xref: /linux/fs/btrfs/block-group.h (revision 5e3992fe72748ed3892be876f09d4d990548b7af)
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 struct btrfs_block_group {
95 	struct btrfs_fs_info *fs_info;
96 	struct inode *inode;
97 	spinlock_t lock;
98 	u64 start;
99 	u64 length;
100 	u64 pinned;
101 	u64 reserved;
102 	u64 used;
103 	u64 delalloc_bytes;
104 	u64 bytes_super;
105 	u64 flags;
106 	u64 cache_generation;
107 	u64 global_root_id;
108 
109 	/*
110 	 * The last committed used bytes of this block group, if the above @used
111 	 * is still the same as @commit_used, we don't need to update block
112 	 * group item of this block group.
113 	 */
114 	u64 commit_used;
115 	/*
116 	 * If the free space extent count exceeds this number, convert the block
117 	 * group to bitmaps.
118 	 */
119 	u32 bitmap_high_thresh;
120 
121 	/*
122 	 * If the free space extent count drops below this number, convert the
123 	 * block group back to extents.
124 	 */
125 	u32 bitmap_low_thresh;
126 
127 	/*
128 	 * It is just used for the delayed data space allocation because
129 	 * only the data space allocation and the relative metadata update
130 	 * can be done cross the transaction.
131 	 */
132 	struct rw_semaphore data_rwsem;
133 
134 	/* For raid56, this is a full stripe, without parity */
135 	unsigned long full_stripe_len;
136 	unsigned long runtime_flags;
137 
138 	unsigned int ro;
139 
140 	int disk_cache_state;
141 
142 	/* Cache tracking stuff */
143 	int cached;
144 	struct btrfs_caching_control *caching_ctl;
145 
146 	struct btrfs_space_info *space_info;
147 
148 	/* Free space cache stuff */
149 	struct btrfs_free_space_ctl *free_space_ctl;
150 
151 	/* Block group cache stuff */
152 	struct rb_node cache_node;
153 
154 	/* For block groups in the same raid type */
155 	struct list_head list;
156 
157 	refcount_t refs;
158 
159 	/*
160 	 * List of struct btrfs_free_clusters for this block group.
161 	 * Today it will only have one thing on it, but that may change
162 	 */
163 	struct list_head cluster_list;
164 
165 	/* For delayed block group creation or deletion of empty block groups */
166 	struct list_head bg_list;
167 
168 	/* For read-only block groups */
169 	struct list_head ro_list;
170 
171 	/*
172 	 * When non-zero it means the block group's logical address and its
173 	 * device extents can not be reused for future block group allocations
174 	 * until the counter goes down to 0. This is to prevent them from being
175 	 * reused while some task is still using the block group after it was
176 	 * deleted - we want to make sure they can only be reused for new block
177 	 * groups after that task is done with the deleted block group.
178 	 */
179 	atomic_t frozen;
180 
181 	/* For discard operations */
182 	struct list_head discard_list;
183 	int discard_index;
184 	u64 discard_eligible_time;
185 	u64 discard_cursor;
186 	enum btrfs_discard_state discard_state;
187 
188 	/* For dirty block groups */
189 	struct list_head dirty_list;
190 	struct list_head io_list;
191 
192 	struct btrfs_io_ctl io_ctl;
193 
194 	/*
195 	 * Incremented when doing extent allocations and holding a read lock
196 	 * on the space_info's groups_sem semaphore.
197 	 * Decremented when an ordered extent that represents an IO against this
198 	 * block group's range is created (after it's added to its inode's
199 	 * root's list of ordered extents) or immediately after the allocation
200 	 * if it's a metadata extent or fallocate extent (for these cases we
201 	 * don't create ordered extents).
202 	 */
203 	atomic_t reservations;
204 
205 	/*
206 	 * Incremented while holding the spinlock *lock* by a task checking if
207 	 * it can perform a nocow write (incremented if the value for the *ro*
208 	 * field is 0). Decremented by such tasks once they create an ordered
209 	 * extent or before that if some error happens before reaching that step.
210 	 * This is to prevent races between block group relocation and nocow
211 	 * writes through direct IO.
212 	 */
213 	atomic_t nocow_writers;
214 
215 	/* Lock for free space tree operations. */
216 	struct mutex free_space_lock;
217 
218 	/*
219 	 * Number of extents in this block group used for swap files.
220 	 * All accesses protected by the spinlock 'lock'.
221 	 */
222 	int swap_extents;
223 
224 	/*
225 	 * Allocation offset for the block group to implement sequential
226 	 * allocation. This is used only on a zoned filesystem.
227 	 */
228 	u64 alloc_offset;
229 	u64 zone_unusable;
230 	u64 zone_capacity;
231 	u64 meta_write_pointer;
232 	struct map_lookup *physical_map;
233 	struct list_head active_bg_list;
234 	struct work_struct zone_finish_work;
235 	struct extent_buffer *last_eb;
236 	enum btrfs_block_group_size_class size_class;
237 };
238 
239 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
240 {
241 	return (block_group->start + block_group->length);
242 }
243 
244 static inline bool btrfs_is_block_group_data_only(
245 					struct btrfs_block_group *block_group)
246 {
247 	/*
248 	 * In mixed mode the fragmentation is expected to be high, lowering the
249 	 * efficiency, so only proper data block groups are considered.
250 	 */
251 	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
252 	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
253 }
254 
255 #ifdef CONFIG_BTRFS_DEBUG
256 int btrfs_should_fragment_free_space(struct btrfs_block_group *block_group);
257 #endif
258 
259 struct btrfs_block_group *btrfs_lookup_first_block_group(
260 		struct btrfs_fs_info *info, u64 bytenr);
261 struct btrfs_block_group *btrfs_lookup_block_group(
262 		struct btrfs_fs_info *info, u64 bytenr);
263 struct btrfs_block_group *btrfs_next_block_group(
264 		struct btrfs_block_group *cache);
265 void btrfs_get_block_group(struct btrfs_block_group *cache);
266 void btrfs_put_block_group(struct btrfs_block_group *cache);
267 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
268 					const u64 start);
269 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
270 struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
271 						  u64 bytenr);
272 void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
273 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
274 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
275 				           u64 num_bytes);
276 int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
277 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
278 struct btrfs_caching_control *btrfs_get_caching_control(
279 		struct btrfs_block_group *cache);
280 u64 add_new_free_space(struct btrfs_block_group *block_group,
281 		       u64 start, u64 end);
282 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
283 				struct btrfs_fs_info *fs_info,
284 				const u64 chunk_offset);
285 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
286 			     u64 group_start, struct extent_map *em);
287 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
288 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
289 void btrfs_reclaim_bgs_work(struct work_struct *work);
290 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
291 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
292 int btrfs_read_block_groups(struct btrfs_fs_info *info);
293 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
294 						 u64 type,
295 						 u64 chunk_offset, u64 size);
296 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
297 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
298 			     bool do_chunk_alloc);
299 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
300 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
301 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
302 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
303 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
304 			     u64 bytenr, u64 num_bytes, bool alloc);
305 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
306 			     u64 ram_bytes, u64 num_bytes, int delalloc,
307 			     bool force_wrong_size_class);
308 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
309 			       u64 num_bytes, int delalloc);
310 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
311 		      enum btrfs_chunk_alloc_enum force);
312 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
313 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
314 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
315 				  bool is_item_insertion);
316 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
317 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
318 int btrfs_free_block_groups(struct btrfs_fs_info *info);
319 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
320 		     u64 physical, u64 **logical, int *naddrs, int *stripe_len);
321 
322 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
323 {
324 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
325 }
326 
327 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
328 {
329 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
330 }
331 
332 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
333 {
334 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
335 }
336 
337 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
338 {
339 	smp_mb();
340 	return cache->cached == BTRFS_CACHE_FINISHED ||
341 		cache->cached == BTRFS_CACHE_ERROR;
342 }
343 
344 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
345 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
346 
347 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
348 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
349 
350 enum btrfs_block_group_size_class btrfs_calc_block_group_size_class(u64 size);
351 int btrfs_use_block_group_size_class(struct btrfs_block_group *bg,
352 				     enum btrfs_block_group_size_class size_class,
353 				     bool force_wrong_size_class);
354 bool btrfs_block_group_should_use_size_class(struct btrfs_block_group *bg);
355 
356 #endif /* BTRFS_BLOCK_GROUP_H */
357