Loading...
1/* SPDX-License-Identifier: GPL-2.0 */
2
3#ifndef BTRFS_EXTENT_IO_TREE_H
4#define BTRFS_EXTENT_IO_TREE_H
5
6#include <linux/rbtree.h>
7#include <linux/spinlock.h>
8#include <linux/refcount.h>
9#include <linux/list.h>
10#include <linux/wait.h>
11#include "misc.h"
12
13struct extent_changeset;
14struct btrfs_fs_info;
15struct btrfs_inode;
16
17/* Bits for the extent state */
18enum {
19 ENUM_BIT(EXTENT_DIRTY),
20 ENUM_BIT(EXTENT_UPTODATE),
21 ENUM_BIT(EXTENT_LOCKED),
22 ENUM_BIT(EXTENT_DIO_LOCKED),
23 ENUM_BIT(EXTENT_NEW),
24 ENUM_BIT(EXTENT_DELALLOC),
25 ENUM_BIT(EXTENT_DEFRAG),
26 ENUM_BIT(EXTENT_BOUNDARY),
27 ENUM_BIT(EXTENT_NODATASUM),
28 ENUM_BIT(EXTENT_CLEAR_META_RESV),
29 ENUM_BIT(EXTENT_NEED_WAIT),
30 ENUM_BIT(EXTENT_NORESERVE),
31 ENUM_BIT(EXTENT_QGROUP_RESERVED),
32 ENUM_BIT(EXTENT_CLEAR_DATA_RESV),
33 /*
34 * Must be cleared only during ordered extent completion or on error
35 * paths if we did not manage to submit bios and create the ordered
36 * extents for the range. Should not be cleared during page release
37 * and page invalidation (if there is an ordered extent in flight),
38 * that is left for the ordered extent completion.
39 */
40 ENUM_BIT(EXTENT_DELALLOC_NEW),
41 /*
42 * When an ordered extent successfully completes for a region marked as
43 * a new delalloc range, use this flag when clearing a new delalloc
44 * range to indicate that the VFS' inode number of bytes should be
45 * incremented and the inode's new delalloc bytes decremented, in an
46 * atomic way to prevent races with stat(2).
47 */
48 ENUM_BIT(EXTENT_ADD_INODE_BYTES),
49 /*
50 * Set during truncate when we're clearing an entire range and we just
51 * want the extent states to go away.
52 */
53 ENUM_BIT(EXTENT_CLEAR_ALL_BITS),
54
55 /*
56 * This must be last.
57 *
58 * Bit not representing a state but a request for NOWAIT semantics,
59 * e.g. when allocating memory, and must be masked out from the other
60 * bits.
61 */
62 ENUM_BIT(EXTENT_NOWAIT)
63};
64
65#define EXTENT_DO_ACCOUNTING (EXTENT_CLEAR_META_RESV | \
66 EXTENT_CLEAR_DATA_RESV)
67#define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | \
68 EXTENT_ADD_INODE_BYTES | \
69 EXTENT_CLEAR_ALL_BITS)
70
71#define EXTENT_LOCK_BITS (EXTENT_LOCKED | EXTENT_DIO_LOCKED)
72
73/*
74 * Redefined bits above which are used only in the device allocation tree,
75 * shouldn't be using EXTENT_LOCKED / EXTENT_BOUNDARY / EXTENT_CLEAR_META_RESV
76 * / EXTENT_CLEAR_DATA_RESV because they have special meaning to the bit
77 * manipulation functions
78 */
79#define CHUNK_ALLOCATED EXTENT_DIRTY
80#define CHUNK_TRIMMED EXTENT_DEFRAG
81#define CHUNK_STATE_MASK (CHUNK_ALLOCATED | \
82 CHUNK_TRIMMED)
83
84enum {
85 IO_TREE_FS_PINNED_EXTENTS,
86 IO_TREE_FS_EXCLUDED_EXTENTS,
87 IO_TREE_BTREE_INODE_IO,
88 IO_TREE_INODE_IO,
89 IO_TREE_RELOC_BLOCKS,
90 IO_TREE_TRANS_DIRTY_PAGES,
91 IO_TREE_ROOT_DIRTY_LOG_PAGES,
92 IO_TREE_INODE_FILE_EXTENT,
93 IO_TREE_LOG_CSUM_RANGE,
94 IO_TREE_SELFTEST,
95 IO_TREE_DEVICE_ALLOC_STATE,
96};
97
98struct extent_io_tree {
99 struct rb_root state;
100 /*
101 * The fs_info is needed for trace points, a tree attached to an inode
102 * needs the inode.
103 *
104 * owner == IO_TREE_INODE_IO - then inode is valid and fs_info can be
105 * accessed as inode->root->fs_info
106 */
107 union {
108 struct btrfs_fs_info *fs_info;
109 struct btrfs_inode *inode;
110 };
111
112 /* Who owns this io tree, should be one of IO_TREE_* */
113 u8 owner;
114
115 spinlock_t lock;
116};
117
118struct extent_state {
119 u64 start;
120 u64 end; /* inclusive */
121 struct rb_node rb_node;
122
123 /* ADD NEW ELEMENTS AFTER THIS */
124 wait_queue_head_t wq;
125 refcount_t refs;
126 u32 state;
127
128#ifdef CONFIG_BTRFS_DEBUG
129 struct list_head leak_list;
130#endif
131};
132
133struct btrfs_inode *extent_io_tree_to_inode(struct extent_io_tree *tree);
134const struct btrfs_inode *extent_io_tree_to_inode_const(const struct extent_io_tree *tree);
135const struct btrfs_fs_info *extent_io_tree_to_fs_info(const struct extent_io_tree *tree);
136
137void extent_io_tree_init(struct btrfs_fs_info *fs_info,
138 struct extent_io_tree *tree, unsigned int owner);
139void extent_io_tree_release(struct extent_io_tree *tree);
140int __lock_extent(struct extent_io_tree *tree, u64 start, u64 end, u32 bits,
141 struct extent_state **cached);
142bool __try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end, u32 bits,
143 struct extent_state **cached);
144
145static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end,
146 struct extent_state **cached)
147{
148 return __lock_extent(tree, start, end, EXTENT_LOCKED, cached);
149}
150
151static inline bool try_lock_extent(struct extent_io_tree *tree, u64 start,
152 u64 end, struct extent_state **cached)
153{
154 return __try_lock_extent(tree, start, end, EXTENT_LOCKED, cached);
155}
156
157int __init extent_state_init_cachep(void);
158void __cold extent_state_free_cachep(void);
159
160u64 count_range_bits(struct extent_io_tree *tree,
161 u64 *start, u64 search_end,
162 u64 max_bytes, u32 bits, int contig,
163 struct extent_state **cached_state);
164
165void free_extent_state(struct extent_state *state);
166bool test_range_bit(struct extent_io_tree *tree, u64 start, u64 end, u32 bit,
167 struct extent_state *cached_state);
168bool test_range_bit_exists(struct extent_io_tree *tree, u64 start, u64 end, u32 bit);
169int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
170 u32 bits, struct extent_changeset *changeset);
171int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
172 u32 bits, struct extent_state **cached,
173 struct extent_changeset *changeset);
174
175static inline int clear_extent_bit(struct extent_io_tree *tree, u64 start,
176 u64 end, u32 bits,
177 struct extent_state **cached)
178{
179 return __clear_extent_bit(tree, start, end, bits, cached, NULL);
180}
181
182static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end,
183 struct extent_state **cached)
184{
185 return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, cached, NULL);
186}
187
188static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start,
189 u64 end, u32 bits)
190{
191 return clear_extent_bit(tree, start, end, bits, NULL);
192}
193
194int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
195 u32 bits, struct extent_changeset *changeset);
196int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
197 u32 bits, struct extent_state **cached_state);
198
199static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
200 u64 end, struct extent_state **cached_state)
201{
202 return __clear_extent_bit(tree, start, end, EXTENT_UPTODATE,
203 cached_state, NULL);
204}
205
206static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start,
207 u64 end, struct extent_state **cached)
208{
209 return clear_extent_bit(tree, start, end,
210 EXTENT_DIRTY | EXTENT_DELALLOC |
211 EXTENT_DO_ACCOUNTING, cached);
212}
213
214int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
215 u32 bits, u32 clear_bits,
216 struct extent_state **cached_state);
217
218bool find_first_extent_bit(struct extent_io_tree *tree, u64 start,
219 u64 *start_ret, u64 *end_ret, u32 bits,
220 struct extent_state **cached_state);
221void find_first_clear_extent_bit(struct extent_io_tree *tree, u64 start,
222 u64 *start_ret, u64 *end_ret, u32 bits);
223int find_contiguous_extent_bit(struct extent_io_tree *tree, u64 start,
224 u64 *start_ret, u64 *end_ret, u32 bits);
225bool btrfs_find_delalloc_range(struct extent_io_tree *tree, u64 *start,
226 u64 *end, u64 max_bytes,
227 struct extent_state **cached_state);
228static inline int lock_dio_extent(struct extent_io_tree *tree, u64 start,
229 u64 end, struct extent_state **cached)
230{
231 return __lock_extent(tree, start, end, EXTENT_DIO_LOCKED, cached);
232}
233
234static inline bool try_lock_dio_extent(struct extent_io_tree *tree, u64 start,
235 u64 end, struct extent_state **cached)
236{
237 return __try_lock_extent(tree, start, end, EXTENT_DIO_LOCKED, cached);
238}
239
240static inline int unlock_dio_extent(struct extent_io_tree *tree, u64 start,
241 u64 end, struct extent_state **cached)
242{
243 return __clear_extent_bit(tree, start, end, EXTENT_DIO_LOCKED, cached, NULL);
244}
245
246#endif /* BTRFS_EXTENT_IO_TREE_H */
1/* SPDX-License-Identifier: GPL-2.0 */
2
3#ifndef BTRFS_EXTENT_IO_TREE_H
4#define BTRFS_EXTENT_IO_TREE_H
5
6struct extent_changeset;
7struct io_failure_record;
8
9/* Bits for the extent state */
10#define EXTENT_DIRTY (1U << 0)
11#define EXTENT_UPTODATE (1U << 1)
12#define EXTENT_LOCKED (1U << 2)
13#define EXTENT_NEW (1U << 3)
14#define EXTENT_DELALLOC (1U << 4)
15#define EXTENT_DEFRAG (1U << 5)
16#define EXTENT_BOUNDARY (1U << 6)
17#define EXTENT_NODATASUM (1U << 7)
18#define EXTENT_CLEAR_META_RESV (1U << 8)
19#define EXTENT_NEED_WAIT (1U << 9)
20#define EXTENT_DAMAGED (1U << 10)
21#define EXTENT_NORESERVE (1U << 11)
22#define EXTENT_QGROUP_RESERVED (1U << 12)
23#define EXTENT_CLEAR_DATA_RESV (1U << 13)
24/*
25 * Must be cleared only during ordered extent completion or on error paths if we
26 * did not manage to submit bios and create the ordered extents for the range.
27 * Should not be cleared during page release and page invalidation (if there is
28 * an ordered extent in flight), that is left for the ordered extent completion.
29 */
30#define EXTENT_DELALLOC_NEW (1U << 14)
31/*
32 * When an ordered extent successfully completes for a region marked as a new
33 * delalloc range, use this flag when clearing a new delalloc range to indicate
34 * that the VFS' inode number of bytes should be incremented and the inode's new
35 * delalloc bytes decremented, in an atomic way to prevent races with stat(2).
36 */
37#define EXTENT_ADD_INODE_BYTES (1U << 15)
38#define EXTENT_DO_ACCOUNTING (EXTENT_CLEAR_META_RESV | \
39 EXTENT_CLEAR_DATA_RESV)
40#define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | \
41 EXTENT_ADD_INODE_BYTES)
42
43/*
44 * Redefined bits above which are used only in the device allocation tree,
45 * shouldn't be using EXTENT_LOCKED / EXTENT_BOUNDARY / EXTENT_CLEAR_META_RESV
46 * / EXTENT_CLEAR_DATA_RESV because they have special meaning to the bit
47 * manipulation functions
48 */
49#define CHUNK_ALLOCATED EXTENT_DIRTY
50#define CHUNK_TRIMMED EXTENT_DEFRAG
51#define CHUNK_STATE_MASK (CHUNK_ALLOCATED | \
52 CHUNK_TRIMMED)
53
54enum {
55 IO_TREE_FS_PINNED_EXTENTS,
56 IO_TREE_FS_EXCLUDED_EXTENTS,
57 IO_TREE_BTREE_INODE_IO,
58 IO_TREE_INODE_IO,
59 IO_TREE_INODE_IO_FAILURE,
60 IO_TREE_RELOC_BLOCKS,
61 IO_TREE_TRANS_DIRTY_PAGES,
62 IO_TREE_ROOT_DIRTY_LOG_PAGES,
63 IO_TREE_INODE_FILE_EXTENT,
64 IO_TREE_LOG_CSUM_RANGE,
65 IO_TREE_SELFTEST,
66 IO_TREE_DEVICE_ALLOC_STATE,
67};
68
69struct extent_io_tree {
70 struct rb_root state;
71 struct btrfs_fs_info *fs_info;
72 void *private_data;
73 u64 dirty_bytes;
74 bool track_uptodate;
75
76 /* Who owns this io tree, should be one of IO_TREE_* */
77 u8 owner;
78
79 spinlock_t lock;
80};
81
82struct extent_state {
83 u64 start;
84 u64 end; /* inclusive */
85 struct rb_node rb_node;
86
87 /* ADD NEW ELEMENTS AFTER THIS */
88 wait_queue_head_t wq;
89 refcount_t refs;
90 u32 state;
91
92 struct io_failure_record *failrec;
93
94#ifdef CONFIG_BTRFS_DEBUG
95 struct list_head leak_list;
96#endif
97};
98
99int __init extent_state_cache_init(void);
100void __cold extent_state_cache_exit(void);
101
102void extent_io_tree_init(struct btrfs_fs_info *fs_info,
103 struct extent_io_tree *tree, unsigned int owner,
104 void *private_data);
105void extent_io_tree_release(struct extent_io_tree *tree);
106
107int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
108 struct extent_state **cached);
109
110static inline int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
111{
112 return lock_extent_bits(tree, start, end, NULL);
113}
114
115int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end);
116
117int __init extent_io_init(void);
118void __cold extent_io_exit(void);
119
120u64 count_range_bits(struct extent_io_tree *tree,
121 u64 *start, u64 search_end,
122 u64 max_bytes, u32 bits, int contig);
123
124void free_extent_state(struct extent_state *state);
125int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
126 u32 bits, int filled, struct extent_state *cached_state);
127int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
128 u32 bits, struct extent_changeset *changeset);
129int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
130 u32 bits, int wake, int delete,
131 struct extent_state **cached);
132int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
133 u32 bits, int wake, int delete,
134 struct extent_state **cached, gfp_t mask,
135 struct extent_changeset *changeset);
136
137static inline int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
138{
139 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL);
140}
141
142static inline int unlock_extent_cached(struct extent_io_tree *tree, u64 start,
143 u64 end, struct extent_state **cached)
144{
145 return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
146 GFP_NOFS, NULL);
147}
148
149static inline int unlock_extent_cached_atomic(struct extent_io_tree *tree,
150 u64 start, u64 end, struct extent_state **cached)
151{
152 return __clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
153 GFP_ATOMIC, NULL);
154}
155
156static inline int clear_extent_bits(struct extent_io_tree *tree, u64 start,
157 u64 end, u32 bits)
158{
159 int wake = 0;
160
161 if (bits & EXTENT_LOCKED)
162 wake = 1;
163
164 return clear_extent_bit(tree, start, end, bits, wake, 0, NULL);
165}
166
167int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
168 u32 bits, struct extent_changeset *changeset);
169int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
170 u32 bits, unsigned exclusive_bits, u64 *failed_start,
171 struct extent_state **cached_state, gfp_t mask,
172 struct extent_changeset *changeset);
173int set_extent_bits_nowait(struct extent_io_tree *tree, u64 start, u64 end,
174 u32 bits);
175
176static inline int set_extent_bits(struct extent_io_tree *tree, u64 start,
177 u64 end, u32 bits)
178{
179 return set_extent_bit(tree, start, end, bits, 0, NULL, NULL, GFP_NOFS,
180 NULL);
181}
182
183static inline int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
184 u64 end, struct extent_state **cached_state)
185{
186 return __clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
187 cached_state, GFP_NOFS, NULL);
188}
189
190static inline int set_extent_dirty(struct extent_io_tree *tree, u64 start,
191 u64 end, gfp_t mask)
192{
193 return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL, NULL,
194 mask, NULL);
195}
196
197static inline int clear_extent_dirty(struct extent_io_tree *tree, u64 start,
198 u64 end, struct extent_state **cached)
199{
200 return clear_extent_bit(tree, start, end,
201 EXTENT_DIRTY | EXTENT_DELALLOC |
202 EXTENT_DO_ACCOUNTING, 0, 0, cached);
203}
204
205int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
206 u32 bits, u32 clear_bits,
207 struct extent_state **cached_state);
208
209static inline int set_extent_delalloc(struct extent_io_tree *tree, u64 start,
210 u64 end, u32 extra_bits,
211 struct extent_state **cached_state)
212{
213 return set_extent_bit(tree, start, end,
214 EXTENT_DELALLOC | EXTENT_UPTODATE | extra_bits,
215 0, NULL, cached_state, GFP_NOFS, NULL);
216}
217
218static inline int set_extent_defrag(struct extent_io_tree *tree, u64 start,
219 u64 end, struct extent_state **cached_state)
220{
221 return set_extent_bit(tree, start, end,
222 EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
223 0, NULL, cached_state, GFP_NOFS, NULL);
224}
225
226static inline int set_extent_new(struct extent_io_tree *tree, u64 start,
227 u64 end)
228{
229 return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL, NULL,
230 GFP_NOFS, NULL);
231}
232
233static inline int set_extent_uptodate(struct extent_io_tree *tree, u64 start,
234 u64 end, struct extent_state **cached_state, gfp_t mask)
235{
236 return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL,
237 cached_state, mask, NULL);
238}
239
240int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
241 u64 *start_ret, u64 *end_ret, u32 bits,
242 struct extent_state **cached_state);
243void find_first_clear_extent_bit(struct extent_io_tree *tree, u64 start,
244 u64 *start_ret, u64 *end_ret, u32 bits);
245int find_contiguous_extent_bit(struct extent_io_tree *tree, u64 start,
246 u64 *start_ret, u64 *end_ret, u32 bits);
247int extent_invalidatepage(struct extent_io_tree *tree,
248 struct page *page, unsigned long offset);
249bool btrfs_find_delalloc_range(struct extent_io_tree *tree, u64 *start,
250 u64 *end, u64 max_bytes,
251 struct extent_state **cached_state);
252
253/* This should be reworked in the future and put elsewhere. */
254struct io_failure_record *get_state_failrec(struct extent_io_tree *tree, u64 start);
255int set_state_failrec(struct extent_io_tree *tree, u64 start,
256 struct io_failure_record *failrec);
257void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start,
258 u64 end);
259int free_io_failure(struct extent_io_tree *failure_tree,
260 struct extent_io_tree *io_tree,
261 struct io_failure_record *rec);
262int clean_io_failure(struct btrfs_fs_info *fs_info,
263 struct extent_io_tree *failure_tree,
264 struct extent_io_tree *io_tree, u64 start,
265 struct page *page, u64 ino, unsigned int pg_offset);
266
267#endif /* BTRFS_EXTENT_IO_TREE_H */