]>
Commit | Line | Data |
---|---|---|
aac0023c JB |
1 | /* SPDX-License-Identifier: GPL-2.0 */ |
2 | ||
3 | #ifndef BTRFS_BLOCK_GROUP_H | |
4 | #define BTRFS_BLOCK_GROUP_H | |
5 | ||
67b61aef DS |
6 | #include "free-space-cache.h" |
7 | ||
aac0023c JB |
8 | enum btrfs_disk_cache_state { |
9 | BTRFS_DC_WRITTEN, | |
10 | BTRFS_DC_ERROR, | |
11 | BTRFS_DC_CLEAR, | |
12 | BTRFS_DC_SETUP, | |
13 | }; | |
14 | ||
2bee7eb8 DZ |
15 | /* |
16 | * This describes the state of the block_group for async discard. This is due | |
17 | * to the two pass nature of it where extent discarding is prioritized over | |
18 | * bitmap discarding. BTRFS_DISCARD_RESET_CURSOR is set when we are resetting | |
19 | * between lists to prevent contention for discard state variables | |
20 | * (eg. discard_cursor). | |
21 | */ | |
22 | enum btrfs_discard_state { | |
23 | BTRFS_DISCARD_EXTENTS, | |
24 | BTRFS_DISCARD_BITMAPS, | |
25 | BTRFS_DISCARD_RESET_CURSOR, | |
26 | }; | |
27 | ||
07730d87 JB |
28 | /* |
29 | * Control flags for do_chunk_alloc's force field CHUNK_ALLOC_NO_FORCE means to | |
30 | * only allocate a chunk if we really need one. | |
31 | * | |
32 | * CHUNK_ALLOC_LIMITED means to only try and allocate one if we have very few | |
33 | * chunks already allocated. This is used as part of the clustering code to | |
34 | * help make sure we have a good pool of storage to cluster in, without filling | |
35 | * the FS with empty chunks | |
36 | * | |
37 | * CHUNK_ALLOC_FORCE means it must try to allocate one | |
38 | */ | |
39 | enum btrfs_chunk_alloc_enum { | |
40 | CHUNK_ALLOC_NO_FORCE, | |
41 | CHUNK_ALLOC_LIMITED, | |
42 | CHUNK_ALLOC_FORCE, | |
43 | }; | |
44 | ||
aac0023c JB |
45 | struct btrfs_caching_control { |
46 | struct list_head list; | |
47 | struct mutex mutex; | |
48 | wait_queue_head_t wait; | |
49 | struct btrfs_work work; | |
32da5386 | 50 | struct btrfs_block_group *block_group; |
aac0023c JB |
51 | u64 progress; |
52 | refcount_t count; | |
53 | }; | |
54 | ||
55 | /* Once caching_thread() finds this much free space, it will wake up waiters. */ | |
56 | #define CACHING_CTL_WAKE_UP SZ_2M | |
57 | ||
32da5386 | 58 | struct btrfs_block_group { |
aac0023c JB |
59 | struct btrfs_fs_info *fs_info; |
60 | struct inode *inode; | |
61 | spinlock_t lock; | |
b3470b5d DS |
62 | u64 start; |
63 | u64 length; | |
aac0023c JB |
64 | u64 pinned; |
65 | u64 reserved; | |
bf38be65 | 66 | u64 used; |
aac0023c JB |
67 | u64 delalloc_bytes; |
68 | u64 bytes_super; | |
69 | u64 flags; | |
70 | u64 cache_generation; | |
71 | ||
72 | /* | |
73 | * If the free space extent count exceeds this number, convert the block | |
74 | * group to bitmaps. | |
75 | */ | |
76 | u32 bitmap_high_thresh; | |
77 | ||
78 | /* | |
79 | * If the free space extent count drops below this number, convert the | |
80 | * block group back to extents. | |
81 | */ | |
82 | u32 bitmap_low_thresh; | |
83 | ||
84 | /* | |
85 | * It is just used for the delayed data space allocation because | |
86 | * only the data space allocation and the relative metadata update | |
87 | * can be done cross the transaction. | |
88 | */ | |
89 | struct rw_semaphore data_rwsem; | |
90 | ||
91 | /* For raid56, this is a full stripe, without parity */ | |
92 | unsigned long full_stripe_len; | |
93 | ||
94 | unsigned int ro; | |
95 | unsigned int iref:1; | |
96 | unsigned int has_caching_ctl:1; | |
97 | unsigned int removed:1; | |
98 | ||
99 | int disk_cache_state; | |
100 | ||
101 | /* Cache tracking stuff */ | |
102 | int cached; | |
103 | struct btrfs_caching_control *caching_ctl; | |
104 | u64 last_byte_to_unpin; | |
105 | ||
106 | struct btrfs_space_info *space_info; | |
107 | ||
108 | /* Free space cache stuff */ | |
109 | struct btrfs_free_space_ctl *free_space_ctl; | |
110 | ||
111 | /* Block group cache stuff */ | |
112 | struct rb_node cache_node; | |
113 | ||
114 | /* For block groups in the same raid type */ | |
115 | struct list_head list; | |
116 | ||
48aaeebe | 117 | refcount_t refs; |
aac0023c JB |
118 | |
119 | /* | |
120 | * List of struct btrfs_free_clusters for this block group. | |
121 | * Today it will only have one thing on it, but that may change | |
122 | */ | |
123 | struct list_head cluster_list; | |
124 | ||
125 | /* For delayed block group creation or deletion of empty block groups */ | |
126 | struct list_head bg_list; | |
127 | ||
128 | /* For read-only block groups */ | |
129 | struct list_head ro_list; | |
130 | ||
6b7304af FM |
131 | /* |
132 | * When non-zero it means the block group's logical address and its | |
133 | * device extents can not be reused for future block group allocations | |
134 | * until the counter goes down to 0. This is to prevent them from being | |
135 | * reused while some task is still using the block group after it was | |
136 | * deleted - we want to make sure they can only be reused for new block | |
137 | * groups after that task is done with the deleted block group. | |
138 | */ | |
139 | atomic_t frozen; | |
140 | ||
b0643e59 | 141 | /* For discard operations */ |
b0643e59 DZ |
142 | struct list_head discard_list; |
143 | int discard_index; | |
144 | u64 discard_eligible_time; | |
2bee7eb8 DZ |
145 | u64 discard_cursor; |
146 | enum btrfs_discard_state discard_state; | |
aac0023c JB |
147 | |
148 | /* For dirty block groups */ | |
149 | struct list_head dirty_list; | |
150 | struct list_head io_list; | |
151 | ||
152 | struct btrfs_io_ctl io_ctl; | |
153 | ||
154 | /* | |
155 | * Incremented when doing extent allocations and holding a read lock | |
156 | * on the space_info's groups_sem semaphore. | |
157 | * Decremented when an ordered extent that represents an IO against this | |
158 | * block group's range is created (after it's added to its inode's | |
159 | * root's list of ordered extents) or immediately after the allocation | |
160 | * if it's a metadata extent or fallocate extent (for these cases we | |
161 | * don't create ordered extents). | |
162 | */ | |
163 | atomic_t reservations; | |
164 | ||
165 | /* | |
166 | * Incremented while holding the spinlock *lock* by a task checking if | |
167 | * it can perform a nocow write (incremented if the value for the *ro* | |
168 | * field is 0). Decremented by such tasks once they create an ordered | |
169 | * extent or before that if some error happens before reaching that step. | |
170 | * This is to prevent races between block group relocation and nocow | |
171 | * writes through direct IO. | |
172 | */ | |
173 | atomic_t nocow_writers; | |
174 | ||
175 | /* Lock for free space tree operations. */ | |
176 | struct mutex free_space_lock; | |
177 | ||
178 | /* | |
179 | * Does the block group need to be added to the free space tree? | |
180 | * Protected by free_space_lock. | |
181 | */ | |
182 | int needs_free_space; | |
183 | ||
184 | /* Record locked full stripes for RAID5/6 block group */ | |
185 | struct btrfs_full_stripe_locks_tree full_stripe_locks_root; | |
186 | }; | |
187 | ||
b0643e59 DZ |
188 | static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group) |
189 | { | |
190 | return (block_group->start + block_group->length); | |
191 | } | |
192 | ||
5cb0724e DZ |
193 | static inline bool btrfs_is_block_group_data_only( |
194 | struct btrfs_block_group *block_group) | |
195 | { | |
196 | /* | |
197 | * In mixed mode the fragmentation is expected to be high, lowering the | |
198 | * efficiency, so only proper data block groups are considered. | |
199 | */ | |
200 | return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) && | |
201 | !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA); | |
202 | } | |
203 | ||
aac0023c JB |
204 | #ifdef CONFIG_BTRFS_DEBUG |
205 | static inline int btrfs_should_fragment_free_space( | |
32da5386 | 206 | struct btrfs_block_group *block_group) |
aac0023c JB |
207 | { |
208 | struct btrfs_fs_info *fs_info = block_group->fs_info; | |
209 | ||
210 | return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) && | |
211 | block_group->flags & BTRFS_BLOCK_GROUP_METADATA) || | |
212 | (btrfs_test_opt(fs_info, FRAGMENT_DATA) && | |
213 | block_group->flags & BTRFS_BLOCK_GROUP_DATA); | |
214 | } | |
215 | #endif | |
216 | ||
32da5386 | 217 | struct btrfs_block_group *btrfs_lookup_first_block_group( |
2e405ad8 | 218 | struct btrfs_fs_info *info, u64 bytenr); |
32da5386 | 219 | struct btrfs_block_group *btrfs_lookup_block_group( |
2e405ad8 | 220 | struct btrfs_fs_info *info, u64 bytenr); |
32da5386 DS |
221 | struct btrfs_block_group *btrfs_next_block_group( |
222 | struct btrfs_block_group *cache); | |
223 | void btrfs_get_block_group(struct btrfs_block_group *cache); | |
224 | void btrfs_put_block_group(struct btrfs_block_group *cache); | |
3eeb3226 JB |
225 | void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info, |
226 | const u64 start); | |
32da5386 | 227 | void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg); |
3eeb3226 JB |
228 | bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr); |
229 | void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr); | |
32da5386 DS |
230 | void btrfs_wait_nocow_writers(struct btrfs_block_group *bg); |
231 | void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache, | |
676f1f75 | 232 | u64 num_bytes); |
32da5386 DS |
233 | int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache); |
234 | int btrfs_cache_block_group(struct btrfs_block_group *cache, | |
676f1f75 | 235 | int load_cache_only); |
e3cb339f JB |
236 | void btrfs_put_caching_control(struct btrfs_caching_control *ctl); |
237 | struct btrfs_caching_control *btrfs_get_caching_control( | |
32da5386 DS |
238 | struct btrfs_block_group *cache); |
239 | u64 add_new_free_space(struct btrfs_block_group *block_group, | |
9f21246d | 240 | u64 start, u64 end); |
e3e0520b JB |
241 | struct btrfs_trans_handle *btrfs_start_trans_remove_block_group( |
242 | struct btrfs_fs_info *fs_info, | |
243 | const u64 chunk_offset); | |
244 | int btrfs_remove_block_group(struct btrfs_trans_handle *trans, | |
245 | u64 group_start, struct extent_map *em); | |
246 | void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info); | |
32da5386 | 247 | void btrfs_mark_bg_unused(struct btrfs_block_group *bg); |
4358d963 JB |
248 | int btrfs_read_block_groups(struct btrfs_fs_info *info); |
249 | int btrfs_make_block_group(struct btrfs_trans_handle *trans, u64 bytes_used, | |
250 | u64 type, u64 chunk_offset, u64 size); | |
251 | void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans); | |
b12de528 QW |
252 | int btrfs_inc_block_group_ro(struct btrfs_block_group *cache, |
253 | bool do_chunk_alloc); | |
32da5386 | 254 | void btrfs_dec_block_group_ro(struct btrfs_block_group *cache); |
77745c05 JB |
255 | int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans); |
256 | int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans); | |
257 | int btrfs_setup_space_cache(struct btrfs_trans_handle *trans); | |
ade4b516 JB |
258 | int btrfs_update_block_group(struct btrfs_trans_handle *trans, |
259 | u64 bytenr, u64 num_bytes, int alloc); | |
32da5386 | 260 | int btrfs_add_reserved_bytes(struct btrfs_block_group *cache, |
ade4b516 | 261 | u64 ram_bytes, u64 num_bytes, int delalloc); |
32da5386 | 262 | void btrfs_free_reserved_bytes(struct btrfs_block_group *cache, |
ade4b516 | 263 | u64 num_bytes, int delalloc); |
07730d87 JB |
264 | int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags, |
265 | enum btrfs_chunk_alloc_enum force); | |
266 | int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type); | |
267 | void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type); | |
878d7b67 | 268 | u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags); |
3e43c279 JB |
269 | void btrfs_put_block_group_cache(struct btrfs_fs_info *info); |
270 | int btrfs_free_block_groups(struct btrfs_fs_info *info); | |
878d7b67 JB |
271 | |
272 | static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info) | |
273 | { | |
274 | return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA); | |
275 | } | |
276 | ||
277 | static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info) | |
278 | { | |
279 | return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA); | |
280 | } | |
281 | ||
282 | static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info) | |
283 | { | |
284 | return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM); | |
285 | } | |
676f1f75 | 286 | |
32da5386 | 287 | static inline int btrfs_block_group_done(struct btrfs_block_group *cache) |
676f1f75 JB |
288 | { |
289 | smp_mb(); | |
290 | return cache->cached == BTRFS_CACHE_FINISHED || | |
291 | cache->cached == BTRFS_CACHE_ERROR; | |
292 | } | |
2e405ad8 | 293 | |
684b752b FM |
294 | void btrfs_freeze_block_group(struct btrfs_block_group *cache); |
295 | void btrfs_unfreeze_block_group(struct btrfs_block_group *cache); | |
296 | ||
96a14336 NB |
297 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
298 | int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start, | |
299 | u64 physical, u64 **logical, int *naddrs, int *stripe_len); | |
300 | #endif | |
301 | ||
aac0023c | 302 | #endif /* BTRFS_BLOCK_GROUP_H */ |