1 /* SPDX-License-Identifier: GPL-2.0 */
2
3 #ifndef BTRFS_BLOCK_GROUP_H
4 #define BTRFS_BLOCK_GROUP_H
5
6 #include "free-space-cache.h"
7
8 enum btrfs_disk_cache_state {
9 BTRFS_DC_WRITTEN,
10 BTRFS_DC_ERROR,
11 BTRFS_DC_CLEAR,
12 BTRFS_DC_SETUP,
13 };
14
15 enum btrfs_block_group_size_class {
16 /* Unset */
17 BTRFS_BG_SZ_NONE,
18 /* 0 < size <= 128K */
19 BTRFS_BG_SZ_SMALL,
20 /* 128K < size <= 8M */
21 BTRFS_BG_SZ_MEDIUM,
22 /* 8M < size < BG_LENGTH */
23 BTRFS_BG_SZ_LARGE,
24 };
25
26 /*
27 * This describes the state of the block_group for async discard. This is due
28 * to the two pass nature of it where extent discarding is prioritized over
29 * bitmap discarding. BTRFS_DISCARD_RESET_CURSOR is set when we are resetting
30 * between lists to prevent contention for discard state variables
31 * (eg. discard_cursor).
32 */
33 enum btrfs_discard_state {
34 BTRFS_DISCARD_EXTENTS,
35 BTRFS_DISCARD_BITMAPS,
36 BTRFS_DISCARD_RESET_CURSOR,
37 };
38
39 /*
40 * Control flags for do_chunk_alloc's force field CHUNK_ALLOC_NO_FORCE means to
41 * only allocate a chunk if we really need one.
42 *
43 * CHUNK_ALLOC_LIMITED means to only try and allocate one if we have very few
44 * chunks already allocated. This is used as part of the clustering code to
45 * help make sure we have a good pool of storage to cluster in, without filling
46 * the FS with empty chunks
47 *
48 * CHUNK_ALLOC_FORCE means it must try to allocate one
49 *
50 * CHUNK_ALLOC_FORCE_FOR_EXTENT like CHUNK_ALLOC_FORCE but called from
51 * find_free_extent() that also activaes the zone
52 */
53 enum btrfs_chunk_alloc_enum {
54 CHUNK_ALLOC_NO_FORCE,
55 CHUNK_ALLOC_LIMITED,
56 CHUNK_ALLOC_FORCE,
57 CHUNK_ALLOC_FORCE_FOR_EXTENT,
58 };
59
60 /* Block group flags set at runtime */
61 enum btrfs_block_group_flags {
62 BLOCK_GROUP_FLAG_IREF,
63 BLOCK_GROUP_FLAG_REMOVED,
64 BLOCK_GROUP_FLAG_TO_COPY,
65 BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
66 BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED,
67 BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
68 BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
69 /* Does the block group need to be added to the free space tree? */
70 BLOCK_GROUP_FLAG_NEEDS_FREE_SPACE,
71 /* Indicate that the block group is placed on a sequential zone */
72 BLOCK_GROUP_FLAG_SEQUENTIAL_ZONE,
73 /*
74 * Indicate that block group is in the list of new block groups of a
75 * transaction.
76 */
77 BLOCK_GROUP_FLAG_NEW,
78 };
79
80 enum btrfs_caching_type {
81 BTRFS_CACHE_NO,
82 BTRFS_CACHE_STARTED,
83 BTRFS_CACHE_FINISHED,
84 BTRFS_CACHE_ERROR,
85 };
86
87 struct btrfs_caching_control {
88 struct list_head list;
89 struct mutex mutex;
90 wait_queue_head_t wait;
91 struct btrfs_work work;
92 struct btrfs_block_group *block_group;
93 /* Track progress of caching during allocation. */
94 atomic_t progress;
95 refcount_t count;
96 };
97
98 /* Once caching_thread() finds this much free space, it will wake up waiters. */
99 #define CACHING_CTL_WAKE_UP SZ_2M
100
101 struct btrfs_block_group {
102 struct btrfs_fs_info *fs_info;
103 struct inode *inode;
104 spinlock_t lock;
105 u64 start;
106 u64 length;
107 u64 pinned;
108 u64 reserved;
109 u64 used;
110 u64 delalloc_bytes;
111 u64 bytes_super;
112 u64 flags;
113 u64 cache_generation;
114 u64 global_root_id;
115
116 /*
117 * The last committed used bytes of this block group, if the above @used
118 * is still the same as @commit_used, we don't need to update block
119 * group item of this block group.
120 */
121 u64 commit_used;
122 /*
123 * If the free space extent count exceeds this number, convert the block
124 * group to bitmaps.
125 */
126 u32 bitmap_high_thresh;
127
128 /*
129 * If the free space extent count drops below this number, convert the
130 * block group back to extents.
131 */
132 u32 bitmap_low_thresh;
133
134 /*
135 * It is just used for the delayed data space allocation because
136 * only the data space allocation and the relative metadata update
137 * can be done cross the transaction.
138 */
139 struct rw_semaphore data_rwsem;
140
141 /* For raid56, this is a full stripe, without parity */
142 unsigned long full_stripe_len;
143 unsigned long runtime_flags;
144
145 unsigned int ro;
146
147 int disk_cache_state;
148
149 /* Cache tracking stuff */
150 int cached;
151 struct btrfs_caching_control *caching_ctl;
152
153 struct btrfs_space_info *space_info;
154
155 /* Free space cache stuff */
156 struct btrfs_free_space_ctl *free_space_ctl;
157
158 /* Block group cache stuff */
159 struct rb_node cache_node;
160
161 /* For block groups in the same raid type */
162 struct list_head list;
163
164 refcount_t refs;
165
166 /*
167 * List of struct btrfs_free_clusters for this block group.
168 * Today it will only have one thing on it, but that may change
169 */
170 struct list_head cluster_list;
171
172 /*
173 * Used for several lists:
174 *
175 * 1) struct btrfs_fs_info::unused_bgs
176 * 2) struct btrfs_fs_info::reclaim_bgs
177 * 3) struct btrfs_transaction::deleted_bgs
178 * 4) struct btrfs_trans_handle::new_bgs
179 */
180 struct list_head bg_list;
181
182 /* For read-only block groups */
183 struct list_head ro_list;
184
185 /*
186 * When non-zero it means the block group's logical address and its
187 * device extents can not be reused for future block group allocations
188 * until the counter goes down to 0. This is to prevent them from being
189 * reused while some task is still using the block group after it was
190 * deleted - we want to make sure they can only be reused for new block
191 * groups after that task is done with the deleted block group.
192 */
193 atomic_t frozen;
194
195 /* For discard operations */
196 struct list_head discard_list;
197 int discard_index;
198 u64 discard_eligible_time;
199 u64 discard_cursor;
200 enum btrfs_discard_state discard_state;
201
202 /* For dirty block groups */
203 struct list_head dirty_list;
204 struct list_head io_list;
205
206 struct btrfs_io_ctl io_ctl;
207
208 /*
209 * Incremented when doing extent allocations and holding a read lock
210 * on the space_info's groups_sem semaphore.
211 * Decremented when an ordered extent that represents an IO against this
212 * block group's range is created (after it's added to its inode's
213 * root's list of ordered extents) or immediately after the allocation
214 * if it's a metadata extent or fallocate extent (for these cases we
215 * don't create ordered extents).
216 */
217 atomic_t reservations;
218
219 /*
220 * Incremented while holding the spinlock *lock* by a task checking if
221 * it can perform a nocow write (incremented if the value for the *ro*
222 * field is 0). Decremented by such tasks once they create an ordered
223 * extent or before that if some error happens before reaching that step.
224 * This is to prevent races between block group relocation and nocow
225 * writes through direct IO.
226 */
227 atomic_t nocow_writers;
228
229 /* Lock for free space tree operations. */
230 struct mutex free_space_lock;
231
232 /*
233 * Number of extents in this block group used for swap files.
234 * All accesses protected by the spinlock 'lock'.
235 */
236 int swap_extents;
237
238 /*
239 * Allocation offset for the block group to implement sequential
240 * allocation. This is used only on a zoned filesystem.
241 */
242 u64 alloc_offset;
243 u64 zone_unusable;
244 u64 zone_capacity;
245 u64 meta_write_pointer;
246 struct map_lookup *physical_map;
247 struct list_head active_bg_list;
248 struct work_struct zone_finish_work;
249 struct extent_buffer *last_eb;
250 enum btrfs_block_group_size_class size_class;
251 };
252
btrfs_block_group_end(struct btrfs_block_group * block_group)253 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
254 {
255 return (block_group->start + block_group->length);
256 }
257
btrfs_is_block_group_used(const struct btrfs_block_group * bg)258 static inline bool btrfs_is_block_group_used(const struct btrfs_block_group *bg)
259 {
260 lockdep_assert_held(&bg->lock);
261
262 return (bg->used > 0 || bg->reserved > 0 || bg->pinned > 0);
263 }
264
btrfs_is_block_group_data_only(struct btrfs_block_group * block_group)265 static inline bool btrfs_is_block_group_data_only(
266 struct btrfs_block_group *block_group)
267 {
268 /*
269 * In mixed mode the fragmentation is expected to be high, lowering the
270 * efficiency, so only proper data block groups are considered.
271 */
272 return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
273 !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
274 }
275
276 #ifdef CONFIG_BTRFS_DEBUG
277 int btrfs_should_fragment_free_space(struct btrfs_block_group *block_group);
278 #endif
279
280 struct btrfs_block_group *btrfs_lookup_first_block_group(
281 struct btrfs_fs_info *info, u64 bytenr);
282 struct btrfs_block_group *btrfs_lookup_block_group(
283 struct btrfs_fs_info *info, u64 bytenr);
284 struct btrfs_block_group *btrfs_next_block_group(
285 struct btrfs_block_group *cache);
286 void btrfs_get_block_group(struct btrfs_block_group *cache);
287 void btrfs_put_block_group(struct btrfs_block_group *cache);
288 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
289 const u64 start);
290 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
291 struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
292 u64 bytenr);
293 void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
294 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
295 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
296 u64 num_bytes);
297 int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
298 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
299 struct btrfs_caching_control *btrfs_get_caching_control(
300 struct btrfs_block_group *cache);
301 int btrfs_add_new_free_space(struct btrfs_block_group *block_group,
302 u64 start, u64 end, u64 *total_added_ret);
303 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
304 struct btrfs_fs_info *fs_info,
305 const u64 chunk_offset);
306 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
307 u64 group_start, struct extent_map *em);
308 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
309 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
310 void btrfs_reclaim_bgs_work(struct work_struct *work);
311 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
312 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
313 int btrfs_read_block_groups(struct btrfs_fs_info *info);
314 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
315 u64 type,
316 u64 chunk_offset, u64 size);
317 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
318 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
319 bool do_chunk_alloc);
320 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
321 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
322 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
323 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
324 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
325 u64 bytenr, u64 num_bytes, bool alloc);
326 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
327 u64 ram_bytes, u64 num_bytes, int delalloc,
328 bool force_wrong_size_class);
329 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
330 u64 num_bytes, int delalloc);
331 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
332 enum btrfs_chunk_alloc_enum force);
333 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
334 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
335 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
336 bool is_item_insertion);
337 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
338 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
339 int btrfs_free_block_groups(struct btrfs_fs_info *info);
340 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
341 u64 physical, u64 **logical, int *naddrs, int *stripe_len);
342
btrfs_data_alloc_profile(struct btrfs_fs_info * fs_info)343 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
344 {
345 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
346 }
347
btrfs_metadata_alloc_profile(struct btrfs_fs_info * fs_info)348 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
349 {
350 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
351 }
352
btrfs_system_alloc_profile(struct btrfs_fs_info * fs_info)353 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
354 {
355 return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
356 }
357
btrfs_block_group_done(struct btrfs_block_group * cache)358 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
359 {
360 smp_mb();
361 return cache->cached == BTRFS_CACHE_FINISHED ||
362 cache->cached == BTRFS_CACHE_ERROR;
363 }
364
365 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
366 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
367
368 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
369 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
370
371 enum btrfs_block_group_size_class btrfs_calc_block_group_size_class(u64 size);
372 int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
373 enum btrfs_block_group_size_class size_class,
374 bool force_wrong_size_class);
375 bool btrfs_block_group_should_use_size_class(struct btrfs_block_group *bg);
376
377 #endif /* BTRFS_BLOCK_GROUP_H */
378