xref: /linux/drivers/md/persistent-data/dm-btree.h (revision 26fbb4c8c7c3ee9a4c3b4de555a8587b5a19154e)
1 /*
2  * Copyright (C) 2011 Red Hat, Inc.
3  *
4  * This file is released under the GPL.
5  */
6 #ifndef _LINUX_DM_BTREE_H
7 #define _LINUX_DM_BTREE_H
8 
9 #include "dm-block-manager.h"
10 
11 struct dm_transaction_manager;
12 
13 /*----------------------------------------------------------------*/
14 
15 /*
16  * Annotations used to check on-disk metadata is handled as little-endian.
17  */
18 #ifdef __CHECKER__
19 #  define __dm_written_to_disk(x) __releases(x)
20 #  define __dm_reads_from_disk(x) __acquires(x)
21 #  define __dm_bless_for_disk(x) __acquire(x)
22 #  define __dm_unbless_for_disk(x) __release(x)
23 #else
24 #  define __dm_written_to_disk(x)
25 #  define __dm_reads_from_disk(x)
26 #  define __dm_bless_for_disk(x)
27 #  define __dm_unbless_for_disk(x)
28 #endif
29 
30 /*----------------------------------------------------------------*/
31 
32 /*
33  * Manipulates hierarchical B+ trees with 64-bit keys and arbitrary-sized
34  * values.
35  */
36 
37 /*
38  * Information about the values stored within the btree.
39  */
40 struct dm_btree_value_type {
41 	void *context;
42 
43 	/*
44 	 * The size in bytes of each value.
45 	 */
46 	uint32_t size;
47 
48 	/*
49 	 * Any of these methods can be safely set to NULL if you do not
50 	 * need the corresponding feature.
51 	 */
52 
53 	/*
54 	 * The btree is making a duplicate of the value, for instance
55 	 * because previously-shared btree nodes have now diverged.
56 	 * @value argument is the new copy that the copy function may modify.
57 	 * (Probably it just wants to increment a reference count
58 	 * somewhere.) This method is _not_ called for insertion of a new
59 	 * value: It is assumed the ref count is already 1.
60 	 */
61 	void (*inc)(void *context, const void *value);
62 
63 	/*
64 	 * This value is being deleted.  The btree takes care of freeing
65 	 * the memory pointed to by @value.  Often the del function just
66 	 * needs to decrement a reference count somewhere.
67 	 */
68 	void (*dec)(void *context, const void *value);
69 
70 	/*
71 	 * A test for equality between two values.  When a value is
72 	 * overwritten with a new one, the old one has the dec method
73 	 * called _unless_ the new and old value are deemed equal.
74 	 */
75 	int (*equal)(void *context, const void *value1, const void *value2);
76 };
77 
78 /*
79  * The shape and contents of a btree.
80  */
81 struct dm_btree_info {
82 	struct dm_transaction_manager *tm;
83 
84 	/*
85 	 * Number of nested btrees. (Not the depth of a single tree.)
86 	 */
87 	unsigned levels;
88 	struct dm_btree_value_type value_type;
89 };
90 
91 /*
92  * Set up an empty tree.  O(1).
93  */
94 int dm_btree_empty(struct dm_btree_info *info, dm_block_t *root);
95 
96 /*
97  * Delete a tree.  O(n) - this is the slow one!  It can also block, so
98  * please don't call it on an IO path.
99  */
100 int dm_btree_del(struct dm_btree_info *info, dm_block_t root);
101 
102 /*
103  * All the lookup functions return -ENODATA if the key cannot be found.
104  */
105 
106 /*
107  * Tries to find a key that matches exactly.  O(ln(n))
108  */
109 int dm_btree_lookup(struct dm_btree_info *info, dm_block_t root,
110 		    uint64_t *keys, void *value_le);
111 
112 /*
113  * Tries to find the first key where the bottom level key is >= to that
114  * given.  Useful for skipping empty sections of the btree.
115  */
116 int dm_btree_lookup_next(struct dm_btree_info *info, dm_block_t root,
117 			 uint64_t *keys, uint64_t *rkey, void *value_le);
118 
119 /*
120  * Insertion (or overwrite an existing value).  O(ln(n))
121  */
122 int dm_btree_insert(struct dm_btree_info *info, dm_block_t root,
123 		    uint64_t *keys, void *value, dm_block_t *new_root)
124 		    __dm_written_to_disk(value);
125 
126 /*
127  * A variant of insert that indicates whether it actually inserted or just
128  * overwrote.  Useful if you're keeping track of the number of entries in a
129  * tree.
130  */
131 int dm_btree_insert_notify(struct dm_btree_info *info, dm_block_t root,
132 			   uint64_t *keys, void *value, dm_block_t *new_root,
133 			   int *inserted)
134 			   __dm_written_to_disk(value);
135 
136 /*
137  * Remove a key if present.  This doesn't remove empty sub trees.  Normally
138  * subtrees represent a separate entity, like a snapshot map, so this is
139  * correct behaviour.  O(ln(n)).
140  */
141 int dm_btree_remove(struct dm_btree_info *info, dm_block_t root,
142 		    uint64_t *keys, dm_block_t *new_root);
143 
144 /*
145  * Removes a _contiguous_ run of values starting from 'keys' and not
146  * reaching keys2 (where keys2 is keys with the final key replaced with
147  * 'end_key').  'end_key' is the one-past-the-end value.  'keys' may be
148  * altered.
149  */
150 int dm_btree_remove_leaves(struct dm_btree_info *info, dm_block_t root,
151 			   uint64_t *keys, uint64_t end_key,
152 			   dm_block_t *new_root, unsigned *nr_removed);
153 
154 /*
155  * Returns < 0 on failure.  Otherwise the number of key entries that have
156  * been filled out.  Remember trees can have zero entries, and as such have
157  * no lowest key.
158  */
159 int dm_btree_find_lowest_key(struct dm_btree_info *info, dm_block_t root,
160 			     uint64_t *result_keys);
161 
162 /*
163  * Returns < 0 on failure.  Otherwise the number of key entries that have
164  * been filled out.  Remember trees can have zero entries, and as such have
165  * no highest key.
166  */
167 int dm_btree_find_highest_key(struct dm_btree_info *info, dm_block_t root,
168 			      uint64_t *result_keys);
169 
170 /*
171  * Iterate through the a btree, calling fn() on each entry.
172  * It only works for single level trees and is internally recursive, so
173  * monitor stack usage carefully.
174  */
175 int dm_btree_walk(struct dm_btree_info *info, dm_block_t root,
176 		  int (*fn)(void *context, uint64_t *keys, void *leaf),
177 		  void *context);
178 
179 
180 /*----------------------------------------------------------------*/
181 
182 /*
183  * Cursor API.  This does not follow the rolling lock convention.  Since we
184  * know the order that values are required we can issue prefetches to speed
185  * up iteration.  Use on a single level btree only.
186  */
187 #define DM_BTREE_CURSOR_MAX_DEPTH 16
188 
189 struct cursor_node {
190 	struct dm_block *b;
191 	unsigned index;
192 };
193 
194 struct dm_btree_cursor {
195 	struct dm_btree_info *info;
196 	dm_block_t root;
197 
198 	bool prefetch_leaves;
199 	unsigned depth;
200 	struct cursor_node nodes[DM_BTREE_CURSOR_MAX_DEPTH];
201 };
202 
203 /*
204  * Creates a fresh cursor.  If prefetch_leaves is set then it is assumed
205  * the btree contains block indexes that will be prefetched.  The cursor is
206  * quite large, so you probably don't want to put it on the stack.
207  */
208 int dm_btree_cursor_begin(struct dm_btree_info *info, dm_block_t root,
209 			  bool prefetch_leaves, struct dm_btree_cursor *c);
210 void dm_btree_cursor_end(struct dm_btree_cursor *c);
211 int dm_btree_cursor_next(struct dm_btree_cursor *c);
212 int dm_btree_cursor_skip(struct dm_btree_cursor *c, uint32_t count);
213 int dm_btree_cursor_get_value(struct dm_btree_cursor *c, uint64_t *key, void *value_le);
214 
215 #endif	/* _LINUX_DM_BTREE_H */
216