xref: /linux/fs/btrfs/block-group.h (revision 497e6b37b0099dc415578488287fd84fb74433eb)
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 /*
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 
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  * CHUNK_ALLOC_FORCE_FOR_EXTENT like CHUNK_ALLOC_FORCE but called from
40  * find_free_extent() that also activaes the zone
41  */
42 enum btrfs_chunk_alloc_enum {
43 	CHUNK_ALLOC_NO_FORCE,
44 	CHUNK_ALLOC_LIMITED,
45 	CHUNK_ALLOC_FORCE,
46 	CHUNK_ALLOC_FORCE_FOR_EXTENT,
47 };
48 
49 /* Block group flags set at runtime */
50 enum btrfs_block_group_flags {
51 	BLOCK_GROUP_FLAG_IREF,
52 	BLOCK_GROUP_FLAG_REMOVED,
53 	BLOCK_GROUP_FLAG_TO_COPY,
54 	BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
55 	BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED,
56 	BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
57 	BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
58 	/* Does the block group need to be added to the free space tree? */
59 	BLOCK_GROUP_FLAG_NEEDS_FREE_SPACE,
60 	/* Indicate that the block group is placed on a sequential zone */
61 	BLOCK_GROUP_FLAG_SEQUENTIAL_ZONE,
62 };
63 
64 enum btrfs_caching_type {
65 	BTRFS_CACHE_NO,
66 	BTRFS_CACHE_STARTED,
67 	BTRFS_CACHE_FINISHED,
68 	BTRFS_CACHE_ERROR,
69 };
70 
71 struct btrfs_caching_control {
72 	struct list_head list;
73 	struct mutex mutex;
74 	wait_queue_head_t wait;
75 	struct btrfs_work work;
76 	struct btrfs_block_group *block_group;
77 	refcount_t count;
78 };
79 
80 /* Once caching_thread() finds this much free space, it will wake up waiters. */
81 #define CACHING_CTL_WAKE_UP SZ_2M
82 
83 /*
84  * Tree to record all locked full stripes of a RAID5/6 block group
85  */
86 struct btrfs_full_stripe_locks_tree {
87 	struct rb_root root;
88 	struct mutex lock;
89 };
90 
91 struct btrfs_block_group {
92 	struct btrfs_fs_info *fs_info;
93 	struct inode *inode;
94 	spinlock_t lock;
95 	u64 start;
96 	u64 length;
97 	u64 pinned;
98 	u64 reserved;
99 	u64 used;
100 	u64 delalloc_bytes;
101 	u64 bytes_super;
102 	u64 flags;
103 	u64 cache_generation;
104 	u64 global_root_id;
105 
106 	/*
107 	 * The last committed used bytes of this block group, if the above @used
108 	 * is still the same as @commit_used, we don't need to update block
109 	 * group item of this block group.
110 	 */
111 	u64 commit_used;
112 	/*
113 	 * If the free space extent count exceeds this number, convert the block
114 	 * group to bitmaps.
115 	 */
116 	u32 bitmap_high_thresh;
117 
118 	/*
119 	 * If the free space extent count drops below this number, convert the
120 	 * block group back to extents.
121 	 */
122 	u32 bitmap_low_thresh;
123 
124 	/*
125 	 * It is just used for the delayed data space allocation because
126 	 * only the data space allocation and the relative metadata update
127 	 * can be done cross the transaction.
128 	 */
129 	struct rw_semaphore data_rwsem;
130 
131 	/* For raid56, this is a full stripe, without parity */
132 	unsigned long full_stripe_len;
133 	unsigned long runtime_flags;
134 
135 	unsigned int ro;
136 
137 	int disk_cache_state;
138 
139 	/* Cache tracking stuff */
140 	int cached;
141 	struct btrfs_caching_control *caching_ctl;
142 
143 	struct btrfs_space_info *space_info;
144 
145 	/* Free space cache stuff */
146 	struct btrfs_free_space_ctl *free_space_ctl;
147 
148 	/* Block group cache stuff */
149 	struct rb_node cache_node;
150 
151 	/* For block groups in the same raid type */
152 	struct list_head list;
153 
154 	refcount_t refs;
155 
156 	/*
157 	 * List of struct btrfs_free_clusters for this block group.
158 	 * Today it will only have one thing on it, but that may change
159 	 */
160 	struct list_head cluster_list;
161 
162 	/* For delayed block group creation or deletion of empty block groups */
163 	struct list_head bg_list;
164 
165 	/* For read-only block groups */
166 	struct list_head ro_list;
167 
168 	/*
169 	 * When non-zero it means the block group's logical address and its
170 	 * device extents can not be reused for future block group allocations
171 	 * until the counter goes down to 0. This is to prevent them from being
172 	 * reused while some task is still using the block group after it was
173 	 * deleted - we want to make sure they can only be reused for new block
174 	 * groups after that task is done with the deleted block group.
175 	 */
176 	atomic_t frozen;
177 
178 	/* For discard operations */
179 	struct list_head discard_list;
180 	int discard_index;
181 	u64 discard_eligible_time;
182 	u64 discard_cursor;
183 	enum btrfs_discard_state discard_state;
184 
185 	/* For dirty block groups */
186 	struct list_head dirty_list;
187 	struct list_head io_list;
188 
189 	struct btrfs_io_ctl io_ctl;
190 
191 	/*
192 	 * Incremented when doing extent allocations and holding a read lock
193 	 * on the space_info's groups_sem semaphore.
194 	 * Decremented when an ordered extent that represents an IO against this
195 	 * block group's range is created (after it's added to its inode's
196 	 * root's list of ordered extents) or immediately after the allocation
197 	 * if it's a metadata extent or fallocate extent (for these cases we
198 	 * don't create ordered extents).
199 	 */
200 	atomic_t reservations;
201 
202 	/*
203 	 * Incremented while holding the spinlock *lock* by a task checking if
204 	 * it can perform a nocow write (incremented if the value for the *ro*
205 	 * field is 0). Decremented by such tasks once they create an ordered
206 	 * extent or before that if some error happens before reaching that step.
207 	 * This is to prevent races between block group relocation and nocow
208 	 * writes through direct IO.
209 	 */
210 	atomic_t nocow_writers;
211 
212 	/* Lock for free space tree operations. */
213 	struct mutex free_space_lock;
214 
215 	/*
216 	 * Number of extents in this block group used for swap files.
217 	 * All accesses protected by the spinlock 'lock'.
218 	 */
219 	int swap_extents;
220 
221 	/* Record locked full stripes for RAID5/6 block group */
222 	struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
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 };
237 
238 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
239 {
240 	return (block_group->start + block_group->length);
241 }
242 
243 static inline bool btrfs_is_block_group_data_only(
244 					struct btrfs_block_group *block_group)
245 {
246 	/*
247 	 * In mixed mode the fragmentation is expected to be high, lowering the
248 	 * efficiency, so only proper data block groups are considered.
249 	 */
250 	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
251 	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
252 }
253 
254 #ifdef CONFIG_BTRFS_DEBUG
255 int btrfs_should_fragment_free_space(struct btrfs_block_group *block_group);
256 #endif
257 
258 struct btrfs_block_group *btrfs_lookup_first_block_group(
259 		struct btrfs_fs_info *info, u64 bytenr);
260 struct btrfs_block_group *btrfs_lookup_block_group(
261 		struct btrfs_fs_info *info, u64 bytenr);
262 struct btrfs_block_group *btrfs_next_block_group(
263 		struct btrfs_block_group *cache);
264 void btrfs_get_block_group(struct btrfs_block_group *cache);
265 void btrfs_put_block_group(struct btrfs_block_group *cache);
266 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
267 					const u64 start);
268 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
269 struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
270 						  u64 bytenr);
271 void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
272 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
273 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
274 				           u64 num_bytes);
275 int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
276 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
277 struct btrfs_caching_control *btrfs_get_caching_control(
278 		struct btrfs_block_group *cache);
279 u64 add_new_free_space(struct btrfs_block_group *block_group,
280 		       u64 start, u64 end);
281 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
282 				struct btrfs_fs_info *fs_info,
283 				const u64 chunk_offset);
284 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
285 			     u64 group_start, struct extent_map *em);
286 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
287 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
288 void btrfs_reclaim_bgs_work(struct work_struct *work);
289 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
290 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
291 int btrfs_read_block_groups(struct btrfs_fs_info *info);
292 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
293 						 u64 bytes_used, u64 type,
294 						 u64 chunk_offset, u64 size);
295 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
296 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
297 			     bool do_chunk_alloc);
298 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
299 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
300 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
301 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
302 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
303 			     u64 bytenr, u64 num_bytes, bool alloc);
304 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
305 			     u64 ram_bytes, u64 num_bytes, int delalloc);
306 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
307 			       u64 num_bytes, int delalloc);
308 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
309 		      enum btrfs_chunk_alloc_enum force);
310 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
311 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
312 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
313 				  bool is_item_insertion);
314 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
315 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
316 int btrfs_free_block_groups(struct btrfs_fs_info *info);
317 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
318 		       struct block_device *bdev, u64 physical, u64 **logical,
319 		       int *naddrs, int *stripe_len);
320 
321 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
322 {
323 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
324 }
325 
326 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
327 {
328 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
329 }
330 
331 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
332 {
333 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
334 }
335 
336 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
337 {
338 	smp_mb();
339 	return cache->cached == BTRFS_CACHE_FINISHED ||
340 		cache->cached == BTRFS_CACHE_ERROR;
341 }
342 
343 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
344 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
345 
346 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
347 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
348 
349 #endif /* BTRFS_BLOCK_GROUP_H */
350