xref: /openbmc/linux/fs/btrfs/block-group.h (revision 519b58bb)
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 };
59 
60 enum btrfs_caching_type {
61 	BTRFS_CACHE_NO,
62 	BTRFS_CACHE_STARTED,
63 	BTRFS_CACHE_FINISHED,
64 	BTRFS_CACHE_ERROR,
65 };
66 
67 struct btrfs_caching_control {
68 	struct list_head list;
69 	struct mutex mutex;
70 	wait_queue_head_t wait;
71 	struct btrfs_work work;
72 	struct btrfs_block_group *block_group;
73 	refcount_t count;
74 };
75 
76 /* Once caching_thread() finds this much free space, it will wake up waiters. */
77 #define CACHING_CTL_WAKE_UP SZ_2M
78 
79 /*
80  * Tree to record all locked full stripes of a RAID5/6 block group
81  */
82 struct btrfs_full_stripe_locks_tree {
83 	struct rb_root root;
84 	struct mutex lock;
85 };
86 
87 struct btrfs_block_group {
88 	struct btrfs_fs_info *fs_info;
89 	struct inode *inode;
90 	spinlock_t lock;
91 	u64 start;
92 	u64 length;
93 	u64 pinned;
94 	u64 reserved;
95 	u64 used;
96 	u64 delalloc_bytes;
97 	u64 bytes_super;
98 	u64 flags;
99 	u64 cache_generation;
100 	u64 global_root_id;
101 
102 	/*
103 	 * If the free space extent count exceeds this number, convert the block
104 	 * group to bitmaps.
105 	 */
106 	u32 bitmap_high_thresh;
107 
108 	/*
109 	 * If the free space extent count drops below this number, convert the
110 	 * block group back to extents.
111 	 */
112 	u32 bitmap_low_thresh;
113 
114 	/*
115 	 * It is just used for the delayed data space allocation because
116 	 * only the data space allocation and the relative metadata update
117 	 * can be done cross the transaction.
118 	 */
119 	struct rw_semaphore data_rwsem;
120 
121 	/* For raid56, this is a full stripe, without parity */
122 	unsigned long full_stripe_len;
123 	unsigned long runtime_flags;
124 
125 	unsigned int ro;
126 
127 	int disk_cache_state;
128 
129 	/* Cache tracking stuff */
130 	int cached;
131 	struct btrfs_caching_control *caching_ctl;
132 
133 	struct btrfs_space_info *space_info;
134 
135 	/* Free space cache stuff */
136 	struct btrfs_free_space_ctl *free_space_ctl;
137 
138 	/* Block group cache stuff */
139 	struct rb_node cache_node;
140 
141 	/* For block groups in the same raid type */
142 	struct list_head list;
143 
144 	refcount_t refs;
145 
146 	/*
147 	 * List of struct btrfs_free_clusters for this block group.
148 	 * Today it will only have one thing on it, but that may change
149 	 */
150 	struct list_head cluster_list;
151 
152 	/* For delayed block group creation or deletion of empty block groups */
153 	struct list_head bg_list;
154 
155 	/* For read-only block groups */
156 	struct list_head ro_list;
157 
158 	/*
159 	 * When non-zero it means the block group's logical address and its
160 	 * device extents can not be reused for future block group allocations
161 	 * until the counter goes down to 0. This is to prevent them from being
162 	 * reused while some task is still using the block group after it was
163 	 * deleted - we want to make sure they can only be reused for new block
164 	 * groups after that task is done with the deleted block group.
165 	 */
166 	atomic_t frozen;
167 
168 	/* For discard operations */
169 	struct list_head discard_list;
170 	int discard_index;
171 	u64 discard_eligible_time;
172 	u64 discard_cursor;
173 	enum btrfs_discard_state discard_state;
174 
175 	/* For dirty block groups */
176 	struct list_head dirty_list;
177 	struct list_head io_list;
178 
179 	struct btrfs_io_ctl io_ctl;
180 
181 	/*
182 	 * Incremented when doing extent allocations and holding a read lock
183 	 * on the space_info's groups_sem semaphore.
184 	 * Decremented when an ordered extent that represents an IO against this
185 	 * block group's range is created (after it's added to its inode's
186 	 * root's list of ordered extents) or immediately after the allocation
187 	 * if it's a metadata extent or fallocate extent (for these cases we
188 	 * don't create ordered extents).
189 	 */
190 	atomic_t reservations;
191 
192 	/*
193 	 * Incremented while holding the spinlock *lock* by a task checking if
194 	 * it can perform a nocow write (incremented if the value for the *ro*
195 	 * field is 0). Decremented by such tasks once they create an ordered
196 	 * extent or before that if some error happens before reaching that step.
197 	 * This is to prevent races between block group relocation and nocow
198 	 * writes through direct IO.
199 	 */
200 	atomic_t nocow_writers;
201 
202 	/* Lock for free space tree operations. */
203 	struct mutex free_space_lock;
204 
205 	/*
206 	 * Does the block group need to be added to the free space tree?
207 	 * Protected by free_space_lock.
208 	 */
209 	int needs_free_space;
210 
211 	/* Flag indicating this block group is placed on a sequential zone */
212 	bool seq_zone;
213 
214 	/*
215 	 * Number of extents in this block group used for swap files.
216 	 * All accesses protected by the spinlock 'lock'.
217 	 */
218 	int swap_extents;
219 
220 	/* Record locked full stripes for RAID5/6 block group */
221 	struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
222 
223 	/*
224 	 * Allocation offset for the block group to implement sequential
225 	 * allocation. This is used only on a zoned filesystem.
226 	 */
227 	u64 alloc_offset;
228 	u64 zone_unusable;
229 	u64 zone_capacity;
230 	u64 meta_write_pointer;
231 	struct map_lookup *physical_map;
232 	struct list_head active_bg_list;
233 	struct work_struct zone_finish_work;
234 	struct extent_buffer *last_eb;
235 };
236 
237 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
238 {
239 	return (block_group->start + block_group->length);
240 }
241 
242 static inline bool btrfs_is_block_group_data_only(
243 					struct btrfs_block_group *block_group)
244 {
245 	/*
246 	 * In mixed mode the fragmentation is expected to be high, lowering the
247 	 * efficiency, so only proper data block groups are considered.
248 	 */
249 	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
250 	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
251 }
252 
253 #ifdef CONFIG_BTRFS_DEBUG
254 static inline int btrfs_should_fragment_free_space(
255 		struct btrfs_block_group *block_group)
256 {
257 	struct btrfs_fs_info *fs_info = block_group->fs_info;
258 
259 	return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
260 		block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
261 	       (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
262 		block_group->flags &  BTRFS_BLOCK_GROUP_DATA);
263 }
264 #endif
265 
266 struct btrfs_block_group *btrfs_lookup_first_block_group(
267 		struct btrfs_fs_info *info, u64 bytenr);
268 struct btrfs_block_group *btrfs_lookup_block_group(
269 		struct btrfs_fs_info *info, u64 bytenr);
270 struct btrfs_block_group *btrfs_next_block_group(
271 		struct btrfs_block_group *cache);
272 void btrfs_get_block_group(struct btrfs_block_group *cache);
273 void btrfs_put_block_group(struct btrfs_block_group *cache);
274 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
275 					const u64 start);
276 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
277 struct btrfs_block_group *btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info,
278 						  u64 bytenr);
279 void btrfs_dec_nocow_writers(struct btrfs_block_group *bg);
280 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
281 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
282 				           u64 num_bytes);
283 int btrfs_cache_block_group(struct btrfs_block_group *cache, bool wait);
284 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
285 struct btrfs_caching_control *btrfs_get_caching_control(
286 		struct btrfs_block_group *cache);
287 u64 add_new_free_space(struct btrfs_block_group *block_group,
288 		       u64 start, u64 end);
289 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
290 				struct btrfs_fs_info *fs_info,
291 				const u64 chunk_offset);
292 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
293 			     u64 group_start, struct extent_map *em);
294 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
295 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
296 void btrfs_reclaim_bgs_work(struct work_struct *work);
297 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
298 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
299 int btrfs_read_block_groups(struct btrfs_fs_info *info);
300 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
301 						 u64 bytes_used, u64 type,
302 						 u64 chunk_offset, u64 size);
303 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
304 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
305 			     bool do_chunk_alloc);
306 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
307 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
308 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
309 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
310 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
311 			     u64 bytenr, u64 num_bytes, bool alloc);
312 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
313 			     u64 ram_bytes, u64 num_bytes, int delalloc);
314 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
315 			       u64 num_bytes, int delalloc);
316 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
317 		      enum btrfs_chunk_alloc_enum force);
318 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
319 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
320 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
321 				  bool is_item_insertion);
322 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
323 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
324 int btrfs_free_block_groups(struct btrfs_fs_info *info);
325 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
326 		       struct block_device *bdev, u64 physical, u64 **logical,
327 		       int *naddrs, int *stripe_len);
328 
329 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
330 {
331 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
332 }
333 
334 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
335 {
336 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
337 }
338 
339 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
340 {
341 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
342 }
343 
344 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
345 {
346 	smp_mb();
347 	return cache->cached == BTRFS_CACHE_FINISHED ||
348 		cache->cached == BTRFS_CACHE_ERROR;
349 }
350 
351 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
352 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
353 
354 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
355 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
356 
357 #endif /* BTRFS_BLOCK_GROUP_H */
358