xref: /illumos-gate/usr/src/uts/common/os/modhash.c (revision a38ddfee9c8c6b6c5a2947ff52fd2338362a4444)
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 (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 /*
29  * mod_hash: flexible hash table implementation.
30  *
31  * This is a reasonably fast, reasonably flexible hash table implementation
32  * which features pluggable hash algorithms to support storing arbitrary keys
33  * and values.  It is designed to handle small (< 100,000 items) amounts of
34  * data.  The hash uses chaining to resolve collisions, and does not feature a
35  * mechanism to grow the hash.  Care must be taken to pick nchains to be large
36  * enough for the application at hand, or lots of time will be wasted searching
37  * hash chains.
38  *
39  * The client of the hash is required to supply a number of items to support
40  * the various hash functions:
41  *
42  * 	- Destructor functions for the key and value being hashed.
43  *	  A destructor is responsible for freeing an object when the hash
44  *	  table is no longer storing it.  Since keys and values can be of
45  *	  arbitrary type, separate destructors for keys & values are used.
46  *	  These may be mod_hash_null_keydtor and mod_hash_null_valdtor if no
47  *	  destructor is needed for either a key or value.
48  *
49  *	- A hashing algorithm which returns a uint_t representing a hash index
50  *	  The number returned need _not_ be between 0 and nchains.  The mod_hash
51  *	  code will take care of doing that.  The second argument (after the
52  *	  key) to the hashing function is a void * that represents
53  *	  hash_alg_data-- this is provided so that the hashing algrorithm can
54  *	  maintain some state across calls, or keep algorithm-specific
55  *	  constants associated with the hash table.
56  *
57  *	  A pointer-hashing and a string-hashing algorithm are supplied in
58  *	  this file.
59  *
60  *	- A key comparator (a la qsort).
61  *	  This is used when searching the hash chain.  The key comparator
62  *	  determines if two keys match.  It should follow the return value
63  *	  semantics of strcmp.
64  *
65  *	  string and pointer comparators are supplied in this file.
66  *
67  * mod_hash_create_strhash() and mod_hash_create_ptrhash() provide good
68  * examples of how to create a customized hash table.
69  *
70  * Basic hash operations:
71  *
72  *   mod_hash_create_strhash(name, nchains, dtor),
73  *	create a hash using strings as keys.
74  *	NOTE: This create a hash which automatically cleans up the string
75  *	      values it is given for keys.
76  *
77  *   mod_hash_create_ptrhash(name, nchains, dtor, key_elem_size):
78  *	create a hash using pointers as keys.
79  *
80  *   mod_hash_create_extended(name, nchains, kdtor, vdtor,
81  *			      hash_alg, hash_alg_data,
82  *			      keycmp, sleep)
83  *	create a customized hash table.
84  *
85  *   mod_hash_destroy_hash(hash):
86  *	destroy the given hash table, calling the key and value destructors
87  *	on each key-value pair stored in the hash.
88  *
89  *   mod_hash_insert(hash, key, val):
90  *	place a key, value pair into the given hash.
91  *	duplicate keys are rejected.
92  *
93  *   mod_hash_insert_reserve(hash, key, val, handle):
94  *	place a key, value pair into the given hash, using handle to indicate
95  *	the reserved storage for the pair.  (no memory allocation is needed
96  *	during a mod_hash_insert_reserve.)  duplicate keys are rejected.
97  *
98  *   mod_hash_reserve(hash, *handle):
99  *      reserve storage for a key-value pair using the memory allocation
100  *      policy of 'hash', returning the storage handle in 'handle'.
101  *
102  *   mod_hash_reserve_nosleep(hash, *handle): reserve storage for a key-value
103  *	pair ignoring the memory allocation policy of 'hash' and always without
104  *	sleep, returning the storage handle in 'handle'.
105  *
106  *   mod_hash_remove(hash, key, *val):
107  *	remove a key-value pair with key 'key' from 'hash', destroying the
108  *	stored key, and returning the value in val.
109  *
110  *   mod_hash_replace(hash, key, val)
111  * 	atomically remove an existing key-value pair from a hash, and replace
112  * 	the key and value with the ones supplied.  The removed key and value
113  * 	(if any) are destroyed.
114  *
115  *   mod_hash_destroy(hash, key):
116  *	remove a key-value pair with key 'key' from 'hash', destroying both
117  *	stored key and stored value.
118  *
119  *   mod_hash_find(hash, key, val):
120  *	find a value in the hash table corresponding to the given key.
121  *
122  *   mod_hash_find_cb(hash, key, val, found_callback)
123  *	find a value in the hash table corresponding to the given key.
124  *	If a value is found, call specified callback passing key and val to it.
125  *      The callback is called with the hash lock held.
126  *	It is intended to be used in situations where the act of locating the
127  *	data must also modify it - such as in reference counting schemes.
128  *
129  *   mod_hash_walk(hash, callback(key, elem, arg), arg)
130  * 	walks all the elements in the hashtable and invokes the callback
131  * 	function with the key/value pair for each element.  the hashtable
132  * 	is locked for readers so the callback function should not attempt
133  * 	to do any updates to the hashable.  the callback function should
134  * 	return MH_WALK_CONTINUE to continue walking the hashtable or
135  * 	MH_WALK_TERMINATE to abort the walk of the hashtable.
136  *
137  *   mod_hash_clear(hash):
138  *	clears the given hash table of entries, calling the key and value
139  *	destructors for every element in the hash.
140  */
141 
142 #include <sys/bitmap.h>
143 #include <sys/debug.h>
144 #include <sys/kmem.h>
145 #include <sys/sunddi.h>
146 
147 #include <sys/modhash_impl.h>
148 
149 /*
150  * MH_KEY_DESTROY()
151  * 	Invoke the key destructor.
152  */
153 #define	MH_KEY_DESTROY(hash, key) ((hash->mh_kdtor)(key))
154 
155 /*
156  * MH_VAL_DESTROY()
157  * 	Invoke the value destructor.
158  */
159 #define	MH_VAL_DESTROY(hash, val) ((hash->mh_vdtor)(val))
160 
161 /*
162  * MH_KEYCMP()
163  * 	Call the key comparator for the given hash keys.
164  */
165 #define	MH_KEYCMP(hash, key1, key2) ((hash->mh_keycmp)(key1, key2))
166 
167 /*
168  * Cache for struct mod_hash_entry
169  */
170 kmem_cache_t *mh_e_cache = NULL;
171 mod_hash_t *mh_head = NULL;
172 kmutex_t mh_head_lock;
173 
174 /*
175  * mod_hash_null_keydtor()
176  * mod_hash_null_valdtor()
177  * 	no-op key and value destructors.
178  */
179 /*ARGSUSED*/
180 void
181 mod_hash_null_keydtor(mod_hash_key_t key)
182 {
183 }
184 
185 /*ARGSUSED*/
186 void
187 mod_hash_null_valdtor(mod_hash_val_t val)
188 {
189 }
190 
191 /*
192  * mod_hash_bystr()
193  * mod_hash_strkey_cmp()
194  * mod_hash_strkey_dtor()
195  * mod_hash_strval_dtor()
196  *	Hash and key comparison routines for hashes with string keys.
197  *
198  * mod_hash_create_strhash()
199  * 	Create a hash using strings as keys
200  *
201  *	The string hashing algorithm is from the "Dragon Book" --
202  *	"Compilers: Principles, Tools & Techniques", by Aho, Sethi, Ullman
203  */
204 
205 /*ARGSUSED*/
206 uint_t
207 mod_hash_bystr(void *hash_data, mod_hash_key_t key)
208 {
209 	uint_t hash = 0;
210 	uint_t g;
211 	char *p, *k = (char *)key;
212 
213 	ASSERT(k);
214 	for (p = k; *p != '\0'; p++) {
215 		hash = (hash << 4) + *p;
216 		if ((g = (hash & 0xf0000000)) != 0) {
217 			hash ^= (g >> 24);
218 			hash ^= g;
219 		}
220 	}
221 	return (hash);
222 }
223 
224 int
225 mod_hash_strkey_cmp(mod_hash_key_t key1, mod_hash_key_t key2)
226 {
227 	return (strcmp((char *)key1, (char *)key2));
228 }
229 
230 void
231 mod_hash_strkey_dtor(mod_hash_key_t key)
232 {
233 	char *c = (char *)key;
234 	kmem_free(c, strlen(c) + 1);
235 }
236 
237 void
238 mod_hash_strval_dtor(mod_hash_val_t val)
239 {
240 	char *c = (char *)val;
241 	kmem_free(c, strlen(c) + 1);
242 }
243 
244 mod_hash_t *
245 mod_hash_create_strhash(char *name, size_t nchains,
246     void (*val_dtor)(mod_hash_val_t))
247 {
248 	return mod_hash_create_extended(name, nchains, mod_hash_strkey_dtor,
249 	    val_dtor, mod_hash_bystr, NULL, mod_hash_strkey_cmp, KM_SLEEP);
250 }
251 
252 void
253 mod_hash_destroy_strhash(mod_hash_t *strhash)
254 {
255 	ASSERT(strhash);
256 	mod_hash_destroy_hash(strhash);
257 }
258 
259 
260 /*
261  * mod_hash_byptr()
262  * mod_hash_ptrkey_cmp()
263  *	Hash and key comparison routines for hashes with pointer keys.
264  *
265  * mod_hash_create_ptrhash()
266  * mod_hash_destroy_ptrhash()
267  * 	Create a hash that uses pointers as keys.  This hash algorithm
268  * 	picks an appropriate set of middle bits in the address to hash on
269  * 	based on the size of the hash table and a hint about the size of
270  * 	the items pointed at.
271  */
272 uint_t
273 mod_hash_byptr(void *hash_data, mod_hash_key_t key)
274 {
275 	uintptr_t k = (uintptr_t)key;
276 	k >>= (int)(uintptr_t)hash_data;
277 
278 	return ((uint_t)k);
279 }
280 
281 int
282 mod_hash_ptrkey_cmp(mod_hash_key_t key1, mod_hash_key_t key2)
283 {
284 	uintptr_t k1 = (uintptr_t)key1;
285 	uintptr_t k2 = (uintptr_t)key2;
286 	if (k1 > k2)
287 		return (-1);
288 	else if (k1 < k2)
289 		return (1);
290 	else
291 		return (0);
292 }
293 
294 mod_hash_t *
295 mod_hash_create_ptrhash(char *name, size_t nchains,
296     void (*val_dtor)(mod_hash_val_t), size_t key_elem_size)
297 {
298 	size_t rshift;
299 
300 	/*
301 	 * We want to hash on the bits in the middle of the address word
302 	 * Bits far to the right in the word have little significance, and
303 	 * are likely to all look the same (for example, an array of
304 	 * 256-byte structures will have the bottom 8 bits of address
305 	 * words the same).  So we want to right-shift each address to
306 	 * ignore the bottom bits.
307 	 *
308 	 * The high bits, which are also unused, will get taken out when
309 	 * mod_hash takes hashkey % nchains.
310 	 */
311 	rshift = highbit(key_elem_size);
312 
313 	return mod_hash_create_extended(name, nchains, mod_hash_null_keydtor,
314 	    val_dtor, mod_hash_byptr, (void *)rshift, mod_hash_ptrkey_cmp,
315 	    KM_SLEEP);
316 }
317 
318 void
319 mod_hash_destroy_ptrhash(mod_hash_t *hash)
320 {
321 	ASSERT(hash);
322 	mod_hash_destroy_hash(hash);
323 }
324 
325 /*
326  * mod_hash_byid()
327  * mod_hash_idkey_cmp()
328  *	Hash and key comparison routines for hashes with 32-bit unsigned keys.
329  *
330  * mod_hash_create_idhash()
331  * mod_hash_destroy_idhash()
332  * mod_hash_iddata_gen()
333  * 	Create a hash that uses numeric keys.
334  *
335  *	The hash algorithm is documented in "Introduction to Algorithms"
336  *	(Cormen, Leiserson, Rivest);  when the hash table is created, it
337  *	attempts to find the next largest prime above the number of hash
338  *	slots.  The hash index is then this number times the key modulo
339  *	the hash size, or (key * prime) % nchains.
340  */
341 uint_t
342 mod_hash_byid(void *hash_data, mod_hash_key_t key)
343 {
344 	uint_t kval = (uint_t)(uintptr_t)hash_data;
345 	return ((uint_t)(uintptr_t)key * (uint_t)kval);
346 }
347 
348 int
349 mod_hash_idkey_cmp(mod_hash_key_t key1, mod_hash_key_t key2)
350 {
351 	return ((uint_t)(uintptr_t)key1 - (uint_t)(uintptr_t)key2);
352 }
353 
354 /*
355  * Generate the next largest prime number greater than nchains; this value
356  * is intended to be later passed in to mod_hash_create_extended() as the
357  * hash_data.
358  */
359 uint_t
360 mod_hash_iddata_gen(size_t nchains)
361 {
362 	uint_t kval, i, prime;
363 
364 	/*
365 	 * Pick the first (odd) prime greater than nchains.  Make sure kval is
366 	 * odd (so start with nchains +1 or +2 as appropriate).
367 	 */
368 	kval = (nchains % 2 == 0) ? nchains + 1 : nchains + 2;
369 
370 	for (;;) {
371 		prime = 1;
372 		for (i = 3; i * i <= kval; i += 2) {
373 			if (kval % i == 0)
374 				prime = 0;
375 		}
376 		if (prime == 1)
377 			break;
378 		kval += 2;
379 	}
380 	return (kval);
381 }
382 
383 mod_hash_t *
384 mod_hash_create_idhash(char *name, size_t nchains,
385     void (*val_dtor)(mod_hash_val_t))
386 {
387 	uint_t kval = mod_hash_iddata_gen(nchains);
388 
389 	return (mod_hash_create_extended(name, nchains, mod_hash_null_keydtor,
390 	    val_dtor, mod_hash_byid, (void *)(uintptr_t)kval,
391 	    mod_hash_idkey_cmp, KM_SLEEP));
392 }
393 
394 void
395 mod_hash_destroy_idhash(mod_hash_t *hash)
396 {
397 	ASSERT(hash);
398 	mod_hash_destroy_hash(hash);
399 }
400 
401 /*
402  * mod_hash_init()
403  * 	sets up globals, etc for mod_hash_*
404  */
405 void
406 mod_hash_init(void)
407 {
408 	ASSERT(mh_e_cache == NULL);
409 	mh_e_cache = kmem_cache_create("mod_hash_entries",
410 	    sizeof (struct mod_hash_entry), 0, NULL, NULL, NULL, NULL,
411 	    NULL, 0);
412 }
413 
414 /*
415  * mod_hash_create_extended()
416  * 	The full-blown hash creation function.
417  *
418  * notes:
419  * 	nchains		- how many hash slots to create.  More hash slots will
420  *			  result in shorter hash chains, but will consume
421  *			  slightly more memory up front.
422  *	sleep		- should be KM_SLEEP or KM_NOSLEEP, to indicate whether
423  *			  to sleep for memory, or fail in low-memory conditions.
424  *
425  * 	Fails only if KM_NOSLEEP was specified, and no memory was available.
426  */
427 mod_hash_t *
428 mod_hash_create_extended(
429     char *hname,			/* descriptive name for hash */
430     size_t nchains,			/* number of hash slots */
431     void (*kdtor)(mod_hash_key_t),	/* key destructor */
432     void (*vdtor)(mod_hash_val_t),	/* value destructor */
433     uint_t (*hash_alg)(void *, mod_hash_key_t), /* hash algorithm */
434     void *hash_alg_data,		/* pass-thru arg for hash_alg */
435     int (*keycmp)(mod_hash_key_t, mod_hash_key_t), /* key comparator */
436     int sleep)				/* whether to sleep for mem */
437 {
438 	mod_hash_t *mod_hash;
439 	ASSERT(hname && keycmp && hash_alg && vdtor && kdtor);
440 
441 	if ((mod_hash = kmem_zalloc(MH_SIZE(nchains), sleep)) == NULL)
442 		return (NULL);
443 
444 	mod_hash->mh_name = kmem_alloc(strlen(hname) + 1, sleep);
445 	if (mod_hash->mh_name == NULL) {
446 		kmem_free(mod_hash, MH_SIZE(nchains));
447 		return (NULL);
448 	}
449 	(void) strcpy(mod_hash->mh_name, hname);
450 
451 	mod_hash->mh_sleep = sleep;
452 	mod_hash->mh_nchains = nchains;
453 	mod_hash->mh_kdtor = kdtor;
454 	mod_hash->mh_vdtor = vdtor;
455 	mod_hash->mh_hashalg = hash_alg;
456 	mod_hash->mh_hashalg_data = hash_alg_data;
457 	mod_hash->mh_keycmp = keycmp;
458 
459 	/*
460 	 * Link the hash up on the list of hashes
461 	 */
462 	mutex_enter(&mh_head_lock);
463 	mod_hash->mh_next = mh_head;
464 	mh_head = mod_hash;
465 	mutex_exit(&mh_head_lock);
466 
467 	return (mod_hash);
468 }
469 
470 /*
471  * mod_hash_destroy_hash()
472  * 	destroy a hash table, destroying all of its stored keys and values
473  * 	as well.
474  */
475 void
476 mod_hash_destroy_hash(mod_hash_t *hash)
477 {
478 	mod_hash_t *mhp, *mhpp;
479 
480 	mutex_enter(&mh_head_lock);
481 	/*
482 	 * Remove the hash from the hash list
483 	 */
484 	if (hash == mh_head) {		/* removing 1st list elem */
485 		mh_head = mh_head->mh_next;
486 	} else {
487 		/*
488 		 * mhpp can start out NULL since we know the 1st elem isn't the
489 		 * droid we're looking for.
490 		 */
491 		mhpp = NULL;
492 		for (mhp = mh_head; mhp != NULL; mhp = mhp->mh_next) {
493 			if (mhp == hash) {
494 				mhpp->mh_next = mhp->mh_next;
495 				break;
496 			}
497 			mhpp = mhp;
498 		}
499 	}
500 	mutex_exit(&mh_head_lock);
501 
502 	/*
503 	 * Clean out keys and values.
504 	 */
505 	mod_hash_clear(hash);
506 
507 	kmem_free(hash->mh_name, strlen(hash->mh_name) + 1);
508 	kmem_free(hash, MH_SIZE(hash->mh_nchains));
509 }
510 
511 /*
512  * i_mod_hash()
513  * 	Call the hashing algorithm for this hash table, with the given key.
514  */
515 uint_t
516 i_mod_hash(mod_hash_t *hash, mod_hash_key_t key)
517 {
518 	uint_t h;
519 	/*
520 	 * Prevent div by 0 problems;
521 	 * Also a nice shortcut when using a hash as a list
522 	 */
523 	if (hash->mh_nchains == 1)
524 		return (0);
525 
526 	h = (hash->mh_hashalg)(hash->mh_hashalg_data, key);
527 	return (h % (hash->mh_nchains - 1));
528 }
529 
530 /*
531  * i_mod_hash_insert_nosync()
532  * mod_hash_insert()
533  * mod_hash_insert_reserve()
534  * 	insert 'val' into the hash table, using 'key' as its key.  If 'key' is
535  * 	already a key in the hash, an error will be returned, and the key-val
536  * 	pair will not be inserted.  i_mod_hash_insert_nosync() supports a simple
537  * 	handle abstraction, allowing hash entry allocation to be separated from
538  * 	the hash insertion.  this abstraction allows simple use of the mod_hash
539  * 	structure in situations where mod_hash_insert() with a KM_SLEEP
540  * 	allocation policy would otherwise be unsafe.
541  */
542 int
543 i_mod_hash_insert_nosync(mod_hash_t *hash, mod_hash_key_t key,
544     mod_hash_val_t val, mod_hash_hndl_t handle)
545 {
546 	uint_t hashidx;
547 	struct mod_hash_entry *entry;
548 
549 	ASSERT(hash);
550 
551 	/*
552 	 * If we've not been given reserved storage, allocate storage directly,
553 	 * using the hash's allocation policy.
554 	 */
555 	if (handle == (mod_hash_hndl_t)0) {
556 		entry = kmem_cache_alloc(mh_e_cache, hash->mh_sleep);
557 		if (entry == NULL) {
558 			hash->mh_stat.mhs_nomem++;
559 			return (MH_ERR_NOMEM);
560 		}
561 	} else {
562 		entry = (struct mod_hash_entry *)handle;
563 	}
564 
565 	hashidx = i_mod_hash(hash, key);
566 	entry->mhe_key = key;
567 	entry->mhe_val = val;
568 	entry->mhe_next = hash->mh_entries[hashidx];
569 
570 	hash->mh_entries[hashidx] = entry;
571 	hash->mh_stat.mhs_nelems++;
572 
573 	return (0);
574 }
575 
576 int
577 mod_hash_insert(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t val)
578 {
579 	int res;
580 	mod_hash_val_t v;
581 
582 	rw_enter(&hash->mh_contents, RW_WRITER);
583 
584 	/*
585 	 * Disallow duplicate keys in the hash
586 	 */
587 	if (i_mod_hash_find_nosync(hash, key, &v) == 0) {
588 		rw_exit(&hash->mh_contents);
589 		hash->mh_stat.mhs_coll++;
590 		return (MH_ERR_DUPLICATE);
591 	}
592 
593 	res = i_mod_hash_insert_nosync(hash, key, val, (mod_hash_hndl_t)0);
594 	rw_exit(&hash->mh_contents);
595 
596 	return (res);
597 }
598 
599 int
600 mod_hash_insert_reserve(mod_hash_t *hash, mod_hash_key_t key,
601     mod_hash_val_t val, mod_hash_hndl_t handle)
602 {
603 	int res;
604 	mod_hash_val_t v;
605 
606 	rw_enter(&hash->mh_contents, RW_WRITER);
607 
608 	/*
609 	 * Disallow duplicate keys in the hash
610 	 */
611 	if (i_mod_hash_find_nosync(hash, key, &v) == 0) {
612 		rw_exit(&hash->mh_contents);
613 		hash->mh_stat.mhs_coll++;
614 		return (MH_ERR_DUPLICATE);
615 	}
616 	res = i_mod_hash_insert_nosync(hash, key, val, handle);
617 	rw_exit(&hash->mh_contents);
618 
619 	return (res);
620 }
621 
622 /*
623  * mod_hash_reserve()
624  * mod_hash_reserve_nosleep()
625  * mod_hash_cancel()
626  *   Make or cancel a mod_hash_entry_t reservation.  Reservations are used in
627  *   mod_hash_insert_reserve() above.
628  */
629 int
630 mod_hash_reserve(mod_hash_t *hash, mod_hash_hndl_t *handlep)
631 {
632 	*handlep = kmem_cache_alloc(mh_e_cache, hash->mh_sleep);
633 	if (*handlep == NULL) {
634 		hash->mh_stat.mhs_nomem++;
635 		return (MH_ERR_NOMEM);
636 	}
637 
638 	return (0);
639 }
640 
641 int
642 mod_hash_reserve_nosleep(mod_hash_t *hash, mod_hash_hndl_t *handlep)
643 {
644 	*handlep = kmem_cache_alloc(mh_e_cache, KM_NOSLEEP);
645 	if (*handlep == NULL) {
646 		hash->mh_stat.mhs_nomem++;
647 		return (MH_ERR_NOMEM);
648 	}
649 
650 	return (0);
651 
652 }
653 
654 /*ARGSUSED*/
655 void
656 mod_hash_cancel(mod_hash_t *hash, mod_hash_hndl_t *handlep)
657 {
658 	kmem_cache_free(mh_e_cache, *handlep);
659 	*handlep = (mod_hash_hndl_t)0;
660 }
661 
662 /*
663  * i_mod_hash_remove_nosync()
664  * mod_hash_remove()
665  * 	Remove an element from the hash table.
666  */
667 int
668 i_mod_hash_remove_nosync(mod_hash_t *hash, mod_hash_key_t key,
669     mod_hash_val_t *val)
670 {
671 	int hashidx;
672 	struct mod_hash_entry *e, *ep;
673 
674 	hashidx = i_mod_hash(hash, key);
675 	ep = NULL; /* e's parent */
676 
677 	for (e = hash->mh_entries[hashidx]; e != NULL; e = e->mhe_next) {
678 		if (MH_KEYCMP(hash, e->mhe_key, key) == 0)
679 			break;
680 		ep = e;
681 	}
682 
683 	if (e == NULL) {	/* not found */
684 		return (MH_ERR_NOTFOUND);
685 	}
686 
687 	if (ep == NULL) 	/* special case 1st element in bucket */
688 		hash->mh_entries[hashidx] = e->mhe_next;
689 	else
690 		ep->mhe_next = e->mhe_next;
691 
692 	/*
693 	 * Clean up resources used by the node's key.
694 	 */
695 	MH_KEY_DESTROY(hash, e->mhe_key);
696 
697 	*val = e->mhe_val;
698 	kmem_cache_free(mh_e_cache, e);
699 	hash->mh_stat.mhs_nelems--;
700 
701 	return (0);
702 }
703 
704 int
705 mod_hash_remove(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t *val)
706 {
707 	int res;
708 
709 	rw_enter(&hash->mh_contents, RW_WRITER);
710 	res = i_mod_hash_remove_nosync(hash, key, val);
711 	rw_exit(&hash->mh_contents);
712 
713 	return (res);
714 }
715 
716 /*
717  * mod_hash_replace()
718  * 	atomically remove an existing key-value pair from a hash, and replace
719  * 	the key and value with the ones supplied.  The removed key and value
720  * 	(if any) are destroyed.
721  */
722 int
723 mod_hash_replace(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t val)
724 {
725 	int res;
726 	mod_hash_val_t v;
727 
728 	rw_enter(&hash->mh_contents, RW_WRITER);
729 
730 	if (i_mod_hash_remove_nosync(hash, key, &v) == 0) {
731 		/*
732 		 * mod_hash_remove() takes care of freeing up the key resources.
733 		 */
734 		MH_VAL_DESTROY(hash, v);
735 	}
736 	res = i_mod_hash_insert_nosync(hash, key, val, (mod_hash_hndl_t)0);
737 
738 	rw_exit(&hash->mh_contents);
739 
740 	return (res);
741 }
742 
743 /*
744  * mod_hash_destroy()
745  * 	Remove an element from the hash table matching 'key', and destroy it.
746  */
747 int
748 mod_hash_destroy(mod_hash_t *hash, mod_hash_key_t key)
749 {
750 	mod_hash_val_t val;
751 	int rv;
752 
753 	rw_enter(&hash->mh_contents, RW_WRITER);
754 
755 	if ((rv = i_mod_hash_remove_nosync(hash, key, &val)) == 0) {
756 		/*
757 		 * mod_hash_remove() takes care of freeing up the key resources.
758 		 */
759 		MH_VAL_DESTROY(hash, val);
760 	}
761 
762 	rw_exit(&hash->mh_contents);
763 	return (rv);
764 }
765 
766 /*
767  * i_mod_hash_find_nosync()
768  * mod_hash_find()
769  * 	Find a value in the hash table corresponding to the given key.
770  */
771 int
772 i_mod_hash_find_nosync(mod_hash_t *hash, mod_hash_key_t key,
773     mod_hash_val_t *val)
774 {
775 	uint_t hashidx;
776 	struct mod_hash_entry *e;
777 
778 	hashidx = i_mod_hash(hash, key);
779 
780 	for (e = hash->mh_entries[hashidx]; e != NULL; e = e->mhe_next) {
781 		if (MH_KEYCMP(hash, e->mhe_key, key) == 0) {
782 			*val = e->mhe_val;
783 			hash->mh_stat.mhs_hit++;
784 			return (0);
785 		}
786 	}
787 	hash->mh_stat.mhs_miss++;
788 	return (MH_ERR_NOTFOUND);
789 }
790 
791 int
792 mod_hash_find(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t *val)
793 {
794 	int res;
795 
796 	rw_enter(&hash->mh_contents, RW_READER);
797 	res = i_mod_hash_find_nosync(hash, key, val);
798 	rw_exit(&hash->mh_contents);
799 
800 	return (res);
801 }
802 
803 int
804 mod_hash_find_cb(mod_hash_t *hash, mod_hash_key_t key, mod_hash_val_t *val,
805     void (*find_cb)(mod_hash_key_t, mod_hash_val_t))
806 {
807 	int res;
808 
809 	rw_enter(&hash->mh_contents, RW_READER);
810 	res = i_mod_hash_find_nosync(hash, key, val);
811 	if (res == 0) {
812 		find_cb(key, *val);
813 	}
814 	rw_exit(&hash->mh_contents);
815 
816 	return (res);
817 }
818 
819 void
820 i_mod_hash_walk_nosync(mod_hash_t *hash,
821     uint_t (*callback)(mod_hash_key_t, mod_hash_val_t *, void *), void *arg)
822 {
823 	struct mod_hash_entry	*e;
824 	uint_t			hashidx;
825 	int			res = MH_WALK_CONTINUE;
826 
827 	for (hashidx = 0;
828 	    (hashidx < (hash->mh_nchains - 1)) && (res == MH_WALK_CONTINUE);
829 	    hashidx++) {
830 		e = hash->mh_entries[hashidx];
831 		while ((e != NULL) && (res == MH_WALK_CONTINUE)) {
832 			res = callback(e->mhe_key, e->mhe_val, arg);
833 			e = e->mhe_next;
834 		}
835 	}
836 }
837 
838 /*
839  * mod_hash_walk()
840  * 	Walks all the elements in the hashtable and invokes the callback
841  * 	function with the key/value pair for each element.  The hashtable
842  * 	is locked for readers so the callback function should not attempt
843  * 	to do any updates to the hashable.  The callback function should
844  * 	return MH_WALK_CONTINUE to continue walking the hashtable or
845  * 	MH_WALK_TERMINATE to abort the walk of the hashtable.
846  */
847 void
848 mod_hash_walk(mod_hash_t *hash,
849     uint_t (*callback)(mod_hash_key_t, mod_hash_val_t *, void *), void *arg)
850 {
851 	rw_enter(&hash->mh_contents, RW_READER);
852 	i_mod_hash_walk_nosync(hash, callback, arg);
853 	rw_exit(&hash->mh_contents);
854 }
855 
856 
857 /*
858  * i_mod_hash_clear_nosync()
859  * mod_hash_clear()
860  *	Clears the given hash table by calling the destructor of every hash
861  *	element and freeing up all mod_hash_entry's.
862  */
863 void
864 i_mod_hash_clear_nosync(mod_hash_t *hash)
865 {
866 	int i;
867 	struct mod_hash_entry *e, *old_e;
868 
869 	for (i = 0; i < hash->mh_nchains; i++) {
870 		e = hash->mh_entries[i];
871 		while (e != NULL) {
872 			MH_KEY_DESTROY(hash, e->mhe_key);
873 			MH_VAL_DESTROY(hash, e->mhe_val);
874 			old_e = e;
875 			e = e->mhe_next;
876 			kmem_cache_free(mh_e_cache, old_e);
877 		}
878 		hash->mh_entries[i] = NULL;
879 	}
880 	hash->mh_stat.mhs_nelems = 0;
881 }
882 
883 void
884 mod_hash_clear(mod_hash_t *hash)
885 {
886 	ASSERT(hash);
887 	rw_enter(&hash->mh_contents, RW_WRITER);
888 	i_mod_hash_clear_nosync(hash);
889 	rw_exit(&hash->mh_contents);
890 }
891