xref: /linux/fs/btrfs/block-group.h (revision 7effbd18e496a954540fb6319db0d87d28a70496)
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 
75 enum btrfs_caching_type {
76 	BTRFS_CACHE_NO,
77 	BTRFS_CACHE_STARTED,
78 	BTRFS_CACHE_FINISHED,
79 	BTRFS_CACHE_ERROR,
80 };
81 
82 struct btrfs_caching_control {
83 	struct list_head list;
84 	struct mutex mutex;
85 	wait_queue_head_t wait;
86 	struct btrfs_work work;
87 	struct btrfs_block_group *block_group;
88 	refcount_t count;
89 };
90 
91 /* Once caching_thread() finds this much free space, it will wake up waiters. */
92 #define CACHING_CTL_WAKE_UP SZ_2M
93 
94 /*
95  * Tree to record all locked full stripes of a RAID5/6 block group
96  */
97 struct btrfs_full_stripe_locks_tree {
98 	struct rb_root root;
99 	struct mutex lock;
100 };
101 
102 struct btrfs_block_group {
103 	struct btrfs_fs_info *fs_info;
104 	struct inode *inode;
105 	spinlock_t lock;
106 	u64 start;
107 	u64 length;
108 	u64 pinned;
109 	u64 reserved;
110 	u64 used;
111 	u64 delalloc_bytes;
112 	u64 bytes_super;
113 	u64 flags;
114 	u64 cache_generation;
115 	u64 global_root_id;
116 
117 	/*
118 	 * The last committed used bytes of this block group, if the above @used
119 	 * is still the same as @commit_used, we don't need to update block
120 	 * group item of this block group.
121 	 */
122 	u64 commit_used;
123 	/*
124 	 * If the free space extent count exceeds this number, convert the block
125 	 * group to bitmaps.
126 	 */
127 	u32 bitmap_high_thresh;
128 
129 	/*
130 	 * If the free space extent count drops below this number, convert the
131 	 * block group back to extents.
132 	 */
133 	u32 bitmap_low_thresh;
134 
135 	/*
136 	 * It is just used for the delayed data space allocation because
137 	 * only the data space allocation and the relative metadata update
138 	 * can be done cross the transaction.
139 	 */
140 	struct rw_semaphore data_rwsem;
141 
142 	/* For raid56, this is a full stripe, without parity */
143 	unsigned long full_stripe_len;
144 	unsigned long runtime_flags;
145 
146 	unsigned int ro;
147 
148 	int disk_cache_state;
149 
150 	/* Cache tracking stuff */
151 	int cached;
152 	struct btrfs_caching_control *caching_ctl;
153 
154 	struct btrfs_space_info *space_info;
155 
156 	/* Free space cache stuff */
157 	struct btrfs_free_space_ctl *free_space_ctl;
158 
159 	/* Block group cache stuff */
160 	struct rb_node cache_node;
161 
162 	/* For block groups in the same raid type */
163 	struct list_head list;
164 
165 	refcount_t refs;
166 
167 	/*
168 	 * List of struct btrfs_free_clusters for this block group.
169 	 * Today it will only have one thing on it, but that may change
170 	 */
171 	struct list_head cluster_list;
172 
173 	/* For delayed block group creation or deletion of empty block groups */
174 	struct list_head bg_list;
175 
176 	/* For read-only block groups */
177 	struct list_head ro_list;
178 
179 	/*
180 	 * When non-zero it means the block group's logical address and its
181 	 * device extents can not be reused for future block group allocations
182 	 * until the counter goes down to 0. This is to prevent them from being
183 	 * reused while some task is still using the block group after it was
184 	 * deleted - we want to make sure they can only be reused for new block
185 	 * groups after that task is done with the deleted block group.
186 	 */
187 	atomic_t frozen;
188 
189 	/* For discard operations */
190 	struct list_head discard_list;
191 	int discard_index;
192 	u64 discard_eligible_time;
193 	u64 discard_cursor;
194 	enum btrfs_discard_state discard_state;
195 
196 	/* For dirty block groups */
197 	struct list_head dirty_list;
198 	struct list_head io_list;
199 
200 	struct btrfs_io_ctl io_ctl;
201 
202 	/*
203 	 * Incremented when doing extent allocations and holding a read lock
204 	 * on the space_info's groups_sem semaphore.
205 	 * Decremented when an ordered extent that represents an IO against this
206 	 * block group's range is created (after it's added to its inode's
207 	 * root's list of ordered extents) or immediately after the allocation
208 	 * if it's a metadata extent or fallocate extent (for these cases we
209 	 * don't create ordered extents).
210 	 */
211 	atomic_t reservations;
212 
213 	/*
214 	 * Incremented while holding the spinlock *lock* by a task checking if
215 	 * it can perform a nocow write (incremented if the value for the *ro*
216 	 * field is 0). Decremented by such tasks once they create an ordered
217 	 * extent or before that if some error happens before reaching that step.
218 	 * This is to prevent races between block group relocation and nocow
219 	 * writes through direct IO.
220 	 */
221 	atomic_t nocow_writers;
222 
223 	/* Lock for free space tree operations. */
224 	struct mutex free_space_lock;
225 
226 	/*
227 	 * Number of extents in this block group used for swap files.
228 	 * All accesses protected by the spinlock 'lock'.
229 	 */
230 	int swap_extents;
231 
232 	/* Record locked full stripes for RAID5/6 block group */
233 	struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
234 
235 	/*
236 	 * Allocation offset for the block group to implement sequential
237 	 * allocation. This is used only on a zoned filesystem.
238 	 */
239 	u64 alloc_offset;
240 	u64 zone_unusable;
241 	u64 zone_capacity;
242 	u64 meta_write_pointer;
243 	struct map_lookup *physical_map;
244 	struct list_head active_bg_list;
245 	struct work_struct zone_finish_work;
246 	struct extent_buffer *last_eb;
247 	enum btrfs_block_group_size_class size_class;
248 };
249 
250 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
251 {
252 	return (block_group->start + block_group->length);
253 }
254 
255 static inline bool btrfs_is_block_group_data_only(
256 					struct btrfs_block_group *block_group)
257 {
258 	/*
259 	 * In mixed mode the fragmentation is expected to be high, lowering the
260 	 * efficiency, so only proper data block groups are considered.
261 	 */
262 	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
263 	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
264 }
265 
266 #ifdef CONFIG_BTRFS_DEBUG
267 int btrfs_should_fragment_free_space(struct btrfs_block_group *block_group);
268 #endif
269 
270 struct btrfs_block_group *btrfs_lookup_first_block_group(
271 		struct btrfs_fs_info *info, u64 bytenr);
272 struct btrfs_block_group *btrfs_lookup_block_group(
273 		struct btrfs_fs_info *info, u64 bytenr);
274 struct btrfs_block_group *btrfs_next_block_group(
275 		struct btrfs_block_group *cache);
276 void btrfs_get_block_group(struct btrfs_block_group *cache);
277 void btrfs_put_block_group(struct btrfs_block_group *cache);
278 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
279 					const u64 start);
280 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
281 struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
282 						  u64 bytenr);
283 void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
284 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
285 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
286 				           u64 num_bytes);
287 int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
288 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
289 struct btrfs_caching_control *btrfs_get_caching_control(
290 		struct btrfs_block_group *cache);
291 u64 add_new_free_space(struct btrfs_block_group *block_group,
292 		       u64 start, u64 end);
293 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
294 				struct btrfs_fs_info *fs_info,
295 				const u64 chunk_offset);
296 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
297 			     u64 group_start, struct extent_map *em);
298 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
299 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
300 void btrfs_reclaim_bgs_work(struct work_struct *work);
301 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
302 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
303 int btrfs_read_block_groups(struct btrfs_fs_info *info);
304 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
305 						 u64 bytes_used, u64 type,
306 						 u64 chunk_offset, u64 size);
307 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
308 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
309 			     bool do_chunk_alloc);
310 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
311 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
312 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
313 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
314 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
315 			     u64 bytenr, u64 num_bytes, bool alloc);
316 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
317 			     u64 ram_bytes, u64 num_bytes, int delalloc,
318 			     bool force_wrong_size_class);
319 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
320 			       u64 num_bytes, int delalloc);
321 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
322 		      enum btrfs_chunk_alloc_enum force);
323 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
324 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
325 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
326 				  bool is_item_insertion);
327 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
328 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
329 int btrfs_free_block_groups(struct btrfs_fs_info *info);
330 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
331 		     u64 physical, u64 **logical, int *naddrs, int *stripe_len);
332 
333 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
334 {
335 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
336 }
337 
338 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
339 {
340 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
341 }
342 
343 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
344 {
345 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
346 }
347 
348 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
349 {
350 	smp_mb();
351 	return cache->cached == BTRFS_CACHE_FINISHED ||
352 		cache->cached == BTRFS_CACHE_ERROR;
353 }
354 
355 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
356 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
357 
358 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
359 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
360 
361 enum btrfs_block_group_size_class btrfs_calc_block_group_size_class(u64 size);
362 int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
363 				     enum btrfs_block_group_size_class size_class,
364 				     bool force_wrong_size_class);
365 bool btrfs_block_group_should_use_size_class(struct btrfs_block_group *bg);
366 
367 #endif /* BTRFS_BLOCK_GROUP_H */
368