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))
84 * @count is the count of every non-NULL element in the ->slots array
85 * whether that is an exceptional entry, a retry entry, a user pointer,
86 * a sibling entry or a pointer to the next level of the tree.
87 * @exceptional is the count of every element in ->slots which is
88 * either radix_tree_exceptional_entry() or is a sibling entry for an
91 struct radix_tree_node {
92 unsigned char shift; /* Bits remaining in each slot */
93 unsigned char offset; /* Slot offset in parent */
94 unsigned char count; /* Total entry count */
95 unsigned char exceptional; /* Exceptional entry count */
96 struct radix_tree_node *parent; /* Used when ascending tree */
97 void *private_data; /* For tree user */
99 struct list_head private_list; /* For tree user */
100 struct rcu_head rcu_head; /* Used when freeing node */
102 void __rcu *slots[RADIX_TREE_MAP_SIZE];
103 unsigned long tags[RADIX_TREE_MAX_TAGS][RADIX_TREE_TAG_LONGS];
106 /* root tags are stored in gfp_mask, shifted by __GFP_BITS_SHIFT */
107 struct radix_tree_root {
109 struct radix_tree_node __rcu *rnode;
112 #define RADIX_TREE_INIT(mask) { \
113 .gfp_mask = (mask), \
117 #define RADIX_TREE(name, mask) \
118 struct radix_tree_root name = RADIX_TREE_INIT(mask)
120 #define INIT_RADIX_TREE(root, mask) \
122 (root)->gfp_mask = (mask); \
123 (root)->rnode = NULL; \
126 static inline bool radix_tree_empty(struct radix_tree_root *root)
128 return root->rnode == NULL;
132 * struct radix_tree_iter - radix tree iterator state
134 * @index: index of current slot
135 * @next_index: one beyond the last index for this chunk
136 * @tags: bit-mask for tag-iterating
137 * @node: node that contains current slot
138 * @shift: shift for the node that holds our slots
140 * This radix tree iterator works in terms of "chunks" of slots. A chunk is a
141 * subinterval of slots contained within one radix tree leaf node. It is
142 * described by a pointer to its first slot and a struct radix_tree_iter
143 * which holds the chunk's position in the tree and its size. For tagged
144 * iteration radix_tree_iter also holds the slots' bit-mask for one chosen
147 struct radix_tree_iter {
149 unsigned long next_index;
151 struct radix_tree_node *node;
152 #ifdef CONFIG_RADIX_TREE_MULTIORDER
157 static inline unsigned int iter_shift(const struct radix_tree_iter *iter)
159 #ifdef CONFIG_RADIX_TREE_MULTIORDER
167 * Radix-tree synchronization
169 * The radix-tree API requires that users provide all synchronisation (with
170 * specific exceptions, noted below).
172 * Synchronization of access to the data items being stored in the tree, and
173 * management of their lifetimes must be completely managed by API users.
175 * For API usage, in general,
176 * - any function _modifying_ the tree or tags (inserting or deleting
177 * items, setting or clearing tags) must exclude other modifications, and
178 * exclude any functions reading the tree.
179 * - any function _reading_ the tree or tags (looking up items or tags,
180 * gang lookups) must exclude modifications to the tree, but may occur
181 * concurrently with other readers.
183 * The notable exceptions to this rule are the following functions:
184 * __radix_tree_lookup
186 * radix_tree_lookup_slot
188 * radix_tree_gang_lookup
189 * radix_tree_gang_lookup_slot
190 * radix_tree_gang_lookup_tag
191 * radix_tree_gang_lookup_tag_slot
194 * The first 8 functions are able to be called locklessly, using RCU. The
195 * caller must ensure calls to these functions are made within rcu_read_lock()
196 * regions. Other readers (lock-free or otherwise) and modifications may be
197 * running concurrently.
199 * It is still required that the caller manage the synchronization and lifetimes
200 * of the items. So if RCU lock-free lookups are used, typically this would mean
201 * that the items have their own locks, or are amenable to lock-free access; and
202 * that the items are freed by RCU (or only freed after having been deleted from
203 * the radix tree *and* a synchronize_rcu() grace period).
205 * (Note, rcu_assign_pointer and rcu_dereference are not needed to control
206 * access to data items when inserting into or looking up from the radix tree)
208 * Note that the value returned by radix_tree_tag_get() may not be relied upon
209 * if only the RCU read lock is held. Functions to set/clear tags and to
210 * delete nodes running concurrently with it may affect its result such that
211 * two consecutive reads in the same locked section may return different
212 * values. If reliability is required, modification functions must also be
213 * excluded from concurrency.
215 * radix_tree_tagged is able to be called without locking or RCU.
219 * radix_tree_deref_slot - dereference a slot
220 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
221 * Returns: item that was stored in that slot with any direct pointer flag
224 * For use with radix_tree_lookup_slot(). Caller must hold tree at least read
225 * locked across slot lookup and dereference. Not required if write lock is
226 * held (ie. items cannot be concurrently inserted).
228 * radix_tree_deref_retry must be used to confirm validity of the pointer if
229 * only the read lock is held.
231 static inline void *radix_tree_deref_slot(void **pslot)
233 return rcu_dereference(*pslot);
237 * radix_tree_deref_slot_protected - dereference a slot without RCU lock but with tree lock held
238 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
239 * Returns: item that was stored in that slot with any direct pointer flag
242 * Similar to radix_tree_deref_slot but only used during migration when a pages
243 * mapping is being moved. The caller does not hold the RCU read lock but it
244 * must hold the tree lock to prevent parallel updates.
246 static inline void *radix_tree_deref_slot_protected(void **pslot,
247 spinlock_t *treelock)
249 return rcu_dereference_protected(*pslot, lockdep_is_held(treelock));
253 * radix_tree_deref_retry - check radix_tree_deref_slot
254 * @arg: pointer returned by radix_tree_deref_slot
255 * Returns: 0 if retry is not required, otherwise retry is required
257 * radix_tree_deref_retry must be used with radix_tree_deref_slot.
259 static inline int radix_tree_deref_retry(void *arg)
261 return unlikely(radix_tree_is_internal_node(arg));
265 * radix_tree_exceptional_entry - radix_tree_deref_slot gave exceptional entry?
266 * @arg: value returned by radix_tree_deref_slot
267 * Returns: 0 if well-aligned pointer, non-0 if exceptional entry.
269 static inline int radix_tree_exceptional_entry(void *arg)
271 /* Not unlikely because radix_tree_exception often tested first */
272 return (unsigned long)arg & RADIX_TREE_EXCEPTIONAL_ENTRY;
276 * radix_tree_exception - radix_tree_deref_slot returned either exception?
277 * @arg: value returned by radix_tree_deref_slot
278 * Returns: 0 if well-aligned pointer, non-0 if either kind of exception.
280 static inline int radix_tree_exception(void *arg)
282 return unlikely((unsigned long)arg & RADIX_TREE_ENTRY_MASK);
285 int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
286 unsigned order, struct radix_tree_node **nodep,
288 int __radix_tree_insert(struct radix_tree_root *, unsigned long index,
289 unsigned order, void *);
290 static inline int radix_tree_insert(struct radix_tree_root *root,
291 unsigned long index, void *entry)
293 return __radix_tree_insert(root, index, 0, entry);
295 void *__radix_tree_lookup(struct radix_tree_root *root, unsigned long index,
296 struct radix_tree_node **nodep, void ***slotp);
297 void *radix_tree_lookup(struct radix_tree_root *, unsigned long);
298 void **radix_tree_lookup_slot(struct radix_tree_root *, unsigned long);
299 typedef void (*radix_tree_update_node_t)(struct radix_tree_node *, void *);
300 void __radix_tree_replace(struct radix_tree_root *root,
301 struct radix_tree_node *node,
302 void **slot, void *item,
303 radix_tree_update_node_t update_node, void *private);
304 void radix_tree_iter_replace(struct radix_tree_root *,
305 const struct radix_tree_iter *, void **slot, void *item);
306 void radix_tree_replace_slot(struct radix_tree_root *root,
307 void **slot, void *item);
308 void __radix_tree_delete_node(struct radix_tree_root *root,
309 struct radix_tree_node *node,
310 radix_tree_update_node_t update_node,
312 void *radix_tree_delete_item(struct radix_tree_root *, unsigned long, void *);
313 void *radix_tree_delete(struct radix_tree_root *, unsigned long);
314 void radix_tree_clear_tags(struct radix_tree_root *root,
315 struct radix_tree_node *node,
317 unsigned int radix_tree_gang_lookup(struct radix_tree_root *root,
318 void **results, unsigned long first_index,
319 unsigned int max_items);
320 unsigned int radix_tree_gang_lookup_slot(struct radix_tree_root *root,
321 void ***results, unsigned long *indices,
322 unsigned long first_index, unsigned int max_items);
323 int radix_tree_preload(gfp_t gfp_mask);
324 int radix_tree_maybe_preload(gfp_t gfp_mask);
325 int radix_tree_maybe_preload_order(gfp_t gfp_mask, int order);
326 void radix_tree_init(void);
327 void *radix_tree_tag_set(struct radix_tree_root *root,
328 unsigned long index, unsigned int tag);
329 void *radix_tree_tag_clear(struct radix_tree_root *root,
330 unsigned long index, unsigned int tag);
331 int radix_tree_tag_get(struct radix_tree_root *root,
332 unsigned long index, unsigned int tag);
333 void radix_tree_iter_tag_set(struct radix_tree_root *root,
334 const struct radix_tree_iter *iter, unsigned int tag);
336 radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
337 unsigned long first_index, unsigned int max_items,
340 radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
341 unsigned long first_index, unsigned int max_items,
343 int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
345 static inline void radix_tree_preload_end(void)
350 int radix_tree_split_preload(unsigned old_order, unsigned new_order, gfp_t);
351 int radix_tree_split(struct radix_tree_root *, unsigned long index,
353 int radix_tree_join(struct radix_tree_root *, unsigned long index,
354 unsigned new_order, void *);
356 #define RADIX_TREE_ITER_TAG_MASK 0x00FF /* tag index in lower byte */
357 #define RADIX_TREE_ITER_TAGGED 0x0100 /* lookup tagged slots */
358 #define RADIX_TREE_ITER_CONTIG 0x0200 /* stop at first hole */
361 * radix_tree_iter_init - initialize radix tree iterator
363 * @iter: pointer to iterator state
364 * @start: iteration starting index
367 static __always_inline void **
368 radix_tree_iter_init(struct radix_tree_iter *iter, unsigned long start)
371 * Leave iter->tags uninitialized. radix_tree_next_chunk() will fill it
372 * in the case of a successful tagged chunk lookup. If the lookup was
373 * unsuccessful or non-tagged then nobody cares about ->tags.
375 * Set index to zero to bypass next_index overflow protection.
376 * See the comment in radix_tree_next_chunk() for details.
379 iter->next_index = start;
384 * radix_tree_next_chunk - find next chunk of slots for iteration
386 * @root: radix tree root
387 * @iter: iterator state
388 * @flags: RADIX_TREE_ITER_* flags and tag index
389 * Returns: pointer to chunk first slot, or NULL if there no more left
391 * This function looks up the next chunk in the radix tree starting from
392 * @iter->next_index. It returns a pointer to the chunk's first slot.
393 * Also it fills @iter with data about chunk: position in the tree (index),
394 * its end (next_index), and constructs a bit mask for tagged iterating (tags).
396 void **radix_tree_next_chunk(struct radix_tree_root *root,
397 struct radix_tree_iter *iter, unsigned flags);
400 * radix_tree_iter_retry - retry this chunk of the iteration
401 * @iter: iterator state
403 * If we iterate over a tree protected only by the RCU lock, a race
404 * against deletion or creation may result in seeing a slot for which
405 * radix_tree_deref_retry() returns true. If so, call this function
406 * and continue the iteration.
408 static inline __must_check
409 void **radix_tree_iter_retry(struct radix_tree_iter *iter)
411 iter->next_index = iter->index;
416 static inline unsigned long
417 __radix_tree_iter_add(struct radix_tree_iter *iter, unsigned long slots)
419 return iter->index + (slots << iter_shift(iter));
423 * radix_tree_iter_resume - resume iterating when the chunk may be invalid
424 * @slot: pointer to current slot
425 * @iter: iterator state
426 * Returns: New slot pointer
428 * If the iterator needs to release then reacquire a lock, the chunk may
429 * have been invalidated by an insertion or deletion. Call this function
430 * before releasing the lock to continue the iteration from the next index.
432 void **__must_check radix_tree_iter_resume(void **slot,
433 struct radix_tree_iter *iter);
436 * radix_tree_chunk_size - get current chunk size
438 * @iter: pointer to radix tree iterator
439 * Returns: current chunk size
441 static __always_inline long
442 radix_tree_chunk_size(struct radix_tree_iter *iter)
444 return (iter->next_index - iter->index) >> iter_shift(iter);
447 #ifdef CONFIG_RADIX_TREE_MULTIORDER
448 void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
451 /* Can't happen without sibling entries, but the compiler can't tell that */
452 static inline void ** __radix_tree_next_slot(void **slot,
453 struct radix_tree_iter *iter, unsigned flags)
460 * radix_tree_next_slot - find next slot in chunk
462 * @slot: pointer to current slot
463 * @iter: pointer to interator state
464 * @flags: RADIX_TREE_ITER_*, should be constant
465 * Returns: pointer to next slot, or NULL if there no more left
467 * This function updates @iter->index in the case of a successful lookup.
468 * For tagged lookup it also eats @iter->tags.
470 * There are several cases where 'slot' can be passed in as NULL to this
471 * function. These cases result from the use of radix_tree_iter_resume() or
472 * radix_tree_iter_retry(). In these cases we don't end up dereferencing
473 * 'slot' because either:
474 * a) we are doing tagged iteration and iter->tags has been set to 0, or
475 * b) we are doing non-tagged iteration, and iter->index and iter->next_index
476 * have been set up so that radix_tree_chunk_size() returns 1 or 0.
478 static __always_inline void **
479 radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
481 if (flags & RADIX_TREE_ITER_TAGGED) {
483 if (unlikely(!iter->tags))
485 if (likely(iter->tags & 1ul)) {
486 iter->index = __radix_tree_iter_add(iter, 1);
490 if (!(flags & RADIX_TREE_ITER_CONTIG)) {
491 unsigned offset = __ffs(iter->tags);
493 iter->tags >>= offset++;
494 iter->index = __radix_tree_iter_add(iter, offset);
499 long count = radix_tree_chunk_size(iter);
501 while (--count > 0) {
503 iter->index = __radix_tree_iter_add(iter, 1);
507 if (flags & RADIX_TREE_ITER_CONTIG) {
508 /* forbid switching to the next chunk */
509 iter->next_index = 0;
517 if (unlikely(radix_tree_is_internal_node(*slot)))
518 return __radix_tree_next_slot(slot, iter, flags);
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 | tag))
572 #endif /* _LINUX_RADIX_TREE_H */