_next;
505 				break;
506 			}
507 			mhpp = mhp;
508 		}
509 	}
510 	mutex_exit(&mh_head_lock);
511 
512 	/*
513 	 * Clean out keys and values.
514 	 */
515 	mod_hash_clear(hash);
516 
517 	kmem_free(hash->mh_name, strlen(hash->mh_name) + 1);
518 	kmem_free(hash, MH_SIZE(hash->mh_nchains));
519 }
520 
521 /*
522  * i_mod_hash()
523  * 	Call the hashing algorithm for this hash table, with the given key.
524  */
525 static uint_t
526 i_mod_hash(mod_hash_t *hash, mod_hash_key_t key)
527 {
528 	uint_t h;
529 	/*
530 	 * Prevent div by 0 problems;
531 	 * Also a nice shortcut when using a hash as a list
532 	 */
533 	if (hash->mh_nchains == 1)
534 		return (0);
535 
536 	h = (hash->mh_hashalg)(hash->mh_hashalg_data, key);
537 	return (h % (hash->mh_nchains - 1));
538 }
539 
540 /*
541  * i_mod_hash_insert_nosync()
542  * mod_hash_insert()
543  * mod_hash_insert_reserve()
544  * 	insert 'val' into the hash table, using 'key' as its key.  If 'key' is
545  * 	already a key in the hash, an error will be returned, and the key-val
546  * 	pair will not be inserted.  i_mod_hash_insert_nosync() supports a simple
547  * 	handle abstraction, allowing hash entry allocation to be separated from
548  * 	the hash insertion.  this abstraction allows simple use of the mod_hash
549  * 	structure in situations where mod_hash_insert() with a KM_SLEEP
550  * 	allocation policy would otherwise be unsafe.
551  */
552 int
553 i_mod_hash_insert_nosync(mod_hash_t *hash, mod_hash_key_t key,
554     mod_hash_val_t val, mod_hash_hndl_t handle)
555 {
556 	uint_t hashidx;
557 	struct mod_hash_entry *entry;
558 
559 	ASSERT(hash);
560 
561 	/*
562 	 * If we've not been given reserved storage, allocate storage directly,
563 	 * using the hash's allocation policy.
564 	 */
565 	if (handle == (mod_hash_hndl_t)0) {
566 		entry = kmem_cache_alloc(mh_e_cache, hash->mh_sleep);
567 		if (entry == NULL) {
568 			hash->mh_stat.mhs_nomem++;
569 			return (MH_ERR_NOMEM);
570 		}
571 	} else {
572 		entry = (struct mod_hash_entry *)handle;
573 	}
574 
575 	hashidx = i_mod_hash(hash, key);
576 	entry->mhe_key = key;
577 	entry->mhe_val = val;
578 	entry->mhe_next = hash->mh_entries[hashidx];
579 
580 	hash->mh_entries[hashidx] = entry;
581 	hash->mh_stat.mhs_nelems++;
582 
583 	return (0);
584 }
585 
586 int
587 mod_hash_insert(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t val)
588 {
589 	int res;
590 	mod_hash_val_t v;
591 
592 	rw_enter(&hash->mh_contents, RW_WRITER);
593 
594 	/*
595 	 * Disallow duplicate keys in the hash
596 	 */
597 	if (i_mod_hash_find_nosync(hash, key, &v) == 0) {
598 		rw_exit(&hash->mh_contents);
599 		hash->mh_stat.mhs_coll++;
600 		return (MH_ERR_DUPLICATE);
601 	}
602 
603 	res = i_mod_hash_insert_nosync(hash, key, val, (mod_hash_hndl_t)0);
604 	rw_exit(&hash->mh_contents);
605 
606 	return (res);
607 }
608 
609 int
610 mod_hash_insert_reserve(mod_hash_t *hash, mod_hash_key_t key,
611     mod_hash_val_t val, mod_hash_hndl_t handle)
612 {
613 	int res;
614 	mod_hash_val_t v;
615 
616 	rw_enter(&hash->mh_contents, RW_WRITER);
617 
618 	/*
619 	 * Disallow duplicate keys in the hash
620 	 */
621 	if (i_mod_hash_find_nosync(hash, key, &v) == 0) {
622 		rw_exit(&hash->mh_contents);
623 		hash->mh_stat.mhs_coll++;
624 		return (MH_ERR_DUPLICATE);
625 	}
626 	res = i_mod_hash_insert_nosync(hash, key, val, handle);
627 	rw_exit(&hash->mh_contents);
628 
629 	return (res);
630 }
631 
632 /*
633  * mod_hash_reserve()
634  * mod_hash_reserve_nosleep()
635  * mod_hash_cancel()
636  *   Make or cancel a mod_hash_entry_t reservation.  Reservations are used in
637  *   mod_hash_insert_reserve() above.
638  */
639 int
640 mod_hash_reserve(mod_hash_t *hash, mod_hash_hndl_t *handlep)
641 {
642 	*handlep = kmem_cache_alloc(mh_e_cache, hash->mh_sleep);
643 	if (*handlep == NULL) {
644 		hash->mh_stat.mhs_nomem++;
645 		return (MH_ERR_NOMEM);
646 	}
647 
648 	return (0);
649 }
650 
651 int
652 mod_hash_reserve_nosleep(mod_hash_t *hash, mod_hash_hndl_t *handlep)
653 {
654 	*handlep = kmem_cache_alloc(mh_e_cache, KM_NOSLEEP);
655 	if (*handlep == NULL) {
656 		hash->mh_stat.mhs_nomem++;
657 		return (MH_ERR_NOMEM);
658 	}
659 
660 	return (0);
661 
662 }
663 
664 /*ARGSUSED*/
665 void
666 mod_hash_cancel(mod_hash_t *hash, mod_hash_hndl_t *handlep)
667 {
668 	kmem_cache_free(mh_e_cache, *handlep);
669 	*handlep = (mod_hash_hndl_t)0;
670 }
671 
672 /*
673  * i_mod_hash_remove_nosync()
674  * mod_hash_remove()
675  * 	Remove an element from the hash table.
676  */
677 int
678 i_mod_hash_remove_nosync(mod_hash_t *hash, mod_hash_key_t key,
679     mod_hash_val_t *val)
680 {
681 	int hashidx;
682 	struct mod_hash_entry *e, *ep;
683 
684 	hashidx = i_mod_hash(hash, key);
685 	ep = NULL; /* e's parent */
686 
687 	for (e = hash->mh_entries[hashidx]; e != NULL; e = e->mhe_next) {
688 		if (MH_KEYCMP(hash, e->mhe_key, key) == 0)
689 			break;
690 		ep = e;
691 	}
692 
693 	if (e == NULL) {	/* not found */
694 		return (MH_ERR_NOTFOUND);
695 	}
696 
697 	if (ep == NULL) 	/* special case 1st element in bucket */
698 		hash->mh_entries[hashidx] = e->mhe_next;
699 	else
700 		ep->mhe_next = e->mhe_next;
701 
702 	/*
703 	 * Clean up resources used by the node's key.
704 	 */
705 	MH_KEY_DESTROY(hash, e->mhe_key);
706 
707 	*val = e->mhe_val;
708 	kmem_cache_free(mh_e_cache, e);
709 	hash->mh_stat.mhs_nelems--;
710 
711 	return (0);
712 }
713 
714 int
715 mod_hash_remove(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t *val)
716 {
717 	int res;
718 
719 	rw_enter(&hash->mh_contents, RW_WRITER);
720 	res = i_mod_hash_remove_nosync(hash, key, val);
721 	rw_exit(&hash->mh_contents);
722 
723 	return (res);
724 }
725 
726 /*
727  * mod_hash_replace()
728  * 	atomically remove an existing key-value pair from a hash, and replace
729  * 	the key and value with the ones supplied.  The removed key and value
730  * 	(if any) are destroyed.
731  */
732 int
733 mod_hash_replace(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t val)
734 {
735 	int res;
736 	mod_hash_val_t v;
737 
738 	rw_enter(&hash->mh_contents, RW_WRITER);
739 
740 	if (i_mod_hash_remove_nosync(hash, key, &v) == 0) {
741 		/*
742 		 * mod_hash_remove() takes care of freeing up the key resources.
743 		 */
744 		MH_VAL_DESTROY(hash, v);
745 	}
746 	res = i_mod_hash_insert_nosync(hash, key, val, (mod_hash_hndl_t)0);
747 
748 	rw_exit(&hash->mh_contents);
749 
750 	return (res);
751 }
752 
753 /*
754  * mod_hash_destroy()
755  * 	Remove an element from the hash table matching 'key', and destroy it.
756  */
757 int
758 mod_hash_destroy(mod_hash_t *hash, mod_hash_key_t key)
759 {
760 	mod_hash_val_t val;
761 	int rv;
762 
763 	rw_enter(&hash->mh_contents, RW_WRITER);
764 
765 	if ((rv = i_mod_hash_remove_nosync(hash, key, &val)) == 0) {
766 		/*
767 		 * mod_hash_remove() takes care of freeing up the key resources.
768 		 */
769 		MH_VAL_DESTROY(hash, val);
770 	}
771 
772 	rw_exit(&hash->mh_contents);
773 	return (rv);
774 }
775 
776 /*
777  * i_mod_hash_find_nosync()
778  * mod_hash_find()
779  * 	Find a value in the hash table corresponding to the given key.
780  */
781 static int
782 i_mod_hash_find_nosync(mod_hash_t *hash, mod_hash_key_t key,
783     mod_hash_val_t *val)
784 {
785 	uint_t hashidx;
786 	struct mod_hash_entry *e;
787 
788 	hashidx = i_mod_hash(hash, key);
789 
790 	for (e = hash->mh_entries[hashidx]; e != NULL; e = e->mhe_next) {
791 		if (MH_KEYCMP(hash, e->mhe_key, key) == 0) {
792 			*val = e->mhe_val;
793 			hash->mh_stat.mhs_hit++;
794 			return (0);
795 		}
796 	}
797 	hash->mh_stat.mhs_miss++;
798 	return (MH_ERR_NOTFOUND);
799 }
800 
801 int
802 mod_hash_find(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t *val)
803 {
804 	int res;
805 
806 	rw_enter(&hash->mh_contents, RW_READER);
807 	res = i_mod_hash_find_nosync(hash, key, val);
808 	rw_exit(&hash->mh_contents);
809 
810 	return (res);
811 }
812 
813 int
814 mod_hash_find_cb(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t *val,
815     void (*find_cb)(mod_hash_key_t, mod_hash_val_t))
816 {
817 	int res;
818 
819 	rw_enter(&hash->mh_contents, RW_READER);
820 	res = i_mod_hash_find_nosync(hash, key, val);
821 	if (res == 0) {
822 		find_cb(key, *val);
823 	}
824 	rw_exit(&hash->mh_contents);
825 
826 	return (res);
827 }
828 
829 static void
830 i_mod_hash_walk_nosync(mod_hash_t *hash,
831     uint_t (*callback)(mod_hash_key_t, mod_hash_val_t *, void *), void *arg)
832 {
833 	struct mod_hash_entry	*e;
834 	uint_t			hashidx;
835 	int			res = MH_WALK_CONTINUE;
836 
837 	for (hashidx = 0;
838 	    (hashidx < (hash->mh_nchains - 1)) && (res == MH_WALK_CONTINUE);
839 	    hashidx++) {
840 		e = hash->mh_entries[hashidx];
841 		while ((e != NULL) && (res == MH_WALK_CONTINUE)) {
842 			res = callback(e->mhe_key, e->mhe_val, arg);
843 			e = e->mhe_next;
844 		}
845 	}
846 }
847 
848 /*
849  * mod_hash_walk()
850  * 	Walks all the elements in the hashtable and invokes the callback
851  * 	function with the key/value pair for each element.  The hashtable
852  * 	is locked for readers so the callback function should not attempt
853  * 	to do any updates to the hashable.  The callback function should
854  * 	return MH_WALK_CONTINUE to continue walking the hashtable or
855  * 	MH_WALK_TERMINATE to abort the walk of the hashtable.
856  */
857 void
858 mod_hash_walk(mod_hash_t *hash,
859     uint_t (*callback)(mod_hash_key_t, mod_hash_val_t *, void *), void *arg)
860 {
861 	rw_enter(&hash->mh_contents, RW_READER);
862 	i_mod_hash_walk_nosync(hash, callback, arg);
863 	rw_exit(&hash->mh_contents);
864 }
865 
866 
867 /*
868  * i_mod_hash_clear_nosync()
869  * mod_hash_clear()
870  *	Clears the given hash table by calling the destructor of every hash
871  *	element and freeing up all mod_hash_entry's.
872  */
873 static void
874 i_mod_hash_clear_nosync(mod_hash_t *hash)
875 {
876 	int i;
877 	struct mod_hash_entry *e, *old_e;
878 
879 	for (i = 0; i < hash->mh_nchains; i++) {
880 		e = hash->mh_entries[i];
881 		while (e != NULL) {
882 			MH_KEY_DESTROY(hash, e->mhe_key);
883 			MH_VAL_DESTROY(hash, e->mhe_val);
884 			old_e = e;
885 			e = e->mhe_next;
886 			kmem_cache_free(mh_e_cache, old_e);
887 		}
888 		hash->mh_entries[i] = NULL;
889 	}
890 	hash->mh_stat.mhs_nelems = 0;
891 }
892 
893 void
894 mod_hash_clear(mod_hash_t *hash)
895 {
896 	ASSERT(hash);
897 	rw_enter(&hash->mh_contents, RW_WRITER);
898 	i_mod_hash_clear_nosync(hash);
899 	rw_exit(&hash->mh_contents);
900 }
901