2 * Copyright (C) 2001 Momchil Velikov
3 * Portions Copyright (C) 2001 Christoph Hellwig
4 * Copyright (C) 2006 Nick Piggin
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License as
8 * published by the Free Software Foundation; either version 2, or (at
9 * your option) any later version.
11 * This program is distributed in the hope that it will be useful, but
12 * WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
20 #ifndef _LINUX_RADIX_TREE_H
21 #define _LINUX_RADIX_TREE_H
23 #include <linux/preempt.h>
24 #include <linux/types.h>
25 #include <linux/bug.h>
26 #include <linux/kernel.h>
27 #include <linux/rcupdate.h>
30 * An indirect pointer (root->rnode pointing to a radix_tree_node, rather
31 * than a data item) is signalled by the low bit set in the root->rnode
34 * In this case root->height is > 0, but the indirect pointer tests are
35 * needed for RCU lookups (because root->height is unreliable). The only
36 * time callers need worry about this is when doing a lookup_slot under
39 * Indirect pointer in fact is also used to tag the last pointer of a node
40 * when it is shrunk, before we rcu free the node. See shrink code for
43 #define RADIX_TREE_INDIRECT_PTR 1
45 * A common use of the radix tree is to store pointers to struct pages;
46 * but shmem/tmpfs needs also to store swap entries in the same tree:
47 * those are marked as exceptional entries to distinguish them.
48 * EXCEPTIONAL_ENTRY tests the bit, EXCEPTIONAL_SHIFT shifts content past it.
50 #define RADIX_TREE_EXCEPTIONAL_ENTRY 2
51 #define RADIX_TREE_EXCEPTIONAL_SHIFT 2
53 static inline int radix_tree_is_indirect_ptr(void *ptr)
55 return (int)((unsigned long)ptr & RADIX_TREE_INDIRECT_PTR);
58 /*** radix-tree API starts here ***/
60 #define RADIX_TREE_MAX_TAGS 3
62 /* root tags are stored in gfp_mask, shifted by __GFP_BITS_SHIFT */
63 struct radix_tree_root {
66 struct radix_tree_node __rcu *rnode;
69 #define RADIX_TREE_INIT(mask) { \
75 #define RADIX_TREE(name, mask) \
76 struct radix_tree_root name = RADIX_TREE_INIT(mask)
78 #define INIT_RADIX_TREE(root, mask) \
81 (root)->gfp_mask = (mask); \
82 (root)->rnode = NULL; \
86 * Radix-tree synchronization
88 * The radix-tree API requires that users provide all synchronisation (with
89 * specific exceptions, noted below).
91 * Synchronization of access to the data items being stored in the tree, and
92 * management of their lifetimes must be completely managed by API users.
94 * For API usage, in general,
95 * - any function _modifying_ the tree or tags (inserting or deleting
96 * items, setting or clearing tags) must exclude other modifications, and
97 * exclude any functions reading the tree.
98 * - any function _reading_ the tree or tags (looking up items or tags,
99 * gang lookups) must exclude modifications to the tree, but may occur
100 * concurrently with other readers.
102 * The notable exceptions to this rule are the following functions:
104 * radix_tree_lookup_slot
106 * radix_tree_gang_lookup
107 * radix_tree_gang_lookup_slot
108 * radix_tree_gang_lookup_tag
109 * radix_tree_gang_lookup_tag_slot
112 * The first 7 functions are able to be called locklessly, using RCU. The
113 * caller must ensure calls to these functions are made within rcu_read_lock()
114 * regions. Other readers (lock-free or otherwise) and modifications may be
115 * running concurrently.
117 * It is still required that the caller manage the synchronization and lifetimes
118 * of the items. So if RCU lock-free lookups are used, typically this would mean
119 * that the items have their own locks, or are amenable to lock-free access; and
120 * that the items are freed by RCU (or only freed after having been deleted from
121 * the radix tree *and* a synchronize_rcu() grace period).
123 * (Note, rcu_assign_pointer and rcu_dereference are not needed to control
124 * access to data items when inserting into or looking up from the radix tree)
126 * Note that the value returned by radix_tree_tag_get() may not be relied upon
127 * if only the RCU read lock is held. Functions to set/clear tags and to
128 * delete nodes running concurrently with it may affect its result such that
129 * two consecutive reads in the same locked section may return different
130 * values. If reliability is required, modification functions must also be
131 * excluded from concurrency.
133 * radix_tree_tagged is able to be called without locking or RCU.
137 * radix_tree_deref_slot - dereference a slot
138 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
139 * Returns: item that was stored in that slot with any direct pointer flag
142 * For use with radix_tree_lookup_slot(). Caller must hold tree at least read
143 * locked across slot lookup and dereference. Not required if write lock is
144 * held (ie. items cannot be concurrently inserted).
146 * radix_tree_deref_retry must be used to confirm validity of the pointer if
147 * only the read lock is held.
149 static inline void *radix_tree_deref_slot(void **pslot)
151 return rcu_dereference(*pslot);
155 * radix_tree_deref_slot_protected - dereference a slot without RCU lock but with tree lock held
156 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
157 * Returns: item that was stored in that slot with any direct pointer flag
160 * Similar to radix_tree_deref_slot but only used during migration when a pages
161 * mapping is being moved. The caller does not hold the RCU read lock but it
162 * must hold the tree lock to prevent parallel updates.
164 static inline void *radix_tree_deref_slot_protected(void **pslot,
165 spinlock_t *treelock)
167 return rcu_dereference_protected(*pslot, lockdep_is_held(treelock));
171 * radix_tree_deref_retry - check radix_tree_deref_slot
172 * @arg: pointer returned by radix_tree_deref_slot
173 * Returns: 0 if retry is not required, otherwise retry is required
175 * radix_tree_deref_retry must be used with radix_tree_deref_slot.
177 static inline int radix_tree_deref_retry(void *arg)
179 return unlikely((unsigned long)arg & RADIX_TREE_INDIRECT_PTR);
183 * radix_tree_exceptional_entry - radix_tree_deref_slot gave exceptional entry?
184 * @arg: value returned by radix_tree_deref_slot
185 * Returns: 0 if well-aligned pointer, non-0 if exceptional entry.
187 static inline int radix_tree_exceptional_entry(void *arg)
189 /* Not unlikely because radix_tree_exception often tested first */
190 return (unsigned long)arg & RADIX_TREE_EXCEPTIONAL_ENTRY;
194 * radix_tree_exception - radix_tree_deref_slot returned either exception?
195 * @arg: value returned by radix_tree_deref_slot
196 * Returns: 0 if well-aligned pointer, non-0 if either kind of exception.
198 static inline int radix_tree_exception(void *arg)
200 return unlikely((unsigned long)arg &
201 (RADIX_TREE_INDIRECT_PTR | RADIX_TREE_EXCEPTIONAL_ENTRY));
205 * radix_tree_replace_slot - replace item in a slot
206 * @pslot: pointer to slot, returned by radix_tree_lookup_slot
207 * @item: new item to store in the slot.
209 * For use with radix_tree_lookup_slot(). Caller must hold tree write locked
210 * across slot lookup and replacement.
212 static inline void radix_tree_replace_slot(void **pslot, void *item)
214 BUG_ON(radix_tree_is_indirect_ptr(item));
215 rcu_assign_pointer(*pslot, item);
218 int radix_tree_insert(struct radix_tree_root *, unsigned long, void *);
219 void *radix_tree_lookup(struct radix_tree_root *, unsigned long);
220 void **radix_tree_lookup_slot(struct radix_tree_root *, unsigned long);
221 void *radix_tree_delete(struct radix_tree_root *, unsigned long);
223 radix_tree_gang_lookup(struct radix_tree_root *root, void **results,
224 unsigned long first_index, unsigned int max_items);
225 unsigned int radix_tree_gang_lookup_slot(struct radix_tree_root *root,
226 void ***results, unsigned long *indices,
227 unsigned long first_index, unsigned int max_items);
228 unsigned long radix_tree_next_hole(struct radix_tree_root *root,
229 unsigned long index, unsigned long max_scan);
230 unsigned long radix_tree_prev_hole(struct radix_tree_root *root,
231 unsigned long index, unsigned long max_scan);
232 int radix_tree_preload(gfp_t gfp_mask);
233 void radix_tree_init(void);
234 void *radix_tree_tag_set(struct radix_tree_root *root,
235 unsigned long index, unsigned int tag);
236 void *radix_tree_tag_clear(struct radix_tree_root *root,
237 unsigned long index, unsigned int tag);
238 int radix_tree_tag_get(struct radix_tree_root *root,
239 unsigned long index, unsigned int tag);
241 radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
242 unsigned long first_index, unsigned int max_items,
245 radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
246 unsigned long first_index, unsigned int max_items,
248 unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
249 unsigned long *first_indexp, unsigned long last_index,
250 unsigned long nr_to_tag,
251 unsigned int fromtag, unsigned int totag);
252 int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
253 unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item);
255 static inline void radix_tree_preload_end(void)
260 #endif /* _LINUX_RADIX_TREE_H */