2 * Copyright (C) 2001 Momchil Velikov
3 * Portions Copyright (C) 2001 Christoph Hellwig
4 * Copyright (C) 2006 Nick Piggin
5 * Copyright (C) 2012 Konstantin Khlebnikov
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License as
9 * published by the Free Software Foundation; either version 2, or (at
10 * your option) any later version.
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 for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 #ifndef _LINUX_RADIX_TREE_H
22 #define _LINUX_RADIX_TREE_H
24 #include <linux/bitops.h>
25 #include <linux/preempt.h>
26 #include <linux/types.h>
27 #include <linux/bug.h>
28 #include <linux/kernel.h>
29 #include <linux/rcupdate.h>
32 * The bottom two bits of the slot determine how the remaining bits in the
33 * slot are interpreted:
37 * 10 - exceptional entry
38 * 11 - this bit combination is currently unused/reserved
40 * The internal entry may be a pointer to the next level in the tree, a
41 * sibling entry, or an indicator that the entry in this slot has been moved
42 * to another location in the tree and the lookup should be restarted. While
43 * NULL fits the 'data pointer' pattern, it means that there is no entry in
44 * the tree for this index (no matter what level of the tree it is found at).
45 * This means that you cannot store NULL in the tree as a value for the index.
47 #define RADIX_TREE_ENTRY_MASK 3UL
48 #define RADIX_TREE_INTERNAL_NODE 1UL
51 * Most users of the radix tree store pointers but shmem/tmpfs stores swap
52 * entries in the same tree. They are marked as exceptional entries to
53 * distinguish them from pointers to struct page.
54 * EXCEPTIONAL_ENTRY tests the bit, EXCEPTIONAL_SHIFT shifts content past it.
56 #define RADIX_TREE_EXCEPTIONAL_ENTRY 2
57 #define RADIX_TREE_EXCEPTIONAL_SHIFT 2
59 static inline bool radix_tree_is_internal_node(void *ptr
)
61 return ((unsigned long)ptr
& RADIX_TREE_ENTRY_MASK
) ==
62 RADIX_TREE_INTERNAL_NODE
;
65 /*** radix-tree API starts here ***/
67 #define RADIX_TREE_MAX_TAGS 3
69 #ifndef RADIX_TREE_MAP_SHIFT
70 #define RADIX_TREE_MAP_SHIFT (CONFIG_BASE_SMALL ? 4 : 6)
73 #define RADIX_TREE_MAP_SIZE (1UL << RADIX_TREE_MAP_SHIFT)
74 #define RADIX_TREE_MAP_MASK (RADIX_TREE_MAP_SIZE-1)
76 #define RADIX_TREE_TAG_LONGS \
77 ((RADIX_TREE_MAP_SIZE + BITS_PER_LONG - 1) / BITS_PER_LONG)
79 #define RADIX_TREE_INDEX_BITS (8 /* CHAR_BIT */ * sizeof(unsigned long))
80 #define RADIX_TREE_MAX_PATH (DIV_ROUND_UP(RADIX_TREE_INDEX_BITS, \
81 RADIX_TREE_MAP_SHIFT))
83 /* Internally used bits of node->count */
84 #define RADIX_TREE_COUNT_SHIFT (RADIX_TREE_MAP_SHIFT + 1)
85 #define RADIX_TREE_COUNT_MASK ((1UL << RADIX_TREE_COUNT_SHIFT) - 1)
87 struct radix_tree_node
{
88 unsigned char shift
; /* Bits remaining in each slot */
89 unsigned char offset
; /* Slot offset in parent */
90 unsigned int count
; /* Total entry count */
91 unsigned char exceptional
; /* Exceptional entry count */
94 /* Used when ascending tree */
95 struct radix_tree_node
*parent
;
99 /* Used when freeing node */
100 struct rcu_head rcu_head
;
103 struct list_head private_list
;
104 void __rcu
*slots
[RADIX_TREE_MAP_SIZE
];
105 unsigned long tags
[RADIX_TREE_MAX_TAGS
][RADIX_TREE_TAG_LONGS
];
108 /* root tags are stored in gfp_mask, shifted by __GFP_BITS_SHIFT */
109 struct radix_tree_root
{
111 struct radix_tree_node __rcu
*rnode
;
114 #define RADIX_TREE_INIT(mask) { \
115 .gfp_mask = (mask), \
119 #define RADIX_TREE(name, mask) \
120 struct radix_tree_root name = RADIX_TREE_INIT(mask)
122 #define INIT_RADIX_TREE(root, mask) \
124 (root)->gfp_mask = (mask); \
125 (root)->rnode = NULL; \
128 static inline bool radix_tree_empty(struct radix_tree_root
*root
)
130 return root
->rnode
== NULL
;
134 * Radix-tree synchronization
136 * The radix-tree API requires that users provide all synchronisation (with
137 * specific exceptions, noted below).
139 * Synchronization of access to the data items being stored in the tree, and
140 * management of their lifetimes must be completely managed by API users.
142 * For API usage, in general,
143 * - any function _modifying_ the tree or tags (inserting or deleting
144 * items, setting or clearing tags) must exclude other modifications, and
145 * exclude any functions reading the tree.
146 * - any function _reading_ the tree or tags (looking up items or tags,
147 * gang lookups) must exclude modifications to the tree, but may occur
148 * concurrently with other readers.
150 * The notable exceptions to this rule are the following functions:
151 * __radix_tree_lookup
153 * radix_tree_lookup_slot
155 * radix_tree_gang_lookup
156 * radix_tree_gang_lookup_slot
157 * radix_tree_gang_lookup_tag
158 * radix_tree_gang_lookup_tag_slot
161 * The first 8 functions are able to be called locklessly, using RCU. The
162 * caller must ensure calls to these functions are made within rcu_read_lock()
163 * regions. Other readers (lock-free or otherwise) and modifications may be
164 * running concurrently.
166 * It is still required that the caller manage the synchronization and lifetimes
167 * of the items. So if RCU lock-free lookups are used, typically this would mean
168 * that the items have their own locks, or are amenable to lock-free access; and
169 * that the items are freed by RCU (or only freed after having been deleted from
170 * the radix tree *and* a synchronize_rcu() grace period).
172 * (Note, rcu_assign_pointer and rcu_dereference are not needed to control
173 * access to data items when inserting into or looking up from the radix tree)
175 * Note that the value returned by radix_tree_tag_get() may not be relied upon
176 * if only the RCU read lock is held. Functions to set/clear tags and to
177 * delete nodes running concurrently with it may affect its result such that
178 * two consecutive reads in the same locked section may return different
179 * values. If reliability is required, modification functions must also be
180 * excluded from concurrency.
182 * radix_tree_tagged is able to be called without locking or RCU.
186 * radix_tree_deref_slot - dereference a slot
187 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
188 * Returns: item that was stored in that slot with any direct pointer flag
191 * For use with radix_tree_lookup_slot(). Caller must hold tree at least read
192 * locked across slot lookup and dereference. Not required if write lock is
193 * held (ie. items cannot be concurrently inserted).
195 * radix_tree_deref_retry must be used to confirm validity of the pointer if
196 * only the read lock is held.
198 static inline void *radix_tree_deref_slot(void **pslot
)
200 return rcu_dereference(*pslot
);
204 * radix_tree_deref_slot_protected - dereference a slot without RCU lock but with tree lock held
205 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
206 * Returns: item that was stored in that slot with any direct pointer flag
209 * Similar to radix_tree_deref_slot but only used during migration when a pages
210 * mapping is being moved. The caller does not hold the RCU read lock but it
211 * must hold the tree lock to prevent parallel updates.
213 static inline void *radix_tree_deref_slot_protected(void **pslot
,
214 spinlock_t
*treelock
)
216 return rcu_dereference_protected(*pslot
, lockdep_is_held(treelock
));
220 * radix_tree_deref_retry - check radix_tree_deref_slot
221 * @arg: pointer returned by radix_tree_deref_slot
222 * Returns: 0 if retry is not required, otherwise retry is required
224 * radix_tree_deref_retry must be used with radix_tree_deref_slot.
226 static inline int radix_tree_deref_retry(void *arg
)
228 return unlikely(radix_tree_is_internal_node(arg
));
232 * radix_tree_exceptional_entry - radix_tree_deref_slot gave exceptional entry?
233 * @arg: value returned by radix_tree_deref_slot
234 * Returns: 0 if well-aligned pointer, non-0 if exceptional entry.
236 static inline int radix_tree_exceptional_entry(void *arg
)
238 /* Not unlikely because radix_tree_exception often tested first */
239 return (unsigned long)arg
& RADIX_TREE_EXCEPTIONAL_ENTRY
;
243 * radix_tree_exception - radix_tree_deref_slot returned either exception?
244 * @arg: value returned by radix_tree_deref_slot
245 * Returns: 0 if well-aligned pointer, non-0 if either kind of exception.
247 static inline int radix_tree_exception(void *arg
)
249 return unlikely((unsigned long)arg
& RADIX_TREE_ENTRY_MASK
);
252 int __radix_tree_create(struct radix_tree_root
*root
, unsigned long index
,
253 unsigned order
, struct radix_tree_node
**nodep
,
255 int __radix_tree_insert(struct radix_tree_root
*, unsigned long index
,
256 unsigned order
, void *);
257 static inline int radix_tree_insert(struct radix_tree_root
*root
,
258 unsigned long index
, void *entry
)
260 return __radix_tree_insert(root
, index
, 0, entry
);
262 void *__radix_tree_lookup(struct radix_tree_root
*root
, unsigned long index
,
263 struct radix_tree_node
**nodep
, void ***slotp
);
264 void *radix_tree_lookup(struct radix_tree_root
*, unsigned long);
265 void **radix_tree_lookup_slot(struct radix_tree_root
*, unsigned long);
266 void __radix_tree_replace(struct radix_tree_root
*root
,
267 struct radix_tree_node
*node
,
268 void **slot
, void *item
);
269 void radix_tree_replace_slot(struct radix_tree_root
*root
,
270 void **slot
, void *item
);
271 bool __radix_tree_delete_node(struct radix_tree_root
*root
,
272 struct radix_tree_node
*node
);
273 void *radix_tree_delete_item(struct radix_tree_root
*, unsigned long, void *);
274 void *radix_tree_delete(struct radix_tree_root
*, unsigned long);
275 void radix_tree_clear_tags(struct radix_tree_root
*root
,
276 struct radix_tree_node
*node
,
278 unsigned int radix_tree_gang_lookup(struct radix_tree_root
*root
,
279 void **results
, unsigned long first_index
,
280 unsigned int max_items
);
281 unsigned int radix_tree_gang_lookup_slot(struct radix_tree_root
*root
,
282 void ***results
, unsigned long *indices
,
283 unsigned long first_index
, unsigned int max_items
);
284 int radix_tree_preload(gfp_t gfp_mask
);
285 int radix_tree_maybe_preload(gfp_t gfp_mask
);
286 int radix_tree_maybe_preload_order(gfp_t gfp_mask
, int order
);
287 void radix_tree_init(void);
288 void *radix_tree_tag_set(struct radix_tree_root
*root
,
289 unsigned long index
, unsigned int tag
);
290 void *radix_tree_tag_clear(struct radix_tree_root
*root
,
291 unsigned long index
, unsigned int tag
);
292 int radix_tree_tag_get(struct radix_tree_root
*root
,
293 unsigned long index
, unsigned int tag
);
295 radix_tree_gang_lookup_tag(struct radix_tree_root
*root
, void **results
,
296 unsigned long first_index
, unsigned int max_items
,
299 radix_tree_gang_lookup_tag_slot(struct radix_tree_root
*root
, void ***results
,
300 unsigned long first_index
, unsigned int max_items
,
302 unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root
*root
,
303 unsigned long *first_indexp
, unsigned long last_index
,
304 unsigned long nr_to_tag
,
305 unsigned int fromtag
, unsigned int totag
);
306 int radix_tree_tagged(struct radix_tree_root
*root
, unsigned int tag
);
307 unsigned long radix_tree_locate_item(struct radix_tree_root
*root
, void *item
);
309 static inline void radix_tree_preload_end(void)
315 * struct radix_tree_iter - radix tree iterator state
317 * @index: index of current slot
318 * @next_index: one beyond the last index for this chunk
319 * @tags: bit-mask for tag-iterating
320 * @shift: shift for the node that holds our slots
322 * This radix tree iterator works in terms of "chunks" of slots. A chunk is a
323 * subinterval of slots contained within one radix tree leaf node. It is
324 * described by a pointer to its first slot and a struct radix_tree_iter
325 * which holds the chunk's position in the tree and its size. For tagged
326 * iteration radix_tree_iter also holds the slots' bit-mask for one chosen
329 struct radix_tree_iter
{
331 unsigned long next_index
;
333 #ifdef CONFIG_RADIX_TREE_MULTIORDER
338 static inline unsigned int iter_shift(struct radix_tree_iter
*iter
)
340 #ifdef CONFIG_RADIX_TREE_MULTIORDER
347 #define RADIX_TREE_ITER_TAG_MASK 0x00FF /* tag index in lower byte */
348 #define RADIX_TREE_ITER_TAGGED 0x0100 /* lookup tagged slots */
349 #define RADIX_TREE_ITER_CONTIG 0x0200 /* stop at first hole */
352 * radix_tree_iter_init - initialize radix tree iterator
354 * @iter: pointer to iterator state
355 * @start: iteration starting index
358 static __always_inline
void **
359 radix_tree_iter_init(struct radix_tree_iter
*iter
, unsigned long start
)
362 * Leave iter->tags uninitialized. radix_tree_next_chunk() will fill it
363 * in the case of a successful tagged chunk lookup. If the lookup was
364 * unsuccessful or non-tagged then nobody cares about ->tags.
366 * Set index to zero to bypass next_index overflow protection.
367 * See the comment in radix_tree_next_chunk() for details.
370 iter
->next_index
= start
;
375 * radix_tree_next_chunk - find next chunk of slots for iteration
377 * @root: radix tree root
378 * @iter: iterator state
379 * @flags: RADIX_TREE_ITER_* flags and tag index
380 * Returns: pointer to chunk first slot, or NULL if there no more left
382 * This function looks up the next chunk in the radix tree starting from
383 * @iter->next_index. It returns a pointer to the chunk's first slot.
384 * Also it fills @iter with data about chunk: position in the tree (index),
385 * its end (next_index), and constructs a bit mask for tagged iterating (tags).
387 void **radix_tree_next_chunk(struct radix_tree_root
*root
,
388 struct radix_tree_iter
*iter
, unsigned flags
);
391 * radix_tree_iter_retry - retry this chunk of the iteration
392 * @iter: iterator state
394 * If we iterate over a tree protected only by the RCU lock, a race
395 * against deletion or creation may result in seeing a slot for which
396 * radix_tree_deref_retry() returns true. If so, call this function
397 * and continue the iteration.
399 static inline __must_check
400 void **radix_tree_iter_retry(struct radix_tree_iter
*iter
)
402 iter
->next_index
= iter
->index
;
407 static inline unsigned long
408 __radix_tree_iter_add(struct radix_tree_iter
*iter
, unsigned long slots
)
410 return iter
->index
+ (slots
<< iter_shift(iter
));
414 * radix_tree_iter_next - resume iterating when the chunk may be invalid
415 * @iter: iterator state
417 * If the iterator needs to release then reacquire a lock, the chunk may
418 * have been invalidated by an insertion or deletion. Call this function
419 * to continue the iteration from the next index.
421 static inline __must_check
422 void **radix_tree_iter_next(struct radix_tree_iter
*iter
)
424 iter
->next_index
= __radix_tree_iter_add(iter
, 1);
430 * radix_tree_chunk_size - get current chunk size
432 * @iter: pointer to radix tree iterator
433 * Returns: current chunk size
435 static __always_inline
long
436 radix_tree_chunk_size(struct radix_tree_iter
*iter
)
438 return (iter
->next_index
- iter
->index
) >> iter_shift(iter
);
441 static inline struct radix_tree_node
*entry_to_node(void *ptr
)
443 return (void *)((unsigned long)ptr
& ~RADIX_TREE_INTERNAL_NODE
);
447 * radix_tree_next_slot - find next slot in chunk
449 * @slot: pointer to current slot
450 * @iter: pointer to interator state
451 * @flags: RADIX_TREE_ITER_*, should be constant
452 * Returns: pointer to next slot, or NULL if there no more left
454 * This function updates @iter->index in the case of a successful lookup.
455 * For tagged lookup it also eats @iter->tags.
457 * There are several cases where 'slot' can be passed in as NULL to this
458 * function. These cases result from the use of radix_tree_iter_next() or
459 * radix_tree_iter_retry(). In these cases we don't end up dereferencing
460 * 'slot' because either:
461 * a) we are doing tagged iteration and iter->tags has been set to 0, or
462 * b) we are doing non-tagged iteration, and iter->index and iter->next_index
463 * have been set up so that radix_tree_chunk_size() returns 1 or 0.
465 static __always_inline
void **
466 radix_tree_next_slot(void **slot
, struct radix_tree_iter
*iter
, unsigned flags
)
468 if (flags
& RADIX_TREE_ITER_TAGGED
) {
472 if (unlikely(!iter
->tags
))
474 while (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER
) &&
475 radix_tree_is_internal_node(slot
[1])) {
476 if (entry_to_node(slot
[1]) == canon
) {
478 iter
->index
= __radix_tree_iter_add(iter
, 1);
482 iter
->next_index
= __radix_tree_iter_add(iter
, 1);
485 if (likely(iter
->tags
& 1ul)) {
486 iter
->index
= __radix_tree_iter_add(iter
, 1);
489 if (!(flags
& RADIX_TREE_ITER_CONTIG
)) {
490 unsigned offset
= __ffs(iter
->tags
);
492 iter
->tags
>>= offset
;
493 iter
->index
= __radix_tree_iter_add(iter
, offset
+ 1);
494 return slot
+ offset
+ 1;
497 long count
= radix_tree_chunk_size(iter
);
500 while (--count
> 0) {
502 iter
->index
= __radix_tree_iter_add(iter
, 1);
504 if (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER
) &&
505 radix_tree_is_internal_node(*slot
)) {
506 if (entry_to_node(*slot
) == canon
)
508 iter
->next_index
= iter
->index
;
514 if (flags
& RADIX_TREE_ITER_CONTIG
) {
515 /* forbid switching to the next chunk */
516 iter
->next_index
= 0;
525 * radix_tree_for_each_slot - iterate over non-empty slots
527 * @slot: the void** variable for pointer to slot
528 * @root: the struct radix_tree_root pointer
529 * @iter: the struct radix_tree_iter pointer
530 * @start: iteration starting index
532 * @slot points to radix tree slot, @iter->index contains its index.
534 #define radix_tree_for_each_slot(slot, root, iter, start) \
535 for (slot = radix_tree_iter_init(iter, start) ; \
536 slot || (slot = radix_tree_next_chunk(root, iter, 0)) ; \
537 slot = radix_tree_next_slot(slot, iter, 0))
540 * radix_tree_for_each_contig - iterate over contiguous slots
542 * @slot: the void** variable for pointer to slot
543 * @root: the struct radix_tree_root pointer
544 * @iter: the struct radix_tree_iter pointer
545 * @start: iteration starting index
547 * @slot points to radix tree slot, @iter->index contains its index.
549 #define radix_tree_for_each_contig(slot, root, iter, start) \
550 for (slot = radix_tree_iter_init(iter, start) ; \
551 slot || (slot = radix_tree_next_chunk(root, iter, \
552 RADIX_TREE_ITER_CONTIG)) ; \
553 slot = radix_tree_next_slot(slot, iter, \
554 RADIX_TREE_ITER_CONTIG))
557 * radix_tree_for_each_tagged - iterate over tagged slots
559 * @slot: the void** variable for pointer to slot
560 * @root: the struct radix_tree_root pointer
561 * @iter: the struct radix_tree_iter pointer
562 * @start: iteration starting index
565 * @slot points to radix tree slot, @iter->index contains its index.
567 #define radix_tree_for_each_tagged(slot, root, iter, start, tag) \
568 for (slot = radix_tree_iter_init(iter, start) ; \
569 slot || (slot = radix_tree_next_chunk(root, iter, \
570 RADIX_TREE_ITER_TAGGED | tag)) ; \
571 slot = radix_tree_next_slot(slot, iter, \
572 RADIX_TREE_ITER_TAGGED))
574 #endif /* _LINUX_RADIX_TREE_H */