via-ircc: Pass PCI device pointer to dma_{alloc, free}_coherent()
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / lib / idr.c
blobe15502e8b21e5bbb47addef5c8fea61b57118508
1 /*
2 * 2002-10-18 written by Jim Houston jim.houston@ccur.com
3 * Copyright (C) 2002 by Concurrent Computer Corporation
4 * Distributed under the GNU GPL license version 2.
6 * Modified by George Anzinger to reuse immediately and to use
7 * find bit instructions. Also removed _irq on spinlocks.
9 * Modified by Nadia Derbey to make it RCU safe.
11 * Small id to pointer translation service.
13 * It uses a radix tree like structure as a sparse array indexed
14 * by the id to obtain the pointer. The bitmap makes allocating
15 * a new id quick.
17 * You call it to allocate an id (an int) an associate with that id a
18 * pointer or what ever, we treat it as a (void *). You can pass this
19 * id to a user for him to pass back at a later time. You then pass
20 * that id to this code and it returns your pointer.
22 * You can release ids at any time. When all ids are released, most of
23 * the memory is returned (we keep IDR_FREE_MAX) in a local pool so we
24 * don't need to go to the memory "store" during an id allocate, just
25 * so you don't need to be too concerned about locking and conflicts
26 * with the slab allocator.
29 #ifndef TEST // to test in user space...
30 #include <linux/slab.h>
31 #include <linux/init.h>
32 #include <linux/module.h>
33 #endif
34 #include <linux/err.h>
35 #include <linux/string.h>
36 #include <linux/idr.h>
38 static struct kmem_cache *idr_layer_cache;
40 static struct idr_layer *get_from_free_list(struct idr *idp)
42 struct idr_layer *p;
43 unsigned long flags;
45 spin_lock_irqsave(&idp->lock, flags);
46 if ((p = idp->id_free)) {
47 idp->id_free = p->ary[0];
48 idp->id_free_cnt--;
49 p->ary[0] = NULL;
51 spin_unlock_irqrestore(&idp->lock, flags);
52 return(p);
55 static void idr_layer_rcu_free(struct rcu_head *head)
57 struct idr_layer *layer;
59 layer = container_of(head, struct idr_layer, rcu_head);
60 kmem_cache_free(idr_layer_cache, layer);
63 static inline void free_layer(struct idr_layer *p)
65 call_rcu(&p->rcu_head, idr_layer_rcu_free);
68 /* only called when idp->lock is held */
69 static void __move_to_free_list(struct idr *idp, struct idr_layer *p)
71 p->ary[0] = idp->id_free;
72 idp->id_free = p;
73 idp->id_free_cnt++;
76 static void move_to_free_list(struct idr *idp, struct idr_layer *p)
78 unsigned long flags;
81 * Depends on the return element being zeroed.
83 spin_lock_irqsave(&idp->lock, flags);
84 __move_to_free_list(idp, p);
85 spin_unlock_irqrestore(&idp->lock, flags);
88 static void idr_mark_full(struct idr_layer **pa, int id)
90 struct idr_layer *p = pa[0];
91 int l = 0;
93 __set_bit(id & IDR_MASK, &p->bitmap);
95 * If this layer is full mark the bit in the layer above to
96 * show that this part of the radix tree is full. This may
97 * complete the layer above and require walking up the radix
98 * tree.
100 while (p->bitmap == IDR_FULL) {
101 if (!(p = pa[++l]))
102 break;
103 id = id >> IDR_BITS;
104 __set_bit((id & IDR_MASK), &p->bitmap);
109 * idr_pre_get - reserve resources for idr allocation
110 * @idp: idr handle
111 * @gfp_mask: memory allocation flags
113 * This function should be called prior to calling the idr_get_new* functions.
114 * It preallocates enough memory to satisfy the worst possible allocation. The
115 * caller should pass in GFP_KERNEL if possible. This of course requires that
116 * no spinning locks be held.
118 * If the system is REALLY out of memory this function returns %0,
119 * otherwise %1.
121 int idr_pre_get(struct idr *idp, gfp_t gfp_mask)
123 while (idp->id_free_cnt < IDR_FREE_MAX) {
124 struct idr_layer *new;
125 new = kmem_cache_zalloc(idr_layer_cache, gfp_mask);
126 if (new == NULL)
127 return (0);
128 move_to_free_list(idp, new);
130 return 1;
132 EXPORT_SYMBOL(idr_pre_get);
134 static int sub_alloc(struct idr *idp, int *starting_id, struct idr_layer **pa)
136 int n, m, sh;
137 struct idr_layer *p, *new;
138 int l, id, oid;
139 unsigned long bm;
141 id = *starting_id;
142 restart:
143 p = idp->top;
144 l = idp->layers;
145 pa[l--] = NULL;
146 while (1) {
148 * We run around this while until we reach the leaf node...
150 n = (id >> (IDR_BITS*l)) & IDR_MASK;
151 bm = ~p->bitmap;
152 m = find_next_bit(&bm, IDR_SIZE, n);
153 if (m == IDR_SIZE) {
154 /* no space available go back to previous layer. */
155 l++;
156 oid = id;
157 id = (id | ((1 << (IDR_BITS * l)) - 1)) + 1;
159 /* if already at the top layer, we need to grow */
160 if (id >= 1 << (idp->layers * IDR_BITS)) {
161 *starting_id = id;
162 return IDR_NEED_TO_GROW;
164 p = pa[l];
165 BUG_ON(!p);
167 /* If we need to go up one layer, continue the
168 * loop; otherwise, restart from the top.
170 sh = IDR_BITS * (l + 1);
171 if (oid >> sh == id >> sh)
172 continue;
173 else
174 goto restart;
176 if (m != n) {
177 sh = IDR_BITS*l;
178 id = ((id >> sh) ^ n ^ m) << sh;
180 if ((id >= MAX_ID_BIT) || (id < 0))
181 return IDR_NOMORE_SPACE;
182 if (l == 0)
183 break;
185 * Create the layer below if it is missing.
187 if (!p->ary[m]) {
188 new = get_from_free_list(idp);
189 if (!new)
190 return -1;
191 new->layer = l-1;
192 rcu_assign_pointer(p->ary[m], new);
193 p->count++;
195 pa[l--] = p;
196 p = p->ary[m];
199 pa[l] = p;
200 return id;
203 static int idr_get_empty_slot(struct idr *idp, int starting_id,
204 struct idr_layer **pa)
206 struct idr_layer *p, *new;
207 int layers, v, id;
208 unsigned long flags;
210 id = starting_id;
211 build_up:
212 p = idp->top;
213 layers = idp->layers;
214 if (unlikely(!p)) {
215 if (!(p = get_from_free_list(idp)))
216 return -1;
217 p->layer = 0;
218 layers = 1;
221 * Add a new layer to the top of the tree if the requested
222 * id is larger than the currently allocated space.
224 while ((layers < (MAX_LEVEL - 1)) && (id >= (1 << (layers*IDR_BITS)))) {
225 layers++;
226 if (!p->count) {
227 /* special case: if the tree is currently empty,
228 * then we grow the tree by moving the top node
229 * upwards.
231 p->layer++;
232 continue;
234 if (!(new = get_from_free_list(idp))) {
236 * The allocation failed. If we built part of
237 * the structure tear it down.
239 spin_lock_irqsave(&idp->lock, flags);
240 for (new = p; p && p != idp->top; new = p) {
241 p = p->ary[0];
242 new->ary[0] = NULL;
243 new->bitmap = new->count = 0;
244 __move_to_free_list(idp, new);
246 spin_unlock_irqrestore(&idp->lock, flags);
247 return -1;
249 new->ary[0] = p;
250 new->count = 1;
251 new->layer = layers-1;
252 if (p->bitmap == IDR_FULL)
253 __set_bit(0, &new->bitmap);
254 p = new;
256 rcu_assign_pointer(idp->top, p);
257 idp->layers = layers;
258 v = sub_alloc(idp, &id, pa);
259 if (v == IDR_NEED_TO_GROW)
260 goto build_up;
261 return(v);
264 static int idr_get_new_above_int(struct idr *idp, void *ptr, int starting_id)
266 struct idr_layer *pa[MAX_LEVEL];
267 int id;
269 id = idr_get_empty_slot(idp, starting_id, pa);
270 if (id >= 0) {
272 * Successfully found an empty slot. Install the user
273 * pointer and mark the slot full.
275 rcu_assign_pointer(pa[0]->ary[id & IDR_MASK],
276 (struct idr_layer *)ptr);
277 pa[0]->count++;
278 idr_mark_full(pa, id);
281 return id;
285 * idr_get_new_above - allocate new idr entry above or equal to a start id
286 * @idp: idr handle
287 * @ptr: pointer you want associated with the id
288 * @starting_id: id to start search at
289 * @id: pointer to the allocated handle
291 * This is the allocate id function. It should be called with any
292 * required locks.
294 * If allocation from IDR's private freelist fails, idr_get_new_above() will
295 * return %-EAGAIN. The caller should retry the idr_pre_get() call to refill
296 * IDR's preallocation and then retry the idr_get_new_above() call.
298 * If the idr is full idr_get_new_above() will return %-ENOSPC.
300 * @id returns a value in the range @starting_id ... %0x7fffffff
302 int idr_get_new_above(struct idr *idp, void *ptr, int starting_id, int *id)
304 int rv;
306 rv = idr_get_new_above_int(idp, ptr, starting_id);
308 * This is a cheap hack until the IDR code can be fixed to
309 * return proper error values.
311 if (rv < 0)
312 return _idr_rc_to_errno(rv);
313 *id = rv;
314 return 0;
316 EXPORT_SYMBOL(idr_get_new_above);
319 * idr_get_new - allocate new idr entry
320 * @idp: idr handle
321 * @ptr: pointer you want associated with the id
322 * @id: pointer to the allocated handle
324 * If allocation from IDR's private freelist fails, idr_get_new_above() will
325 * return %-EAGAIN. The caller should retry the idr_pre_get() call to refill
326 * IDR's preallocation and then retry the idr_get_new_above() call.
328 * If the idr is full idr_get_new_above() will return %-ENOSPC.
330 * @id returns a value in the range %0 ... %0x7fffffff
332 int idr_get_new(struct idr *idp, void *ptr, int *id)
334 int rv;
336 rv = idr_get_new_above_int(idp, ptr, 0);
338 * This is a cheap hack until the IDR code can be fixed to
339 * return proper error values.
341 if (rv < 0)
342 return _idr_rc_to_errno(rv);
343 *id = rv;
344 return 0;
346 EXPORT_SYMBOL(idr_get_new);
348 static void idr_remove_warning(int id)
350 printk(KERN_WARNING
351 "idr_remove called for id=%d which is not allocated.\n", id);
352 dump_stack();
355 static void sub_remove(struct idr *idp, int shift, int id)
357 struct idr_layer *p = idp->top;
358 struct idr_layer **pa[MAX_LEVEL];
359 struct idr_layer ***paa = &pa[0];
360 struct idr_layer *to_free;
361 int n;
363 *paa = NULL;
364 *++paa = &idp->top;
366 while ((shift > 0) && p) {
367 n = (id >> shift) & IDR_MASK;
368 __clear_bit(n, &p->bitmap);
369 *++paa = &p->ary[n];
370 p = p->ary[n];
371 shift -= IDR_BITS;
373 n = id & IDR_MASK;
374 if (likely(p != NULL && test_bit(n, &p->bitmap))){
375 __clear_bit(n, &p->bitmap);
376 rcu_assign_pointer(p->ary[n], NULL);
377 to_free = NULL;
378 while(*paa && ! --((**paa)->count)){
379 if (to_free)
380 free_layer(to_free);
381 to_free = **paa;
382 **paa-- = NULL;
384 if (!*paa)
385 idp->layers = 0;
386 if (to_free)
387 free_layer(to_free);
388 } else
389 idr_remove_warning(id);
393 * idr_remove - remove the given id and free its slot
394 * @idp: idr handle
395 * @id: unique key
397 void idr_remove(struct idr *idp, int id)
399 struct idr_layer *p;
400 struct idr_layer *to_free;
402 /* Mask off upper bits we don't use for the search. */
403 id &= MAX_ID_MASK;
405 sub_remove(idp, (idp->layers - 1) * IDR_BITS, id);
406 if (idp->top && idp->top->count == 1 && (idp->layers > 1) &&
407 idp->top->ary[0]) {
409 * Single child at leftmost slot: we can shrink the tree.
410 * This level is not needed anymore since when layers are
411 * inserted, they are inserted at the top of the existing
412 * tree.
414 to_free = idp->top;
415 p = idp->top->ary[0];
416 rcu_assign_pointer(idp->top, p);
417 --idp->layers;
418 to_free->bitmap = to_free->count = 0;
419 free_layer(to_free);
421 while (idp->id_free_cnt >= IDR_FREE_MAX) {
422 p = get_from_free_list(idp);
424 * Note: we don't call the rcu callback here, since the only
425 * layers that fall into the freelist are those that have been
426 * preallocated.
428 kmem_cache_free(idr_layer_cache, p);
430 return;
432 EXPORT_SYMBOL(idr_remove);
435 * idr_remove_all - remove all ids from the given idr tree
436 * @idp: idr handle
438 * idr_destroy() only frees up unused, cached idp_layers, but this
439 * function will remove all id mappings and leave all idp_layers
440 * unused.
442 * A typical clean-up sequence for objects stored in an idr tree will
443 * use idr_for_each() to free all objects, if necessay, then
444 * idr_remove_all() to remove all ids, and idr_destroy() to free
445 * up the cached idr_layers.
447 void idr_remove_all(struct idr *idp)
449 int n, id, max;
450 int bt_mask;
451 struct idr_layer *p;
452 struct idr_layer *pa[MAX_LEVEL];
453 struct idr_layer **paa = &pa[0];
455 n = idp->layers * IDR_BITS;
456 p = idp->top;
457 rcu_assign_pointer(idp->top, NULL);
458 max = 1 << n;
460 id = 0;
461 while (id < max) {
462 while (n > IDR_BITS && p) {
463 n -= IDR_BITS;
464 *paa++ = p;
465 p = p->ary[(id >> n) & IDR_MASK];
468 bt_mask = id;
469 id += 1 << n;
470 /* Get the highest bit that the above add changed from 0->1. */
471 while (n < fls(id ^ bt_mask)) {
472 if (p)
473 free_layer(p);
474 n += IDR_BITS;
475 p = *--paa;
478 idp->layers = 0;
480 EXPORT_SYMBOL(idr_remove_all);
483 * idr_destroy - release all cached layers within an idr tree
484 * @idp: idr handle
486 void idr_destroy(struct idr *idp)
488 while (idp->id_free_cnt) {
489 struct idr_layer *p = get_from_free_list(idp);
490 kmem_cache_free(idr_layer_cache, p);
493 EXPORT_SYMBOL(idr_destroy);
496 * idr_find - return pointer for given id
497 * @idp: idr handle
498 * @id: lookup key
500 * Return the pointer given the id it has been registered with. A %NULL
501 * return indicates that @id is not valid or you passed %NULL in
502 * idr_get_new().
504 * This function can be called under rcu_read_lock(), given that the leaf
505 * pointers lifetimes are correctly managed.
507 void *idr_find(struct idr *idp, int id)
509 int n;
510 struct idr_layer *p;
512 p = rcu_dereference_raw(idp->top);
513 if (!p)
514 return NULL;
515 n = (p->layer+1) * IDR_BITS;
517 /* Mask off upper bits we don't use for the search. */
518 id &= MAX_ID_MASK;
520 if (id >= (1 << n))
521 return NULL;
522 BUG_ON(n == 0);
524 while (n > 0 && p) {
525 n -= IDR_BITS;
526 BUG_ON(n != p->layer*IDR_BITS);
527 p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
529 return((void *)p);
531 EXPORT_SYMBOL(idr_find);
534 * idr_for_each - iterate through all stored pointers
535 * @idp: idr handle
536 * @fn: function to be called for each pointer
537 * @data: data passed back to callback function
539 * Iterate over the pointers registered with the given idr. The
540 * callback function will be called for each pointer currently
541 * registered, passing the id, the pointer and the data pointer passed
542 * to this function. It is not safe to modify the idr tree while in
543 * the callback, so functions such as idr_get_new and idr_remove are
544 * not allowed.
546 * We check the return of @fn each time. If it returns anything other
547 * than %0, we break out and return that value.
549 * The caller must serialize idr_for_each() vs idr_get_new() and idr_remove().
551 int idr_for_each(struct idr *idp,
552 int (*fn)(int id, void *p, void *data), void *data)
554 int n, id, max, error = 0;
555 struct idr_layer *p;
556 struct idr_layer *pa[MAX_LEVEL];
557 struct idr_layer **paa = &pa[0];
559 n = idp->layers * IDR_BITS;
560 p = rcu_dereference_raw(idp->top);
561 max = 1 << n;
563 id = 0;
564 while (id < max) {
565 while (n > 0 && p) {
566 n -= IDR_BITS;
567 *paa++ = p;
568 p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
571 if (p) {
572 error = fn(id, (void *)p, data);
573 if (error)
574 break;
577 id += 1 << n;
578 while (n < fls(id)) {
579 n += IDR_BITS;
580 p = *--paa;
584 return error;
586 EXPORT_SYMBOL(idr_for_each);
589 * idr_get_next - lookup next object of id to given id.
590 * @idp: idr handle
591 * @nextidp: pointer to lookup key
593 * Returns pointer to registered object with id, which is next number to
594 * given id. After being looked up, *@nextidp will be updated for the next
595 * iteration.
598 void *idr_get_next(struct idr *idp, int *nextidp)
600 struct idr_layer *p, *pa[MAX_LEVEL];
601 struct idr_layer **paa = &pa[0];
602 int id = *nextidp;
603 int n, max;
605 /* find first ent */
606 n = idp->layers * IDR_BITS;
607 max = 1 << n;
608 p = rcu_dereference_raw(idp->top);
609 if (!p)
610 return NULL;
612 while (id < max) {
613 while (n > 0 && p) {
614 n -= IDR_BITS;
615 *paa++ = p;
616 p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
619 if (p) {
620 *nextidp = id;
621 return p;
624 id += 1 << n;
625 while (n < fls(id)) {
626 n += IDR_BITS;
627 p = *--paa;
630 return NULL;
632 EXPORT_SYMBOL(idr_get_next);
636 * idr_replace - replace pointer for given id
637 * @idp: idr handle
638 * @ptr: pointer you want associated with the id
639 * @id: lookup key
641 * Replace the pointer registered with an id and return the old value.
642 * A %-ENOENT return indicates that @id was not found.
643 * A %-EINVAL return indicates that @id was not within valid constraints.
645 * The caller must serialize with writers.
647 void *idr_replace(struct idr *idp, void *ptr, int id)
649 int n;
650 struct idr_layer *p, *old_p;
652 p = idp->top;
653 if (!p)
654 return ERR_PTR(-EINVAL);
656 n = (p->layer+1) * IDR_BITS;
658 id &= MAX_ID_MASK;
660 if (id >= (1 << n))
661 return ERR_PTR(-EINVAL);
663 n -= IDR_BITS;
664 while ((n > 0) && p) {
665 p = p->ary[(id >> n) & IDR_MASK];
666 n -= IDR_BITS;
669 n = id & IDR_MASK;
670 if (unlikely(p == NULL || !test_bit(n, &p->bitmap)))
671 return ERR_PTR(-ENOENT);
673 old_p = p->ary[n];
674 rcu_assign_pointer(p->ary[n], ptr);
676 return old_p;
678 EXPORT_SYMBOL(idr_replace);
680 void __init idr_init_cache(void)
682 idr_layer_cache = kmem_cache_create("idr_layer_cache",
683 sizeof(struct idr_layer), 0, SLAB_PANIC, NULL);
687 * idr_init - initialize idr handle
688 * @idp: idr handle
690 * This function is use to set up the handle (@idp) that you will pass
691 * to the rest of the functions.
693 void idr_init(struct idr *idp)
695 memset(idp, 0, sizeof(struct idr));
696 spin_lock_init(&idp->lock);
698 EXPORT_SYMBOL(idr_init);
702 * DOC: IDA description
703 * IDA - IDR based ID allocator
705 * This is id allocator without id -> pointer translation. Memory
706 * usage is much lower than full blown idr because each id only
707 * occupies a bit. ida uses a custom leaf node which contains
708 * IDA_BITMAP_BITS slots.
710 * 2007-04-25 written by Tejun Heo <htejun@gmail.com>
713 static void free_bitmap(struct ida *ida, struct ida_bitmap *bitmap)
715 unsigned long flags;
717 if (!ida->free_bitmap) {
718 spin_lock_irqsave(&ida->idr.lock, flags);
719 if (!ida->free_bitmap) {
720 ida->free_bitmap = bitmap;
721 bitmap = NULL;
723 spin_unlock_irqrestore(&ida->idr.lock, flags);
726 kfree(bitmap);
730 * ida_pre_get - reserve resources for ida allocation
731 * @ida: ida handle
732 * @gfp_mask: memory allocation flag
734 * This function should be called prior to locking and calling the
735 * following function. It preallocates enough memory to satisfy the
736 * worst possible allocation.
738 * If the system is REALLY out of memory this function returns %0,
739 * otherwise %1.
741 int ida_pre_get(struct ida *ida, gfp_t gfp_mask)
743 /* allocate idr_layers */
744 if (!idr_pre_get(&ida->idr, gfp_mask))
745 return 0;
747 /* allocate free_bitmap */
748 if (!ida->free_bitmap) {
749 struct ida_bitmap *bitmap;
751 bitmap = kmalloc(sizeof(struct ida_bitmap), gfp_mask);
752 if (!bitmap)
753 return 0;
755 free_bitmap(ida, bitmap);
758 return 1;
760 EXPORT_SYMBOL(ida_pre_get);
763 * ida_get_new_above - allocate new ID above or equal to a start id
764 * @ida: ida handle
765 * @starting_id: id to start search at
766 * @p_id: pointer to the allocated handle
768 * Allocate new ID above or equal to @ida. It should be called with
769 * any required locks.
771 * If memory is required, it will return %-EAGAIN, you should unlock
772 * and go back to the ida_pre_get() call. If the ida is full, it will
773 * return %-ENOSPC.
775 * @p_id returns a value in the range @starting_id ... %0x7fffffff.
777 int ida_get_new_above(struct ida *ida, int starting_id, int *p_id)
779 struct idr_layer *pa[MAX_LEVEL];
780 struct ida_bitmap *bitmap;
781 unsigned long flags;
782 int idr_id = starting_id / IDA_BITMAP_BITS;
783 int offset = starting_id % IDA_BITMAP_BITS;
784 int t, id;
786 restart:
787 /* get vacant slot */
788 t = idr_get_empty_slot(&ida->idr, idr_id, pa);
789 if (t < 0)
790 return _idr_rc_to_errno(t);
792 if (t * IDA_BITMAP_BITS >= MAX_ID_BIT)
793 return -ENOSPC;
795 if (t != idr_id)
796 offset = 0;
797 idr_id = t;
799 /* if bitmap isn't there, create a new one */
800 bitmap = (void *)pa[0]->ary[idr_id & IDR_MASK];
801 if (!bitmap) {
802 spin_lock_irqsave(&ida->idr.lock, flags);
803 bitmap = ida->free_bitmap;
804 ida->free_bitmap = NULL;
805 spin_unlock_irqrestore(&ida->idr.lock, flags);
807 if (!bitmap)
808 return -EAGAIN;
810 memset(bitmap, 0, sizeof(struct ida_bitmap));
811 rcu_assign_pointer(pa[0]->ary[idr_id & IDR_MASK],
812 (void *)bitmap);
813 pa[0]->count++;
816 /* lookup for empty slot */
817 t = find_next_zero_bit(bitmap->bitmap, IDA_BITMAP_BITS, offset);
818 if (t == IDA_BITMAP_BITS) {
819 /* no empty slot after offset, continue to the next chunk */
820 idr_id++;
821 offset = 0;
822 goto restart;
825 id = idr_id * IDA_BITMAP_BITS + t;
826 if (id >= MAX_ID_BIT)
827 return -ENOSPC;
829 __set_bit(t, bitmap->bitmap);
830 if (++bitmap->nr_busy == IDA_BITMAP_BITS)
831 idr_mark_full(pa, idr_id);
833 *p_id = id;
835 /* Each leaf node can handle nearly a thousand slots and the
836 * whole idea of ida is to have small memory foot print.
837 * Throw away extra resources one by one after each successful
838 * allocation.
840 if (ida->idr.id_free_cnt || ida->free_bitmap) {
841 struct idr_layer *p = get_from_free_list(&ida->idr);
842 if (p)
843 kmem_cache_free(idr_layer_cache, p);
846 return 0;
848 EXPORT_SYMBOL(ida_get_new_above);
851 * ida_get_new - allocate new ID
852 * @ida: idr handle
853 * @p_id: pointer to the allocated handle
855 * Allocate new ID. It should be called with any required locks.
857 * If memory is required, it will return %-EAGAIN, you should unlock
858 * and go back to the idr_pre_get() call. If the idr is full, it will
859 * return %-ENOSPC.
861 * @id returns a value in the range %0 ... %0x7fffffff.
863 int ida_get_new(struct ida *ida, int *p_id)
865 return ida_get_new_above(ida, 0, p_id);
867 EXPORT_SYMBOL(ida_get_new);
870 * ida_remove - remove the given ID
871 * @ida: ida handle
872 * @id: ID to free
874 void ida_remove(struct ida *ida, int id)
876 struct idr_layer *p = ida->idr.top;
877 int shift = (ida->idr.layers - 1) * IDR_BITS;
878 int idr_id = id / IDA_BITMAP_BITS;
879 int offset = id % IDA_BITMAP_BITS;
880 int n;
881 struct ida_bitmap *bitmap;
883 /* clear full bits while looking up the leaf idr_layer */
884 while ((shift > 0) && p) {
885 n = (idr_id >> shift) & IDR_MASK;
886 __clear_bit(n, &p->bitmap);
887 p = p->ary[n];
888 shift -= IDR_BITS;
891 if (p == NULL)
892 goto err;
894 n = idr_id & IDR_MASK;
895 __clear_bit(n, &p->bitmap);
897 bitmap = (void *)p->ary[n];
898 if (!test_bit(offset, bitmap->bitmap))
899 goto err;
901 /* update bitmap and remove it if empty */
902 __clear_bit(offset, bitmap->bitmap);
903 if (--bitmap->nr_busy == 0) {
904 __set_bit(n, &p->bitmap); /* to please idr_remove() */
905 idr_remove(&ida->idr, idr_id);
906 free_bitmap(ida, bitmap);
909 return;
911 err:
912 printk(KERN_WARNING
913 "ida_remove called for id=%d which is not allocated.\n", id);
915 EXPORT_SYMBOL(ida_remove);
918 * ida_destroy - release all cached layers within an ida tree
919 * @ida: ida handle
921 void ida_destroy(struct ida *ida)
923 idr_destroy(&ida->idr);
924 kfree(ida->free_bitmap);
926 EXPORT_SYMBOL(ida_destroy);
929 * ida_init - initialize ida handle
930 * @ida: ida handle
932 * This function is use to set up the handle (@ida) that you will pass
933 * to the rest of the functions.
935 void ida_init(struct ida *ida)
937 memset(ida, 0, sizeof(struct ida));
938 idr_init(&ida->idr);
941 EXPORT_SYMBOL(ida_init);