xref: /openbmc/linux/fs/btrfs/block-group.h (revision 240e6d25)
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 enum btrfs_chunk_alloc_enum {
40 	CHUNK_ALLOC_NO_FORCE,
41 	CHUNK_ALLOC_LIMITED,
42 	CHUNK_ALLOC_FORCE,
43 };
44 
45 struct btrfs_caching_control {
46 	struct list_head list;
47 	struct mutex mutex;
48 	wait_queue_head_t wait;
49 	struct btrfs_work work;
50 	struct btrfs_block_group *block_group;
51 	u64 progress;
52 	refcount_t count;
53 };
54 
55 /* Once caching_thread() finds this much free space, it will wake up waiters. */
56 #define CACHING_CTL_WAKE_UP SZ_2M
57 
58 struct btrfs_block_group {
59 	struct btrfs_fs_info *fs_info;
60 	struct inode *inode;
61 	spinlock_t lock;
62 	u64 start;
63 	u64 length;
64 	u64 pinned;
65 	u64 reserved;
66 	u64 used;
67 	u64 delalloc_bytes;
68 	u64 bytes_super;
69 	u64 flags;
70 	u64 cache_generation;
71 
72 	/*
73 	 * If the free space extent count exceeds this number, convert the block
74 	 * group to bitmaps.
75 	 */
76 	u32 bitmap_high_thresh;
77 
78 	/*
79 	 * If the free space extent count drops below this number, convert the
80 	 * block group back to extents.
81 	 */
82 	u32 bitmap_low_thresh;
83 
84 	/*
85 	 * It is just used for the delayed data space allocation because
86 	 * only the data space allocation and the relative metadata update
87 	 * can be done cross the transaction.
88 	 */
89 	struct rw_semaphore data_rwsem;
90 
91 	/* For raid56, this is a full stripe, without parity */
92 	unsigned long full_stripe_len;
93 
94 	unsigned int ro;
95 	unsigned int iref:1;
96 	unsigned int has_caching_ctl:1;
97 	unsigned int removed:1;
98 	unsigned int to_copy:1;
99 	unsigned int relocating_repair:1;
100 	unsigned int chunk_item_inserted:1;
101 	unsigned int zone_is_active:1;
102 
103 	int disk_cache_state;
104 
105 	/* Cache tracking stuff */
106 	int cached;
107 	struct btrfs_caching_control *caching_ctl;
108 	u64 last_byte_to_unpin;
109 
110 	struct btrfs_space_info *space_info;
111 
112 	/* Free space cache stuff */
113 	struct btrfs_free_space_ctl *free_space_ctl;
114 
115 	/* Block group cache stuff */
116 	struct rb_node cache_node;
117 
118 	/* For block groups in the same raid type */
119 	struct list_head list;
120 
121 	refcount_t refs;
122 
123 	/*
124 	 * List of struct btrfs_free_clusters for this block group.
125 	 * Today it will only have one thing on it, but that may change
126 	 */
127 	struct list_head cluster_list;
128 
129 	/* For delayed block group creation or deletion of empty block groups */
130 	struct list_head bg_list;
131 
132 	/* For read-only block groups */
133 	struct list_head ro_list;
134 
135 	/*
136 	 * When non-zero it means the block group's logical address and its
137 	 * device extents can not be reused for future block group allocations
138 	 * until the counter goes down to 0. This is to prevent them from being
139 	 * reused while some task is still using the block group after it was
140 	 * deleted - we want to make sure they can only be reused for new block
141 	 * groups after that task is done with the deleted block group.
142 	 */
143 	atomic_t frozen;
144 
145 	/* For discard operations */
146 	struct list_head discard_list;
147 	int discard_index;
148 	u64 discard_eligible_time;
149 	u64 discard_cursor;
150 	enum btrfs_discard_state discard_state;
151 
152 	/* For dirty block groups */
153 	struct list_head dirty_list;
154 	struct list_head io_list;
155 
156 	struct btrfs_io_ctl io_ctl;
157 
158 	/*
159 	 * Incremented when doing extent allocations and holding a read lock
160 	 * on the space_info's groups_sem semaphore.
161 	 * Decremented when an ordered extent that represents an IO against this
162 	 * block group's range is created (after it's added to its inode's
163 	 * root's list of ordered extents) or immediately after the allocation
164 	 * if it's a metadata extent or fallocate extent (for these cases we
165 	 * don't create ordered extents).
166 	 */
167 	atomic_t reservations;
168 
169 	/*
170 	 * Incremented while holding the spinlock *lock* by a task checking if
171 	 * it can perform a nocow write (incremented if the value for the *ro*
172 	 * field is 0). Decremented by such tasks once they create an ordered
173 	 * extent or before that if some error happens before reaching that step.
174 	 * This is to prevent races between block group relocation and nocow
175 	 * writes through direct IO.
176 	 */
177 	atomic_t nocow_writers;
178 
179 	/* Lock for free space tree operations. */
180 	struct mutex free_space_lock;
181 
182 	/*
183 	 * Does the block group need to be added to the free space tree?
184 	 * Protected by free_space_lock.
185 	 */
186 	int needs_free_space;
187 
188 	/* Flag indicating this block group is placed on a sequential zone */
189 	bool seq_zone;
190 
191 	/*
192 	 * Number of extents in this block group used for swap files.
193 	 * All accesses protected by the spinlock 'lock'.
194 	 */
195 	int swap_extents;
196 
197 	/* Record locked full stripes for RAID5/6 block group */
198 	struct btrfs_full_stripe_locks_tree full_stripe_locks_root;
199 
200 	/*
201 	 * Allocation offset for the block group to implement sequential
202 	 * allocation. This is used only on a zoned filesystem.
203 	 */
204 	u64 alloc_offset;
205 	u64 zone_unusable;
206 	u64 zone_capacity;
207 	u64 meta_write_pointer;
208 	struct map_lookup *physical_map;
209 	struct list_head active_bg_list;
210 };
211 
212 static inline u64 btrfs_block_group_end(struct btrfs_block_group *block_group)
213 {
214 	return (block_group->start + block_group->length);
215 }
216 
217 static inline bool btrfs_is_block_group_data_only(
218 					struct btrfs_block_group *block_group)
219 {
220 	/*
221 	 * In mixed mode the fragmentation is expected to be high, lowering the
222 	 * efficiency, so only proper data block groups are considered.
223 	 */
224 	return (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
225 	       !(block_group->flags & BTRFS_BLOCK_GROUP_METADATA);
226 }
227 
228 #ifdef CONFIG_BTRFS_DEBUG
229 static inline int btrfs_should_fragment_free_space(
230 		struct btrfs_block_group *block_group)
231 {
232 	struct btrfs_fs_info *fs_info = block_group->fs_info;
233 
234 	return (btrfs_test_opt(fs_info, FRAGMENT_METADATA) &&
235 		block_group->flags & BTRFS_BLOCK_GROUP_METADATA) ||
236 	       (btrfs_test_opt(fs_info, FRAGMENT_DATA) &&
237 		block_group->flags &  BTRFS_BLOCK_GROUP_DATA);
238 }
239 #endif
240 
241 struct btrfs_block_group *btrfs_lookup_first_block_group(
242 		struct btrfs_fs_info *info, u64 bytenr);
243 struct btrfs_block_group *btrfs_lookup_block_group(
244 		struct btrfs_fs_info *info, u64 bytenr);
245 struct btrfs_block_group *btrfs_next_block_group(
246 		struct btrfs_block_group *cache);
247 void btrfs_get_block_group(struct btrfs_block_group *cache);
248 void btrfs_put_block_group(struct btrfs_block_group *cache);
249 void btrfs_dec_block_group_reservations(struct btrfs_fs_info *fs_info,
250 					const u64 start);
251 void btrfs_wait_block_group_reservations(struct btrfs_block_group *bg);
252 bool btrfs_inc_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
253 void btrfs_dec_nocow_writers(struct btrfs_fs_info *fs_info, u64 bytenr);
254 void btrfs_wait_nocow_writers(struct btrfs_block_group *bg);
255 void btrfs_wait_block_group_cache_progress(struct btrfs_block_group *cache,
256 				           u64 num_bytes);
257 int btrfs_wait_block_group_cache_done(struct btrfs_block_group *cache);
258 int btrfs_cache_block_group(struct btrfs_block_group *cache,
259 			    int load_cache_only);
260 void btrfs_put_caching_control(struct btrfs_caching_control *ctl);
261 struct btrfs_caching_control *btrfs_get_caching_control(
262 		struct btrfs_block_group *cache);
263 u64 add_new_free_space(struct btrfs_block_group *block_group,
264 		       u64 start, u64 end);
265 struct btrfs_trans_handle *btrfs_start_trans_remove_block_group(
266 				struct btrfs_fs_info *fs_info,
267 				const u64 chunk_offset);
268 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
269 			     u64 group_start, struct extent_map *em);
270 void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info);
271 void btrfs_mark_bg_unused(struct btrfs_block_group *bg);
272 void btrfs_reclaim_bgs_work(struct work_struct *work);
273 void btrfs_reclaim_bgs(struct btrfs_fs_info *fs_info);
274 void btrfs_mark_bg_to_reclaim(struct btrfs_block_group *bg);
275 int btrfs_read_block_groups(struct btrfs_fs_info *info);
276 struct btrfs_block_group *btrfs_make_block_group(struct btrfs_trans_handle *trans,
277 						 u64 bytes_used, u64 type,
278 						 u64 chunk_offset, u64 size);
279 void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans);
280 int btrfs_inc_block_group_ro(struct btrfs_block_group *cache,
281 			     bool do_chunk_alloc);
282 void btrfs_dec_block_group_ro(struct btrfs_block_group *cache);
283 int btrfs_start_dirty_block_groups(struct btrfs_trans_handle *trans);
284 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans);
285 int btrfs_setup_space_cache(struct btrfs_trans_handle *trans);
286 int btrfs_update_block_group(struct btrfs_trans_handle *trans,
287 			     u64 bytenr, u64 num_bytes, bool alloc);
288 int btrfs_add_reserved_bytes(struct btrfs_block_group *cache,
289 			     u64 ram_bytes, u64 num_bytes, int delalloc);
290 void btrfs_free_reserved_bytes(struct btrfs_block_group *cache,
291 			       u64 num_bytes, int delalloc);
292 int btrfs_chunk_alloc(struct btrfs_trans_handle *trans, u64 flags,
293 		      enum btrfs_chunk_alloc_enum force);
294 int btrfs_force_chunk_alloc(struct btrfs_trans_handle *trans, u64 type);
295 void check_system_chunk(struct btrfs_trans_handle *trans, const u64 type);
296 void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
297 				  bool is_item_insertion);
298 u64 btrfs_get_alloc_profile(struct btrfs_fs_info *fs_info, u64 orig_flags);
299 void btrfs_put_block_group_cache(struct btrfs_fs_info *info);
300 int btrfs_free_block_groups(struct btrfs_fs_info *info);
301 void btrfs_wait_space_cache_v1_finished(struct btrfs_block_group *cache,
302 				struct btrfs_caching_control *caching_ctl);
303 int btrfs_rmap_block(struct btrfs_fs_info *fs_info, u64 chunk_start,
304 		       struct block_device *bdev, u64 physical, u64 **logical,
305 		       int *naddrs, int *stripe_len);
306 
307 static inline u64 btrfs_data_alloc_profile(struct btrfs_fs_info *fs_info)
308 {
309 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_DATA);
310 }
311 
312 static inline u64 btrfs_metadata_alloc_profile(struct btrfs_fs_info *fs_info)
313 {
314 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_METADATA);
315 }
316 
317 static inline u64 btrfs_system_alloc_profile(struct btrfs_fs_info *fs_info)
318 {
319 	return btrfs_get_alloc_profile(fs_info, BTRFS_BLOCK_GROUP_SYSTEM);
320 }
321 
322 static inline int btrfs_block_group_done(struct btrfs_block_group *cache)
323 {
324 	smp_mb();
325 	return cache->cached == BTRFS_CACHE_FINISHED ||
326 		cache->cached == BTRFS_CACHE_ERROR;
327 }
328 
329 void btrfs_freeze_block_group(struct btrfs_block_group *cache);
330 void btrfs_unfreeze_block_group(struct btrfs_block_group *cache);
331 
332 bool btrfs_inc_block_group_swap_extents(struct btrfs_block_group *bg);
333 void btrfs_dec_block_group_swap_extents(struct btrfs_block_group *bg, int amount);
334 
335 #endif /* BTRFS_BLOCK_GROUP_H */
336