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 * Entries in the radix tree have the low bit set if they refer to a
33 * radix_tree_node. If the low bit is clear then the entry is user data.
35 * We also use the low bit to indicate that the slot will be freed in the
36 * next RCU idle period, and users need to re-walk the tree to find the
37 * new slot for the index that they were looking for. See the comment in
38 * radix_tree_shrink() for details.
40 #define RADIX_TREE_INTERNAL_NODE 1
43 * A common use of the radix tree is to store pointers to struct pages;
44 * but shmem/tmpfs needs also to store swap entries in the same tree:
45 * those are marked as exceptional entries to distinguish them.
46 * EXCEPTIONAL_ENTRY tests the bit, EXCEPTIONAL_SHIFT shifts content past it.
48 #define RADIX_TREE_EXCEPTIONAL_ENTRY 2
49 #define RADIX_TREE_EXCEPTIONAL_SHIFT 2
51 static inline int radix_tree_is_internal_node(void *ptr)
53 return (int)((unsigned long)ptr & RADIX_TREE_INTERNAL_NODE);
56 /*** radix-tree API starts here ***/
58 #define RADIX_TREE_MAX_TAGS 3
60 #ifndef RADIX_TREE_MAP_SHIFT
61 #define RADIX_TREE_MAP_SHIFT (CONFIG_BASE_SMALL ? 4 : 6)
64 #define RADIX_TREE_MAP_SIZE (1UL << RADIX_TREE_MAP_SHIFT)
65 #define RADIX_TREE_MAP_MASK (RADIX_TREE_MAP_SIZE-1)
67 #define RADIX_TREE_TAG_LONGS \
68 ((RADIX_TREE_MAP_SIZE + BITS_PER_LONG - 1) / BITS_PER_LONG)
70 #define RADIX_TREE_INDEX_BITS (8 /* CHAR_BIT */ * sizeof(unsigned long))
71 #define RADIX_TREE_MAX_PATH (DIV_ROUND_UP(RADIX_TREE_INDEX_BITS, \
72 RADIX_TREE_MAP_SHIFT))
74 /* Internally used bits of node->count */
75 #define RADIX_TREE_COUNT_SHIFT (RADIX_TREE_MAP_SHIFT + 1)
76 #define RADIX_TREE_COUNT_MASK ((1UL << RADIX_TREE_COUNT_SHIFT) - 1)
78 struct radix_tree_node {
79 unsigned char shift; /* Bits remaining in each slot */
80 unsigned char offset; /* Slot offset in parent */
84 /* Used when ascending tree */
85 struct radix_tree_node *parent;
89 /* Used when freeing node */
90 struct rcu_head rcu_head;
93 struct list_head private_list;
94 void __rcu *slots[RADIX_TREE_MAP_SIZE];
95 unsigned long tags[RADIX_TREE_MAX_TAGS][RADIX_TREE_TAG_LONGS];
98 /* root tags are stored in gfp_mask, shifted by __GFP_BITS_SHIFT */
99 struct radix_tree_root {
101 struct radix_tree_node __rcu *rnode;
104 #define RADIX_TREE_INIT(mask) { \
105 .gfp_mask = (mask), \
109 #define RADIX_TREE(name, mask) \
110 struct radix_tree_root name = RADIX_TREE_INIT(mask)
112 #define INIT_RADIX_TREE(root, mask) \
114 (root)->gfp_mask = (mask); \
115 (root)->rnode = NULL; \
118 static inline bool radix_tree_empty(struct radix_tree_root *root)
120 return root->rnode == NULL;
124 * Radix-tree synchronization
126 * The radix-tree API requires that users provide all synchronisation (with
127 * specific exceptions, noted below).
129 * Synchronization of access to the data items being stored in the tree, and
130 * management of their lifetimes must be completely managed by API users.
132 * For API usage, in general,
133 * - any function _modifying_ the tree or tags (inserting or deleting
134 * items, setting or clearing tags) must exclude other modifications, and
135 * exclude any functions reading the tree.
136 * - any function _reading_ the tree or tags (looking up items or tags,
137 * gang lookups) must exclude modifications to the tree, but may occur
138 * concurrently with other readers.
140 * The notable exceptions to this rule are the following functions:
141 * __radix_tree_lookup
143 * radix_tree_lookup_slot
145 * radix_tree_gang_lookup
146 * radix_tree_gang_lookup_slot
147 * radix_tree_gang_lookup_tag
148 * radix_tree_gang_lookup_tag_slot
151 * The first 8 functions are able to be called locklessly, using RCU. The
152 * caller must ensure calls to these functions are made within rcu_read_lock()
153 * regions. Other readers (lock-free or otherwise) and modifications may be
154 * running concurrently.
156 * It is still required that the caller manage the synchronization and lifetimes
157 * of the items. So if RCU lock-free lookups are used, typically this would mean
158 * that the items have their own locks, or are amenable to lock-free access; and
159 * that the items are freed by RCU (or only freed after having been deleted from
160 * the radix tree *and* a synchronize_rcu() grace period).
162 * (Note, rcu_assign_pointer and rcu_dereference are not needed to control
163 * access to data items when inserting into or looking up from the radix tree)
165 * Note that the value returned by radix_tree_tag_get() may not be relied upon
166 * if only the RCU read lock is held. Functions to set/clear tags and to
167 * delete nodes running concurrently with it may affect its result such that
168 * two consecutive reads in the same locked section may return different
169 * values. If reliability is required, modification functions must also be
170 * excluded from concurrency.
172 * radix_tree_tagged is able to be called without locking or RCU.
176 * radix_tree_deref_slot - dereference a slot
177 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
178 * Returns: item that was stored in that slot with any direct pointer flag
181 * For use with radix_tree_lookup_slot(). Caller must hold tree at least read
182 * locked across slot lookup and dereference. Not required if write lock is
183 * held (ie. items cannot be concurrently inserted).
185 * radix_tree_deref_retry must be used to confirm validity of the pointer if
186 * only the read lock is held.
188 static inline void *radix_tree_deref_slot(void **pslot)
190 return rcu_dereference(*pslot);
194 * radix_tree_deref_slot_protected - dereference a slot without RCU lock but with tree lock held
195 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
196 * Returns: item that was stored in that slot with any direct pointer flag
199 * Similar to radix_tree_deref_slot but only used during migration when a pages
200 * mapping is being moved. The caller does not hold the RCU read lock but it
201 * must hold the tree lock to prevent parallel updates.
203 static inline void *radix_tree_deref_slot_protected(void **pslot,
204 spinlock_t *treelock)
206 return rcu_dereference_protected(*pslot, lockdep_is_held(treelock));
210 * radix_tree_deref_retry - check radix_tree_deref_slot
211 * @arg: pointer returned by radix_tree_deref_slot
212 * Returns: 0 if retry is not required, otherwise retry is required
214 * radix_tree_deref_retry must be used with radix_tree_deref_slot.
216 static inline int radix_tree_deref_retry(void *arg)
218 return unlikely(radix_tree_is_internal_node(arg));
222 * radix_tree_exceptional_entry - radix_tree_deref_slot gave exceptional entry?
223 * @arg: value returned by radix_tree_deref_slot
224 * Returns: 0 if well-aligned pointer, non-0 if exceptional entry.
226 static inline int radix_tree_exceptional_entry(void *arg)
228 /* Not unlikely because radix_tree_exception often tested first */
229 return (unsigned long)arg & RADIX_TREE_EXCEPTIONAL_ENTRY;
233 * radix_tree_exception - radix_tree_deref_slot returned either exception?
234 * @arg: value returned by radix_tree_deref_slot
235 * Returns: 0 if well-aligned pointer, non-0 if either kind of exception.
237 static inline int radix_tree_exception(void *arg)
239 return unlikely((unsigned long)arg &
240 (RADIX_TREE_INTERNAL_NODE | RADIX_TREE_EXCEPTIONAL_ENTRY));
244 * radix_tree_replace_slot - replace item in a slot
245 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
246 * @item: new item to store in the slot.
248 * For use with radix_tree_lookup_slot(). Caller must hold tree write locked
249 * across slot lookup and replacement.
251 static inline void radix_tree_replace_slot(void **pslot, void *item)
253 BUG_ON(radix_tree_is_internal_node(item));
254 rcu_assign_pointer(*pslot, item);
257 int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
258 unsigned order, struct radix_tree_node **nodep,
260 int __radix_tree_insert(struct radix_tree_root *, unsigned long index,
261 unsigned order, void *);
262 static inline int radix_tree_insert(struct radix_tree_root *root,
263 unsigned long index, void *entry)
265 return __radix_tree_insert(root, index, 0, entry);
267 void *__radix_tree_lookup(struct radix_tree_root *root, unsigned long index,
268 struct radix_tree_node **nodep, void ***slotp);
269 void *radix_tree_lookup(struct radix_tree_root *, unsigned long);
270 void **radix_tree_lookup_slot(struct radix_tree_root *, unsigned long);
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 struct radix_tree_node *radix_tree_replace_clear_tags(
276 struct radix_tree_root *root,
277 unsigned long index, void *entry);
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 void radix_tree_init(void);
287 void *radix_tree_tag_set(struct radix_tree_root *root,
288 unsigned long index, unsigned int tag);
289 void *radix_tree_tag_clear(struct radix_tree_root *root,
290 unsigned long index, unsigned int tag);
291 int radix_tree_tag_get(struct radix_tree_root *root,
292 unsigned long index, unsigned int tag);
294 radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
295 unsigned long first_index, unsigned int max_items,
298 radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
299 unsigned long first_index, unsigned int max_items,
301 unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
302 unsigned long *first_indexp, unsigned long last_index,
303 unsigned long nr_to_tag,
304 unsigned int fromtag, unsigned int totag);
305 int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
306 unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item);
308 static inline void radix_tree_preload_end(void)
314 * struct radix_tree_iter - radix tree iterator state
316 * @index: index of current slot
317 * @next_index: one beyond the last index for this chunk
318 * @tags: bit-mask for tag-iterating
319 * @shift: shift for the node that holds our slots
321 * This radix tree iterator works in terms of "chunks" of slots. A chunk is a
322 * subinterval of slots contained within one radix tree leaf node. It is
323 * described by a pointer to its first slot and a struct radix_tree_iter
324 * which holds the chunk's position in the tree and its size. For tagged
325 * iteration radix_tree_iter also holds the slots' bit-mask for one chosen
328 struct radix_tree_iter {
330 unsigned long next_index;
332 #ifdef CONFIG_RADIX_TREE_MULTIORDER
337 static inline unsigned int iter_shift(struct radix_tree_iter *iter)
339 #ifdef CONFIG_RADIX_TREE_MULTIORDER
346 #define RADIX_TREE_ITER_TAG_MASK 0x00FF /* tag index in lower byte */
347 #define RADIX_TREE_ITER_TAGGED 0x0100 /* lookup tagged slots */
348 #define RADIX_TREE_ITER_CONTIG 0x0200 /* stop at first hole */
351 * radix_tree_iter_init - initialize radix tree iterator
353 * @iter: pointer to iterator state
354 * @start: iteration starting index
357 static __always_inline void **
358 radix_tree_iter_init(struct radix_tree_iter *iter, unsigned long start)
361 * Leave iter->tags uninitialized. radix_tree_next_chunk() will fill it
362 * in the case of a successful tagged chunk lookup. If the lookup was
363 * unsuccessful or non-tagged then nobody cares about ->tags.
365 * Set index to zero to bypass next_index overflow protection.
366 * See the comment in radix_tree_next_chunk() for details.
369 iter->next_index = start;
374 * radix_tree_next_chunk - find next chunk of slots for iteration
376 * @root: radix tree root
377 * @iter: iterator state
378 * @flags: RADIX_TREE_ITER_* flags and tag index
379 * Returns: pointer to chunk first slot, or NULL if there no more left
381 * This function looks up the next chunk in the radix tree starting from
382 * @iter->next_index. It returns a pointer to the chunk's first slot.
383 * Also it fills @iter with data about chunk: position in the tree (index),
384 * its end (next_index), and constructs a bit mask for tagged iterating (tags).
386 void **radix_tree_next_chunk(struct radix_tree_root *root,
387 struct radix_tree_iter *iter, unsigned flags);
390 * radix_tree_iter_retry - retry this chunk of the iteration
391 * @iter: iterator state
393 * If we iterate over a tree protected only by the RCU lock, a race
394 * against deletion or creation may result in seeing a slot for which
395 * radix_tree_deref_retry() returns true. If so, call this function
396 * and continue the iteration.
398 static inline __must_check
399 void **radix_tree_iter_retry(struct radix_tree_iter *iter)
401 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 static __always_inline void **
456 radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
458 if (flags & RADIX_TREE_ITER_TAGGED) {
462 if (unlikely(!iter->tags))
464 while (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
465 radix_tree_is_internal_node(slot[1])) {
466 if (entry_to_node(slot[1]) == canon) {
468 iter->index = __radix_tree_iter_add(iter, 1);
472 iter->next_index = __radix_tree_iter_add(iter, 1);
475 if (likely(iter->tags & 1ul)) {
476 iter->index = __radix_tree_iter_add(iter, 1);
479 if (!(flags & RADIX_TREE_ITER_CONTIG)) {
480 unsigned offset = __ffs(iter->tags);
482 iter->tags >>= offset;
483 iter->index = __radix_tree_iter_add(iter, offset + 1);
484 return slot + offset + 1;
487 long count = radix_tree_chunk_size(iter);
490 while (--count > 0) {
492 iter->index = __radix_tree_iter_add(iter, 1);
494 if (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
495 radix_tree_is_internal_node(*slot)) {
496 if (entry_to_node(*slot) == canon)
498 iter->next_index = iter->index;
504 if (flags & RADIX_TREE_ITER_CONTIG) {
505 /* forbid switching to the next chunk */
506 iter->next_index = 0;
515 * radix_tree_for_each_slot - iterate over non-empty slots
517 * @slot: the void** variable for pointer to slot
518 * @root: the struct radix_tree_root pointer
519 * @iter: the struct radix_tree_iter pointer
520 * @start: iteration starting index
522 * @slot points to radix tree slot, @iter->index contains its index.
524 #define radix_tree_for_each_slot(slot, root, iter, start) \
525 for (slot = radix_tree_iter_init(iter, start) ; \
526 slot || (slot = radix_tree_next_chunk(root, iter, 0)) ; \
527 slot = radix_tree_next_slot(slot, iter, 0))
530 * radix_tree_for_each_contig - iterate over contiguous slots
532 * @slot: the void** variable for pointer to slot
533 * @root: the struct radix_tree_root pointer
534 * @iter: the struct radix_tree_iter pointer
535 * @start: iteration starting index
537 * @slot points to radix tree slot, @iter->index contains its index.
539 #define radix_tree_for_each_contig(slot, root, iter, start) \
540 for (slot = radix_tree_iter_init(iter, start) ; \
541 slot || (slot = radix_tree_next_chunk(root, iter, \
542 RADIX_TREE_ITER_CONTIG)) ; \
543 slot = radix_tree_next_slot(slot, iter, \
544 RADIX_TREE_ITER_CONTIG))
547 * radix_tree_for_each_tagged - iterate over tagged slots
549 * @slot: the void** variable for pointer to slot
550 * @root: the struct radix_tree_root pointer
551 * @iter: the struct radix_tree_iter pointer
552 * @start: iteration starting index
555 * @slot points to radix tree slot, @iter->index contains its index.
557 #define radix_tree_for_each_tagged(slot, root, iter, start, tag) \
558 for (slot = radix_tree_iter_init(iter, start) ; \
559 slot || (slot = radix_tree_next_chunk(root, iter, \
560 RADIX_TREE_ITER_TAGGED | tag)) ; \
561 slot = radix_tree_next_slot(slot, iter, \
562 RADIX_TREE_ITER_TAGGED))
564 #endif /* _LINUX_RADIX_TREE_H */