xref: /linux/fs/btrfs/block-group.h (revision be239684b18e1cdcafcf8c7face4a2f562c745ad)
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 struct btrfs_chunk_map;
9 
10 enum btrfs_disk_cache_state {
11 	BTRFS_DC_WRITTEN,
12 	BTRFS_DC_ERROR,
13 	BTRFS_DC_CLEAR,
14 	BTRFS_DC_SETUP,
15 };
16 
17 enum btrfs_block_group_size_class {
18 	/* Unset */
19 	BTRFS_BG_SZ_NONE,
20 	/* 0 < size <= 128K */
21 	BTRFS_BG_SZ_SMALL,
22 	/* 128K < size <= 8M */
23 	BTRFS_BG_SZ_MEDIUM,
24 	/* 8M < size < BG_LENGTH */
25 	BTRFS_BG_SZ_LARGE,
26 };
27 
28 /*
29  * This describes the state of the block_group for async discard.  This is due
30  * to the two pass nature of it where extent discarding is prioritized over
31  * bitmap discarding.  BTRFS_DISCARD_RESET_CURSOR is set when we are resetting
32  * between lists to prevent contention for discard state variables
33  * (eg. discard_cursor).
34  */
35 enum btrfs_discard_state {
36 	BTRFS_DISCARD_EXTENTS,
37 	BTRFS_DISCARD_BITMAPS,
38 	BTRFS_DISCARD_RESET_CURSOR,
39 };
40 
41 /*
42  * Control flags for do_chunk_alloc's force field CHUNK_ALLOC_NO_FORCE means to
43  * only allocate a chunk if we really need one.
44  *
45  * CHUNK_ALLOC_LIMITED means to only try and allocate one if we have very few
46  * chunks already allocated.  This is used as part of the clustering code to
47  * help make sure we have a good pool of storage to cluster in, without filling
48  * the FS with empty chunks
49  *
50  * CHUNK_ALLOC_FORCE means it must try to allocate one
51  *
52  * CHUNK_ALLOC_FORCE_FOR_EXTENT like CHUNK_ALLOC_FORCE but called from
53  * find_free_extent() that also activaes the zone
54  */
55 enum btrfs_chunk_alloc_enum {
56 	CHUNK_ALLOC_NO_FORCE,
57 	CHUNK_ALLOC_LIMITED,
58 	CHUNK_ALLOC_FORCE,
59 	CHUNK_ALLOC_FORCE_FOR_EXTENT,
60 };
61 
62 /* Block group flags set at runtime */
63 enum btrfs_block_group_flags {
64 	BLOCK_GROUP_FLAG_IREF,
65 	BLOCK_GROUP_FLAG_REMOVED,
66 	BLOCK_GROUP_FLAG_TO_COPY,
67 	BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
68 	BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED,
69 	BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
70 	BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
71 	/* Does the block group need to be added to the free space tree? */
72 	BLOCK_GROUP_FLAG_NEEDS_FREE_SPACE,
73 	/* Indicate that the block group is placed on a sequential zone */
74 	BLOCK_GROUP_FLAG_SEQUENTIAL_ZONE,
75 	/*
76 	 * Indicate that block group is in the list of new block groups of a
77 	 * transaction.
78 	 */
79 	BLOCK_GROUP_FLAG_NEW,
80 };
81 
82 enum btrfs_caching_type {
83 	BTRFS_CACHE_NO,
84 	BTRFS_CACHE_STARTED,
85 	BTRFS_CACHE_FINISHED,
86 	BTRFS_CACHE_ERROR,
87 };
88 
89 struct btrfs_caching_control {
90 	struct list_head list;
91 	struct mutex mutex;
92 	wait_queue_head_t wait;
93 	struct btrfs_work work;
94 	struct btrfs_block_group *block_group;
95 	/* Track progress of caching during allocation. */
96 	atomic_t progress;
97 	refcount_t count;
98 };
99 
100 /* Once caching_thread() finds this much free space, it will wake up waiters. */
101 #define CACHING_CTL_WAKE_UP SZ_2M
102 
103 struct btrfs_block_group {
104 	struct btrfs_fs_info *fs_info;
105 	struct inode *inode;
106 	spinlock_t lock;
107 	u64 start;
108 	u64 length;
109 	u64 pinned;
110 	u64 reserved;
111 	u64 used;
112 	u64 delalloc_bytes;
113 	u64 bytes_super;
114 	u64 flags;
115 	u64 cache_generation;
116 	u64 global_root_id;
117 
118 	/*
119 	 * The last committed used bytes of this block group, if the above @used
120 	 * is still the same as @commit_used, we don't need to update block
121 	 * group item of this block group.
122 	 */
123 	u64 commit_used;
124 	/*
125 	 * If the free space extent count exceeds this number, convert the block
126 	 * group to bitmaps.
127 	 */
128 	u32 bitmap_high_thresh;
129 
130 	/*
131 	 * If the free space extent count drops below this number, convert the
132 	 * block group back to extents.
133 	 */
134 	u32 bitmap_low_thresh;
135 
136 	/*
137 	 * It is just used for the delayed data space allocation because
138 	 * only the data space allocation and the relative metadata update
139 	 * can be done cross the transaction.
140 	 */
141 	struct rw_semaphore data_rwsem;
142 
143 	/* For raid56, this is a full stripe, without parity */
144 	unsigned long full_stripe_len;
145 	unsigned long runtime_flags;
146 
147 	unsigned int ro;
148 
149 	int disk_cache_state;
150 
151 	/* Cache tracking stuff */
152 	int cached;
153 	struct btrfs_caching_control *caching_ctl;
154 
155 	struct btrfs_space_info *space_info;
156 
157 	/* Free space cache stuff */
158 	struct btrfs_free_space_ctl *free_space_ctl;
159 
160 	/* Block group cache stuff */
161 	struct rb_node cache_node;
162 
163 	/* For block groups in the same raid type */
164 	struct list_head list;
165 
166 	refcount_t refs;
167 
168 	/*
169 	 * List of struct btrfs_free_clusters for this block group.
170 	 * Today it will only have one thing on it, but that may change
171 	 */
172 	struct list_head cluster_list;
173 
174 	/*
175 	 * Used for several lists:
176 	 *
177 	 * 1) struct btrfs_fs_info::unused_bgs
178 	 * 2) struct btrfs_fs_info::reclaim_bgs
179 	 * 3) struct btrfs_transaction::deleted_bgs
180 	 * 4) struct btrfs_trans_handle::new_bgs
181 	 */
182 	struct list_head bg_list;
183 
184 	/* For read-only block groups */
185 	struct list_head ro_list;
186 
187 	/*
188 	 * When non-zero it means the block group's logical address and its
189 	 * device extents can not be reused for future block group allocations
190 	 * until the counter goes down to 0. This is to prevent them from being
191 	 * reused while some task is still using the block group after it was
192 	 * deleted - we want to make sure they can only be reused for new block
193 	 * groups after that task is done with the deleted block group.
194 	 */
195 	atomic_t frozen;
196 
197 	/* For discard operations */
198 	struct list_head discard_list;
199 	int discard_index;
200 	u64 discard_eligible_time;
201 	u64 discard_cursor;
202 	enum btrfs_discard_state discard_state;
203 
204 	/* For dirty block groups */
205 	struct list_head dirty_list;
206 	struct list_head io_list;
207 
208 	struct btrfs_io_ctl io_ctl;
209 
210 	/*
211 	 * Incremented when doing extent allocations and holding a read lock
212 	 * on the space_info's groups_sem semaphore.
213 	 * Decremented when an ordered extent that represents an IO against this
214 	 * block group's range is created (after it's added to its inode's
215 	 * root's list of ordered extents) or immediately after the allocation
216 	 * if it's a metadata extent or fallocate extent (for these cases we
217 	 * don't create ordered extents).
218 	 */
219 	atomic_t reservations;
220 
221 	/*
222 	 * Incremented while holding the spinlock *lock* by a task checking if
223 	 * it can perform a nocow write (incremented if the value for the *ro*
224 	 * field is 0). Decremented by such tasks once they create an ordered
225 	 * extent or before that if some error happens before reaching that step.
226 	 * This is to prevent races between block group relocation and nocow
227 	 * writes through direct IO.
228 	 */
229 	atomic_t nocow_writers;
230 
231 	/* Lock for free space tree operations. */
232 	struct mutex free_space_lock;
233 
234 	/*
235 	 * Number of extents in this block group used for swap files.
236 	 * All accesses protected by the spinlock 'lock'.
237 	 */
238 	int swap_extents;
239 
240 	/*
241 	 * Allocation offset for the block group to implement sequential
242 	 * allocation. This is used only on a zoned filesystem.
243 	 */
244 	u64 alloc_offset;
245 	u64 zone_unusable;
246 	u64 zone_capacity;
247 	u64 meta_write_pointer;
248 	struct btrfs_chunk_map *physical_map;
249 	struct list_head active_bg_list;
250 	struct work_struct zone_finish_work;
251 	struct extent_buffer *last_eb;
252 	enum btrfs_block_group_size_class size_class;
253 };
254 
255 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
256 {
257 	return (block_group->start + block_group->length);
258 }
259 
260 static inline bool btrfs_is_block_group_used(const struct btrfs_block_group *bg)
261 {
262 	lockdep_assert_held(&bg->lock);
263 
264 	return (bg->used > 0 || bg->reserved > 0 || bg->pinned > 0);
265 }
266 
267 static inline bool btrfs_is_block_group_data_only(
268 					struct btrfs_block_group *block_group)
269 {
270 	/*
271 	 * In mixed mode the fragmentation is expected to be high, lowering the
272 	 * efficiency, so only proper data block groups are considered.
273 	 */
274 	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
275 	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
276 }
277 
278 #ifdef CONFIG_BTRFS_DEBUG
279 int btrfs_should_fragment_free_space(struct btrfs_block_group *block_group);
280 #endif
281 
282 struct btrfs_block_group *btrfs_lookup_first_block_group(
283 		struct btrfs_fs_info *info, u64 bytenr);
284 struct btrfs_block_group *btrfs_lookup_block_group(
285 		struct btrfs_fs_info *info, u64 bytenr);
286 struct btrfs_block_group *btrfs_next_block_group(
287 		struct btrfs_block_group *cache);
288 void btrfs_get_block_group(struct btrfs_block_group *cache);
289 void btrfs_put_block_group(struct btrfs_block_group *cache);
290 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
291 					const u64 start);
292 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
293 struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
294 						  u64 bytenr);
295 void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
296 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
297 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
298 				           u64 num_bytes);
299 int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
300 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
301 struct btrfs_caching_control *btrfs_get_caching_control(
302 		struct btrfs_block_group *cache);
303 int btrfs_add_new_free_space(struct btrfs_block_group *block_group,
304 			     u64 start, u64 end, u64 *total_added_ret);
305 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
306 				struct btrfs_fs_info *fs_info,
307 				const u64 chunk_offset);
308 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
309 			     struct btrfs_chunk_map *map);
310 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
311 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
312 void btrfs_reclaim_bgs_work(struct work_struct *work);
313 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
314 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
315 int btrfs_read_block_groups(struct btrfs_fs_info *info);
316 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
317 						 u64 type,
318 						 u64 chunk_offset, u64 size);
319 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
320 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
321 			     bool do_chunk_alloc);
322 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
323 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
324 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
325 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
326 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
327 			     u64 bytenr, u64 num_bytes, bool alloc);
328 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
329 			     u64 ram_bytes, u64 num_bytes, int delalloc,
330 			     bool force_wrong_size_class);
331 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
332 			       u64 num_bytes, int delalloc);
333 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
334 		      enum btrfs_chunk_alloc_enum force);
335 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
336 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
337 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
338 				  bool is_item_insertion);
339 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
340 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
341 int btrfs_free_block_groups(struct btrfs_fs_info *info);
342 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
343 		     u64 physical, u64 **logical, int *naddrs, int *stripe_len);
344 
345 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
346 {
347 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
348 }
349 
350 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
351 {
352 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
353 }
354 
355 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
356 {
357 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
358 }
359 
360 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
361 {
362 	smp_mb();
363 	return cache->cached == BTRFS_CACHE_FINISHED ||
364 		cache->cached == BTRFS_CACHE_ERROR;
365 }
366 
367 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
368 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
369 
370 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
371 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
372 
373 enum btrfs_block_group_size_class btrfs_calc_block_group_size_class(u64 size);
374 int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
375 				     enum btrfs_block_group_size_class size_class,
376 				     bool force_wrong_size_class);
377 bool btrfs_block_group_should_use_size_class(struct btrfs_block_group *bg);
378 
379 #endif /* BTRFS_BLOCK_GROUP_H */
380