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 struct radix_tree_node {
84 unsigned char shift; /* Bits remaining in each slot */
85 unsigned char offset; /* Slot offset in parent */
86 unsigned char count; /* Total entry count */
87 unsigned char exceptional; /* Exceptional entry count */
90 /* Used when ascending tree */
91 struct radix_tree_node *parent;
95 /* Used when freeing node */
96 struct rcu_head rcu_head;
99 struct list_head private_list;
100 void __rcu *slots[RADIX_TREE_MAP_SIZE];
101 unsigned long tags[RADIX_TREE_MAX_TAGS][RADIX_TREE_TAG_LONGS];
104 /* root tags are stored in gfp_mask, shifted by __GFP_BITS_SHIFT */
105 struct radix_tree_root {
107 struct radix_tree_node __rcu *rnode;
110 #define RADIX_TREE_INIT(mask) { \
111 .gfp_mask = (mask), \
115 #define RADIX_TREE(name, mask) \
116 struct radix_tree_root name = RADIX_TREE_INIT(mask)
118 #define INIT_RADIX_TREE(root, mask) \
120 (root)->gfp_mask = (mask); \
121 (root)->rnode = NULL; \
124 static inline bool radix_tree_empty(struct radix_tree_root *root)
126 return root->rnode == NULL;
130 * Radix-tree synchronization
132 * The radix-tree API requires that users provide all synchronisation (with
133 * specific exceptions, noted below).
135 * Synchronization of access to the data items being stored in the tree, and
136 * management of their lifetimes must be completely managed by API users.
138 * For API usage, in general,
139 * - any function _modifying_ the tree or tags (inserting or deleting
140 * items, setting or clearing tags) must exclude other modifications, and
141 * exclude any functions reading the tree.
142 * - any function _reading_ the tree or tags (looking up items or tags,
143 * gang lookups) must exclude modifications to the tree, but may occur
144 * concurrently with other readers.
146 * The notable exceptions to this rule are the following functions:
147 * __radix_tree_lookup
149 * radix_tree_lookup_slot
151 * radix_tree_gang_lookup
152 * radix_tree_gang_lookup_slot
153 * radix_tree_gang_lookup_tag
154 * radix_tree_gang_lookup_tag_slot
157 * The first 8 functions are able to be called locklessly, using RCU. The
158 * caller must ensure calls to these functions are made within rcu_read_lock()
159 * regions. Other readers (lock-free or otherwise) and modifications may be
160 * running concurrently.
162 * It is still required that the caller manage the synchronization and lifetimes
163 * of the items. So if RCU lock-free lookups are used, typically this would mean
164 * that the items have their own locks, or are amenable to lock-free access; and
165 * that the items are freed by RCU (or only freed after having been deleted from
166 * the radix tree *and* a synchronize_rcu() grace period).
168 * (Note, rcu_assign_pointer and rcu_dereference are not needed to control
169 * access to data items when inserting into or looking up from the radix tree)
171 * Note that the value returned by radix_tree_tag_get() may not be relied upon
172 * if only the RCU read lock is held. Functions to set/clear tags and to
173 * delete nodes running concurrently with it may affect its result such that
174 * two consecutive reads in the same locked section may return different
175 * values. If reliability is required, modification functions must also be
176 * excluded from concurrency.
178 * radix_tree_tagged is able to be called without locking or RCU.
182 * radix_tree_deref_slot - dereference a slot
183 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
184 * Returns: item that was stored in that slot with any direct pointer flag
187 * For use with radix_tree_lookup_slot(). Caller must hold tree at least read
188 * locked across slot lookup and dereference. Not required if write lock is
189 * held (ie. items cannot be concurrently inserted).
191 * radix_tree_deref_retry must be used to confirm validity of the pointer if
192 * only the read lock is held.
194 static inline void *radix_tree_deref_slot(void **pslot)
196 return rcu_dereference(*pslot);
200 * radix_tree_deref_slot_protected - dereference a slot without RCU lock but with tree lock held
201 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
202 * Returns: item that was stored in that slot with any direct pointer flag
205 * Similar to radix_tree_deref_slot but only used during migration when a pages
206 * mapping is being moved. The caller does not hold the RCU read lock but it
207 * must hold the tree lock to prevent parallel updates.
209 static inline void *radix_tree_deref_slot_protected(void **pslot,
210 spinlock_t *treelock)
212 return rcu_dereference_protected(*pslot, lockdep_is_held(treelock));
216 * radix_tree_deref_retry - check radix_tree_deref_slot
217 * @arg: pointer returned by radix_tree_deref_slot
218 * Returns: 0 if retry is not required, otherwise retry is required
220 * radix_tree_deref_retry must be used with radix_tree_deref_slot.
222 static inline int radix_tree_deref_retry(void *arg)
224 return unlikely(radix_tree_is_internal_node(arg));
228 * radix_tree_exceptional_entry - radix_tree_deref_slot gave exceptional entry?
229 * @arg: value returned by radix_tree_deref_slot
230 * Returns: 0 if well-aligned pointer, non-0 if exceptional entry.
232 static inline int radix_tree_exceptional_entry(void *arg)
234 /* Not unlikely because radix_tree_exception often tested first */
235 return (unsigned long)arg & RADIX_TREE_EXCEPTIONAL_ENTRY;
239 * radix_tree_exception - radix_tree_deref_slot returned either exception?
240 * @arg: value returned by radix_tree_deref_slot
241 * Returns: 0 if well-aligned pointer, non-0 if either kind of exception.
243 static inline int radix_tree_exception(void *arg)
245 return unlikely((unsigned long)arg & RADIX_TREE_ENTRY_MASK);
248 int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
249 unsigned order, struct radix_tree_node **nodep,
251 int __radix_tree_insert(struct radix_tree_root *, unsigned long index,
252 unsigned order, void *);
253 static inline int radix_tree_insert(struct radix_tree_root *root,
254 unsigned long index, void *entry)
256 return __radix_tree_insert(root, index, 0, entry);
258 void *__radix_tree_lookup(struct radix_tree_root *root, unsigned long index,
259 struct radix_tree_node **nodep, void ***slotp);
260 void *radix_tree_lookup(struct radix_tree_root *, unsigned long);
261 void **radix_tree_lookup_slot(struct radix_tree_root *, unsigned long);
262 typedef void (*radix_tree_update_node_t)(struct radix_tree_node *, void *);
263 void __radix_tree_replace(struct radix_tree_root *root,
264 struct radix_tree_node *node,
265 void **slot, void *item,
266 radix_tree_update_node_t update_node, void *private);
267 void radix_tree_replace_slot(struct radix_tree_root *root,
268 void **slot, void *item);
269 void __radix_tree_delete_node(struct radix_tree_root *root,
270 struct radix_tree_node *node);
271 void *radix_tree_delete_item(struct radix_tree_root *, unsigned long, void *);
272 void *radix_tree_delete(struct radix_tree_root *, unsigned long);
273 void radix_tree_clear_tags(struct radix_tree_root *root,
274 struct radix_tree_node *node,
276 unsigned int radix_tree_gang_lookup(struct radix_tree_root *root,
277 void **results, unsigned long first_index,
278 unsigned int max_items);
279 unsigned int radix_tree_gang_lookup_slot(struct radix_tree_root *root,
280 void ***results, unsigned long *indices,
281 unsigned long first_index, unsigned int max_items);
282 int radix_tree_preload(gfp_t gfp_mask);
283 int radix_tree_maybe_preload(gfp_t gfp_mask);
284 int radix_tree_maybe_preload_order(gfp_t gfp_mask, int order);
285 void radix_tree_init(void);
286 void *radix_tree_tag_set(struct radix_tree_root *root,
287 unsigned long index, unsigned int tag);
288 void *radix_tree_tag_clear(struct radix_tree_root *root,
289 unsigned long index, unsigned int tag);
290 int radix_tree_tag_get(struct radix_tree_root *root,
291 unsigned long index, unsigned int tag);
293 radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
294 unsigned long first_index, unsigned int max_items,
297 radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
298 unsigned long first_index, unsigned int max_items,
300 unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
301 unsigned long *first_indexp, unsigned long last_index,
302 unsigned long nr_to_tag,
303 unsigned int fromtag, unsigned int totag);
304 int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
305 unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item);
307 static inline void radix_tree_preload_end(void)
313 * struct radix_tree_iter - radix tree iterator state
315 * @index: index of current slot
316 * @next_index: one beyond the last index for this chunk
317 * @tags: bit-mask for tag-iterating
318 * @shift: shift for the node that holds our slots
320 * This radix tree iterator works in terms of "chunks" of slots. A chunk is a
321 * subinterval of slots contained within one radix tree leaf node. It is
322 * described by a pointer to its first slot and a struct radix_tree_iter
323 * which holds the chunk's position in the tree and its size. For tagged
324 * iteration radix_tree_iter also holds the slots' bit-mask for one chosen
327 struct radix_tree_iter {
329 unsigned long next_index;
331 #ifdef CONFIG_RADIX_TREE_MULTIORDER
336 static inline unsigned int iter_shift(struct radix_tree_iter *iter)
338 #ifdef CONFIG_RADIX_TREE_MULTIORDER
345 #define RADIX_TREE_ITER_TAG_MASK 0x00FF /* tag index in lower byte */
346 #define RADIX_TREE_ITER_TAGGED 0x0100 /* lookup tagged slots */
347 #define RADIX_TREE_ITER_CONTIG 0x0200 /* stop at first hole */
350 * radix_tree_iter_init - initialize radix tree iterator
352 * @iter: pointer to iterator state
353 * @start: iteration starting index
356 static __always_inline void **
357 radix_tree_iter_init(struct radix_tree_iter *iter, unsigned long start)
360 * Leave iter->tags uninitialized. radix_tree_next_chunk() will fill it
361 * in the case of a successful tagged chunk lookup. If the lookup was
362 * unsuccessful or non-tagged then nobody cares about ->tags.
364 * Set index to zero to bypass next_index overflow protection.
365 * See the comment in radix_tree_next_chunk() for details.
368 iter->next_index = start;
373 * radix_tree_next_chunk - find next chunk of slots for iteration
375 * @root: radix tree root
376 * @iter: iterator state
377 * @flags: RADIX_TREE_ITER_* flags and tag index
378 * Returns: pointer to chunk first slot, or NULL if there no more left
380 * This function looks up the next chunk in the radix tree starting from
381 * @iter->next_index. It returns a pointer to the chunk's first slot.
382 * Also it fills @iter with data about chunk: position in the tree (index),
383 * its end (next_index), and constructs a bit mask for tagged iterating (tags).
385 void **radix_tree_next_chunk(struct radix_tree_root *root,
386 struct radix_tree_iter *iter, unsigned flags);
389 * radix_tree_iter_retry - retry this chunk of the iteration
390 * @iter: iterator state
392 * If we iterate over a tree protected only by the RCU lock, a race
393 * against deletion or creation may result in seeing a slot for which
394 * radix_tree_deref_retry() returns true. If so, call this function
395 * and continue the iteration.
397 static inline __must_check
398 void **radix_tree_iter_retry(struct radix_tree_iter *iter)
400 iter->next_index = iter->index;
405 static inline unsigned long
406 __radix_tree_iter_add(struct radix_tree_iter *iter, unsigned long slots)
408 return iter->index + (slots << iter_shift(iter));
412 * radix_tree_iter_next - resume iterating when the chunk may be invalid
413 * @iter: iterator state
415 * If the iterator needs to release then reacquire a lock, the chunk may
416 * have been invalidated by an insertion or deletion. Call this function
417 * to continue the iteration from the next index.
419 static inline __must_check
420 void **radix_tree_iter_next(struct radix_tree_iter *iter)
422 iter->next_index = __radix_tree_iter_add(iter, 1);
428 * radix_tree_chunk_size - get current chunk size
430 * @iter: pointer to radix tree iterator
431 * Returns: current chunk size
433 static __always_inline long
434 radix_tree_chunk_size(struct radix_tree_iter *iter)
436 return (iter->next_index - iter->index) >> iter_shift(iter);
439 static inline struct radix_tree_node *entry_to_node(void *ptr)
441 return (void *)((unsigned long)ptr & ~RADIX_TREE_INTERNAL_NODE);
445 * radix_tree_next_slot - find next slot in chunk
447 * @slot: pointer to current slot
448 * @iter: pointer to interator state
449 * @flags: RADIX_TREE_ITER_*, should be constant
450 * Returns: pointer to next slot, or NULL if there no more left
452 * This function updates @iter->index in the case of a successful lookup.
453 * For tagged lookup it also eats @iter->tags.
455 * There are several cases where 'slot' can be passed in as NULL to this
456 * function. These cases result from the use of radix_tree_iter_next() or
457 * radix_tree_iter_retry(). In these cases we don't end up dereferencing
458 * 'slot' because either:
459 * a) we are doing tagged iteration and iter->tags has been set to 0, or
460 * b) we are doing non-tagged iteration, and iter->index and iter->next_index
461 * have been set up so that radix_tree_chunk_size() returns 1 or 0.
463 static __always_inline void **
464 radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
466 if (flags & RADIX_TREE_ITER_TAGGED) {
470 if (unlikely(!iter->tags))
472 while (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
473 radix_tree_is_internal_node(slot[1])) {
474 if (entry_to_node(slot[1]) == canon) {
476 iter->index = __radix_tree_iter_add(iter, 1);
480 iter->next_index = __radix_tree_iter_add(iter, 1);
483 if (likely(iter->tags & 1ul)) {
484 iter->index = __radix_tree_iter_add(iter, 1);
487 if (!(flags & RADIX_TREE_ITER_CONTIG)) {
488 unsigned offset = __ffs(iter->tags);
490 iter->tags >>= offset;
491 iter->index = __radix_tree_iter_add(iter, offset + 1);
492 return slot + offset + 1;
495 long count = radix_tree_chunk_size(iter);
498 while (--count > 0) {
500 iter->index = __radix_tree_iter_add(iter, 1);
502 if (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
503 radix_tree_is_internal_node(*slot)) {
504 if (entry_to_node(*slot) == canon)
506 iter->next_index = iter->index;
512 if (flags & RADIX_TREE_ITER_CONTIG) {
513 /* forbid switching to the next chunk */
514 iter->next_index = 0;
523 * radix_tree_for_each_slot - iterate over non-empty slots
525 * @slot: the void** variable for pointer to slot
526 * @root: the struct radix_tree_root pointer
527 * @iter: the struct radix_tree_iter pointer
528 * @start: iteration starting index
530 * @slot points to radix tree slot, @iter->index contains its index.
532 #define radix_tree_for_each_slot(slot, root, iter, start) \
533 for (slot = radix_tree_iter_init(iter, start) ; \
534 slot || (slot = radix_tree_next_chunk(root, iter, 0)) ; \
535 slot = radix_tree_next_slot(slot, iter, 0))
538 * radix_tree_for_each_contig - iterate over contiguous slots
540 * @slot: the void** variable for pointer to slot
541 * @root: the struct radix_tree_root pointer
542 * @iter: the struct radix_tree_iter pointer
543 * @start: iteration starting index
545 * @slot points to radix tree slot, @iter->index contains its index.
547 #define radix_tree_for_each_contig(slot, root, iter, start) \
548 for (slot = radix_tree_iter_init(iter, start) ; \
549 slot || (slot = radix_tree_next_chunk(root, iter, \
550 RADIX_TREE_ITER_CONTIG)) ; \
551 slot = radix_tree_next_slot(slot, iter, \
552 RADIX_TREE_ITER_CONTIG))
555 * radix_tree_for_each_tagged - iterate over tagged slots
557 * @slot: the void** variable for pointer to slot
558 * @root: the struct radix_tree_root pointer
559 * @iter: the struct radix_tree_iter pointer
560 * @start: iteration starting index
563 * @slot points to radix tree slot, @iter->index contains its index.
565 #define radix_tree_for_each_tagged(slot, root, iter, start, tag) \
566 for (slot = radix_tree_iter_init(iter, start) ; \
567 slot || (slot = radix_tree_next_chunk(root, iter, \
568 RADIX_TREE_ITER_TAGGED | tag)) ; \
569 slot = radix_tree_next_slot(slot, iter, \
570 RADIX_TREE_ITER_TAGGED))
572 #endif /* _LINUX_RADIX_TREE_H */